Information about increasing mean annual and seasonal air temperature is relevant in the determination of suitable species range for terrestrial, freshwater and intertidal species (Thomas et al., 2004; Elith et al., 2010; Hincapie and Caicedo, 2013; Cooper, 2014; Krist et al., 2014; Lindner et al., 2014; Saintilan et al., 2014; Lenoir and Svenning, 2015; Myers-Smith et al., 2015; Urban, 2015; Thorne et al., 2017). Ocean ecosystems are affected by the ocean temperature CID (described in Section 12.3.6.1). Species redistribution and extinction studies also need information about climate velocity, a comparison of the pace of warming to geographical temperature gradients that indicates the rate at which a species would have to move to maintain its climatological temperature (Thomas et al., 2004; Loarie et al., 2009; Dobrowski et al., 2013; Burrows et al., 2014; Dobrowski and Parks, 2016; Sittaro et al., 2017) with some studies incorporating additional variables beyond temperature (Hamann et al., 2015). Many freshwater ecosystems are strongly constrained by stream and lake temperatures (Scheurer et al., 2009; Comte and Grenouillet, 2013; Contador et al., 2014; Knouft and Ficklin, 2017). Warmer and more stratified lake temperatures are more conducive to cyanobacteria blooms with implications for ecosystem health and water resource quality (Whitehead et al., 2009; Moss et al., 2011; Jones and Brett, 2014; Chapra et al., 2017; Shatwell et al., 2019). Consideration of nighttime and daytime temperature trends also elucidates different biophysical effects on vegetation (Peng et al., 2013). Changes in the seasonal timing caused by warming trends are critical to species ranges and ecosystem function (Pearce-Higgins et al., 2015; Hughes et al., 2017b), and indices that characterize the onset of spring shed light on plant emergence and development (Ault et al., 2015).

Mean air temperature dictates many aspects of crop cultivation, livestock production, agroforestry and output from freshwater aquaculture and fisheries, as well as the potential for food contamination. Mean warming alters suitable cultivation zones for crop species (Bragança et al., 2016; Gendron St-Marseille et al., 2019; IPCC, 2019c) and tree species (Hanewinkel et al., 2013; Fei et al., 2017). Crop and ecosystem service productivity often responds directly to mean temperatures, although this is dependent on farming systems (Bassu et al., 2014; Challinor et al., 2014; Lobell and Tebaldi, 2014; Rosenzweig et al., 2014; Asseng et al., 2015; Li et al., 2015; Fleisher et al., 2017; Zhao et al., 2017; Smith and Fazil, 2019). Many studies relate plant development (phenology), insect generation cycles and pest outbreaks to growing degree days, an aggregation of daily thermal units above a threshold (e.g., T_mean>5°C) that accelerates with warmer conditions (Hof and Svahlin, 2016; Ruosteenoja et al., 2016; Tripathi et al., 2016). Many plants respond to changes in nighttime temperatures that affect respiration and transpiration rates (Narayanan et al., 2015; X. Chen et al., 2019), and warming of the soil column is also relevant to determine plant sprouting (Grotjahn, 2021). A number of indices have been developed to represent the length of the viable local growing season, including a count of days where T_max>5°C (Mueller et al., 2015) or the period between a year’s first and last set of five consecutive days with a weighted T_mean≥10°C (G. Li et al., 2018). Warmer conditions and altered seasonality modify the range and metabolism of some pollinators, pests, diseases and weeds (Wolfe et al., 2008; Bebber, 2015; Aljaryian and Kumar, 2016; IPBES, 2016; Ramesh et al., 2017; Deutsch et al., 2018; Nyangiwe et al., 2018) and may reduce the effectiveness of winter storage for farmers and caching species (Sutton et al., 2016).

Warming raises accumulated seasonal heat indices used in livestock production, especially when humidity is high (Key et al., 2014; Lallo et al., 2018), determines aquaculture suitability and is important for wild fish species migration (Tripathi et al., 2016; Brander et al., 2017). Agricultural planners may also calculate how overall warming trends alter the accumulation of vernalization units or chilling hours for agricultural or horticultural crops (often accumulated temperature deficit below a given daily or hourly threshold; Dennis and Peacock, 2009; Luedeling, 2012; Tripathi et al., 2016; Grotjahn, 2021). Warming in the post-harvest is also important for the determination of spoilage and waste (Stathers et al., 2013) as well as food-borne diseases (Kovats et al., 2004; Mbow et al., 2019).

Warming affects road degradation rates (Chinowsky and Arndt, 2012; Espinet et al., 2016) and warming rates inform designs for long-term energy efficiency of buildings (Kalvelage et al., 2014). Mean temperature drives seasonal energy demand, often expressed using winter heating degree days (the accumulated deficit of daily temperatures below a ‘comfortable’ indoor temperature, e.g., 15.5°C) and summer cooling degree days (the accumulated excess of temperature above a ‘comfortable’ level, e.g., 18°C; Spinoni et al., 2015; Arnell et al., 2019). Energy resources may also need information on warming trends to determine suitable zones and overall productivity for biofuels and solar panels, the efficiency of which decreases with higher temperatures (Schaeffer et al., 2012; Wild et al., 2015; Solaun and Cerdá, 2019).

Health impacts and risk studies compare seasonal temperature conditions to limiting thresholds to understand range shifts and incubation rates for pathogens, disease vectors and zoonotic hosts (e.g., mosquitoes, ticks; Caminade et al., 2012, 2014; Eisen and Moore, 2013; Lima et al., 2016; Ogden, 2017; Monaghan et al., 2018) and warming of surface ocean and lake waters conducive to bacterial outbreaks (Baker-Austin et al., 2013; Jacobs et al., 2015; Vezzulli et al., 2015). Warmer conditions can also affect tourism (Kovács et al., 2017) and impact human health by lengthening the allergy season and increasing pollen concentration (Hamaoui-Laguel et al., 2015; Kinney et al., 2015a; Lake et al., 2017; Upperman et al., 2017; Sapkota et al., 2019; Ziska et al., 2019).

Impacts and risk assessments utilize a large variety of indices and approaches tailored to evaluate heat impacts on human health (Sanderson et al., 2017; C. Gao et al., 2018; McGregor and Vanos, 2018; Staiger et al., 2019; J. Zhu et al., 2019; Schwingshackl et al., 2021). A mixture of simple and complex heat stress indices often combine extreme temperatures and high humidity to capture human health challenges (Aström et al., 2013; Chow et al., 2016; Dahl et al., 2017a; Im et al., 2017; Coffel et al., 2018; J. Li et al., 2018; Vanos et al., 2020). Different optimum temperatures and extreme heat thresholds based on local distributions are needed to reflect acclimation of different locations and populations (Hajat et al., 2014; WHO, 2014; Kinney et al., 2015b; Russo et al., 2015; Petitti et al., 2016; Dosio, 2017; Cheng et al., 2018; Lay et al., 2018; Schwingshackl et al., 2021). Hot and humid heat episodes can be deadly (Mora et al., 2017), are associated with elevated hospital intake (Goldie et al., 2017) and lower safety and productivity of outdoor labourers (Dunne et al., 2013; Graff Zivin and Neidell, 2014; Kjellstrom et al., 2016; Pal and Eltahir, 2016; Y. Zhao et al., 2016; Mora et al., 2017; Watts et al., 2018; Orlov et al., 2019). Elevated nighttime temperatures prevent the human body from experiencing relief from heat stress (Zhang et al., 2012) and can be tracked over extended periods of sequential day and night heat extremes (Murage et al., 2017; Mukherjee and Mishra, 2018). Extreme heat also exacerbates asthma, respiratory difficulties and response to airborne allergens such as hay fever (Upperman et al., 2017). Extreme heat affects outdoor exercise such as the use of bike-share facilities (Heaney et al., 2019; Vanos et al., 2020). Large-scale recreational and sporting events such as marathons and tennis tournaments monitor heat extremes when determining the viability of host cities (Smith et al., 2016, 2018).

Short-term exposure of crops to temperatures beyond a critical temperature threshold can lead to lower yields and above a limiting temperature threshold, crops may fail altogether (Schlenker and Roberts, 2009; Lobell et al., 2012, 2013; Gourdji et al., 2013; Deryng et al., 2014; Schauberger et al., 2017; Tesfaye et al., 2017; Vogel et al., 2019). The exact level of these thresholds depends on species, cultivar and farm management (Hatfield and Prueger, 2015; Hatfield et al., 2015; Bisbis et al., 2018; Grotjahn, 2021). The timing of heatwaves is particularly important, as extreme heat is more damaging during critical phenological stages (Teixeira et al., 2013; Eyshi Rezaei et al., 2015; Fontana et al., 2015; B. Wang et al., 2017; Mäkinen et al., 2018). Extreme canopy temperatures, rather than 2 m air temperatures, may be a more robust biophysical indicator of heat impacts on crop production (Siebert et al., 2017). Heat stress indices based upon temperature and humidity determine livestock productivity as well as conception and mortality rates (Key et al., 2014; Dash et al., 2016; Pragna et al., 2016; Rojas-Downing et al., 2017).

Heat extremes factor in mortality, morbidity and the range of some thermally sensitive ecosystem species (Smith and Nagy, 2015; Ratnayake et al., 2019; Thomsen et al., 2019). Combined heat and drought stress can reduce forest and grassland primary productivity (Ciais et al., 2005; De Boeck et al., 2018) and even cause tree mortality at higher extremes (Teskey et al., 2015).

Extreme heat events raise temperatures in buildings and cities already warmed by the urban heat island effect (Gaffin et al., 2012; Oleson et al., 2018; Zhao et al., 2018; Mauree et al., 2019; Box 10.3) and can induce disruptions in critical infrastructure networks (Chapman et al., 2013). Heat affects transportation infrastructure by warping roads and airport runways (Chinowsky and Arndt, 2012) or buckling railways (Dobney et al., 2010; Dépoues, 2017; Chinowsky et al., 2019), and high temperatures reduce air density leading to aircraft take-off weight restrictions (Coffel et al., 2017; Palko, 2017; T. Zhou et al., 2018). Heat extremes increase peak cooling demand and challenge transmission and transformer capacity (Sathaye et al., 2013; Russo et al., 2016; Craig et al., 2018; X. Gao et al., 2018) and may cause transmission lines to sag or fail (Gupta et al., 2012). Thermal and nuclear electricity plants may be challenged when using warmer river waters for cooling or when mixing waste waters back into waterways without causing ecosystem impacts (Kopytko and Perkins, 2011; van Vliet et al., 2016; Tobin et al., 2018). Extreme temperature can also reduce photovoltaic panel efficiency (Jerez et al., 2015).

The magnitude and timing (relative to developmental stages) of cold extremes (such as the typical coldest day of the year) set limits in the range of species habitat for ecosystems as well as for agricultural and forest pests (Osland et al., 2013; Cavanaugh et al., 2014; Parker and Abatzoglou, 2016; Brunner et al., 2018; Unterberger et al., 2018). Cold air outbreaks can lead to chilling injuries for crops (even above 0°C) and may kill outdoor livestock (particularly young animals; Mader et al., 2010; Liu et al., 2013; Grotjahn, 2021), but are often necessary for crop chill requirements (Dennis and Peacock, 2009).

Increases in human mortality can occur on exceptionally cold days (e.g., <1st percentile of temperatures in winter) although thresholds and human-perceived temperatures linked to wind speed (i.e., ‘wind chill’) vary geographically due to acclimatization (Li et al., 2013; Gao et al., 2015; J. Li et al., 2018; J. Zhu et al., 2019). The timing of ‘unseasonal’ cold spells also affect human health (Kinney et al., 2015b). Extreme cold can increase heat and electricity demand (Stuivenvolt-Allen and Wang, 2019), cause water pipes to burst, and mechanically alter roads, railroads and buildings (Underwood et al., 2017).

Frost (T_min<0°C) is a natural and fundamental aspect of many ecosystems, with more extreme conditions defined as ice (or icing) days (T_max<0°C) (L.A. Vincent et al., 2018). Agricultural systems planning (e.g., planting calendars, seed selection or the opportunity to double-crop) requires information about the start and end of the frost-free season (Wypych et al., 2017; Wolfe et al., 2018). Crops and wild plants can be directly damaged by frost, but hard or killing frosts (at a threshold several degrees below freezing) can kill crops or lower harvest quality depending on duration (which relates to soil temperature penetration) and plant developmental stage (Crimp et al., 2016a; Cradock-Henry, 2017; G. Li et al., 2018; Mäkinen et al., 2018; Grotjahn, 2021). Earlier disappearance of snow cover reduces natural insulation that protects plants and burrowing animals from hard frost damages (Trnka et al., 2014; Mäkinen et al., 2018). In some cases an early season warm spell may reduce plant hardiness or induce fruit tree flowering that exposes plants to devastating subsequent frost impacts (Hufkens et al., 2012; Hatfield et al., 2014; Tripathi et al., 2016; Brunner et al., 2018; DeGaetano, 2018; Unterberger et al., 2018; Wolfe et al., 2018). Shifts in the seasonality of frozen soils also affect groundwater recharge and surface streamflow for water resource applications, particularly when peak precipitation is shifted to a season that no longer has frozen soils (Jyrkama and Sykes, 2007).

Regional information about the spring and autumn seasonal periods in which freeze-thaw cycles are common (such as the dates of first spring thaw and last spring frost, or the number of days where T_max>0°C and T_min<0°C) are particularly useful in estimating the rate of potential road and building damage or determining seasonal truck weight restrictions (Kvande and Lisø, 2009; Chinowsky and Arndt, 2012; Palko, 2017; Daniel et al., 2018). The altitude of the freezing level also identifies portions of mountain slopes where freeze/thaw transitions or changes in snowpack condition can influence landslide and snow avalanche hazards (Coe et al., 2018). The geographical distribution of frost is also a determining factor in the range of vectors for human diseases such as malaria (X. Zhao et al., 2016; Smith et al., 2020).

Figure 12.3 illustrates how successive heat and cold hazards can potentially affect important natural and human systems, with climatic pressures reaching new sectoral assets or becoming increasingly severe as conditions become more extreme. While the precise value of any CID threshold may depend strongly on local environmental and system characteristics, there are common patterns and interdependencies in the types of thresholds encountered. Changes in the regional profile of CIDs can thus substantially alter threshold exceedance likelihoods.

Figure 12.3 | Conceptual illustration of representative climatic impact-driver thresholds showing how graduating thresholds affect successive sectoral assets and lead to potentially more acute hazards as conditions become more extreme (exact values are not shown as these must be tailored to reflect diverse vulnerabilities of regional assets). Representative threshold definitions (T = instantaneous temperature; T= mean temperature): Cities and Infrastructures: T_trans= temperature at which energy transmission lines efficiency reduced; T_aircraft = temperature at which aircraft become weight-restricted for takeoff; T_hotroads= temperature above which roads begin to warp; T_stream= temperature at which streams are not capable of adequately cooling thermal plants; CDD_min= minimum temperature for calculating cooling degree days; HDD_max= maximum temperature for calculating heating degree days; T_ice= temperature at which ice threatens transportation; T_permafrost = mean seasonal temperature above which permafrost thaws at critical depths; T_coldroads= temperature below which road asphalt performance suffers. Health: T_deadly= temperature above which prolonged exposure may be deadly (often combined with humidity for heat indices); T_severe= temperature above which prolonged exposure may cause elevated morbidity; T_blooms= mean temperature for harmful algal or cyanobacteria blooms; T_danger= level of dangerous cold temperatures (often combined with wind for chill indices); T_overwinter= temperature below which disease vector species cannot survive winter. Ecosystems (CID indices for air and ocean temperature): T_hotlim and T_coldlim= limiting hot and cold temperatures for a given species range; T_frost = frost threshold; T_max and T_min= maximum and minimum suitable annual mean temperatures for a given species; T_crit = critical temperature above which a given species is stressed. Agriculture: T_hotlim= temperature above which a crop or livestock species dies; T_hotpest = maximum (or ‘lethal’) temperature above which an agricultural pest/disease/weed cannot survive; T_crit = temperature at which productivity for a given crop is depressed; T_opt = optimal mean temperature for a given plant’s productivity; GDD_min= threshold temperature for growing degree days determining plant development; T_chill= temperature below which chilling units are accumulated; T_frost = temperature below which frost occurs; T_hfrost = temperature below which a hard frost threatens crops or livestock; T_coldpest = minimum winter temperature below which a given agricultural pest cannot survive; T_coldlim= minimum temperature below which a given crop cannot survive.

Open figure

Changes in mean precipitation alter total water resources and long-term surface, snowpack and groundwater reservoirs (Schewe et al., 2014). Annual and seasonal wet trends can alter the suitable geographic range of species, with implications for biodiversity and vector-borne diseases (Knouft and Ficklin, 2017; Smith et al., 2020). The rate at which higher total streamflow increases river erosion and changes sediment loading is relevant for fish breeding (Scheurer et al., 2009), the location of riverine salt fronts that affect coastal agriculture and ecosystems (Chun et al., 2018; Vu et al., 2018), coastal freshwater stratification (Baker-Austin et al., 2013; Bell et al., 2013), and the accretion of sediment in estuaries and beaches (Syvitski and Milliman, 2007). Wetter conditions may shift tourist appeal (Kovács et al., 2017) and alter the pace of degradation for paved and especially unpaved roads (Chinowsky and Arndt, 2012).

Many agricultural systems require minimum rainfall totals or rely upon irrigation (Mbow et al., 2019). The length of the wet season helps determine the potential for multiple cropping seasons, but inconsistency of wet season arrival times poses challenges for farm management (Waha et al., 2020). Wetter growing season conditions increase the chance of waterlogging, which can delay planting or damage planted seeds (Rosenzweig et al., 2002; Ben-Ari et al., 2018; Mäkinen et al., 2018; Wolfe et al., 2018; Kolberg et al., 2019; Grotjahn, 2021). Tomasek et al. (2017) calculated ‘workable days’ for agricultural machinery around planting and harvest time set in part by limits in soil moisture saturation below which farmers can utilize critical machinery with less rutting or soil compaction. Wetter conditions may also increase canopy moisture that is conducive to crop pathogens (Garrett et al., 2006; Kilroy, 2015; Grotjahn, 2021).

A large variety of climate indices and models are utilized to understand how river flooding affects both natural or built environments with highly variable hazard thresholds, given unique local topography and engineered defences such as dams and polders (Arnell and Gosling, 2016; Ekström et al., 2018). Key transportation routes, built infrastructure and agricultural lands are threatened when floods exceed design standards commonly based around flood magnitudes of a given historic return period (e.g., 1-in-100-year flood event), an annual exceedance probability or precipitation intensity-duration-frequency relationships with key indices (e.g., 10-day cumulative precipitation) related to catchment size and properties (Hirabayashi et al., 2013; Arnell and Lloyd-Hughes, 2014; Kundzewicz et al., 2014; Arnell and Gosling, 2016; Dikanski et al., 2016; Gosling and Arnell, 2016; Forzieri et al., 2017; Fluixá-Sanmartín et al., 2018; Koks et al., 2019). Floods and high-flow events can scour river beds and elevate silt loads, reducing water quality and accelerating deposition in estuaries and reservoirs (Khan et al., 2018; Parasiewicz et al., 2019). Floods can knock down, drown or wash away crops and livestock, and partially submerged plants can have yield reduction depending on water turbidity and their development stage (Ruane et al., 2013; Shrestha et al., 2019). Basin snowpack properties may also be important during heavy rain events, as rain-on-snow events can lead to rapid acceleration of flood stages that threaten wildlife and society (Hansen et al., 2014).

Heavy downpours can lead to pluvial flooding in cities, roadways, farmland, subway tunnels and buildings (particularly those with basements; Grahn and Nyberg, 2017; Palko, 2017; Pregnolato et al., 2017; Orr et al., 2018). Heavy precipitation may overwhelm city transportation and storm water drainage systems, which are typically designed using intensity-duration-frequency information such as the return periods for 1-, 6- or 24-hour rainfall totals (Kermanshah et al., 2017; Depietri and McPhearson, 2018; Rosenzweig et al., 2018; Courty et al., 2019). Heavy rain events can directly cause leaf loss and damage, or knock over crops, also driving pollutant entrainment and erosion hazards in terrestrial ecosystems and farmland, with downstream ramifications for water quality (Hatfield et al., 2014; Segura et al., 2014; Li and Fang, 2016; Chhetri et al., 2019). The proportion of total precipitation that falls in heavy events also affects the percentage that is retained in the soil column, altering groundwater recharge and deep soil moisture content for agricultural use (Fishman, 2016; Lesk et al., 2020).

Landslides, mudslides, rockfalls and other mass movements can lead to fatalities, destroy infrastructure and housing stock, and block critical transportation routes. Climate models cannot resolve these complex slope failure processes (nor triggering mechanisms such as earthquakes), so most studies rely on proxies or conditions conducive to slope failure (Gariano and Guzzetti, 2016; Ho et al., 2017). Common indices include precipitation intensity-duration thresholds (Brunetti et al., 2010; Khan et al., 2012; Melchiorre and Frattini, 2012) and thresholds related to antecedent wet periods and extreme rainfall intensities (Alvioli et al., 2018; Monsieurs et al., 2019). Landslides and rockfalls may also be exacerbated by permafrost thaw and receding glaciers in polar and mountain areas (Cook et al., 2016; Haeberli et al., 2017; Patton et al., 2019).

Aridity indices may track long-term changes in precipitation, evapotranspiration demand, surface water, groundwater or soil moisture (Sherwood and Fu, 2014; Herrera-Pantoja and Hiscock, 2015; B.I. Cook et al., 2020). Changes in soil moisture and surface water can shift the rate of carbon uptake by ecosystems (Humphrey et al., 2018) and alter suitable climate zones for wild species and agricultural cultivation (Feng and Fu, 2013; Garcia et al., 2014; Huang et al., 2016a; Schlaepfer et al., 2017; Fatemi et al., 2018; IPCC, 2019c) as well as the prevalence of related pests and pathogen-carrying vectors (Paritsis and Veblen, 2011; Smith et al., 2020). Water table depth, in relation to rooting depth, is also important for farms and forests under dry conditions (Feng et al., 2006). A reduction in water availability (via aridity or hydrological drought) challenges water supplies needed for for municipal, industrial, agriculture and hydropower use (Schaeffer et al., 2012; Arnell and Lloyd-Hughes, 2014; Schewe et al., 2014; Gosling and Arnell, 2016; van Vliet et al., 2016).

Water managers often utilize a variety of hydrological drought indices and hydrological models to characterize water resources, low flow conditions and the potential for irrigation (Wanders and Wada, 2015; Mukherjee et al., 2018). Low flow volume and intermittency thresholds can indicate reductions in dissolved oxygen, more concentrated pollutants, and higher stream temperatures relevant for ecosystems, water resource quality and thermal power plant cooling (Feeley et al., 2008; Döll and Schmied, 2012; Schaeffer et al., 2012; Prudhomme et al., 2014; van Vliet et al., 2016). Low water levels may also restrict waterway navigation for commerce and recreation (Forzieri et al., 2018).

Agricultural and ecological drought indices relate to the ability of plants to meet growth and transpiration needs (Table 11.3; Zargar et al., 2011; Lobell et al., 2015; Pedro-Monzonís et al., 2015; Bachmair et al., 2016; Wehner et al., 2017; Naumann et al., 2018) and the timing and duration of droughts can lead to substantially different impacts (Peña-Gallardo et al., 2019). Drought stress for agriculture and ecosystems is difficult to directly observe, and therefore scientists use a variety of drought indices (Table 11.3), proxy information about changes in precipitation supply and reference evapotranspiration demand, the ratio of actual/potential evapotranspiration or a deficit in available soil water content, particularly at rooting level (Park Williams et al., 2013; Trnka et al., 2014; C.D. Allen et al., 2015; Svoboda and Fuchs, 2017; Mäkinen et al., 2018; Otkin et al., 2018). Severe water stress can lead to crop failure, in particular when droughts persist for an extended period or occur during key plant developmental stages (Hatfield et al., 2014; Jolly et al., 2015; Leng and Hall, 2019). Projections of high wind speed and low humidity (even for just a portion of the day) can also inform studies examining fruit desiccation and rice cracking (Grotjahn, 2021). Drought also raises disease infection rates for West Nile virus (Paull et al., 2017), and the alternation of dry and wet spells induces swelling and shrinkage of clay soils that can lead to sinkholes and destabilize buildings (Hadji et al., 2014).

Complex fire weather indices shed light on conditions that increase the likelihood of wildfire and shifts in the fire season (Flannigan et al., 2013; Bedia et al., 2015; Jolly et al., 2015; Harvey, 2016; Littell et al., 2016; Westerling, 2016; Abatzoglou et al., 2019), which pose particularly acute challenges for indigenous communities (Christianson and McGee, 2019). Projection of future lightning frequency provides information on an important natural triggering mechanism, particularly when coupled with long-term warming and drying trends (Romps et al., 2014; Jin et al., 2015; Veraverbeke et al., 2017). Fuel aridity metrics also help determine vegetative fuel desiccation and therefore the ignitability, flammability and spread of fires when they occur (Abatzoglou and Williams, 2016). The presence of snow cover can influence the length of the fire season and the penetration of fire danger into new portions of the Arctic tundra (Young et al., 2017; Abatzoglou et al., 2019). Data on the changing characteristics of local wind circulations like the Santa Ana in California shed light on future intensity and spread patterns for fires (Jin et al., 2015). Fires also produce smoke plumes that reduce air and water quality (via deposition), adversely affecting health, visibility and water resources both near and far downwind (Dennekamp and Abramson, 2011; McKenzie et al., 2014; Dreessen et al., 2016; Liu et al., 2016; Martin, 2016).

Changes in the speed and direction of prevailing winds can alter the profile of seed dispersal, windblown pest and disease vectors, animal activities, and dust or pollen dispersal affecting ecosystems, agriculture and human health (Reid and Gamble, 2009; Bullock et al., 2012; Hellberg and Chu, 2016; Nourani et al., 2017). Seasonal winds influence algal blooms, ecosystems and fisheries via lake mixing, ocean currents and coastal upwelling (Bakun et al., 2015; Townhill et al., 2018; Woolway et al., 2020). Changes to wind density also modify a region’s wind and wave renewable energy endowment (Schaeffer et al., 2012; Sierra et al., 2017; Craig et al., 2018; Devis et al., 2018; Tobin et al., 2018; Yalew et al., 2020). D. Li et al. (2020) and Karnauskas et al. (2018a) evaluated wind thresholds at turbine height (about 80–100 m above ground) including periods outside of cut-in (2.5–3 m s^–1) and cut-out (about 25 m s^–1) levels beyond which given turbines could not operate.

High winds associated with severe storms can destroy trees and houses, break plant stems and knock fruits, nuts and grains to the ground, with tolerance thresholds depending on crop species and developmental stage (Seidl et al., 2017; Lai, 2018; Elsner et al., 2019; Grotjahn, 2021). Severe storms particularly threaten energy infrastructure, with maximum wind speed associated with treefall and breaking of above-ground electrical transmission lines (Ward, 2013; Nik et al., 2020). The profile of heavy wind gusts is also required in the design of skyscrapers (C.-H. Wang et al., 2013) and bridges (Mondoro et al., 2018). Severe storms are difficult to simulate at the relatively coarse spatial scales of Earth system models, thus scientists often project changes by noting areas with increased convective available potential energy (CAPE) and strong low-level wind shear as these are conducive to tornado formation (Diffenbaugh et al., 2013; Tippett et al., 2016; Glazer et al., 2021).

Tropical cyclones and severe coastal storms can deliver wind, water and coastal hazards with the potential for widespread mortality and damages to cities, housing, transportation and energy infrastructure, ecosystems and agricultural lands (Burkett, 2011; NASEM, 2012; Bell et al., 2013; Wehof et al., 2014; Ward et al., 2016; Cheal et al., 2017; Godoi et al., 2018; Koks et al., 2019; Pinnegar et al., 2019). Storm planning is often tied to the Saffir –Simpson scale related to peak sustained wind speed (Izaguirre et al., 2021), with several indices focusing on storms’ overall power and energy, size and translation speed to anticipate destructive potential (Knutson et al., 2015; Wang and Toumi, 2016; Parker et al., 2018; Hassanzadeh et al., 2020).

Sand and dust storms erode soils, damage crops and induce problems for health, transportation, mechanical equipment and built infrastructure corresponding to the magnitude and duration of high winds and particulate matter concentrations (Goudie, 2014; O’Loingsigh et al., 2014; Crooks et al., 2016; Barreau et al., 2017; Bhattachan et al., 2018; Al Ameri et al., 2019; Middleton et al., 2019). Dust events may be represented as the number of dust hours per year and by particulate matter (PM) concentrations (Leys et al., 2011; Spickett et al., 2011; Hand et al., 2016). Photovoltaic panels can lose energy production efficiency with dust accumulation (Patt et al., 2013; Javed et al., 2017). It is also useful to track dust storm deposition of nutrients necessary for coral and tropical forest systems, but they may also feed algal blooms harming lake and coastal ecosystems, health and recreation (Jickells et al., 2005; Hallegraeff et al., 2014; Gabric et al., 2016). Dust storms also cause air pollution and redistribute the soil-based fungus associated with Valley fever (Barreau et al., 2017; Coopersmith et al., 2017; Tong et al., 2017; Gorris et al., 2018).

A large number of indices have been used in water resource and ecosystem studies to track changes in snow under current and future climate conditions, including measurements of the snow water equivalent at key seasonal dates, the fraction of precipitation falling as snow, the first and last days of snow cover, and cold season temperatures (Mills et al., 2013; Pierce and Cayan, 2013; Berghuijs et al., 2014; Klos et al., 2014; Musselman et al., 2017; Rhoades et al., 2018). Impact studies also examine shifts in seasonal streamflow for snow-fed river basins (Mote et al., 2005; Pederson et al., 2011; Beniston and Stoffel, 2014; Coppola et al., 2014b, 2018; Fyfe et al., 2017; Islam et al., 2017; Knouft and Ficklin, 2017) as well as the geographic extent of snow cover and the depth of frosts when snow cover’s natural insulation is absent (Scheurer et al., 2009; Millar and Stephenson, 2015). Studies examining the impact of snow changes on winter recreation and transportation have used thresholds of about 30 cm snow depth or snow water equivalent >10 cm to determine the length of the season for alpine and cross-country skiing and snowmobiling (Damm et al., 2017; Wobus et al., 2017b; Spandre et al., 2019; Steiger et al., 2019; Abegg et al., 2021). Changes in snow quality also affect recreational activities (Rutty et al., 2017), and artificial snowmaking can augment recreational snowpack depending on the number of suitable snowmaking hours (e.g., where wet bulb globe temperature (WBGT) <–2.2°C; Wobus et al., 2017b). Local detail may also be provided by tracking the seasonal rain–snow transition line across space and elevation (Berghuijs et al., 2014) (Pierce and Cayan, 2013; Berghuijs et al., 2014; Klos et al., 2014; Musselman et al., 2017).

Change in ice sheet and glacier spatial extent and surface mass balance is relevant for polar and high mountain ecosystems and downstream assets that rely on glacial water resources (J.R. Lee et al., 2017; Milner et al., 2017; Huss and Hock, 2018; Schaefli et al., 2019). The loss of glaciers reduces the thermal consistency of cold streams suitable for some freshwater species (Giersch et al., 2017), and parks and recreation areas may lose appeal as glaciers and seasonal snow cover retreat (Gonzalez et al., 2018; Wang and Zhou, 2019). Rapid glacial retreat can lead to glacial lakes and outburst floods that endanger downstream communities (Carrivick and Tweed, 2016; Cook et al., 2016; Harrison et al., 2018).

Changes in permafrost temperature, extent and active layer thickness are metrics that track how permafrost thaw below, for example, roads, airstrips, rails and building foundations in high-latitude and mountain regions may destabilize settlements and critical infrastructure (Pendakur, 2016; Derksen et al., 2018; Duvillard et al., 2019; Olsson et al., 2019; Streletskiy et al., 2019). Warmer conditions can also affect ecosystems, built infrastructure and water resources through thawing of especially ice-rich permafrost (≥20% ice content) and by thawing of ice wedges (Shiklomanov et al., 2017; Hjort et al., 2018), creation of thermokarst ponds and increased subsurface drainage for polar and high-mountain wetlands (Walvoord and Kurylyk, 2016; Farquharson et al., 2019) and the release of water pollutants such as mercury (Burkett, 2011; Schaeffer et al., 2012; Schuster et al., 2018).

Reductions in the duration of thick sea, lake and river ice influence ecosystems as well as ice fishing, hunting, dog sledding and snowmobiling, which are recreation activities for some but vital aspects of many traditional indigenous communities (Durkalec et al., 2015; AMAP, 2017; Baztan et al., 2017; Arp et al., 2018; Rokaya et al., 2018; Knoll et al., 2019; Meredith et al., 2019; Sharma et al., 2019). The seasonal extent of thin ice and iceberg density also determines the viability of shipping lanes and seasonal roads (Valsson and Ulfarsson, 2011; Pizzolato et al., 2016; AMAP, 2017; Mullan et al., 2017; Sturm et al., 2017), oil and gas exploration timing (Schaeffer et al., 2012) and the seasonality of phytoplankton blooms (Oziel et al., 2017). Sea ice is a critical aspect of some ecosystems and fisheries (Massom and Stammerjohn, 2010; Jenouvrier et al., 2014; Bindoff et al., 2019; Meredith et al., 2019). Various definitions of ‘ice free’ Arctic Ocean conditions can be tailored to represent transportation needs, including thresholds of ice coverage (<5% or <30% or <1 million km²) in September or over a four-month period (Laliberté et al., 2016; Jahn, 2018).

Heavy snowfall is a substantial concern for cities, settlements and key transportation and energy infrastructure (Ward, 2013; Palko, 2017; Janoski et al., 2018; Collins et al., 2019). Heavy snowfall can interfere with transportation (Herring et al., 2018) and cause a loss of both work and school days depending on local snow removal infrastructure. Freezing rain and ice storms can be treacherous for road and air travel (Tamerius et al., 2016), and can knock down power and telecommunication lines if ice accumulation is high (Degelia et al., 2016). Rain-on-snow events can create a solid barrier that hinders wildlife and livestock grazing that is important to indigenous communities (Forbes et al., 2016). Shifts in the frequency, seasonal timing and regions susceptible to ice storms alter risks for agriculture and infrastructure (Lambert and Hansen, 2011; Klima and Morgan, 2015; Ning and Bradley, 2015; Groisman et al., 2016).

Information on the changing frequency and size distribution of hail can help stakeholders build resilience for agriculture, vehicles, transportation infrastructure and buildings, solar panels and wild species that see critical damage at particular hail size thresholds (Dessens et al., 2007; Webb et al., 2009; Patt et al., 2013; Fiss et al., 2019). Most climate models do not directly resolve hail and therefore studies often examine proxies associated with severe mesoscale storms (Tippett et al., 2015; Prein and Holland, 2018), although some regional studies now utilize hail-resolving models (Mahoney et al., 2012; Brimelow et al., 2017).

Information about the changing frequency and seasonal timing of snow avalanches is important to assess threats to transportation routes, infrastructure, recreational skiing and people living in alpine communities (Lazar and Williams, 2008; Mock et al., 2017; Ballesteros-Cánovas et al., 2018; Hock et al., 2019). Like landslides and other mass movements, snow avalanches are not directly resolved by climate models and are thus tracked using proxy climate information describing snow avalanche susceptibility, particularly the snow water equivalent, and triggering mechanisms such as warm spells, high winds, rain-on-snow and heavy precipitation (Hock et al., 2019). The quality of snow also provides insight into avalanche hazards (Mock et al., 2017), with the seasonal altitude of wet snowpack (>0.5% liquid water by volume) particularly important in determining characteristics of potential avalanches (Castebrunet et al., 2014).

Sea level rise hazards for coastal ecosystems, infrastructure, farmland, cities and settlements in a particular region are often driven by regional changes in relative sea level (RSL) that account for land uplift or subsidence and thus represent local asset vulnerability better than global mean sea level (Box 9.1; Hallegatte et al., 2013; Hinkel et al., 2013; McInnes et al., 2016; Weatherdon et al., 2016; Brown et al., 2018; IPCC, 2019b; Rasoulkhani et al., 2020). Vertical land motion (i.e., land subsidence) caused by local fluid (gas or groundwater) extraction can also have a large influence on relative sea levels (Minderhoud et al., 2020). Several indices have been suggested to signify coastal inundation, including a threshold when the local land elevation falls below the local mean higher high water (MHHW) that is close to the ‘high tide’ level (Kulp and Strauss, 2019) or a threshold when flooding occurs about once every two weeks (Sweet and Park, 2014; Dahl et al., 2017b). RSL rise (or RSLR) can drive increased inland penetration of above-ground and subterranean salt water fronts (i.e., salinity intrusion) affecting coastal ecosystems, agriculture and water resources (Ferguson and Gleeson, 2012; Kirwan and Megonigal, 2013; Rotzoll and Fletcher, 2013; Chen et al., 2016; Colombani et al., 2016; Holding et al., 2016; Sawyer et al., 2016; Mohammed and Scholz, 2018). The rate of RSLR can determine the survival and net pressure on niche coastal ecosystems such as mangroves, tidal flats, sea grasses and coral reefs (Hubbard et al., 2008; Craft et al., 2009; Bell et al., 2013; Kirwan and Megonigal, 2013; Alongi, 2015; Ellison, 2015; Lovelock et al., 2015; Ward et al., 2016; Lee et al., 2018).

Episodic coastal flooding of coastal communities, farmland, buildings, transportation routes, industry and other infrastructure is caused by extreme total water levels (ETWL), which is the combination of RSL, tides, storm surge and high wave setup at the shoreline (Vitousek et al., 2017; Melet et al., 2018; Vousdoukas et al., 2018, 2020a; Koks et al., 2019; Kirezci et al., 2020). Coastal settlement and infrastructure design often uses coastal flooding metrics such as the ETWL frequency distribution or the 100-year average return interval storm tide (storm surge + high tide) level (McInnes et al., 2016; Mills et al., 2016; Walsh et al., 2016b; Zheng et al., 2017). The duration of floods that overtop coastal protection, due to extreme coastal water levels (ECWL), is important for port and harbour operations and coastal energy infrastructure thresholds (Bilskie et al., 2016; Camus et al., 2017). Frequent inundation by salt water can also have significant impacts on water resources, crops, aquaculture and transportation systems due to corrosion and undercutting of coastal roads, bridges and rails (Zimmerman and Faris, 2010; N. Ahmed et al., 2019; Gopalakrishnan et al., 2019).

Effective management of coastal ecosystems, cities, settlements, beaches and infrastructure requires information about coastal erosion driven by storm surge, waves and sea level rise (Dawson et al., 2009; Hinkel et al., 2013; Harley et al., 2017; Mentaschi et al., 2017). Coastal erosion is generally accompanied by shoreline retreat, which can occur as a gradual process (e.g., due to sea level rise) or as an episodic event due to storm surge and/or extreme waves, especially when combined with high tide (Ranasinghe, 2016). The most commonly used shoreline retreat index is the magnitude of shoreline retreat by a pre-determined planning horizon such as 50 or 100 years into the future. Commonly used metrics for episodic coastal erosion include the beach erosion volume due to the 100-year recurrence storm wave height, the full exceedance probability distribution of coastal erosion volume (Li et al., 2014a; Pender et al., 2015; Ranasinghe and Callaghan, 2017) and the cumulative storm energy and storm power index (Godoi et al., 2018). The destruction or overtopping of barrier islands may lead to irreversible changes in the physical system as well as in coastal ecosystems (Carrasco et al., 2016; Zinnert et al., 2019). Shoreline position change rates along inlet-interrupted coasts may also be affected by changes in river flows and fluvial sediment supply (Hinkel et al., 2013; Bamunawala et al., 2018; Ranasinghe et al., 2019). Permafrost thaw and Arctic sea ice decline also reduce natural coastal protection from wave erosion for communities and industry (Forbes, 2011; Melvin et al., 2017).

Shifts in thermal zones affect the suitability of fisheries and marine and coastal species habitat and migration routes (Hoegh-Guldberg and Bruno, 2010; Doney et al., 2012; Burrows et al., 2014; Urban, 2015; Hixson and Arts, 2016; Tripathi et al., 2016; N. Ahmed et al., 2019; Bindoff et al., 2019). Intertidal species are particularly dependent on suitable conditions for both air and sea surface temperatures (Monaco and McQuaid, 2019). The structure of ocean warming also affects the intensity of upper-ocean stratification and the timing and strength of coastal upwelling (driven also by mean wind changes), which alters the vertical transport of oxygen- and nutrient-rich waters affecting fishery and marine ecosystem productivity (D. Wang et al., 2015).

Marine heatwaves (MHW) push water temperatures above key thresholds and have been associated with coral bleaching episodes, species shifts and harmful algal blooms that can disrupt ecosystems, tourism and human health (Box 9.2; Wernberg et al., 2016; Arias-Ortiz et al., 2018; Oliver et al., 2018; Frölicher, 2019; Smale et al., 2019; Sully et al., 2019). The duration and return period of marine heatwaves provide insight into aggregate stresses on marine species, fisheries and ecosystems, with various indices gauging cumulative intensity or the number of days, weeks or months exceeding critical thresholds (Frieler et al., 2013; Frölicher et al., 2018; Hughes et al., 2018b; Cheung and Frölicher, 2020). Hobday et al. (2016) defined marine heatwaves as the exceedance of the 90th percentile of the sea surface temperature (SST) distribution on a given Julian day during five or more consecutive days, while Box 9.2, Figure 1 shows MHW as an exceedance of 99th-percentile 11-day de-seasonalized SSTs. The return period of marine heatwaves is also critical in determining a coral system’s ability to recover before the next event (Hughes et al., 2018a).

Uptake of atmospheric CO₂ and subsequent increases in dissolved CO₂ lowers ocean pH and can reduce carbonate ion concentrations below critical calcium carbonate saturation thresholds for marine and aquatic organisms growth, reproduction and/or survival, with extended implications for marine ecosystems including fisheries (Bell et al., 2013; Kroeker et al., 2013; Barange et al., 2014; Dutkiewicz et al., 2015; Ekstrom et al., 2015; Gattuso et al., 2015; Mathis et al., 2015a; Nagelkerken and Connell, 2015; Behrenfeld et al., 2016; Nagelkerken and Munday, 2016; Tripathi et al., 2016; Jiang et al., 2018; Weiss et al., 2018; N. Ahmed et al., 2019; Bindoff et al., 2019). Lower pH may provide more favourable conditions for toxic algal blooms (Riebesell et al., 2018) and can interact with hypoxic zones to impact ecosystems (Gobler and Baumann, 2016; Cai et al., 2017).

Changes in currents, sea ice brine rejection and net freshwater flux in the ocean can alter salinity with effects on mixed layer structure, density stratification and the vertical movement of nutrients and marine organisms (Freeland, 2013; Haumann et al., 2016).

Ocean warming and increased stratification decrease the oxygen content of the ocean (Griffiths et al., 2017; Schmidtko et al., 2017; Bindoff et al., 2019), lead to an expansion of oxygen minimum zones in the open ocean (Stramma et al., 2012; Zhang et al., 2013) and exacerbate the creation of anoxic ‘dead zones’ in the coastal oceans (Breitburg et al., 2018). Such a decline (characterized by successive dissolved oxygen concentration thresholds) could affect a wide range of marine organisms and reduce marine habitats (Chan et al., 2008; Vaquer-Sunyer and Duarte, 2008; Hoegh-Guldberg and Bruno, 2010; Altieri and Gedan, 2015; Breitburg et al., 2018) and can also lead to further local acidification (Zhang and Gao, 2016; Laurent et al., 2017).

Although future air pollution will be strongly driven by air quality policies, anthropogenically-driven changes to temperature, humidity, precipitation and synoptic patterns have the potential to affect the emissions, production, concentration and transport of particulate matter (e.g., from dust, fires, pollen) and gaseous pollutants such as sulphur dioxide, tropospheric ozone and nitrogen dioxide (Section 6.5) with resulting impacts on human health, agriculture and ecosystems (Ren et al., 2011; Fiore et al., 2015; Kinney et al., 2015a; Tian et al., 2016; Orru et al., 2017; Emberson et al., 2018; Hayes et al., 2020). Information about conditions leading to poor air quality is also important for visibility in natural parks and tourist locations (Yue et al., 2013; Val Martin et al., 2015), as well as the efficiency of solar photovoltaic panels (Sweerts et al., 2019). Relevant information about conditions favouring air pollution includes tracking warmer conditions that accelerate ozone formation (Peel et al., 2013; Schnell et al., 2016) and the frequency and duration of stagnant air events (Horton et al., 2014; Fann et al., 2015; Lelieveld et al., 2015; Vautard et al., 2018), although no regional index has proven sufficient to capture regional changes or acute events (Kerr and Waugh, 2018; Schnell et al., 2018). By contrast, precipitation and moister air tend to reduce pollution (Section 6.5).

Carbon dioxide (CO₂) is a well-mixed greenhouse gas with global repercussions on Earth’s energy balance; however, atmospheric CO₂ concentration changes at the land surface also affect plant functions within ecosystems and agriculture (see also Chapter 5). High CO₂ concentration can increase photosynthesis rates and primary production within natural ecosystems (Norby et al., 2010; Ratliff et al., 2015; Zhu et al., 2016) and agricultural crops (Hatfield et al., 2011; Leakey et al., 2012; Bell et al., 2013; Glenn et al., 2014; Nagelkerken and Connell, 2015; Behrenfeld et al., 2016; Deryng et al., 2016; Kimball, 2016). High CO₂ concentration affects total biomass and plant sugar content important to bioenergy production (Schaeffer et al., 2012), but also helps some pests and weeds flourish (Hamilton et al., 2005; Wolfe et al., 2008; Valerio et al., 2013; Korres et al., 2016; Stinson et al., 2016; Ramesh et al., 2017), while potentially shifting the effectiveness of herbicides (Varanasi et al., 2016; Refatti et al., 2019). Higher CO₂ concentration reduces transpiration losses during drought conditions (Cammarano et al., 2016; Deryng et al., 2016; Swann et al., 2016; Durand et al., 2018), which also changes the energy balance within the plant canopy (Webber et al., 2017). Higher CO₂ reduces the nutritional density of crops and forage lands (Loladze, 2014; Müller et al., 2014; Myers et al., 2014, 2017; X. Li et al., 2016; M.A. Lee et al., 2017; Smith and Myers, 2018; Zhu et al., 2018; Beach et al., 2019) and can increase the production of toxins (Ziska et al., 2007) and allergenic pollen (Schmidt, 2016).

Changes in surface solar and longwave radiation fluxes alter photosynthesis rates and potential evapotranspiration for natural ecosystems and food, fibre and energy crops (Mäkinen et al., 2018); changes in radiation fluxes can also shift solar energy resources (Schaeffer et al., 2012; Jerez et al., 2015; Wild et al., 2015; Fant et al., 2016; Craig et al., 2018). Plants and aquatic systems particularly respond to changes in photosynthetically active radiation (PAR) and the fraction of diffuse radiation (Proctor et al., 2018; Ren et al., 2018; Ryu et al., 2018). Increases in ultraviolet radiation can also detrimentally affect ecosystems and human health (Barnes et al., 2019).

Additional CIDs may be relevant for regional studies but are not the focus of assessment in this Report. For example, information about changes in the frequency and seasonal timing of fog helps anticipate airport delays and cool beach days, and is also important for water delivery and retention in coastal ecological and agricultural systems (Torregrosa et al., 2014).

Threats to many sectoral assets and associated systems may also be compounded when multiple hazards occur simultaneously in the same place, affect multiple regions at the same time, or occur in a sequence that may amplify overall impact (Section 11.8; IPCC, 2012; Clarke et al., 2018; Zscheischler et al., 2018; Raymond et al., 2020). There is emerging literature on many connected extremes and their associated hazards (e.g., climatic conditions that could drive multi-breadbasket failures; Trnka et al., 2019; Kornhuber et al., 2020), but a full accounting is not practical here especially considering the many possible CID combinations and the need to assess how exposed systems would be vulnerable to compound CIDs (assessed in WGII). Table 12.2 is once again instructive here in considering hazard-related storylines, as the multiple CIDs affecting a given sectoral asset (assessing across a row of Table 12.2) point to potentially dangerous hazard combinations. Similarly, change in a single CID has the potential to affect multiple sectoral assets (assessing down a column of Table 12.2) in a manner with broader systemic implications (AR6 WGII).

Recent literature defines CID indices to represent trends and thresholds that influence sectoral assets, albeit with considerable variation owing to the unique characteristics of regional and sectoral assets. Indices include direct information about the CID’s profile (magnitude, frequency, duration, timing, spatial extent) or utilize atmospheric conditions as a proxy for CIDs that are more difficult to directly observe or simulate. Each sector is affected by multiple CIDs, and each CID affects multiple sectors. Assets within the same sector may require different or tailored indices even for the same CID. These indices may be defined to capture graduated thresholds associated with tipping points or inflection points in a particular sectoral vulnerability, with commonalities in the types of processes these thresholds represent even as their precise magnitude may vary by specific sectoral system and asset.

Mean air temperature: The African continent has experienced increased warming since the beginning of the 20th century in regions where measurements allow a sufficient homogeneous observation coverage to estimate trends (high confidence) (Figure Atlas.11). This warming is very likely attributable to human influence (Chapter 3 and Atlas.4.2) at continental scale. Mean annual temperatures have increased at a high rate since the mid-20th century, reaching 0.2°C–0.5°C per decade in some regions such as north, north-eastern, west and south-western Africa (high confidence) (Atlas.4.2 and Figure Atlas.11).

It is very likely that temperatures will increase in all future emissions scenarios and all regions of Africa (Atlas.4.4). By the end of the century under RCP8.5 or SSP5-8.5, all African regions will very likely experience a warming larger than 3°C except Central Africa, where warming is very likely expected above 2.5°C, while under RCP2.6 or SSP1-2.6, the warming remains very likely limited to below 2°C (Figure Atlas.12). A very likely warming with ranges between 0.5°C and 2.5°C is projected by the mid-century for all scenarios depending on the region (high confidence). Mean temperatures for all regions are projected to increase with increasing global warming (virtually certain) (Figure Atlas.12).

Extreme heat: Warm extremes have increased in most of the regions (high confidence), NEAF, SEAF and MDG (medium confidence) and with low confidence in CAF (Table 11.4). Despite the increasing mean temperature, there is low confidence (limited evidence) that Africa has experienced increased extreme heat stress trend for agriculture or human health in the last two decades of the 20th century in a few regions such as West Africa, South Africa and North Africa considering the period from 1973 to 2012 (Knutson and Ploshay, 2016).

A substantial increase in heatwave magnitude and frequency over most of the Africa domain is projected for even 2°C global warming (high confidence) (Sections 11.3 and 11.9, and Table 11.4), with potential effects on health and agriculture. The number of days with maximum temperature exceeding 35°C is projected to increase (Coppola et al., 2021b) in the range of 50–100 days by 2050 under SSP5-8.5 in WAF, ESAF and WSAF and NEAF (high confidence) (Figure 12.4b). Under SSP1-2.6, the change in the number of exceedance days remains limited to about 40–50 days per year at the end of the century in these regions, while it increases by 150 days or more in WAF, CAF for SSP5-8.5 (Figure 12.4a,c; Figure 12.SM.1).

Mortality-related heat stress levels and deadly temperatures are very likely to become more frequent in the future in RCP8.5/SSP5-8.5 and RCP4.5/SSP2-4.5 and for a 2°C global warming (Mora et al., 2017; Nangombe et al., 2018; Sylla et al., 2018a; Rohat et al., 2019; Q. Sun et al., 2019). In particular the equatorial regions where heat is combined with higher humidity levels, but also North Africa, the Sahel and Southern Africa (Figure 12.4d–f) are among the regions with the largest increases of heat stress (Zhao et al., 2015; Ahmadalipour and Moradkhani, 2018; Coffel et al., 2018). Mitigation scenarios make a large difference in frequency of exceedance of high heat stress indices thresholds (e.g., HI > 41°C) by the end of the century (Figure 12.4d–f; Schwingshackl et al., 2021). In West Africa and Central Africa, under SSP5-8.5, the expected number of days per year with HI > 41°C will increase by around 200 days while in SSP1-2.6 such exceedances are expected to increase by less than 50 days per year (Figure 12.4; Figure 12.SM.2).

Cold spell and frost: Africa experiences cold events and frost days that can affect agriculture, infrastructure, health and ecosystems, especially in Southern and North Africa, which have marked cold seasons and mountainous areas. Cold spells have likely decreased in frequency over subtropical areas. In particular, in North and Southern Africa, the frequency of cold events has likely decreased in the last few decades (Sections 11.3 and 11.9). There is a high confidence that cold spells and low target temperatures will decrease in future climates under all scenarios in West, Central and East Africa. Heating degree days will have a substantial decrease by the end of the century for up to about one month under RCP8.5 in North and Southern Africa (high confidence) (Coppola et al., 2021b).

There is high confidence that extreme heat has increased in frequency and intensity in most African regions. Heatwaves and deadly heat stress and the frequency of exceedance of hot temperature thresholds (e.g., 35°C) will drastically increase by the end of the century (high confidence) under SSP5-8.5, but limited increases are expected in SSP1-2.6. Dangerous heat stress thresholds (HI > 41°C) are projected to be crossed more than 200 days more in West and Central Africa under SSP5-8.5, while this increase remains limited to a few tens of days more for SSP1-2.6. Cold spells and frost days are projected to occur less frequently in all scenarios.

Figure 12.5 | Projected changes in selected climatic impact-driver indices for Africa. Mean change in 1-in-100-year river discharge per unit catchment area (Q100, m³s^–1km^–2) from CORDEX-Africa models for 2041–2060 relative to 1995–2014 for RCP8.5. (b) Shoreline position change along sandy coasts by the year 2100 relative to 2010 for RCP8.5 (metres; negative values indicate shoreline retreat) from the CMIP5-based dataset presented by Vousdoukas et al. (2020b). (c) Bar plots for Q100 (m³s^–1km^–2) averaged over land areas for the AR6 WGI Reference Regions (defined in Chapter 1). The left-hand column within each panel (associated with the left-hand y-axis) shows the ‘recent past’ (1995–2014) Q100 absolute values in grey shades. The other columns (associated with the right-hand y-axis) show the Q100 changes relative to the recent past values for two time periods (‘mid’ 2041–2060 and ‘long’ 2081–2100) and for three global warming levels (GWLs, defined relative to the pre-industrial period 1850–1900): 1.5°C (purple), 2°C (yellow) and 4°C (brown). The bars show the median (dots) and the 10–90th percentile range of model ensemble values across each model ensemble. CMIP6 is shown by the darkest colours, CMIP5 by medium, and CORDEX by light. SSP5-8.5/RCP8.5 is shown in red and SSP1-2.6/RCP2.6 in blue. (d) Bar plots for shoreline position change show CMIP5-based projections of shoreline position change along sandy coasts for 2050 and 2100 relative to 2010 for RCP8.5 (red) and RCP4.5 (blue) from Vousdoukas et al. (2020b). Dots indicate regional mean change estimates and bars show the 5–95th percentile range of associated uncertainty. Note that these shoreline position change projections assume that there are no additional sediment sinks/sources or any physical barriers to shoreline retreat. See Technical (Annex VI for details of indices. Further details on data sources and processing are available in the chapter data table (Table 12.SM.1).

Open figure

Mean precipitation: Since the mid-20th century, precipitation trends have varied in Africa but notable drying trends are found in eastern, central and north-eastern parts of Southern Africa, Central Africa and in the Horn of Africa (Atlas.4.2).

There is high confidence in projected mean precipitation decreases in North Africa and West Southern Africa and medium confidence in East Southern Africa by the end of the 21st century (Dosio et al., 2019; Gebrechorkos et al., 2019; Teichmann et al., 2021; Atlas.4.5). The Western Africa region features a gradient in which precipitation decreases in the west and increases in the east and increase is also projected over Eastern Africa (medium confidence) (Atlas.4.5), with trends in Western Africa affecting the boreal summer monsoon (Chen et al., 2020). Increasing precipitation for 1.5°C and 2°C GWLs are found in central and eastern Sahel with low confidence and the wet signal is getting stronger and more extended for a 3°C and 4°C warmer world (Atlas.4.4).

A change in monsoon seasonality is also reported in Western Africa and Sahel (low confidence) with a forward shift in time (later onset and end; Section 8.2; Mariotti et al., 2011; Seth et al., 2013; Ashfaq et al., 2021). This shift has been associated with a precipitation decrease during the monsoon season attributed to a decrease of African easterly wave activity in the 6–9-day regime (Mariotti et al., 2014) and a soil precipitation feedback reported in Mariotti et al. (2011).

River flood: Generally in Africa from 1990 through 2014, annual flood frequencies have fluctuated and there is medium confidence in an upward trend in flood events occurrences (C.-J. Li et al., 2016). In particular, over Western Africa, upward trends in hydrological extremes such as maximum peak discharge have likely occurred during the last few decades (i.e., after 1980) and have caused increased flood events in riparian countries of rivers such as Niger, Senegal and Volta (high confidence) (Nka et al., 2015; Aich et al., 2016a; Wilcox et al., 2018; Tramblay et al., 2020). In Southern Africa, trends in flood occurrences were decreasing prior to 1980 and increasing afterwards (medium confidence) (Tramblay et al., 2020).

Under future climate scenarios, the extreme river discharge as characterized by the 30-year return period of 5-day average peak flow is projected to increase by the end of the century for RCP8.5 (more than 10% relative to the 1971–2000 period) for most of the tropical African river basins (Dankers et al., 2014) and a consistent increase of flood magnitude is projected across humid tropical Africa by 2050 for the A1B scenario (medium confidence) (Figure 12.5; Arnell and Gosling, 2013). Specifically, in Western Africa there is not a univocal pattern of change for future projections (Roudier et al., 2014); However, under RCP8.5, there is medium confidence of a projected increase of 20-year flood magnitudes by 2050 in countries within the Niger River basin (Aich et al., 2016b) and low confidence (limited evidence) of an increase in extreme peak flows and their duration in countries of the Volta River basin by 2050 and 2090 (Jin et al., 2018). A significant median change of flood magnitude for the Gambia River (–4.5%) and for the Sessandra (+14.4%) and Niger (+6.1%) are projected under several scenarios between mid- and end-of-century (Roudier et al., 2014). In East Africa, extreme flows are projected to increase for regions within the Blue Nile basin with low confidence (limited evidence) (Aich et al., 2014). However, uncertainty due to the climate scenario dominates the projection of extreme flows (Aich et al., 2014; Krysanova et al., 2017) for the Blue Nile and Niger River basins. Averaged over the African continent for different levels of global warming, the present-day 100-year return period flood levels will have a return period of 40 years in 1.5°C and 2°C (Alfieri et al., 2017) and 21 years for 4°C warmer climate (Hirabayashi et al., 2013; Alfieri et al., 2017).

Heavy precipitation and pluvial flood: Chapter 11 found that heavy precipitation intensity and frequency has likely increased over West and East Southern Africa but there is no evidence due to a lack of studies that any significant trend is observed in any other region. In addition, East Africa has experienced strong precipitation variability and intense wet spells leading to widespread pluvial flooding events hitting most countries including Ethiopia, Somalia, Kenya and Tanzania (medium confidence). Finally, with respect to Southern Africa, heavy precipitations events have increased in frequency (medium confidence).

In West Africa and Central Africa, there is high confidence that the intensity of extreme precipitation will increase in a future climate under both RCP4.5 and RCP8.5 scenarios and 1.5°C and 2°C GWLs threatening widespread flood occurrences before, during and after the mature monsoon season (Chapter 11). Extreme precipitation intensity is also increasing in several other regions, such as SAH, NEAF, SEAF, ESAF and MDG (high confidence) for 2°C GWL and higher (Chapter 11).

Landslides: There is an increase in reported landslides in WAF, CAF, NEAF and SEAF in the past decades but with limited evidence of significant trends (Gariano and Guzzetti, 2016; Haque et al., 2019). There is low confidence (limited evidence) of a future increase in landslides in central-eastern Africa, and literature is largely missing to assess this important hazard (Gariano and Guzzetti, 2016).

Aridity: Section 11.9 assesses medium confidence in observed long-term declines of soil moisture and aridity indices in several African regions (NAF, WAF). Trends in East Africa are not definitive given uncertain balances between precipitation and potential evaporation (Kew et al., 2021). Projected declines in precipitation and soil moisture trends indicate high confidence in increased aridity over the 21st century in NAF, WSAF and ESAF but low confidence elsewhere in Africa (Section 11.9; see also Figure 12.4j–l; Gizaw and Gan, 2017). A growing number of studies provide further regional context on expanding aridity in several places in East and West Africa, respectively (Sylla et al., 2016a; Liu et al., 2018b; Haile et al., 2020).

Hydrological drought: Section 11.9 noted observed decreases in hydrological drought over the Mediterranean (high confidence) and diminished summer river flows in West Africa (medium confidence). Recent regional modelling studies project substantial increases in hydrological drought affecting major West African river basins under 1.5°C and 2°C GWLs and RCP4.5 and RCP8.5 scenarios (Oguntunde et al. 2018, 2020; Sylla et al., 2018b); however, there remains low confidence in future projections given disagreement with global model runoff projections (e.g., B.I. Cook et al., 2020). There is high confidence that a 2°C GWL would see an increase in hydrological droughts in the Mediterranean region, and medium confidence in increasing hydrological drought conditions in the Southern Africa regions (Section 11.9).

Agricultural and ecological drought: Farmers and food security experts in East Africa have noted spatial extensions in seasonal agricultural droughts in recent decades (Elagib, 2014), but it is difficult to disentangle these trends from climate variability. In Ethiopia, past severe agricultural drought conditions in the northern regions are moderately common events in recent years (Zeleke et al., 2017). In Southern Africa, the number of ‘flash’ droughts (with rapid onset and durations from a few days to couple of months) have increased by 220% between 1961 and 2016 as a result of anthropogenic warming (Yuan et al., 2018). Section 11.9 notes medium confidence increases in agricultural and ecological drought trends in North, Western and Central Africa as well as both Southern Africa regions. The most striking drought is the Western Cape drought in 2015–2018, a prolonged drought that resulted in acute water shortages (Wolski, 2018; Burls et al., 2019; Section 10.6.2). Anthropogenic climate change caused a threefold increase in the probability of such a drought to occur (Chapters 10 and 11; Botai et al., 2017; Otto et al., 2018). Section 11.9 assesses increases in agricultural and ecological drought at 2°C GWL for North Africa and West Southern Africa (high confidence) and for East Southern Africa and Madagascar (medium confidence), with confidence generally rising for higher emissions scenarios (Sylla et al., 2016b; Zhao and Dai, 2017; Diedhiou et al., 2018; Abiodun et al., 2019; Todzo et al., 2020; Coppola et al., 2021b). Liu et al. (2018b) identified the Southern Africa region as the drought ‘hottest spot’ in Africa in 1.5°C and 2°C global warming scenarios.

Fire weather: There is low confidence (low agreement) in recent reductions in fire activity given soil moisture increases in some regions and substantial land use changes (Andela et al., 2017; Forkel et al., 2019; Zubkova et al., 2019). Days prone to fire conditions are going to increase in all extratropical Africa until the end of the century and fire weather indices are projected to largely increase in North and Southern Africa, where increasing aridity trends occur (high confidence), with an emerging signal well before the middle of the century where drought and heat increase will combine (Chapter 11; Engelbrecht et al., 2015; Abatzoglou et al., 2019). There is low confidence (limited evidence) of fire weather changes for other African regions.

Total precipitation is projected to decrease in the northernmost (high confidence) and southernmost regions of Africa (medium confidence), with West and East Africa regions each having a west-to-east pattern of decreasing-to-increasing precipitation (medium confidence). Most African regions will undergo an increase in heavy precipitation that can lead to pluvial floods (high confidence), even as increasing dry climatic impact-drivers (aridity, hydrological, agricultural and ecological droughts, fire weather) are generally projected in the North Africa and Southern African regions (high confidence) and western portions of West Africa (medium confidence).

Mean wind speed: Decreasing trends in wind speeds have occurred in many parts of Africa (low confidence due to observations with limited homogeneity) (McVicar et al., 2012; AR5 WGI). There is high confidence in climate change-induced future decreasing mean wind, wind energy potential and strong winds in North Africa and Mediterranean regions as a consequence of the poleward shift of the Hadley cell (Karnauskas et al., 2018a; Kjellström et al., 2018; Sivakumar and Lucio, 2018; Tobin et al., 2018; Jung and Schindler, 2019) in the RCP4.5 and RCP8.5 scenarios by the middle of the century or beyond, and for a GWL of 2°C or higher. Over Western Africa and Southern Africa a future significant increase in wind speeds and wind energy potential is expected (medium confidence) (Figure 12.4m–o; Karnauskas et al., 2018a; Jung and Schindler, 2019).

Severe wind storm: A limited number of studies allow an assessment of past trends in wind storms. In West Africa and specifically in the Sahel band, more intense storms have occurred since the 1980s (low confidence, limited evidence). A persistent and large increase of frequency of Sahelian mesoscale convective storms has been found in several studies (Panthou et al., 2014; C.M. Taylor et al., 2017), with consequences for extreme rainfalls, and potentially extreme winds (low confidence, limited evidence). There is low confidence of a general increasing trend in extreme winds across Western, Central, Eastern and Southern Africa in a majority of regions by the middle of the century even in high-end scenarios. The frequency of Mediterranean wind storms reaching North Africa, including Medicanes, is projected to decrease, but their intensities are projected to increase, by the mid-century and beyond under SRES A1B, SRES A2 and RCP8.5 (medium confidence) (Chapter 11; Cavicchia et al., 2014; Walsh et al., 2014; Tous et al., 2016; Romera et al., 2017; Romero and Emanuel, 2017; González-Alemán et al., 2019).

Tropical cyclone: In the South Indian Ocean, an increase in Category 5 cyclones has been observed in recent decades (Fitchett, 2018) as in other basins (Section 11.7). However, there is a projected decrease in the frequency of tropical cyclones making landfall over Madagascar, South Eastern Africa and East Southern Africa in a 1°C, 2°C and 3°C warmer world (medium confidence) (Malherbe et al., 2013; Roberts et al., 2015, 2020; Muthige et al., 2018; Knutson et al., 2020). There is medium confidence in general increasing intensities for cyclones in such studies for African regions.

Sand and dust storm: North Africa and the Sahel, and to a lesser extent Southern Africa, are prone to dust storms, having consequences on health (Querol et al., 2019), transmission of infectious diseases (Agier et al., 2013; Wu et al., 2016), and solar power generation and related maintenance costs. There is limited evidence and low agreement of secular 20th century trends in wind speeds or dust emissions (limited length of data records, large variability). Dust variations are controlled by changes in surface winds, precipitation and vegetation, which in turn are modulated at multiple time scales by dominant modes of internal climate variability (Chapter 10). In North Africa, wind variability explains both the observed high concentrations between the 1970s and 1980s and lower concentrations thereafter (Ridley et al., 2014; Evan et al., 2016). Yet, the effect of vegetation changes may not be negligible (Pu and Ginoux, 2017, 2018).

Changes to the frequency and intensity of dust storms also remain largely uncertain due to uncertainty in future regional wind and precipitation as the climate warms, CO₂ fertilization effects on vegetation (Huang et al., 2017), and anthropogenic land use and land-cover change due to land management and invasive species (Ginoux et al., 2012; Webb and Pierre, 2018). Dust loadings and related air pollution hazards (from fine particles that affect health) are projected to generally decrease in many regions of the Sahara and Sahel due to the changing winds (Evan et al., 2016) and slightly increase over the Guinea coast and West Africa (low confidence) (Ji et al., 2018).

In summary, there is high confidence of a decrease in mean wind speed and wind energy potential in North Africa and medium confidence of an increase in Southern and Western Africa, by the middle of the century regardless of climate scenario or global warming level equal or superior to 2°C, high confidence of a decrease in frequency of cyclones landing in SEAF, ESAF and MDG, and low confidence of a general increase in wind storms in most African regions located south of the Sahel. The evolution of dust storms remains largely uncertain.

Snow and glacier: African glaciers are located in East Africa and more specifically on Mount Kenya, the Rwenzori Mountains and Mount Kilimanjaro, with glaciers reducing substantially in each region (high confidence) (Taylor et al., 2006; Cullen et al., 2013; Chen et al., 2018; Prinz et al., 2018; Wang and Zhou, 2019). Observation and future projection of African glacier mass changes are assessed in Section 9.5.1 within the low-latitude glacier region, which is one of the regions with the largest mass loss even under low-emissions scenarios (assessment of this region is dominated by glaciers in the South American Andes, however) (high confidence). Glaciers in the low-latitude region will lose 67 ± 42%, 86 ± 24% and 94 ± 13% of their mass by the end of the century for RCP2.6, RCP4.5 and RCP8.5 scenarios respectively (Marzeion et al., 2020). Cullen et al. (2013) calculated that even imbalances between the Mount Kilimanjaro glaciers and present-day climate would be enough to eliminate the mountain’s glaciers by 2060. Snow water equivalent and snow cover season duration also decline in the East African mountains, Ethiopian Highlands and Atlas Mountains with climate change (high confidence) (López-Moreno et al., 2017).

In conclusion, there is high confidence that African snow and glaciers have very significantly decreased in the last decades and that this trend will continue over the 21st century.

Relative sea level: Around Africa, from 1900 to 2018, a new tide gauge-based reconstruction finds a regional mean RSL change of 2.07 [1.36 to 2.77] mm yr^–1 in the South Atlantic and 1.33 [0.80 to 1.86] mm yr^–1 in the Indian Ocean (Frederikse et al., 2020), compared to a GMSL change of around 1.7 mm yr^–1Section 2.3.3.3 and Table 9.5). For the period 1993–2018, these RSLR rates, based on satellite altimetry, increased to 3.45 [3.04 to 3.86] mm yr^–1 and 3.65 [3.23 to 4.08] mm yr^–1 respectively (Frederikse et al., 2020), compared to a GMSL change of 3.25 mm yr^–1Section 2.3.3.3 and Table 9.5).

Relative sea level rise is virtually certain to continue in the oceans around Africa. Regional mean RSLR projections for the oceans around Africa range from 0.4–0.5 m under SSP1-2.6 to 0.8–0.9 m under SSP5-8.5 for 2081–2100 relative to 1995–2014 (median values), which is within the range of projected GMSL change (Section 9.6.3.3). These RSLR projections may, however, be underestimated due to potential partial representation of land subsidence in their assessment (Section 9.6.3.2).

Coastal flood: The present-day 1-in-100-year extreme total water level is between 0.1 and 1.2 m around Africa, with values around 1 m or above along the south-west, south-east and central east coasts (Vousdoukas et al., 2018).

Extreme total water level (ETWL) magnitude and occurrence frequency are expected to increase throughout the region (high confidence) (Figure 12.4p–r and Figure 12.SM.6). Across the continent, the 5th–95th percentile range of the 1-in-100-year ETWL is projected to increase (relative to 1980–2014) by 7–36 cm and by 14–42 cm by 2050 under RCP4.5 and RCP8.5 respectively. By 2100, this range is projected to be 28–86 cm and 43–190 cm under RCP4.5 and RCP8.5 respectively (Vousdoukas et al., 2018; Kirezci et al., 2020). In terms of ETWL occurrence frequencies, the present-day 1-in-100-year ETWL is projected to have median return periods of around 1-in-10-years to 1-in-20-years by 2050 and 1-in-1-year to 1-in-5-years by 2100 in southern and North Africa and occur more than once per year by 2050 and 2100 in most of East and West Africa under RCP4.5 (Vousdoukas et al., 2018). The present-day 1-in-50-year ETWL is projected to occur around three times a year by 2100 with an SLR of 1 m in Africa (Vitousek et al., 2017).

Coastal erosion: Shoreline retreat rates up to 1 m yr^–1have been observed around the continent during 1984–2015, except in ESAF, which has experienced a shoreline progradation rate of 0.1 m/r over the same period (Luijendijk et al., 2018; Mentaschi et al., 2018). Mentaschi et al. (2018) report a coastal area losses of 160 km² and 460 km² over a 30-year period (1984–2015) along the Atlantic and Indian Ocean coasts of the continent. At the more regional level, in Ghana along the Gulf of Guinea about 79% of the shoreline was found to be retreating while 21% was found to be stable or prograding over the period 1974–1996 (Addo and Addo, 2016).

Projections indicate that a vast majority of sandy coasts in the region will experience shoreline retreat throughout the 21st century (high confidence), while parts of the ESAF and western MDG coastline are projected to prograde over the 21st century, if present ambient trends continue. Median shoreline change projections (CMIP5), relative to 2010, presented by Vousdoukas et al. (2020b) show that, under RCP4.5, shorelines in Africa will retreat by between 30 m (SAH, NEAF, WSAF, ESAF, MDG) and 55 m (WAF, CAF), by mid-century. By the same period but under RCP8.5, the median shoreline retreat is projected to be between 35 m (SAH, NEAF, WSAF, ESAF) and 65 m (WAF, CAF). By 2100, more than 100 m of median retreat is projected in WAF, CAF and SEAF under RCP4.5, while under RCP8.5, more than 100 m of shoreline retreat is projected in all regions except NEAF and WSAF. Under RCP8.5 especially, the projected retreat by 2100 is greater than 150 m in WAF and CAF. The total length of sandy coasts in Africa that is projected to retreat by more than a median of 100 m by 2100 under RCP4.5 and RCP8.5 is about 13,000 km and 17,000 km respectively, an increase of approximately 33%.

Marine heatwave (MHW): From 1982 to 2016, the coastal oceans of Africa have experienced on average 2–3 MHWs per year, with the coastal oceans around the southern half of the continent experiencing on average 2.5–3 MHWs per year. The average duration was between 5 and 15 days (Oliver et al., 2018). Changes over the 20th century, derived from MHW proxies, show an increase in frequency between 0.5 and 2 MHWs per decade over the region, especially off the Horn of Africa; an increase in intensity per event around Southern Africa; and an increase in MHW duration along the North African coastlines (Oliver et al., 2018).

There is high confidence that MHWs will increase around Africa. Mean SST, a common proxy for MHWs, is projected to increase by 1°C (2°C) around Africa by 2100, with a hotspot of around 2°C (5°C) along the coastlines of South Africa under RCP4.5 (RCP8.5; Interactive Atlas). Under global warming conditions, MHW intensity and duration will increase in the coastal zones of all sub-regions of Africa (Frölicher et al., 2018). Projections for SSP1-2.6 and SSP5-8.5 both show an increase in MHWs around Australasia by 2081–2100, relative to 1985–2014 (Box 9.2, Figure 1).

In general, there is high confidence that most coastal- and ocean-related hazards in Africa will increase over the 21st century. Relative sea level rise is virtually certain to continue around Africa, contributing to increased coastal flooding in low-lying areas (high confidence) and shoreline retreat along most sandy coasts (high confidence). Marine heatwaves are also expected to increase around the region over the 21st century (high confidence).

The assessed direction of change in CIDs for Africa and associated confidence levels are illustrated in Table 12.3. No relevant literature could be found for permafrost and hail, although these phenomena may be relevant in parts of the continent.

Table 12.3 | Summary of confidence in direction of projected change in climatic impact-drivers in Africa, representing their aggregate characteristic changes for mid-century for scenarios RCP4.5, SSP2-4.5, SRES A1B, or above within each AR6 region (defined in Chapter 1), approximately corresponding (for CIDs that are independent of sea level rise) to global warming levels between 2°C and 2.4°C (see Section 12.4 for more details of the assessment method). The table also includes the assessment of observed or projected time-of-emergence of the CID change signal from the natural interannual variability if found with at least medium confidence in Section 12.5.2.

Mean air temperature: A long-term warming trend in annual mean surface temperature has been observed across Asia during 1960–2015, and the warming accelerated after the 1970s (high confidence) (Davi et al., 2015; Aich et al., 2017; Cheong et al., 2018; S. Dong et al., 2018; IPCC, 2018; Krishnan et al., 2019; M. Zhang et al., 2019). Records also indicate a higher rate of warming in minimum temperatures than maximum temperatures in Asia, leading to more frequent warm nights and warm days, and less frequent cold days and cold nights (high confidence) (Supari et al., 2017; Akperov et al., 2018; Cheong et al., 2018; Rahimi et al., 2018; Khan et al., 2019a; L. Li et al., 2019; M. Zhang et al., 2019).

Projections show continued warming over Asia in the future with contrasted regional patterns across the continent (high confidence) (Figure 4.19). For RCP8.5/SSP5-8.5 at the end of the century, the mean estimated warming exceeds 5°C in WSB, ESB and RFE and 7°C in some parts (high confidence). In most areas of ARP and WCA, 5°C is exceeded (Ozturk et al., 2017), but EAS, SAS and SEA have a lower projected warming of less than 5°C (Basha et al., 2017; C. Lu et al., 2019; Almazroui et al., 2020; Atlas.5). Under SSP1-2.6, the warming remains limited to 2°C in most areas except Arctic regions, where it exceeds 2°C (Figure 4.19).

Extreme heat: There is increased evidence and high confidence of more frequent heat extremes in the recent decades than in previous ones in most of Asia (Acar Deniz and Gönençgil, 2015; Rohini et al., 2016; Mishra et al., 2017; You et al., 2017; Imada et al., 2018; Khan et al., 2019b; Krishnan et al., 2019; Rahimi et al., 2019; Yin et al., 2019; Chapter 11) due to the effects of anthropogenic global warming, El Niño and urbanization (Luo and Lau, 2017; Thirumalai et al., 2017; Imada et al., 2019; Y. Sun et al., 2019; Zhou et al., 2019). But there is medium confidence of heat extremes increasing in frequency in many parts of India (Rohini et al., 2016; Mazdiyasni et al., 2017; van Oldenborgh et al., 2018; Sen Roy, 2019; Kumar et al., 2020) partly due to the alleviation of anthropogenic warming by increased air pollution with aerosols and expanding irrigation (van Oldenborgh et al., 2018; Thiery et al., 2020).

Extreme heat events are very likely to become more intense and/or more frequent in SAS, WCA, ARP, EAS, and SEA by the end of 21st century, especially under RCP6.0 and RCP8.5 (Figure 12.4a–c and Chapter 11; Lelieveld et al., 2016; Pal and Eltahir, 2016; Guo et al., 2017; Mishra et al., 2017; Dosio et al., 2018; Lin et al., 2018; Nasim et al., 2018; Shin et al., 2018; Hong et al., 2019; Su and Dong, 2019; Khan et al., 2020; Kumar et al., 2020). The exceedance of the dangerous heat stress 41°C threshold of the HI is expected to increase by about 250 days in SEA and by 50–150 days in SAS, WCA, ARP and EAS for SSP5-8.5 at the end of the century (Figure 12.4d–f and Figure 12.SM.2). Under SSP1-2.6, the increase would be restricted to less than 30 days in many of these regions except SEA, where the number of exceedance days increases by about 100 days in some areas. Such increases are already present in the middle of the century (Figure 12.4d–f; Schwingshackl et al., 2021). In these regions, the increase in number of days with exceedance of 35°C of high heat stress is also expected to increase substantially for the mid-century under SSP5-8.5 (typically by 10–50 days except in Arctic and Siberian regions), and by more than 60 days in areas of SEA, and a large difference is found between low- and high-end scenarios in the end of the century (high confidence) (Figure 12.4b). Over WSB, ESB and RFE also, an increase of extreme heat durations and frequency is expected in all scenarios (high confidence) (Kattsov et al., 2017; Khlebnikova et al., 2019b).

Cold spell and frost: Cold spells intensity and frequency, as well as the number of frost days, in most Asian regions have been decreasing since the beginning of the 20th century (high confidence) (Chapter 11; Sheikh et al., 2015; Donat et al., 2016; Erlat and Türkeş, 2016; S. Dong et al., 2018; Liao et al., 2018, 2020; Lu et al., 2018; van Oldenborgh et al., 2019), except for the central Eurasian regions, where there was a cooling trend during 1995–2014, which is linked to sea ice loss in the Barents–Kara Seas (medium confidence) (Atlas.5.2; Wegmann et al., 2018; Blackport et al., 2019; Mori et al., 2019).

It is very likely that cold spells will have a decreasing frequency in all future scenarios across Asian regions (J. Guo et al., 2018; Sui et al., 2018; L. Li et al., 2019), as well as frost days (L. Wang et al., 2017; Fallah-Ghalhari et al., 2019) except in tropical Asia (Chapter 11).

In Asia, temperatures have warmed during the last century (high confidence). Extreme heat episodes have become more frequent in most regions (high confidence), and are very likely to increase in all regions of Asia under all warming scenarios during this century. Dangerous heat stress thresholds such as HI > 41°C will be crossed much more often (typically 5 0–1 50 days per year more than the recent past) in many southern Asia regions at the end of the century under SSP5-8.5 while these numbers should remain limited to a few tens under SSP1-2.6 (high confidence). It is very likely that cold spells and frost days will decrease in frequency in all future scenarios across Asian regions during the century.

Mean precipitation: The most prominent features about changes in precipitation over Asia (1901–2010) are the increasing precipitation trends across higher latitudes, along with some scattered smaller regions of detectable increases and decreases (Knutson and Zeng, 2018); however, spatial variability remains high (W. Wang et al., 2015; Limsakul and Singhruck, 2016; Supari et al., 2017; Rahimi et al., 2018, 2019; Sein et al., 2018; Kumar et al., 2019; H. Wang et al., 2019; see Atlas.5) (medium confidence).

Mean precipitation is likely to increase in most areas of northern (WSB, ESB, RFE), southern (ECA, TIB, SAS) and East Asia (EAS) in different scenarios (high confidence) (Huang et al., 2014; Xu et al., 2017; Kusunoki, 2018; Mandapaka and Lo, 2018; Luo et al., 2019; Wu et al., 2019; X. Zhu et al., 2019; Almazroui et al., 2020; Jiang et al., 2020; Rai et al., 2020; see Atlas.5). Monsoon circulation will also increase seasonal contrasts, with SAS seeing wetter wet seasons and drier dry seasons (Atlas.5.3). Higher uncertainty between CMIP5 and CMIP6 as well as spatial differences lend low confidence to model projections in ARP and WCA (Atlas.5.5), with large seasonal differences (Zhu et al., 2020) and some models projecting decreases in precipitation in Central Asia (Ozturk et al., 2017), Pakistan (Nabeel and Athar, 2020) and SEA (Supari et al., 2020).

River flood: Flood risk has grown in many places in China from 1961 to 2017 (Kundzewicz et al., 2019) (low confidence). In SAS, the numbers of flood events and human fatalities have increased in India during 1978–2006 (Singh and Kumar, 2013), whereas the average country-wide inundation depth has been decreasing during 2002–2010 in Bangladesh, attributed to improved flood management (low confidence) (Sciance and Nooner, 2018).

Given the increase of heavy precipitation in most Asian regions, the river flood frequency and intensities will change consequently in Asia. Over China floods will increase with different levels under different warming scenarios (medium confidence) (Lin et al., 2018; Kundzewicz et al., 2019; Liang et al., 2019; Gu et al., 2020). Monsoon floods will be more intense in SAS (medium confidence) (Nowreen et al., 2015; Babur et al., 2016; Mohammed et al., 2018). The total flood damage will increase greatly in river basins in SEA countries under the conditions of climate change and rapid urbanization in the near future (Dahal et al., 2018; Kefi et al., 2020). A changing snowmelt regime in the mountains may contribute to a shift of spring floods to earlier periods in Central Asia in future (medium confidence) (Reyer et al., 2017b). The annual maximum river discharge can almost double by the mid-21st century in major Siberian rivers, and annual maximum flood area is projected to increase across Siberia mostly by 2–5% relative to the baseline period (1990–1999) under RCP8.5 scenario (medium confidence) (Shkolnik et al., 2018).

Heavy precipitation and pluvial flood: Pluvial floods are driven by extreme precipitation and land use. Observed changes in extreme precipitation vary considerably by region (Chapter 11). Heavy precipitation is very likely to become more intense and frequent in all areas of Asia except in ARP (medium confidence) for a 2°C GWL or higher (Chapter 11).

Landslide: The majority of non-seismic fatal landslide events were triggered by rainfall, and Asia is the dominant geographical area of landslide distribution (Froude and Petley, 2018). Floods and landslides are the most frequently occurring natural hazards in the eastern Himalayas and hilly regions, particularly caused by torrential rain during the monsoon season (Gaire et al., 2015; Syed and Al Amin, 2016). They accounted for nearly half of the events recorded in the countries of the HKH region (Vaidya et al., 2019). Intense monsoon rainfall in northern India and western Nepal in 2013, which led to landslides and one of the worst floods in history, has been linked to increased loading of GHG and aerosols (Cho et al., 2016). Due to an increase of heavy precipitation and permafrost thawing, an increase in landslides is expected in some areas of Asia, such as northern Taiwan (China), some South Korean mountains, Himalayan mountains, and permafrost territories of Siberia, and the increase is expected to be the greatest over areas covered by current glaciers and glacial lakes (medium confidence, medium evidence) (Kim et al., 2015; Kharuk et al., 2016; C.-W. Chen et al., 2019; Kirschbaum et al., 2020).

Aridity: Aridity in West Central Asia and parts of South Asia increased in recent decades (medium confidence), as documented in Afghanistan (Qutbudin et al., 2019), Iran (Zarei et al., 2016; Zolfaghari et al., 2016; Pour et al., 2020), most parts of Pakistan (K. Ahmed et al., 2018, 2019), and many parts of India (Roxy et al., 2015; Mallya et al., 2016; Matin and Behera, 2017; Ramarao et al., 2019). Some spatial and seasonal differences within these regions remain, with Ambika and Mishra (2020) noting significant aridity declines over the Indo–Gangetic Plain in India during 1979–2018 due in part to the effect of irrigation, and Araghi et al. (2018) found that many parts of Iran show no significant trends in aridity. There was a drying tendency in the dry season and significant wetting in the wet season in the Philippines during 1951–2010 (Villafuerte et al., 2014), and slight wetting in Vietnam during 1980–2017 (Stojanovic et al., 2020) (low confidence). In EAS there is low confidence of broad aridity changes, as the frequency of droughts have increased (especially in spring) along a strip extending from south-west China to the western part of north-east China; however, there is no evidence of a significant increase in drought severity over China as a whole and many parts in the arid north-west China got wetter during 1961–2012 (W. Wang et al., 2015; Zhai et al., 2017; H. Wang et al., 2019; Zhang and Shen, 2019). In Siberia, the number of dry days has decreased for much of the region, but increased in its southern parts (Khlebnikova et al., 2019a).

The counteracting factors of projected increases in precipitation and temperature across most of Asia (Section 11.9 and Atlas.5) leads to low confidence (limited evidence, inconsistent trends) for broad, long-term aridity changes with medium confidence only for aridity increases in West Central Asia and East Asia. A growing number of studies highlight the potential for more localized aridity trends, including projection ensembles indicating significant increase in aridity and more frequent and intense droughts in most parts of China (Y. Li et al., 2019; Yao et al., 2020) and India under RCP4.5 and RCP8.5 for the 2020–2100 period (Gupta and Jain, 2018; Bisht et al., 2019; Preethi et al., 2019).

Hydrological drought: Section 11.9 indicates that limited evidence and inconsistent regional trends gives low confidence to observed and projected changes in hydrological drought in all Asian regions at a 2°C GWL (approximately mid-century), although West Central Asia hydrological droughts increase at the 4°C GWL (approximately end-of-century under higher emissions scenarios) (medium confidence). Human activities such as reservoir operation and water abstraction have had a profound effect on low river flow characteristics and drought impacts in many Asian regions (Kazemzadeh and Malekian, 2016; Yang et al., 2020b). There was no observed overall long-term change of both meteorological droughts and hydrological droughts over India during 1870–2018 (Mishra, 2020), but there were strong trends towards drying of soil moisture in north-central India (Ganeshi et al., 2020) and intensified droughts in north-west India, parts of Peninsular India, and Myanmar (Malik et al., 2016). The frequency of water scarcity connected with hydrological droughts has increased significantly in southern Russia since the beginning of the 21st century (Frolova et al., 2017). Higher future temperatures are expected to alter the seasonal profile of hydrologic droughts given reduced summer snowmelt (medium confidence) downstream of mountains such as the Himalayas and the Tibetan Plateau (Sorg et al., 2014). Several studies project more severe future hydrological drought in the Weihe River basin in northern China (Yuan et al., 2016; Sun and Zhou, 2020).

Agricultural and ecological drought: Section 11.9 assesses medium confidence in observed increases to agricultural and ecological droughts in West Central Asia, East Central Asia, and East Asia. Persistent droughts were the main factor for grassland degradation and desertification in Central Asia in the early 21st century (G. Zhang et al., 2018; Emadodin et al., 2019). Compound meteorological drought and heat events, which lead to water stress conditions for agricultural and ecological systems, have become more frequent, widespread and persistent in China especially since the late 1990s (Yu and Zhai, 2020). There were more agricultural droughts in northern China than in southern China, and the intensity of agricultural drought increased during 1951–2018 (Zhao et al., 2021).

Studies examining a 2°C GWL give low confidence for projected broad changes to agricultural and ecological drought across all Asia regions, although at 4°C GWL agricultural and ecological drought increases are projected for West Central Asia and East Asia along with a decrease in South Asia (medium confidence) (Section 11.9). Summer temperature increase will enhance evapotranspiration, facilitating ecological and agricultural drought over Central Asia towards the latter half of this century (Chapter 11; see also Figure 12.4 for soil moisture and DF indices; Ozturk et al., 2017; Reyer et al., 2017b; Senatore et al., 2019). However, broader changes in droughts could not be determined in Asia due to the mixture of total precipitation signals together with temperature increase patterns (Section 11.9 and Atlas.5).

Fire weather: Under the global warming scenario of 2°C, the magnitude of length and frequency of fire seasons are projected to increase with strong effects in India, China and Russia (medium confidence) (Q. Sun et al., 2019). Abatzoglou et al. (2019) found that higher fire weather conditions due to climate change emerge in the first part of the 21st century in South China, WCA as well as in boreal areas of Siberia and RFE. The potential burned areas in five Central Asian countries (Kazakhstan, Kyrgyzstan, Tajikistan, Uzbekistan and Turkmenistan) will increase by 2–8% in the 2030s and 3–13% in the 2080s compared with the baseline (medium confidence) (1971–2000; Zong et al., 2020).

In conclusion, there is medium confidence that extreme precipitation, mean precipitation and river floods will increase across most Asian regions. There is low confidence for projected changes in aridity and drought given overall increases in precipitation and regional inconsistencies, with medium increases for West Central Asia and East Asia especially beyond the middle of the century and global warming levels beyond 2°C. Fire weather seasons are projected to lengthen and intensify particularly in the northern regions (medium confidence).

Mean wind speed: There is high confidence of the slowdown in terrestrial near-surface wind speed (SWS) in Asia by approximately –0.1 m s^–1 per decade since the 1950s based on observations and reanalysis data, with the significant decreases in Central Asia among the highest in the world followed by EAS and SAS (J. Wu et al., 2018; Tian et al., 2019; R. Zhang et al., 2019). But a short-term strengthening in SWS was observed during the winter since 2000 in eastern China (medium confidence) (Zeng et al., 2019; Zha et al., 2019).

There is medium confidence of future declining mean SWS in Asia, except in SAS and SEA, as global projections indicate a decreasing trend in all climate scenarios for most of northern Asia, TIB and East Asia by the mid-century (Karnauskas et al., 2018a; Fedotova, 2019; Jung and Schindler, 2019; Ohba, 2019; J. Wu et al., 2020; Zha et al., 2020; Figure 12.4m–o), with negative effects on wind energy potential. Decreases in North Asia are generally modest, not exceeding 10% for the mid-century and 20% for the end of century for the RCP8.5 and RCP4.5 scenarios (Figure 12.4m–o).

Severe wind storms: Consistent with the general mean decreasing surface winds, there is medium confidence that strong winds declined faster than weak winds in the past few decades in Asia in general (Vautard et al., 2010; Tian et al., 2019), but evidence is lacking for spatial patterns. There is low confidence that extra-tropical cyclones will decline in number in future climate scenarios over WCA, TIB, WSB and ESB, and intensify over the Arctic regions as a result of the poleward shift of storm tracks (Basu et al., 2018; Chapter 11). There is limited evidence for projection of changes in severe winds occurring in convective storms in Asia.

Tropical cyclone: There was an increase in the number and intensification rate of intense tropical cyclones (TC), such as Category 4–5 (wind speeds >58 m s^–1), in the Western North Pacific (WNP) and Bay of Bengal since the mid-1980s (medium confidence) (Section 11.7; Kim et al., 2016; Mei and Xie, 2016; Walsh et al., 2016a; Knutson et al., 2019). There is medium confidence that there has been a significant north-westward shift in TC tracks and a poleward shift in the average latitude where TCs reach their peak intensity in the WNP since the 1980s (Knutson et al., 2019; J. Sun et al., 2019; Lee et al., 2020), increasing exposure to TC passage and more destructive landfall over eastern China, Japan, and Korea in the last few decades (Kossin et al., 2016; Li et al., 2017; Altman et al., 2018; Liu and Chan, 2019), and decreasing exposure in the region of SAS and southern China (Cinco et al., 2016; Kossin et al., 2016; see Chapter 11). However, while the analysis shows fewer typhoons, more extreme TCs have affected the Philippines (low confidence) (Takagi and Esteban, 2016). The frequency and duration of tropical cyclones has significantly increased over time over the Arabian Sea and insignificantly decreased over the Bay of Bengal during 1977–2018 (low confidence) (Fan et al., 2020).

There is medium confidence that future TC numbers will decrease but the maximum TC wind intensities will increase in the western Pacific as elsewhere (Chapter 11, see Figure 11.24; Choi et al., 2019; Cha et al., 2020; Knutson et al., 2020). The simulations for the late 21st century for the RCP8.5 scenario yield considerably more TCs in the WNP that exceed 49.4 m s⁻¹ (Category 3) intensity (Mclay et al., 2019). There is medium confidence that the average location of the maximum wind will migrate poleward (see Chapter 11), and TC translation speeds at the higher latitudes would decrease (Yamaguchi et al., 2020). As a consequence, the intensity of TCs affecting the Japan Islands would increase in the future under the RCP8.5 scenario (Yoshida et al., 2017), whereas the frequency of TCs affecting the Philippine region and Vietnam is projected to decrease (Kieu-Thi et al., 2016; C. Wang et al., 2017; Gallo et al., 2019) (medium confidence).

Sand and dust storm: The Asia-Pacific region contributes 26.8 per cent to global dust emissions as of 2012 (UNESCAP, 2018). In West Asia, the frequency of dust events has increased markedly in some areas (east and north-east of Saudi Arabia, north-west of Iraq and east of Syria) from 1980 to the present (Nabavi et al., 2016; Alobaidi et al., 2017). This marked dust increase has been associated with drought conditions in the Fertile Crescent (Notaro et al., 2015; Yu et al., 2015), likely amplified by anthropogenic warming (Kelley et al., 2015; Chapter 10). Dust storm frequency in most regions of northern China show a decreasing trend since the 1960s due to the decrease in surface wind speed (medium confidence) (Guan et al., 2017).

While dust activity has decreased greatly over EAS, current climate models are unable to reproduce the trends (Guan et al., 2015, 2017; Zha et al., 2017; C. Wu et al., 2018). Thus, there is limited evidence for future trends of sand and dust storms in Asia.

In conclusion, surface wind speeds have been decreasing in Asia (high confidence), but there is a large uncertainty in future trends. There is medium confidence that mean wind speeds will decrease in Central and northern Asia, and that tropical cyclones will have decreasing frequency and increasing intensity overall.

Snow: There is no significant interannual trend of total snow cover from 2000 to 2016 over Eurasia (X. Wang et al., 2017a; Sun et al., 2020). Observations do show significant changes in the seasonal timing of Eurasian snow cover extent (especially for earlier spring snowmelt) since the 1970s, with seasonal changes expected to continue in the future (high confidence) (Yeo et al., 2017; Zhong et al., 2021). By 2100, snowline elevations are projected to rise between 400 and 900 m (4.4 to 10.0 m yr^–1) in the Indus, Ganges and Brahmaputra basins under the RCP8.5 scenario (Viste and Sorteberg, 2015).

Glacier: Observation and future projection of glacier mass changes in Asia are assessed in Section 9.5.1 grouped in three main regions: northern Asia, High Mountains of Asia, and Caucuses and Middle East. All regions show continuing decline in glacier mass and area in the coming century (high confidence). Under RCP2.6 the pace of glacier loss slows, but glacier losses increase in RCP8.5 and peak in the mid to late 21st century. GlacierMIP projections indicate that glaciers in the High Mountains of Asia lose 42 ± 25%, 56 ± 24% and 71 ± 21% of their 2015 mass by the end of the century for RCP2.6, RCP4.5 and RCP8.5 scenarios respectively. Under the same scenarios, glaciers in North Asia would lose 57 ± 40%, 72 ± 38% and 85 ± 30% of their mass, and glaciers in the Caucuses and the Middle East would lose 68 ± 32%, 83 ± 19% and 94 ± 13% of their mass (see also Kraaijenbrink et al., 2017; Rounce et al., 2020).

Although enhanced meltwater from snow and glaciers largely offsets hydrological drought-like conditions (Pritchard, 2019), this effect is unsustainable and may reverse as these cryospheric buffers disappear (medium confidence) (Gan et al., 2015; W. Dong et al., 2018; Huss and Hock, 2018). In the Himalayas and the TIB region higher temperatures will lead to higher glacier melt rates and significant glacier shrinkage and a summer runoff decrease (medium confidence) (Sorg et al., 2014). Glacier runoff in the Asian high mountains will increase up to mid-century, and after that runoff might decrease due to the loss of glacier storage (medium confidence) (Lutz et al., 2014; Huss and Hock, 2018; Rounce et al., 2020).

Compared with the 1990s, the number of lakes in TIB in the 2010s decreased by 2%, whereas total lake area expanded by 25% (S. Wang et al., 2020) due to the joint effect of precipitation increase and glacier retreat. Many new lakes are predicted to form as a consequence of continued glacier retreat in the Himalaya-Karakoram region (Linsbauer et al., 2016). As many of these lakes will develop at the immediate foot of steep icy peaks with degrading permafrost and decreasing slope stability, the risk of glacier lake outburst floods and floods from landslides into moraine-dammed lakes is increasing in Asian high mountains (high confidence) (Haeberli et al., 2017; Kapitsa et al., 2017; Bajracharya et al., 2018; Narama et al., 2018; S. Wang et al., 2020).

Permafrost : Permafrost is thawing in Asia (high confidence). Temperatures in the cold continuous permafrost of north-eastern East Siberia rose from the 1980s up to 2017, and the active layer thicknesses in Siberia and Russian Far East generally increased from late 1990s to 2017 (Romanovsky et al., 2018). The change in mean annual ground temperature for northern Siberia is about +0.1 to +0.3°C per decade since 2000 (Romanovsky et al., 2018). Ground temperature in the permafrost regions of TIB (taking 40% of TIB currently) increased (0.02–0.26°C per decade for different boreholes) during 1980 to 2018, and the active layer thickened at a rate of 19.5 cm per decade (L. Zhao et al., 2020). There is high confidence that permafrost in Asian high mountains will continue to thaw and the active layer thickness will increase (Bolch et al., 2019). The permafrost area is projected to decline by 13.4–27.7% and 60–90% in TIB (L. Zhao et al., 2020) and 32% ± 11% and 76% ± 12% in Russia (Guo and Wang, 2016) by the end of the 21st century under the RCP2.6 and RCP8.5 scenarios respectively (high confidence).

Lake and river ice: Lake ice cover duration got shorter in many lakes in TIB (Yao et al., 2016; Cai et al., 2019; Guo et al., 2020) and some other areas such as north-west China (Cai et al., 2020) and north-east China (Yang et al., 2019) in the last two decades (high confidence). River ice cover extent decreased in TIB as well (H. Li et al., 2020; Yang et al., 2020a). Climate warming also leads to a significant reduction in the period with ice phenomena and the decrement of ice regime hazard in Russian lowland rivers (Agafonova et al., 2017), and the Inner Mongolia reach of the Yellow River in northern China (Wan et al., 2020) (high confidence). Lake ice and river ice in Asia are expected to decline with projected increases in surface air temperature towards the end of this century (high confidence) (Guo et al., 2020; Yang et al., 2020a).

Heavy snowfall and ice storm: Observed trends in heavy snowfall and ice storms are uncertain. Annual maximum snow depth decreased for the period between 1962 and 2016 on the western side of both eastern and western Japan, at rates of 12.3% and 14.6% per decade respectively (MOE et al., 2018). Observational results generally show a decrease in the frequency and an increase in the mean intensity of snowfalls in most Chinese regions (medium confidence) (B. Zhou et al., 2018). Because of the decrease in the snow frequency, the occurrence of large-scale snow disasters in TIB decreased (low confidence) (Qiu et al., 2018; S. Wang et al., 2019). Large parts of northern high-latitude continents (including Siberia and RFE) have experienced cold snaps and heavy snowfalls in the past few winters, and the reduction of Arctic sea ice would increase the chance of heavy snowfall events in those regions in the coming decades (medium confidence) (Song and Liu, 2017). Heavy snowfall is projected to occur more frequently in Japan’s Northern Alps, the inland areas of Honshu Island and Hokkaido Island (Kawase et al., 2016, 2020; MOE et al., 2018), and the heavy wet snowfall can be enhanced over the mountainous regions in central Japan and northern part of Japan (Ohba and Sugimoto, 2020) (medium confidence).

Hail: The hailstorm in the Asian region shows a decreasing trend in several regions (low confidence, limited evidence). In China severe weather days including thunderstorms, hail and/or damaging wind have decreased by 50% from 1961 to 2010 (M. Li et al., 2016; Zhang et al., 2017), and the hail size decreased since 1980 (Ni et al., 2017). A rate of decrease of 0.214 hail days per decade has also been reported for Mongolia between 1984–2013, where the annual number of hail days averaged is 0.74 (Lkhamjav et al., 2017).

Snow avalanche: There is as yet limited evidence for the evolution of avalanches in Asia. Tree-ring-based snow avalanche reconstructions in the Indian Himalayas show an increase in avalanche occurrence and runout distances in recent decades (Ballesteros-Cánovas et al., 2018).

In summary, snowpack and glaciers are projected to continue decreasing and permafrost to continue thawing in Asia (high confidence). There is medium confidence of increasing heavy snowfall in some regions, but limited evidence on future changes in hail and snow avalanches.

Relative sea level: Around Asia, from 1900–2018, a new tide gauge-based reconstruction finds a regional mean RSL change of 1.33 [0.80 to 1.86] mm yr^–1 in the Indian Ocean–Southern Pacific and 1.68 [1.27 to 2.09] mm yr^–1 in the North-west Pacific (Frederikse et al., 2020), compared to a GMSL change of around 1.7 mm yr^–1Section 2.3.3.3 and Table 9.5). For the period 1993–2018, the RSLR rates around Asia, based on satellite altimetry, increased to 3.65 [3.23 to 4.08] mm yr^–1 and 3.53 [2.64 to 4.45] mm yr^–1 respectively (Frederikse et al., 2020), compared to a GMSL change of 3.25 mm yr^–1Section 2.3.3.3 and Table 9.5). The rate of RSLR along the coastline of China ranges from –2.3 ± 1.9 to +5.7 ± 0.4 mm yr^–1 during 1980–2016; after removing the vertical land movement, the average rate of sea level rise is 2.9 ± 0.8 mm yr^–1 over 1980–2016 and 3.2 ± 1.1 mm yr^–1 since 1993 (Qu et al., 2019). However, the rates of land subsidence reported by Minderhoud et al. (2017) are substantially higher than those reported by Qu et al. (2019). RSL change in many coastal areas in Asia, especially in EAS, is affected by land subsidence due to sediment compaction under building mass and groundwater extraction (high confidence) (Erban et al., 2014; Nicholls, 2015; Minderhoud et al., 2019; Qu et al., 2019). During 1991–2016, the Mekong Delta in Vietnam sank on average about 18 cm as a consequence of groundwater withdrawal, and the subsidence related to groundwater extraction has gradually increased with highest sinking rates estimated to be 11 mm yr^–1 in 2015 (Minderhoud et al., 2017).

Relative sea level rise is very likely to continue in the oceans around Asia. Regional mean RSLR projections for the oceans around Asia range from 0.3–0.5 m under SSP1-2.6 to 0.7–0.8 m under SSP5-8.5 for 2081–2100 relative to 1995–2014 (median values), which is within the range of projected GMSL change (Section 9.6.3.3). These RSLR projections may, however, be underestimated due to potential partial representation of land subsidence in their assessment (Section 9.6.3.2).

Coastal flood: The present-day 1-in-100-year ETWL is between 0.5–8 m around Asia, with values above 2.5 m or above common along the coasts of Central and north-eastern Asia (Vousdoukas et al., 2018; Kirezci et al., 2020). Sea level rise and land subsidence will jointly lead to more flooding in delta areas in Asia (high confidence) (Takagi et al., 2016; J. Wang et al., 2018).

Extreme total water level magnitude and occurrence frequency are expected to increase throughout the region (high confidence) (Figure 12.4p–r and Figure 12.SM.6). Across the region, the 5–95th percentile range of the 1-in-100-year ETWL is projected to increase (relative to 1980–2014) by 7–44 cm and by 10–52 cm by 2050 under RCP4.5 and RCP8.5 respectively. By 2100, this range is projected to be 11–91 cm and 28–187 cm under RCP4.5 and RCP8.5 respectively (Vousdoukas et al., 2018; Kirezci et al., 2020). Furthermore, the present-day 1-in-100-year ETWL is projected to have median return periods of around 1-in-50-years by 2050 and 1-in-10-years by 2100 under RCP4.5 in most of Asia, except SEA and ARP, in which the present-day 1-in-100-year ETWL is projected occur once per year or more, both by 2050 and 2100 (Vousdoukas et al., 2018). The present-day 1-in-50-year ETWL is projected to occur around three times a year by 2100 with a SLR of 1 m across Asia (Vitousek et al., 2017). Compound impacts of precipitation change, land subsidence, sea level rise, upstream hydropower development, and local water infrastructure development may lead to larger flood extent and prolonged inundation in the Vietnamese Mekong Delta (Triet et al., 2020).

Coastal erosion: Over the past 30 years, South, South East and East Asia exhibit the most pronounced delta changes globally due to strong human-induced changes to the fluvial sediment flux (Nienhuis et al., 2020). Satellite derived shoreline change estimates over 1984–2015 indicate shoreline retreat rates between 0.5 m yr^–1 and 1 m yr^–1along the coasts of WCA and ARP, increasing to 3 m yr^–1 in SAS. Over the same period, shoreline progradation has been observed along the coasts of RFE (0.2 m yr^–1), SEA (0.1 m yr^–1) and EAS (0.5 m yr^–1) (Luijendijk et al., 2018; Mentaschi et al., 2018). Meanwhile, there has been a gross coastal area loss of 3,590 km² in South Asia, and a loss of 2,350 km² in Pacific Asia, over a 30-year period (1984–2015) (Mentaschi et al., 2018).

Projections indicate that a majority of sandy coasts in the Asia region will experience shoreline retreat (high confidence) (Udo and Takeda, 2017; Ritphring et al., 2018; Vousdoukas et al., 2020b), while parts of the RFE, EAS, SEA and WCA coastline are projected to prograde over the 21st century, if present ambient shoreline change trends continue. Median shoreline change projections (CMIP5), relative to 2010, presented by Vousdoukas et al. (2020b) show that, by mid-century, sandy shorelines in Asia will retreat by between 10–50 m, except in SAS where shoreline retreat is projected to exceed 100 m, under both RCP4.5 and RCP8.5. By 2100, and under RCP4.5, shoreline retreats of around 85, 100 and 300 m are projected along the sandy coastlines of SEA and WCA, ARP and SAS respectively (50 m or less in other Asian regions), while under RCP8.5, over the same period, sandy shorelines along all regions with coastlines, except RFE and EAS, are projected to retreat by more than 100 m, with the retreat in SAS reaching 350 m (2100 RCP8.5 projections for RFE and EAS are about 60 m and about 85 m respectively; Figure 12.6).

Marine heatwave: There have been frequent marine heatwaves (MHW) in the coastal oceans of Asia, connected to the increase between 0.25°C and 1°C in mean SST of the coastal oceans since 1982–1998 (Oliver et al., 2018). There is high confidence that MHWs will increase around most of Asia. Mean SST is projected to increase by 1°C (2°C) around Asia by 2100, with a hotspot of around 2°C (5°C) along the coastlines of the Sea of Japan and the RFE under RCP4.5 (RCP8.5; Interactive Atlas). Under global warming conditions, MHW intensity and duration are projected to increase in the coastal zones of all sub-regions of Asia, but most notably in SEA and SAS (Frölicher et al., 2018). Projections for SSP1-2.6 and SSP5-8.5 both show an increase in MHWs around Asia by 2081–2100, relative to 1985–2014 (Box 9.2, Figure 1).

In general, there is high confidence that most coastal/ocean-related climatic impact-drivers in Asia will increase over the 21st century. Relative sea level rise is very likely to continue around Asia, contributing to increased coastal flooding in low-lying areas (high confidence) and shoreline retreat along most sandy coasts (high confidence). Marine heatwaves are also expected to increase around the region over the 21st century (high confidence).

The assessed direction of change in climatic impact-drivers for Asia and associated confidence levels are illustrated in Table 12.4.

Table 12.4 | Summary of confidence in direction of projected change in climatic impact-drivers in Asia, representing their aggregate characteristic changes for mid-century for scenarios RCP4.5, SSP2-4.5, SRES A1B or above within each AR6 region (defined in Chapter 1), approximately corresponding (for CIDs that are independent of sea level rise) to global warming levels between 2°C and 2.4°C (see Section 12.4 for more details of the assessment method). The table also includes the assessment of observed or projected time-of-emergence of the CID change signal from the natural interannual variability if found with at least medium confidence in Section 12.5.2.

Mean air temperature: Across Australia mean temperatures have increased by 1.44°C ± 0.24°C during the period 1910–2019, with most of the warming occurring since 1950 (Atlas.6.2; CSIRO and BOM, 2020; Trewin et al., 2020). In New Zealand, an increase of 1.1°C has been measured from 1909–2016 (Atlas.6.2; MfE and Stats NZ, 2020). In the period 1980–2014 a rate of increase of 0.1°C–0.3°C per decade has been observed (Atlas.11 and Atlas.20).

Mean temperature in Australasia is projected to continue to rise through the 21st century (virtually certain) (Atlas.6.4). Projections for Australia indicate that the average temperature will increase by +1.1°C (0.84–1.52°C 10–90th percentile range) by 2041–2060 (mid-century), and by +1.9°C (1.29–2.58°C) by 2081–2100 (end-century), relative to the baseline period of 1995–2014, under SSP2-4.5 (Interactive Atlas). For SSP5-8.5, the projected changes are up to +1.5°C (1.17–1.96°C) and +3.7°C (2.75–4.91°C) for mid- and end-century respectively. For SSP1-2.6, mean temperature is projected to rise by +0.9°C (0.55–1.26°C) and +1.0°C (0.55–1.54°C) relative to 1995–2014 by mid- and end-century respectively (Interactive Atlas). In New Zealand, an increase of mean temperature of +1.0°C (0.60–1.32°C) relative to 1995–2014 is projected by mid-century, and an increase of +1.6°C (1.03–2.26°C) by end-century under SSP2-4.5. For SSP5-8.5, the projected increase in mean temperature is +1.3°C (0.91–1.66°C 10–90th percentile range) and +3.1°C (2.20–4.05°C) relative to 1995–2014 by mid- and end-century respectively. For SSP1-2.6, the projected increase in mean temperature is +0.75°C (0.39–1.06°C) and +0.8°C (0.47–1.46°C) relative to 1995–2014 by mid- and end-century, respectively (Interactive Atlas).

Extreme heat: The region has a very likely trend of increasing frequency and severity of hot extremes since the 1950s (Table 11.10). Extreme minimum temperatures have increased in all seasons over most of Australia and exceeds the increase in extreme maximum temperatures (X.L. Wang et al., 2013; Jakob and Walland, 2016). Heatwave characteristics and hot extremes have increased across many Australian regions since the mid-20th century (Table 11.10; CSIRO and BOM, 2020). The number of days per year with maximum temperature greater than 35°C has increased over most parts of Australia from 1957–2015, with the largest increasing trends of 0.4–1 days/year occurring in north-western, Northern, north-eastern Australia and parts of Central Australia (CSIRO and BOM, 2016). Long-term changes of hot extremes in Australia have been attributed to anthropogenic influence (Table 11.10). In New Zealand, the number of annual heatwave days increased at 18 of 30 sites during the period 1972–2019 (MfE and Stats NZ, 2020).

More frequent hot extremes and heatwaves are expected over the 21st century in Australia (virtually certain) (Table 11.10). Heat thresholds potentially affecting agriculture and health, such as 35°C or 40°C, are projected to be exceeded more frequently over the 21st century in Australia under all RCPs (high confidence). By 2090 under RCP4.5, the average number of days per year with maximum temperatures above 35°C is highly spatially variable and is expected to increase by 50–100%, while the number of days per year with maximum temperatures above 40°C is expected to increase by 200%, relative to 1985–2005 (CSIRO and BOM, 2015). Under RCP8.5 the corresponding projected increases are even greater, with a greater than 100% increase in most of Australia, and far greater increases in Central and Northern Australia (up to a 20-fold increase in Darwin). Projections for New Zealand indicate more frequent hot extremes (virtually certain) (Table 11.10). Figure 12.4b, c shows CMIP6 projections of mean number of days per year with maximum temperature exceeding 35°C under SSP5-8.5, which are consistent with the above assessed literature and across the two CMIP generations, and indicate a strong difference depending on the mitigation scenario (e.g., over 100 days more per year under SSP5-8.5 in NAU, but, in general, less than 60 days more per year under SSP1-2.6 in NAU; Figure 12.SM.1).

The projected frequency of exceeding dangerous humid heat thresholds is increasing in Australia, with a strong increase in Northern Australia for RCP8.5 (high confidence) (Zhao et al., 2015; Mora et al., 2017; Brouillet and Joussaume, 2019), consistently across CMIP5, CMIP6 and CORDEX simulations (Figure 12.4d–f and Figure 12.SM.2). Using the HI index, by end-century, the average number of days exceeding 41°C is projected to increase in NAU by about 100 days and by about 25 days under SSP5-8.5 and SSP1-2.6, respectively. The projections for New Zealand indicate no appreciable increase in the number of days with HI > 41°C across SSPs, time periods and CMIP generations (Figure 12.4d–f and Figure 12.SM.2).

Cold spell and frost: Excepting parts of Southern Australia, the Australasian region has a significant trend of decreasing frequency in cold extremes since the 1950s (high confidence) (Table 11.10) and there is high confidence that such trends are attributable to anthropogenic influence (Table 11.10). The number of frost days per year in Australia has on average declined at a rate of 0.15 days/decade in the past century (Alexander and Arblaster, 2017), except in some regions of Southern Australia, where an increase in both number and season length has been reported (Dittus et al., 2014; Crimp et al., 2016b). The number of frost days has decreased at 12 of 30 monitoring sites around New Zealand over the period 1972–2019 (MfE and Stats NZ, 2020).

Less frequent cold extremes are virtually certain in Australasia (Table 11.10) while a decrease of frost days is projected with high confidence for the region. Projections, relative to 1986–2005, for the number of frost days per year in Australia indicate declines of 0.9 days by mid-century and 1.1 days by end-century for RCP4.5, while for RCP8.5, the projected declines are 1.0 days and 1.3 days by mid- and end-century respectively (Alexander and Arblaster, 2017; Herold et al., 2018). Projections for New Zealand indicate that the number of frost days will decrease by 30% (RCP2.6) to 50% (RCP8.5) by 2040, relative to 1986–2005. By 2090, the decrease ranges from 30% (RCP2.6) to 90% (RCP8.5) (MfE and Stats NZ, 2017).

In general, there is high confidence that most heat hazards in Australasia will increase and that cold hazards will decrease over the 21st century. The mean temperature in Australasia is virtually certain to continue to rise through the 21st century, accompanied by less frequent cold extremes (virtually certain) and frost days (high confidence), and more frequent hot extremes (virtually certain). Heat stress is projected to increase in Australia (high confidence).

Mean precipitation: Here, only increases in precipitation (under ‘Wet’) are addressed, with decreases (under ‘Dry’) addressed in ‘Aridity’ below.

In terms of wet climatic impact-drivers, detectable anthropogenic increases in precipitation in Australia have been reported particularly for north-central Australia for the period 1901–2010 (Knutson and Zeng, 2018). Figure Atlas.11 indicates no significant trend in precipitation over the region during the baseline period 1960–2015, except for the Global Precipitation Climatology Project (GPCP) dataset, which shows an increasing trend in north-central Australia. In New Zealand, increases in annual rainfall have been observed between 1960–2019 in the south and west of the South Island and east of the North Island. Note however, for the most part, the above reported trends in New Zealand have been classified as statistically not significant (Figure Atlas.20).

Annual mean precipitation is projected to increase in Central and north-east Australia (low confidence) and in the south and west of New Zealand (medium confidence) (Atlas.6.4). Liu et al. (2018a) show that under 1.5°C warming, Central and north-east Australia will become wetter. Projected patterns in annual precipitation exhibit increases in the west and south of New Zealand (Atlas.6.4; Liu et al., 2018a) and project that the South Island will be wetter under both 1.5°C and 2°C warming. However, there is limited model agreement for projected rainfall changes in Australasia as shown in the Atlas.

River flood: Streamflow observations in Australia have shown that negative trends dominate in annual maximum flow and that stations with significant negative trends were mostly located in the south-east and south-west (Gu et al., 2020). The observed peak flow trend in Southern Australia is attributed to the decrease of soil moisture, although an increase of flood magnitude is possible for very rare events. For the more frequent flood events, the increase of extreme precipitation is balanced by the decrease of soil moisture. (Wasko and Nathan, 2019).

While median annual runoff is projected to decrease in most of Australia (Chiew et al., 2017), consistent with projected decreases in average rainfall (CSIRO and BOM, 2015; Alexander and Arblaster, 2017), river floods are projected to increase due to more intense extreme rainfall events and associated increase in runoff (medium confidence). Asadieh and Krakauer (2017) found a decrease in the value of the 95% percentile of mean streamflow with RCP8.5 by the end of the century in all of Australia, except in a small part in centre of the country. In terms of relative increases, flooding is expected to increase more in Northern Australia (driven by convective rainfall systems) than in Southern Australia (where more intense extreme rainfall may be compensated by drier antecedent moisture conditions; Alexander and Arblaster, 2017; Dey et al., 2019) with flood frequency increasing in Northern Australia and along parts of the east coast and decreasing in south-western Western Australia (Hirabayashi et al., 2013). Gu et al. (2020) project larger flood magnitude and volumes under both RCP2.6 and RCP8.5 in Northern Australia, and smaller flood magnitudes and volumes in Southern Australia under the same RCPs. These findings are in general agreement with the patterns in peak flow, corresponding to the 1-in-100-year return period streamflow, shown in Figure 12.7a,c for mid-21st century under RCP8.5.

There is medium confidence that river flooding will increase in New Zealand. Projections for New Zealand indicate that the 1-in-50-year and 1-in-100-year flood peaks for rivers in many parts of the country may increase by 5 to 10% by 2050 and more by 2100 (with large variation between models and emissions scenarios), with a corresponding decrease in return periods for specific flood levels (Gray et al., 2005; Carey-Smith et al., 2010; McMillan et al., 2010, 2012; Ballinger et al., 2011).

Heavy precipitation and pluvial flood: Rainfall extremes have been detected to increase in Australasia, with low confidence (Table 11.10). There is high confidence that R × 1 day and R × 5 day precipitation extremes will increase for 2°C or lower warming for the region as a whole, but on a sub-regional basis there is only medium confidence of increases in NAU and CAU and low confidence of increases on EAU, SAU and NZ. For warming levels exceeding 2°C, these extremes are very likely to increase in NAU and CAU and they are likely to increase elsewhere in the region (Section 11.9).

Landslide: Based on local slope characteristics, lithology and seismic activity, the South Island and the eastern half of the North Island of New Zealand are vulnerable to landslide occurrence (Broeckx et al., 2020). The potential for land and rockslides increases with, amongst other factors, total precipitation rates, precipitation intensity, mountain permafrost thaw rates, glacier retreat and air temperature (Crozier, 2010; Allen and Huggel, 2013; Gariano and Guzzetti, 2016; IPCC, 2019a). Given the increase of the magnitude of these physical variables in areas that are already highly susceptible to mass movements (MfE, 2018), there is low confidence that the occurrence of landslides will increase under future climate conditions.

Aridity: In terms of dry climatic impact-drivers, a substantial decrease in precipitation has been observed across Southern Australia during the cool season (April–October) (medium confidence). The drying trend has been particularly strong over south-west Western Australia between May and July, with rainfall since 1970 being around 20% less than the 1900–1969 average (CSIRO and BOM, 2020). Detectable decreases in mean precipitation, attributable at least in part to anthropogenic forcing, have been reported for parts of south-west Australia (Delworth and Zeng, 2014; Knutson and Zeng, 2018), south-east Australia, and Tasmania (Knutson and Zeng, 2018). In New Zealand, the north-east of the South Island and western and the northern parts of the North Island show decreasing precipitation trends during 1960–2019 (MfE and Stats NZ, 2020).

Aridity is projected to increase, especially during winter and spring, with medium confidence in SAU but with high confidence in south-west Western Australia (Table 11.11 and Atlas.6.4). In EAU and in the north and east of NZ, aridity is projected to increase with medium confidence, while a decrease is projected with medium confidence in the south and west of NZ (Atlas.6.4). Although there is only low confidence in the projected decrease of mean annual precipitation in south-western and eastern Australia and the north and east of New Zealand, there is high confidence of reduced winter and spring precipitation in Australia in future, mostly in south-western and eastern Australia (Atlas.6.4). Liu et al. (2018b) show that under 2°C warming, most of Australia is projected to become drier based on the Palmer Drought Severity Index (PDSI), with the exception of the tropical north-east. Ferguson et al. (2018) project that between 1976–2005 and 2070–2099, winters will become drier (mainly in Southern Australia) under RCP8.5. Liu et al. (2018b) project that the North Island of New Zealand will be drier under both 1.5°C and 2°C warming.

Hydrological drought: There is low confidence of observed changes in hydrological droughts in Australasia, except in SAU where there is medium confidence of an observed increase in the south-east and south-west. Future projections indicate medium confidence in further hydrological drought increases for Southern Australia for warming levels of 2°C or higher (Section 11.9). Mean annual runoff in far south-east and far south-west Australia are projected to decline by median values of 20 and 50% respectively, by mid-century under RCP8.5 (Chiew et al., 2017). Prudhomme et al. (2014) assess changes in the Drought Index (DI), defined as areal runoff less than the 10th percentile over the reference period 1976–2005, and project DI increases for both Australia and New Zealand by 10–20% by 2070–2099 under RCP8.5, with the greatest effects being in the southern parts of the Australian continent. These projections are consistent with the trends shown in Figure 12.4g–i (Figure 12.SM.3). The SPI drought frequency is projected to increase in SAU and particularly in south-west Western Australia by mid-century, while by the end of the century SPI drought frequency is projected to increase all over Australia, and particularly strongly in south-west Western Australia as well as southern Victoria (see Figure 12.4g–i). For the Murray–Darling basin, Ferguson et al. (2018) project effectively no change (–1%) in mean precipitation, a 27% decrease in P–E, and 30% increase in runoff in 2070–2099 relative to 1976–2005 with RCP8.5.

Agricultural and ecological drought: There is medium confidence in observations of agricultural and ecological droughts increasing in SAU and decreasing in NAU, while there is low confidence of changes elsewhere in the region (Section 11.9). More regional studies have observed an increase in agricultural and ecological drought intensity in south-west Australia and an increase in drought intensity in parts of south-east Australia, while the length of droughts therein has increased (Section 11.9). In New Zealand, since 1972–73, soils at 7 of 30 monitored sites became drier, while the 2012–13 drought was one of the most extreme in the previous 41 years (MfE and Stats NZ, 2017). Future evaporative demand is projected to lead to medium confidence increases in agricultural and ecological droughts for 2°C of global warming in SAU and EAU and low confidence for changes in CAU, NAU and NZ, although there is medium confidence of increases in CAU with 4°C of global warming (Section 11.9). There is medium confidence for more time in agricultural and ecological drought in SAU by mid-21st century (Coppola et al., 2021b) as well as by the end of the 21st century (Herold et al., 2018). The Standardized Precipitation Evapotranspiration Index (SPEI) shows a springtime intensification in SAU with moderate and severe droughts in the south-west and moderate droughts in the south-east (Herold et al., 2018). There is consensus among the different model ensembles (CORDEX-CORE, CMIP5 and CMIP6) that the drought frequency (DF), one of several proxies for agricultural and ecological drought, will increase in all four Australian regions for both mid-century (NAU 0.2–2 DF increase, CAU 0.5–2 DF increase, SAU 1–3 DF increase and EAU 0.8–3 DF increase) and end-century (0.8–2.7 DF increase for NAU, 1.2–2 DF increase for CAU, 2.2–3.8 for SAU and 0.2–3 for EAU) for both RCP8.5 and SSP5-8.5, with CMIP6 showing the lowest increase (Figure 12.4g–l and Figure 12.SM.4; Coppola et al., 2021b).

Fire weather: Dowdy and Pepler (2018) examined atmospheric conditions conducive to pyroconvection in the period 1979–2016, and found an increased risk in south-east Australia during spring and summer, due to changes in vertical atmospheric stability and humidity, in combination with adverse near-surface fire weather conditions. CSIRO and BOM (2018) and Dowdy (2018) found that the annual 90th percentile daily Forest Fire Danger Index (FFDI) has increased from 1950–2016 in parts of Australia, especially in Southern Australia (1–2.5 per decade) and in spring and summer. These studies indicate an increase in the frequency and magnitude of FFDI extreme quantiles, as well as a shift of the fire season start towards spring, lengthening the fire season. The unprecedented large fires of austral spring and summer of 2019 in south-east Australia were a result of extreme hot and dry weather in significantly drier than average conditions that had persisted since 2017, in combination with consistently stronger than average winds, resulting in above average to highest on record FFDI values in much of the country (Abram et al., 2021). These fires have been attributed to climate change through the temperature component of fire weather indices (van Oldenborgh et al., 2021). In New Zealand, days with very high and extreme fire weather increased in 12 out of 28 monitored sites, and decreased in 8, in the period 1997–2019 (MfE and Stats NZ, 2020). Attribution studies indicate that there is medium confidence of an anthropogenically driven past increase in fire weather conditions, essentially due to increase in frequency of extreme heat waves. (Hope et al., 2019; Lewis et al., 2020; van Oldenborgh et al., 2021).

Fire weather indices are projected to increase in most of Australia (high confidence) and many parts of New Zealand (medium confidence), in particular with respect to extreme fire and induced pyroconvection (Dowdy et al., 2019b). Increasing mean temperature, cool season rainfall decline, and changes in tropical climate variability all contribute to a future increase in extreme fire risk in Australia (Abram et al., 2021). Projections indicate that the annual cumulative FFDI will increase by 31–33% in Southern and Eastern Australia, and by 17–25% in Northern Australia and the Rangelands by 2090 (relative to 1995) under RCP8.5 (CSIRO and BOM, 2015). Using a CMIP5 ensemble of 17 models, Abatzoglou et al. (2019) found a statistically significant positive trend for fire weather intensity and fire season length for future mid-century conditions under RCP8.5, including a detectable anthropogenic influence on fire risk magnitude and fire season length by 2040 in Western Australia and along the Queensland coastline. Using the C-Haines and FFDI indices with A2 and RCP8.5 respectively, Di Virgilio et al. (2019) and Clarke et al. (2019) have shown that extreme fire weather frequency will increase in south-eastern Australia by the end of the 21st century. Most of these projections indicate that the biggest increases in fire weather conditions will be in late spring, effectively resulting in longer (stronger) fire seasons in areas where spring is the shoulder (peak) season. In New Zealand, Watt et al. (2019) projected that the number of days with very high to extreme fire risk will increase by 71% by 2040, and by a further 12% by 2090, for the A1B scenario, with fire risk increase all along the east coast. The most marked relative changes by 2090 were projected for Wellington and Dunedin, where very high to extreme fire risk is projected to increase by, respectively, 89% to 32 days and 207% to 18 days, compared to the baseline period 1970–1999.

Annual mean precipitation is projected to increase in Central and north-east Australia (low confidence) and in the south and west of New Zealand (medium confidence), while it is projected to decrease in Southern Australia (medium confidence), albeit with high confidence in south-west Western Australia, in Eastern Australia (medium confidence), and in the north and east of New Zealand (medium confidence). Heavy precipitation and pluvial flooding are projected to increase with medium confidence in Northern Australia and Central Australia. There is medium confidence that river flooding will increase in New Zealand and Australia, with higher increases in Northern Australia. Aridity is projected to increase with medium confidence in Southern Australia (high confidence in south-west Western Australia), Eastern Australia (medium confidence), and in the north and east of New Zealand (medium confidence). Hydrological droughts are projected to increase in Southern Australia (medium confidence), while agricultural and ecological droughts are projected to increase with medium confidence in Southern Australia and Eastern Australia. Fire weather is projected to increase throughout Australia (high confidence) and New Zealand (medium confidence).

Mean wind speed: There is low confidence of a mean wind speed trend in the last decades (low agreement) (McVicar et al., 2012; Troccoli et al., 2012; Azorin-Molina et al., 2018; J. Wu et al., 2018), as long-term measurements are not homogeneous.

In future climate scenarios wind speed trends in Australia exhibit generally weak amplitudes with low agreement among models (Figure 12.4m–o and Figure 12.SM.5) with uncertain consequences on wind power potential (CSIRO and BOM, 2015; Karnauskas et al., 2018a; Jung and Schindler, 2019). However, there is medium confidence that, by the end of the century, annual mean wind power will significantly increase in north-eastern Australia under RCP8.5, but there is low confidence of an increase by end-century under RCP4.5, and for any scenario by mid-century (Karnauskas et al., 2018a). In New Zealand, mean wind patterns are projected to become more north-easterly in summer, and westerlies to become more intense in winter (low confidence), in agreement with the strengthening of the Southern Hemisphere storm tracks (Section 4.5.1).

Severe wind storm: There is generally low confidence in observed changes in extreme winds and extratropical storms in Australasia (Section 11.7.2). CMIP5 projections of severe winds indicate a general increase in north-eastern Australia, and decreases in some parts in Southern and Central Australia (medium confidence) by the end of the century under RCP8.5 (CSIRO and BOM, 2015; Kumar et al., 2015; Jung and Schindler, 2019). Elsewhere trends are diverse and vary across simulations with low agreement. Projections of changes in the 1-in-25-year return period winds (based on annual maxima) for 2074–2100 relative to 1979–2005 for RCP8.5 show an increase in tropical areas of Northern Australia (Kumar et al., 2015).

In New Zealand, the frequency and magnitude of extreme winds have decreased (from 1980–2019) at 12 of 14 monitored sites and increased at two monitored sites (MfE and Stats NZ, 2020). Due to the intensification and the shift of the austral storm track by the end of the century (Yin, 2005), increases in extreme wind speed in New Zealand are projected over the South Island and the southern part of the North Island by mid- and end-century for all RCPs (low confidence) (MfE, 2018).

Tropical cyclone: In Australia, the number of TCs has generally declined since 1982, and the frequency of intense TCs that make landfall in north-eastern Australia has declined significantly since the 19th century (medium confidence) (Kuleshov et al., 2010; Callaghan and Power, 2011; Holland and Bruyère, 2014; Knutson et al., 2019; CSIRO and BOM, 2020). There is high confidence that cyclones making landfall along north-eastern and northern Australian coastlines will decrease in number and low confidence of an increase in their intensities for 2°C of global warming as well as for the mid-century period with scenarios RCP4.5 and above (Roberts et al., 2015, 2020; Bacmeister et al., 2018; Knutson et al., 2020), with the amplitude of changes increasing from RCP4.5 to RCP8.5 (Bacmeister et al., 2018). Decreases in frequency are projected for ‘east coast lows’ (Walsh et al., 2016b; Dowdy et al., 2019a).

Sand and dust storm: Australia is recognized to be the largest dust source in the Southern Hemisphere (Zheng et al., 2016). Land-use and land-cover change have increased dust emissions in Australia in the past 200 years (Marx et al., 2014). While projections suggest a decrease in severe winds in Central and Southern Australia, changes in vegetation due to increased aridity and hydrological drought could be expected to result in increased wind erosion and dust emission across the country (medium confidence) (Webb et al., 2020).

In Australasia, there is low confidence in projected mean wind speeds and wind power potential, with a medium confidence increase projected only in north-eastern Australia under high emissions scenarios and by the end of the 21st century. Tropical cyclones in north-eastern and North Australia are projected to decrease in number (high confidence) while their intensity is projected to increase (low confidence).

Snow: The snow season length in Australia has decreased by 5% during 2000–2013 relative to 1954–1999, especially in spring (Pepler et al., 2015). A shift in the date of peak snowfall has also been observed with an 11-day advance over the same period (Pepler et al., 2015). A decreasing trend in maximum snow depth has been observed for Australian alpine regions since the late 1950s, with the largest declines during spring and at lower altitudes. Maximum snow depth is highly variable and is strongly influenced by rare heavy snowfall days, which have no observed trends in frequency (CSIRO and BOM, 2020).

Projections for Southern Australia and New Zealand show a continuing reduction in snowfall during the 21st century (high confidence). The magnitude of decrease varies with the altitude of the region and the emissions scenario. At elevations lower than 1500 m, years without snowfall are projected from 2030 in some models. By 2090, and under RCP8.5, such years are projected to become common (CSIRO and BOM, 2015). The number of annual snow days in New Zealand is projected to decrease under all RCPs, by up to 30 days or more by 2090 under RCP8.5, relative to 1986–2005 (MfE, 2018).

Glacier: Glacier mass and areal extent in New Zealand is projected to continue to decease over the 21st century (high confidence) (Section 9.5.1.3). Glacier ice volume from 1977–2018 in New Zealand has decreased from 26.6 to 17.9 km³ (a loss of 33%; Salinger et al., 2019). Relative to 2015, glaciers in New Zealand are projected to lose 36 ± 44%, 53 ± 33% and 77 ± 27% of their mass by the end of the century under RCP2.6, RCP4.5 and RCP8.5 respectively, with the loss rates decreasing over time under RCP2.6 and increasing under RCP8.5 (Marzeion et al., 2020).

In summary, snowfall is expected to decrease throughout the region at high altitudes in both Australia (high confidence) and New Zealand (medium confidence). In New Zealand, glacier ice mass and extent are expected to decrease over the 21st century for all scenarios (high confidence).

Relative sea level: Around Australasia, from 1900–2018, a new tide gauge-based reconstruction finds a regional mean RSL change of 1.33 [0.80 to 1.86] mm yr^–1 in the Indian Ocean–South Pacific region (Frederikse et al., 2020), compared to a GMSL change of around 1.7 mm yr^–1Section 2.3.3.3 and Table 9.5). For the period 1993–2018, the RSLR rates, based on satellite altimetry, increased to 3.65 [3.23 to 4.08] mm yr^–1 (Frederikse et al., 2020), compared to a GMSL change of 3.25 mm yr^–1Section 2.3.3.3 and Table 9.5).

Relative sea level is virtually certain to increase throughout the region over the 21st century (Section 9.6.3, Figure 9.28). Regional mean RSLR projections for the oceans around Australasia range from 0.4–0.5 m under SSP1-2.6 to 0.7–0.9 m under SSP5-8.5 for 2081–2100 relative to 1995–2014 (median values), which means local RSL change falls within the range of mean projected GMSL change (Section 9.6.3.1). However these RSLR projections may be underestimated due to potential partial representation of land subsidence (Section 9.6.3.2).

Coastal flood: The most commonly used index for episodic coastal inundation in Australia is the summation of a high end SLR and the 1-in-100-year storm tide level (the combined sea level due to storm surge and tide) (CSIRO and BOM, 2016; McInnes et al., 2016). However, episodic coastal flooding is caused by extreme total water levels (ETWL), which is the combination of SLR, tides, surge and wave setup (Section 12.3.5.2). The present-day 1-in-100-year ETWL is between 0.5–2.5 m around most of Australia, except the north-western coast where 1-in-100-year ETWL can be as large as 6–7 m (Vousdoukas et al., 2018; O’Grady et al., 2019; Kirezci et al., 2020).

Extreme total water level magnitude and occurrence frequency are expected to increase throughout the region (high confidence) (Figure 12.4p–r and Figure 12.SM.6). Across the region, the 5–95th percentile range of the 1-in-100-year ETWL is projected increase (relative to 1980–2014) by 5–35 cm and by 10–40 cm by 2050 under RCP4.5 and RCP8.5 respectively (Figure 12.4q). By 2100 (Figure 12.4p,r), this range is projected to be 25–80 cm and 50–190 cm under RCP4.5 and RCP8.5 respectively (Vousdoukas et al., 2018; Kirezci et al., 2020). Furthermore, the present-day 1-in-100-year ETWL is projected to have median return periods of around 1-in-20-years by 2050 and 1-in-1-year by 2100 in SAU and NZ and return periods of around 1-in-50-years by 2050 and 1-in-20-years by 2100 in NAU under RCP4.5 (Vousdoukas et al., 2018), while the present-day 1-in-50-year ETWL is projected to occur around three times a year by 2100 with a SLR of 1 m around Australasia (Vitousek et al., 2017).

Coastal erosion: Satellite derived shoreline retreat rates for the period between 1984–2015 show retreat rates between 0.5 and 1 m yr^–1around the region, except in SAU where a shoreline progradation rate of 0.1 m yr^–1has been observed (Luijendijk et al., 2018; Mentaschi et al., 2018). Mentaschi et al. (2018) report a coastal area loss of 350 km² over the same period in Western Australia from satellite observations.

Projections indicate that a majority of sandy coasts in the region will experience shoreline retreat, throughout the 21st century (high confidence) (Figure 12.7b,d). Median shoreline change projections (CMIP5) under both RCP4.5 and RCP8.5 presented by Vousdoukas et al. (2020b) show that, by mid-century, sandy shorelines will retreat (relative to 2010) by between 50 and 80 m all around Australasia, except in SAU and NZ where the projected retreat (relative to 2010) is between 35 and 50 m. By 2100, median shoreline retreats exceeding 100 m (relative to 2010) are projected along the sandy coasts of NAU (about 150 m), CAU (about 160 m), and EAU (about 110 m) under RCP4.5m, while projections for SAU and NZ are around 80–90 m. Under RCP8.5, shoreline retreat exceeding 100 m is projected all around the region by 2100 (relative to 2010) with retreats as high as 220 m in NAU and CAU (about 170 m in EAU and about 130 m in SAU and NZ; Figure 12.7b,d). The total length of sandy coasts in Australasia that is projected to retreat by more than a median of 100 m by 2100 under RCP4.5 and RCP8.5 is about 12,500 and 16,000 km respectively, an increase of approximately 30%.

Distinct from long-term coastline recession, storms and storm surges also result in episodic coastal erosion. In general, the historically measured maximum episodic coastal erosion (either eroded volume or coastline retreat distance) or that due to a 1-in-100-year return period storm wave height is used as a design criterion for coastal zone management and planning in Australia (Wainwright et al., 2014; Mortlock et al., 2017).

While there is wide recognition in Australia that the combined effect of SLR, changing storm surge and wave climates will directly affect future episodic coastal erosion (McInnes et al., 2016; Ranasinghe, 2016; Harley et al., 2017) only a few projections of how this hazard may evolve are available for Australia. In one such study, Jongejan et al. (2016) provide projections of how the full exceedance probability curve of the maximum erosion per year may evolve over the 21st century (due to the combined action of SLR, storm surge and storm waves). Their results show that, for example, the 0.01 exceedance probability maximum coastline retreat in 2025 will have an exceedance probability of 0.015 by 2050 and 0.07 by 2100.

Marine heatwave: The mean SST of the ocean around Australia and east of New Zealand has warmed at a rate of about 0.22°C per decade between 1992 and 2016 (Wijffels et al., 2018), which is higher than the global average SST increase of 0.16°C per decade (Oliver et al., 2018). This mean ocean surface warming is connected to longer and more frequent marine heatwaves in the region (Oliver et al., 2018). Over the period 1982–2016, the coastal ocean of Australia experienced on average more than 1.5 marine heatwaves (MHWs) per year, with the north coast of Western Australia and the Tasman Sea experiencing on average 2.5–3 MHWs per year. The average duration was between 10 and 15 days, with somewhat longer and hotter MHWs in the Tasman Sea. In New Zealand, the south-east coast of South Island experiences the most MHWs (2.5–3 per year). The duration of MHW in New Zealand is on average 10–15 days (Oliver et al., 2018). Changes around Australasia over the 20th century, derived from MHW proxies, show an increase in frequency between 0.3 and 1.5 MHW per decade, except along the south-east coast of New Zealand (Box 9.2); an increase in duration per event; and the total number of MHW days per decade, with the change being stronger in the Tasman Sea than elsewhere (Oliver et al., 2018).

There is high confidence that MHWs will increase around most of Australasia. Under RCP4.5 and RCP8.5 respectively, mean SST is projected to increase by 1°C and 2°C around Australia by 2100, with a hotspot of around 2°C for RCP4.5 and of 4°C for RCP8.5 along the south-east coast between Sydney and Tasmania (Interactive Atlas). Under all RCPs, the mean SST around Australia is expected to increase in the future, with median values of around 0.4°C–1.0°C by 2030 under RCP4.5, and 2°C–4°C by 2090 under RCP8.5 (CSIRO and BOM, 2015). Warming is expected to be largest along the north-west coast of Australia, southern Western Australia, and along the east coast of Tasmania (CSIRO and BOM, 2018). More frequent, extensive, intense and longer lasting MHWs are projected around Australia and New Zealand for GWLs of 1.5°C, 2°C and 3.5°C relative to the modelled reference value for 1861–1880 (Frölicher et al., 2018). Projections for SSP1-2.6 and SSP5-8.5 both show an increase in MHWs around Australasia by 2081–2100, relative to 1985–2014 (Box 9.2, Figure 1).

In general, there is high confidence that most coastal/ocean-related hazards in Australasia will increase over the 21st century. Relative sea level rise is virtually certain to continue in the oceans around Australasia, contributing to increased coastal flooding in low-lying areas (high confidence) and shoreline retreat along most sandy coasts (high confidence). Marine heatwaves are also expected to increase around the region over the 21st century (high confidence).

The assessed direction of change in climatic impact-drivers for Australasia and associated confidence levels are illustrated in Table 12.5, together with emergence time information (Section 12.5.2). No assessable literature could be found for hail and snow avalanches, although these phenomena may be relevant in parts of the region.

Table 12.5 | Summary of confidence in direction of projected change in climatic impact-drivers in Australasia, representing their aggregate characteristic changes for mid-century for scenarios RCP4.5, SSP2-4.5, SRES A1B, or above within each AR6 region (defined in Chapter 1), approximately corresponding (for CIDs that are independent of sea level rise) to global warming levels between 2°C and 2.4°C (see Section 12.4 for more details of the assessment method). The table also includes the assessment of observed or projected time-of-emergence of the CID change signal from the natural interannual variability if found with at least medium confidence in Section 12.5.2.

Mean air temperature: New literature confirms a continuous warming since the beginning of the 20th century in the majority of the eight sub-regions (Atlas.7). However, observational datasets in several areas are still short and trend estimation is hindered by year-to-year and interannual variability. Atlas projections point to a virtually certain warming across all sub-regions, with the largest increases taking place in the Amazon basin (NSA and SAM; Atlas.7.2.4). A consistent increase in temperature-related indices linked to several climate-sensitive sectors (e.g., growing degree days, cooling degree days) is found across CMIP5, CMIP6 and CORDEX-CORE projections, with smaller increases for cooling degree days in mid-latitude regions than in SCA and the Amazon (Coppola et al., 2021b). Daily mean temperature exceedances of a typical 21.5°C threshold for a successful incubation of disease pathogens inside many mosquito vectors (Lambrechts et al., 2011; Blanford et al., 2013; Mordecai et al., 2013, 2017) will be crossed much more frequently, potentially driving increases in the incidence of vector-borne diseases (Laporta et al., 2015; Messina et al., 2019).

Extreme heat: Chapter 11 found high confidence of increased heatwaves in all regions except SSA over the past decades. There is evidence of increasing heat stress over summer in much of SES and SWS using the wet bulb globe temperature (WBGT) index for the period 1973–2012, and this has been attributed to human influence on the climate system (Knutson and Ploshay, 2016). Climate change projections point to major increases in several heat indices across the region for all scenarios (high confidence). Largest increases in the frequency of hot days (maximum temperatures, Tx > 35°C) are projected for the Amazon basin under SSP5-8.5 with more than 200 days per year at the end of the century under SSP5-8.5 relative to 1995-2014, while such increases remain moderate (50–100 days) in SSP1-2.6. For the dangerous heat threshold of HI > 41°C, increases in frequency are similar to that in Tx > 35°C (Figure 12.4 and Figures 12.SM.1 and 12.SM.2; Coppola et al., 2021b; Schwingshackl et al., 2021).

Cold spell and frost: A decreasing frequency of cold days and nights has been observed in many sub-regions (Table 11.13). There is medium confidence (limited agreement) of a decrease in frost days in SWS, SES and SSA. Projections consistently suggest a general decrease in the frequency of cold spells and frost days in the region as indicated by several indices based on minimum temperature (Chou et al., 2014; López-Franca et al., 2016; C. Li et al., 2021). Heating degree days are consistently projected to decrease by 5 degree days per year in the Amazon region, and up to 20–30 degree days per year in NWS, SWS and SES, under RCP8.5/SSP5-8.5 by mid century (Coppola et al., 2021b).

In conclusion, it is virtually certain that warming will continue everywhere in Central and South America and there is high confidence that by the end of the century most regions will undergo extreme heat stress conditions much more often than in recent past (e.g., increase of dangerous heat with HI > 41°C, or Tx > 35°C) with more than 200 additional days per year under SSP5-8.5, while such conditions will be met typically 5 0–1 00 more days per year under SSP1-2.6 over the same regions. Cold spells and frost days will have a decreasing trend (high confidence).

Mean precipitation: The (Atlas documents diverse historical precipitation trends in the region, including a small but not significant increasing trend in SCA, a decreasing trend in south-eastern and north-eastern Brazil, and an increasing trend in SSA. Projections indicate a drying signal for SCA (medium confidence) (Coppola et al., 2014a; Nakaegawa et al., 2014), NES and SWS (high confidence) (Atlas.7.2.5) and the well-known dipole for South America, meaning increasing precipitation over subtropical regions like the Río de La Plata basin (SES) (high confidence) and decreasing precipitation in the Amazon (NSA) (medium confidence) (Chou et al., 2014; Llopart et al., 2014; Reboita et al., 2014; Sánchez et al., 2015; Teichmann et al., 2021). These features are consistent with observations (Sena et al., 2018) and are evident in regional and global model projections by mid- and end-of-century for both RCP4.5 and RCP8.5 (Jones and Carvalho, 2013).

River flood: Emerging literature in the region documents ongoing changes in river floods. Mernild et al. (2018) report decreases and increases in annual runoff west of the Andes Cordillera’s continental divide, with the greatest decreases in the number of low (<10th percentile) runoff conditions and the greatest increases in high (>90th percentile) runoff conditions. In coastal north-east Peru, extreme precipitation events recently caused devastating river floods and landslides (Son et al., 2020). In Brazil, floods are becoming more frequent and intense in wet regions but less frequent and intense in drier regions (Bartiko et al., 2019; Borges de Amorim and Chaffe, 2019), with higher propagation of hydrological changes through anthropogenically modified agricultural basins (Chagas and Chaffe, 2018). Record, catastrophic, unprecedented, and once-in-a-century flooding events have also been reported in recent decades in the tributaries of the Amazon River or along its mainstream (Sena et al., 2012; Espinoza et al., 2013; Marengo et al., 2013; Filizola et al., 2014), in Argentinean rural and urban areas (Barros et al., 2015), in the lower reaches of the Atrato, Cauca and Magdalena rivers in Colombia (Hoyos et al., 2013; Ávila et al., 2019), in basins whose mainstreams flow through important metropolitan areas such as Concepción, Chile (Rojas et al., 2017), and even in one of Earth’s driest regions, the Atacama Desert (Wilcox et al., 2016). In the Amazon basin, the significant increase in extreme flow is associated with the strengthening of the Walker circulation (Barichivich et al., 2018).

Available projections for the region show increases in river floods in SES and SAM (medium confidence). Projections indicate that SES and the coasts of Ecuador and Peru will experience a tendency towards wetter conditions that can be a proxy for longer periods of flooding and enhanced river discharges (Zaninelli et al., 2019). CORDEX models project the strongest changes for the peak flow with a return period of 100 years in SES by mid-century and under RCP8.5 (Figure 12.8). At the continental scale, on the contrary, Alfieri et al. (2017) suggest that 100-year river floods will decrease under RCP8.5. Regional projections of river floods have high uncertainty, however, owing to differences in hydrological models (low confidence) (Reyer et al., 2017a). Fábrega et al. (2013) projected increases in surface runoff for Panama, while Zulkafli et al. (2016) identified increases in 100-year floods of 7.5 and 12.0% in projections for the Peruvian Amazon wet season under RCPs 4.5 and 8.5 respectively. Wetter conditions and ±20% variations in annual mean streamflow are also projected for the Río de La Plata under the warming levels of 1.5°C, 2°C and 3°C above pre‐industrial conditions (Montroull et al., 2018). In central Chile, 50-year peak flows are expected to be greater by mid-century than 100-year peak flows observed over the reference period (Bozkurt et al., 2018).

Figure 12.8 | Projected changes in selected climatic impact-driver indices for Central and South America.

Open figure

Heavy precipitation and pluvial flood: Table 11.14 indicated that there is low confidence due to limited evidence of extreme precipitation trends in almost all Central and South America, except in SES where increases in the magnitude and frequency of heavy precipitation have been observed (high confidence). In general, data scarcity persists for a representative continental assessment. Chapter 11 projections indicate low confidence of increase, compared to the modern period, in the intensity and frequency of heavy precipitation in SCA and SWS for all GWLs, and medium confidence of increase in NSA, NES, SSA, SAM and SES for GWL of 4°C. In NWS, a wide range of changes is projected (low confidence).

Landslide: Several regions in Central America, as well as Colombia and south-eastern Brazil, are considered areas of high incidence of observed fatal landslides. In these areas, El Niño–Southern Oscillation (ENSO)-driven fluctuations in rainfall amounts (Sepúlveda and Petley, 2015) and climate change (Nehren et al., 2019) seem to be key factors. Rockfalls, ice- and rock-ice avalanches, lahars and landslides have been reported frequently in the southern, extratropical Andes in the last decades (Gariano and Guzzetti, 2016). A large number of ice- and moraine-dammed lakes have consequently failed, causing floods that rank amongst the largest events ever recorded (Iribarren Anacona et al., 2015). However, published literature is largely missing for a reliable assessment of past and future trends of such hazards.

Aridity: Several regional studies suggest increasing trends in the frequency and length of droughts in the region, such as: over NWS (Domínguez-Castro et al., 2018), NSA (Marengo and Espinoza, 2016; Cunha et al., 2019) and NES (Marengo and Bernasconi, 2015), over southern Amazonia (Fu et al., 2013; Boisier et al., 2015), in the São Francisco River basin and the capital city Distrito Federal in Brazil (Borges et al., 2018; Bezerra et al., 2019), in the southern Andes (Vera and Díaz, 2015), in central southern Chile (Boisier et al., 2018), in SES (Rivera and Penalba, 2014) and, during recent years, in SSA (Rivera and Penalba, 2014). Chapter 8 indicated medium confidence of anthropogenic forcing on observed drying trends in central Chile. Additional discussion on droughts and aridity trends in South America is presented in Sections 8.3.1.6, 8.4.1.6 and 8.6.2.1.

Sections 8.3.2.4 and 8.4.1.6 point to two important drying hotspots in South America with long-term soil moisture decline and precipitation declines: the Amazon basin (SAM and NSA) and SWS (medium confidence) (Figure 12.4). End-of-century RCP8.5 projections show a longer dry season in the central part of South America and decreased precipitation over the Amazon and central Brazil (Teichmann et al., 2013; Coppola et al., 2014a; Giorgi et al., 2014; Llopart et al., 2014; Atlas.7). Seasonal changes are also projected by end century under RCP 8.5, with decreases in June–July–August (JJA) rainfall projected for NSA, the coastal region of SES, SAM and the southern portion of SWS (Marengo et al., 2016). Decreases in December –January–February (DJF) rainfall are projected for the central part of South America in the near term (Kitoh et al., 2011; Chou et al., 2014; Cabré et al., 2016). Regional projections for Central and South America also indicate an increase in dryness in SCA and NES by mid- to end-century (medium confidence) (Chou et al., 2014; Marengo and Bernasconi, 2015).

Hydrological drought: Chapter 11 assessed mostly low confidence in observed changes in hydrological droughts given a lack of studies and clear evidence, with medium confidence only for a streamflow decrease in sub-regions of SWS (Table 11.15). Some trends are becoming more clear, such as the ones reported for Colombia (NWS) by Carmona and Poveda (2014), who indicated that 62% of the 25- to 50-year-long monthly average streamflow time series exhibited significant decreasing trends. However, studies of discharge changes indicate that uncertainty is still large, as argued by Pabón-Caicedo et al. (2020) for the full extent of the Andes.

A number of studies project decreases in runoff and river discharge for SCA, Colombia, Brazil and the southern part of South America by the end of this century (Nakaegawa and Vergara, 2010; Arnell and Gosling, 2013; van Vliet et al., 2013). Section 11.9 assessed high confidence in projections of increases in hydrological droughts in NSA, SAM, SWS, and SSA under for 4°C GWL, medium confidence in SCA, and low confidence in the rest of the sub-regions given insufficient evidence, lack of signal or mixed signals among the available studies. Signals are much more uncertain for the middle of the century (or for a 2°C GWL).

Agricultural and ecological drought: Section 11.9 assessed low confidence in observed changes in agricultural and ecological drought across Central and South America due to regional heterogeneity and differences depending on the drought metrics used, except in NES, which has seen a dominant increase in drought severity (medium confidence).

NSA and SAM are the two regions where the strongest signal of increasing number of dry days (NDD) and drought frequency (DF) is projected compared to other regions of the world (Coppola et al., 2021b). By the end of this century and under RCP8.5, the NSA area average value for NDD reaches 43, 32 and 27, within the CORDEX-CORE, CMIP5 and CMIP6 ensembles respectively. For the frequency of droughts, the NSA area average value is of about 4.6, 3.4 and 3.8. For the same period and scenario, the SAM region shows NDD and DF values of 29, 20 and 21, and of 4, 3 and 3.5 respectively (Figure 12.4j–l). In Central America, a significantly drier northern region and a wetter southern region are projected for mid-century by (Hidalgo et al., 2017), whilst Fuentes-Franco et al. (2015) pointed to more pronounced dry periods during the rainy season in SCA by the end of this century under RCP8.5. Increases in the frequency of meteorological droughts that may initiate other drought types are projected for the eastern part of the Amazon and the opposite for the west under RCP8.5 (Duffy et al., 2015). In central Chile, the occurrence of extended droughts, such as the recently experienced 2010–2015 megadrought (which is still driving impacts), is projected to increase from one to up to five events per 100 years under RCP8.5 (Bozkurt et al., 2018). Section 11.9 highlights change in confidence in increases in drought severity in SCA, NSA, NES, SAM, SWS, and SSA from low to highunder the three GWLs of 1.5°C, 2°C and 4°C. NES and SES change from low confidence to medium confidence increases in agricultural and ecological drought severity by 4°C GWL with different metrics and high agreement between studies. Only SAM and SSA have projections of agricultural and ecological drought increasing with high confidence for the middle of the century, or for a 2°C GWL, and NSA, NES and SCA are projected to increase with medium confidence.

Fire weather: There is evidence of increases in forest fire activity (number of fires, burned area and fire duration) in central and south-central Chile, where more conducive fire weather conditions have been proposed as the main driver (González et al., 2018; Urrutia-Jalabert et al., 2018). Projections indicate that the Amazon will be one of the regions in the world with the highest increase in fire weather indices over the 21st century and under all RCPs (high confidence) (Betts et al., 2015; Abatzoglou et al., 2019; Q. Sun et al., 2019). This is consistent with the large increase in the frequency of joint occurrence of extreme hot and dry days projected for a 2°C warming level or more (Vogel et al., 2020). Projections of fire weather indices also show an increased risk in SWS (high confidence), SSA and SCA (medium confidence). However, wildfires highly depend on land use and appropriate management may help mitigate future increases in fire risk (Fonseca et al., 2019).

Mean precipitation is projected to change in a dipole pattern with increases in NWS and SES and decreases in NES and SWS (high confidence) with further decreases in NSA and SCA (medium confidence). There is medium confidence of an increase in river floods in SAM and SES. There is high confidence of an increase in drought duration in NES, an in the number of dry days and drought frequency in NSA and SAM. Dry climatic impact-drivers are projected to increase at the regional level with higher global warming levels. The strongest signal of future increase in agricultural and ecological drought, aridity and fire weather is over the Amazon region (high confidence).

Mean wind speed: Due to the lack of long-term homogeneous records or limited observations in the region, past wind speed trends are difficult to establish. Global climate models project an increase in wind speeds, under all future scenarios, augmenting wind power potential in most parts of Central and South America, especially in NES, where changes lie in the range 0–20% by 2050 under RCP8.5 and 0–40% under RCP8.5 (medium confidence) (Karnauskas et al., 2018a; Reboita et al., 2018; Jung and Schindler, 2019). In Patagonia, wind speeds are projected to decrease. For RCP4.5 changes remain marginal and have low confidence (low agreement) (Figure ­12.4m–o).

Severe wind storm: Similar observational limitations inhibit an assessment of long-term extreme wind trends. However, Pes et al. (2017) found extreme wind increases in most of Brazil over the past decades. Future projections indicate a slight decrease in the number of extratropical cyclones in mid-latitudes (limited evidence, low confidence) (Reboita et al., 2018), and an increase of extreme winds in tropical areas (limited evidence, low confidence) (Kumar et al., 2015). Climate models project a shift and an intensification of southern storm tracks, with most effects offshore over the Southern Ocean (Chapter 4), with low confidence (low agreement) of significant extreme wind changes over land and coastal areas across the 21st century (Chang, 2017; Augusto Sanabria and Carril, 2018; Reboita et al., 2021).

Tropical cyclone: CMIP5 and CMIP6 simulations, including the new High Resolution Model Intercomparison Project (HighResMIP), project a decrease in the frequency of tropical cyclones in the Atlantic and Pacific coasts of Central America for the mid-century or under a 2°C GWL, accompanied with an increased frequency of intense cyclones (medium confidence) (Section 11.7.1.5; Diro et al., 2014; Knutson et al., 2020; Roberts et al., 2020).

In summary, there is limited evidence of current trends in observed wind speed and wind storms in Central and South America. Climate projections indicate a decrease in frequency of tropical cyclones in Central America accompanied with an increased frequency of intense cyclones, and an increase in mean wind speed and wind power potential in most parts of Central and South America (medium confidence).

Snow: Historical studies of seasonal snow cover are limited and restricted to the Andes Cordillera. Mernild et al. (2017) indicated that much of the area north of 23°S experienced a decrease in the number of snow cover days, while the southern half of the Andes Cordillera experienced the opposite. A reduction in snow cover of about 15% was simulated for areas with altitudes in the range of 3000–5000 m, whereas in regions with altitude below 1000 m (Patagonia) snow cover extent increased. The reduced snowfall over the Chilean and Argentinean Andes mountains, which has resulted in unprecedented reductions in river flow, reservoir volumes and groundwater levels, led to the most severe and long-lasting hydrological drought (2010–2015) reported in the adjacent semi-arid foothills of the central Andes (Garreaud et al., 2017; Rivera et al., 2017; Masiokas et al., 2020). Projections (based on process understanding) in Section 9.5.3.3 point to decreases in seasonal snow cover extent and duration across South America as global climate continues to warm (high confidence).

Glacier: Observation and future projection of Central and South America glacier mass changes are assessed in Section 9.5.1, grouped in two main regions: Low Latitude region (98% of which is glaciers in the Andes) and the Southern Andes region. An increase in the number and areal extent of glacial lakes in the Southern Andes was reported for the period 1986–2016 (Wilson et al., 2018). Similar changes are being observed in the central Andes (Colonia et al., 2017). Since 1800 at least 15 ice‐dammed lakes and 16 moraine‐dammed lakes have failed in the extratropical Andes, causing high-magnitude glacial lake outburst floods (Rojas et al., 2014; Cook et al., 2016; Wilson et al., 2018; Drenkhan et al., 2019). Partially due to glacier shrinkage and lake growth, the frequency of outburst floods has increased in the last 30–40 years (Carey et al., 2012; Iribarren Anacona et al., 2015).

Glaciers across South America are expected to continue to lose mass and glacier area in the coming century (high confidence) (Section 9.5). In terms of their mass, glaciers in the Low Latitude region are projected by GlacierMIP to lose 67 ± 42%, 86 ± 24% and 94 ± 13%, of their 2015 baseline mass by the end of the century under RCP2.6, RCP4.5 and RCP8.5 respectively (Marzeion et al., 2020). Glaciers in the Southern Andes show decreasing mass loss rates for RCP2.6, and increasing rates for RCP8.5, which peak in the mid to late 21st century. Glaciers in the Southern Andes are projected to lose 26 ± 27%, 33 ± 26% and 47 ± 26% of their 2015 mass by the end of the century under RCP2.6, RCP4.5 and RCP8.5 respectively (Section 9.5.1.3).

The central Andes will experience the highest disturbance to the thermal regime of the 21st century. As a consequence, in the Argentinean Andes up to 95% of rock glaciers in the southern desert Andes and in the central Andes will rest in areas above 0°C under the worst case scenario of warming (the freezing level might move up more than twice as much as during the entire Holocene; Drewes et al., 2018)

Permafrost: There is limited information on the ongoing changes and projections of permafrost conditions in the region. Based on model projections under the IPCC A1B scenario, permafrost areas in the Bolivian Andes will eventually be lost, but this could take years to decades or longer depending on permafrost thickness (Rangecroft et al., 2016).

In conclusion, glacier volume loss and permafrost thawing will continue in the Andes Cordillera under all climate scenarios (high confidence), causing important reductions in river flow and potentially high-magnitude glacial lake outburst floods.

Relative sea level: Around Central and South America, over 1900–2018, a new tide gauge-based reconstruction finds a regional mean relative sea level (RSL) change of 2.07 [1.36 to 2.77] mm yr^–1 in the South Atlantic, 2.49 [1.89 to 3.06] mm yr^–1 in the subtropical North Atlantic and 1.20 [0.76 to 1.62] mm yr^–1 in the East Pacific (Frederikse et al., 2020), compared to a global mean sea level (GMSL) change of around 1.7 mm yr^–1Section 2.3.3.3 and Table 9.5). For the period 1993–2018, these RSLR rates, based on satellite altimetry, increased to 3.45 [3.04 to 3.86)] mm yr^–1, 4.04 [2.77 to 5.24] mm yr^–1 and 2.35 [0.70 to 4.06] mm yr^–1 respectively (Frederikse et al., 2020), compared to a GMSL change of 3.25 mm yr^–1Section 2.3.3.3 and Table 9.5).

Relative sea level rise is extremely likely to continue in the oceans around Central and South America. Regional mean RSLR projections for the oceans around Central and South America range from 0.3–0.5 m under SSP1-2.6 to 0.5–0.9 m under SSP5-8.5 for 2081–2100 relative to 1995–2014 (median values), which is around the projected GMSL change (Section 9.6.3.3). These RSLR projections may, however, be underestimated due to potential partial representation of land subsidence in their assessment (Section 9.6.3.2).

Coastal flood: The present-day 1-in-100-year ETWL ranges from 0.5 to 2.5 m around most of Central and South America, except in SSA and SWS, where 1-in-100-year ETWLs can be as large as 5 to 6 m (Vousdoukas et al., 2018). ETWL magnitude and occurrence frequency are expected to increase throughout the region (high confidence) (Figure 12.4p–r and Figure 12.SM.6). Across the region, the 5–95th percentile range of the 1-in-100-year ETWL is projected to increase (relative to 1980–2014) by 8–34 cm and by 10–43 cm by 2050 under RCP4.5 and RCP8.5 respectively (Vousdoukas et al., 2018; Kirezci et al., 2020). By 2100, this range is projected to be 21–93 cm and 34–190 cm under RCP4.5 and RCP8.5 respectively (Vousdoukas et al., 2018; Kirezci et al., 2020). Furthermore, under RCP4.5, the present-day 1-in-100-year ETWL is projected to have median return periods of 1-in-10-years to 1-in-50-years by 2050 and 1-in-1-year by 2100 in SES, SSA and SWS. In other regions of Central and South America, the present-day 1-in-100-year ETWL is projected to occur once per year or more by both 2050 and 2100 under RCP4.5. The present-day 1-in-50-year ETWL is projected to occur around three times a year by 2100 with a SLR of 1 m (Vitousek et al., 2017).

Coastal erosion: According to satellite data, shoreline retreat rates of around 1 m yr^–1have been observed along the sandy coasts of SCA, SES and SSA over the period 1984–2015, while shoreline progradation rates of around 0.25 m yr^–1has been observed in NWS and NSA. The sandy shorelines in NES and SWS have remained more or less stable over the same period (Luijendijk et al., 2018; Mentaschi et al., 2018). Using satellite observations, Mentaschi et al. (2018) report a coastal area loss of 250 km² over a 30-year period (1984–2015) along the Pacific coast of South America, and of 780 km² along the Atlantic coastlines.

Projections indicate that a majority of sandy coasts in the region will experience shoreline retreat throughout the 21st century (high confidence). Median shoreline change projections (CMIP5) for the mid-century period show that, relative to 2010, sandy shorelines will retreat by between 30 and 75 m in SCA, NES, SES and SSA under both RCP4.5 and RCP8.5, while the projected mid-century retreats are less than 30 m in NSA, NWS and SWS for both RCPs (Vousdoukas et al., 2020b). Parts of the coastline in these latter three regions are projected to prograde over the 21st century, if present ambient shoreline change trends continue (Vousdoukas et al., 2020b). By 2100, median retreats of more than 100 m are projected in SCA, NES, SES and SSA under both RCPs, while retreats between 50–100 m are projected for NSA, NWS and SWS under both RCPs (Figure 12.8). Notably, the projected shoreline retreats in SCA and SES approach 150 m by 2100 under RCP4.5 and 200 m under RCP8.5. The total length of sandy coasts in Central and South America that is projected to retreat by more than a median of 100 m by 2100 under RCP4.5 and RCP8.5 is about 12,000 and 15,000 km respectively, an increase of approximately 30%.

Marine heatwave: The mean sea surface temperature (SST) of the Atlantic Ocean and the Caribbean around Central and South America increased from 0.25°C to 1°C over the period 1982–1998 (Oliver et al., 2018). This mean ocean surface warming is connected to longer and more frequent marine heatwaves (MHW) in the region (Oliver et al., 2018). Over the period 1982–2016, the coastlines experienced on average more than 1.0 MHW per year, with the Pacific coast of Northern Central America and the coast of SES (Atlantic) experiencing on average 2.5–3 MHWs per year. The average duration was between 10 and 15 days, with the notable exception of the equatorial Pacific coastline, which experiences MHWs with >30 days average duration related to ENSO conditions. In the south-western Atlantic shelf (32–38°S), more than half of the days with MHWs have occurred since 2014, and the most intense event (1.7°C above previous maximum) took place in the austral summer of 2017 (Manta et al., 2018). Changes over the 20th century, derived from MHW proxies, show an increase in frequency between 0.5 and 2 MHW per decade in the South Atlantic, the Caribbean and the Pacific coast of Northern Central America, an increase in intensity per event in the South Atlantic, and a decrease along the equatorial Pacific coastline. The total number of MHW days per year increases around Central and South America, with the exception of the equatorial Pacific coastline (Oliver et al., 2018).

There is high confidence that MHWs will increase around Central and South America. Mean SST is projected to increase by 1°C (2°C) by 2100, with a hotspot of about 2°C (4°C) along the coast of South-Eastern South America and North-Western South America under RCP4.5 (RCP8.5; Interactive Atlas). More frequent MHWs are projected around the region for GWLs of 1.5°C, 2°C and 3.5°C relative to the modelled reference value for 1861–1880 (Frölicher et al., 2018). Projections for SSP1-2.6 and SSP5-8.5 both show an increase in MHWs around Central and South America by 2081–2100, relative to 1985–2014 (Box 9.2, Figure 1).

In summary, relative sea level rise is extremely likely to continue in the oceans around Central and South America, contributing to increased coastal flooding in low-lying areas (high confidence) and shoreline retreat along most sandy coasts (high confidence). Marine heatwaves are also expected to increase around the region over the 21st century (high confidence).

The assessed direction of change in climatic impact-drivers for Central and South America and associated confidence levels are illustrated in Table 12.6. No assessable literature could be found for sand and dust storm, lake and sea ice, heavy snowfall and ice storms, hail and snow avalanches, although these phenomena may be relevant in parts of the region.

Table 12.6 | Summary of confidence in direction of projected change in climatic impact-drivers in Central and South America, representing their aggregate characteristic changes for mid-century for scenarios RCP4.5, SSP2-4.5, SRES A1B, or above within each AR6 region (defined in Chapter 1), approximately corresponding (for CIDs that are independent of sea level rise) to global warming levels between 2 and 2.4°C (see Section 12.4 for more details of the assessment method). The table also includes the assessment of observed or projected time-of-emergence of the CID change signal from the natural interannual variability if found with at least medium confidence in Section 12.5.2.

Mean air temperature: Since AR5, studies have confirmed that the mean warming trend in Europe is increasing (Atlas.8.2). The observed warming trend patterns are largely consistent with those simulated by global and regional climate models and it is very likely that such trends are, in large part, due to human influence on climate (Section 3.3.1).

All temperature trends are very likely to continue for a global warming of 1.5°C or 2°C and 3°C (Atlas.8.4). Future warming leads to the exceedance of different temperature thresholds relevant for vector-borne diseases (medium confidence) (Caminade et al., 2012; Medlock et al., 2013), invasive allergens (medium confidence) (Storkey et al., 2014; Hamaoui-Laguel et al., 2015), SST thresholds in the Mediterranean (likely to exceed 20°C), or relevant for the Vibriobacteria development (Vezzulli et al., 2015). Future warming is also projected to lead to the exceedance of cooling degree day index (>22°C) thresholds, characterizing a potential increase in energy demand for cooling in southern Europe with increases likely exceeding 40% in some areas (Spinoni et al., 2015) by 2050 under RCP8.5 (high confidence) (Section 12.3 and Atlas.8; Coppola et al., 2021a).

Extreme heat: The frequency of heatwaves observed in Europe has very likely increased in recent decades due to human-induced change in atmospheric composition (Section 11.3) and a detectable anthropogenic increase in a summer heat stress index over all regions of Europe has been identified based on WBGT index trends for 1973–2012 (medium confidence, limited evidence) (Knutson and Ploshay, 2016).

It is very likely that the frequency of heatwaves will increase during the 21st century regardless of the emissions scenario in each European region, and for 1.5°C and 2°C GWLs (Section 11.3.5). Heat stress due to both high temperature and humidity, affecting morbidity, mortality and labour capacity (Section 12.3) is projected to increase under all emissions scenarios and GWLs by the middle of the century (Figure 12.4a–f). Under RCP8.5, the expected number of days with WBGT higher than 31°C is about 25, 30 and 40 days per year, as projected by EURO-CORDEX, CMIP5 and CMIP6 respectively on average over the Mediterranean region, and around 30, 40 and 60 days per year in low coastal plain areas such as the Po Valley, the Italian, Greek and Spanish coasts, and the Mediterranean islands (Coppola et al., 2021a). An average increase of a few days per year of maximum daily temperature exceeding 35°C, a typical critical threshold for crop productivity, is expected by the mid-century in central Europe, and an increase of 10–20 days is expected for the Mediterranean areas (Figure 12.4b; Coppola et al., 2021a). By contrast, under SSP1-2.6, the increase in this number of days remains limited to less than about 10 days, and confined to the Mediterranean regions. Mitigation is expected to have a strong effect, with the dangerous heat threshold of HI > 41°C projected to be crossed 5–10 days more per year in the Mediterranean regions and a few days per year more in WCE and EEU under SSP5-8.5, while such increases would be virtually absent under SSP1-2.6 (Figure 12.4d–f).

Cold spell and frost: Temperature observations for winter cold spells in Europe show a long-term decreasing frequency (Brunner et al., 2018), with their probability of occurrence projected to decrease in the future (high confidence) and virtually disappear by the end of the century (Section 11.3). The frequency of frost days will very likely decrease for all scenarios and all time horizons (Lindner et al., 2014; Coppola et al., 2021a) with consequences for agriculture and forests. A simple heating degree day index, characterizing heating demand, shows a large observed decreasing trend for winter heating energy demand in Europe (Spinoni et al., 2015). This trend is very likely to continue through the 21st century, with decreases in the range of 20–30% for Northern Europe, about 20% for central Europe and 35% for southern Europe, by mid-century under RCP8.5 (Spinoni et al., 2018b; Coppola et al., 2021a; Interactive Atlas).

In summary, irrespective of the scenario, it is virtually certain that warming will continue in Europe, and there is high confidence that the observed increase in heat extremes is due to human activities. It is very likely that the frequency of heat extremes will increase over the 21st century with an increasing gradient toward the southern regions. Extreme heat will exceed critical thresholds for health, agriculture and other sectors more frequently (high confidence), with strong differences between mitigation scenarios. It is very likely that the frequency of cold spells and frost days will keep decreasing over the course of this century and it is likely that cold spells will virtually disappear towards the end of the century.

Mean precipitation: Precipitation has generally increased in northern Europe and decreased in southern Europe, especially in winter (Fischer and Knutti, 2016; Knutson and Zeng, 2018) but in the latter, precipitation trends are strongly dependent on the examined period (Atlas.8). These trends in precipitation increases in the north and decreases in the south are also represented by global and regional climate simulations (Jacob et al., 2014; Rajczak and Schär, 2017; Lionello and Scarascia, 2018; Coppola et al., 2021a; Atlas.8.2) and have been attributed to climate change (Sections 3.3.2, 8.3.1).

Studies since AR5, together with EURO-CORDEX and MED-CORDEX experiments and the latest CMIP6 ensemble, have increased confidence in regional projections of mean and extreme precipitation (Prein et al., 2016) despite their wet bias, and show that it is very likely that precipitation will increase in Northern Europe in DJF and decrease in the Mediterranean in JJA under all climate scenarios except RCP2.6/SSP1-2.6 and for both mid- and end-century periods (Coppola et al., 2021a; Atlas.8.5).

River flood: There is high confidence of an observed increasing trend of river floods in Western and Central Europe (WCE) and medium confidence of a decrease in Northern (NEU) and southern Europe (MED).

The SR1.5 shows evidence of an increase in reported floods in the UK over the period 1884–2013, and increasing trends in annual maximum daily streamflow data over 1966–2005 in parts of Europe. Although high flow does not show uniform trends for the entire region (Hall et al., 2014; Mediero et al., 2015) or specific regions (Mudersbach et al., 2017; Vicente-Serrano et al., 2017; Tramblay et al., 2019), regional patterns of significant flood trends do exist. Based on the most extended river flow database spanning the period 1960–2010, an increase in floods frequency in north-western Europe, decreasing in medium and large catchments in southern Europe and decreasing floods in Eastern Europe has been detected (Blöschl et al., 2019) in line with Mediero et al. (2014), Arheimer and Lindström (2015), Gudmundsson et al. (2017), Krysanova et al. (2017), Kundzewicz et al. (2018) and Mangini et al. (2018).

There is high confidence of river floods increasing in Western and Central Europe (WCE) and medium confidence of a decrease in Northern (NEU), Eastern (EEU) and southern Europe (MED) for mid- and end-century under RCP8.5 and low confidence under RCP2.6. The projected increase in WCE is roughly 10% (18% by end of century) and the projected decrease in NEU is 5% (11% by end of century) for the peak flow with a return period of 100 years for mid-century, under RCP8.5 (Di Sante et al., 2021; Figure 12.9a for mid-century (Q100) projections of flood discharges per unit catchment area (Blöschl et al., 2019) based on EURO-CORDEX models).

Figure 12.9 | Projected changes in selected climatic impact-driver indices for Europe. (a) Mean change in 1-in-100-year river discharge per unit catchment area (Q100, m³s^–1km^–2), and (b) median change in the number of days with snow water equivalent (SWE) over 100 mm (from November to March), from EURO-CORDEX models for 2041–2060 relative to 1995–2014 and RCP8.5. Diagonal lines indicate where less than 80% of models agree on the sign (direction) of change. (c) Bar plots for Q100 (m³s^–1km^–2) averaged over land areas for the AR6 WGI Reference Regions (defined in Chapter 1). The left-hand column within each panel (associated with the left-hand y-axis) shows the ‘recent past’ (1995–2014) Q100 absolute values in grey shades. The other columns (associated with the right-hand y-axis) show the Q100 changes relative to the recent past values for two time periods (‘mid’ 2041–2060 and ‘long’ 2081–2100) and for three global warming levels (defined relative to the pre-industrial period 1850–1900): 1.5°C (purple), 2°C (yellow) and 4°C (brown). The bars show the median (dots) and the 10–90th percentile range of model ensemble values across each model ensemble. CMIP6 is shown by the darkest colours, CMIP5 by medium, and CORDEX by light. SSP5-8.5/RCP8.5 is shown in red and SSP1-2.6/RCP2.6 in blue. (d) As for (c) but showing absolute values for number of days with SWE > 100mm, masked to grid cells with at least 14 such days in the recent past. See Technical (Annex VI for details of indices. Further details on data sources and processing are available in the chapter data table (Table 12.SM.1).

Open figure

Using frequency analysis of extreme peak flow events above a 100-year return period as a threshold, which is the average protection level of the European river network (Rojas et al., 2013), Alfieri et al. (2017) and Alfieri et al. (2015) show that Europe is one of the regions where the largest increases in flood risk may occur, with only few countries in Eastern Europe showing a decrease (Poland, Lithuania, Belarus) (Osuch et al., 2017). They find a significant increase of events with peak discharge above 100-year return period (Q100) in most of Europe in line with Rojas et al. (2012), Hirabayashi et al. (2013), Dankers et al. (2014), Forzieri et al. (2016), Roudier et al. (2016), Thober et al. (2018), and an increase in the magnitude of floods in southern Europe, although Giuntoli et al. (2015) projects no change. A modest but significant decrease in the 100-year return period river flood is projected for southern (due to reduction of precipitation) and north-eastern European regions, the latter because of the strong reduction in snowmelt induced river floods (Thober et al., 2018; Di Sante et al., 2021).

Heavy precipitation and pluvial flood: Heavy precipitation frequency trends have been detected in Europe with high confidence for the NEU and Alpine regions and with medium confidence in WCE, and also attributed to climate change with high confidence in NEU (Section 11.9). Guerreiro et al. (2017), based on observations, showed that 20% of city areas in WCE and MED regions are affected by pluvial flooding and less than 10% of city areas in the northern and western coastal cities.

Projections based on multiple lines of evidence from global to convective permitting model scales show high confidence in extreme precipitation increase in the northern, central and eastern European regions (NEU, WCE, EEU) and in the Alpine area. Increases with medium confidence are projected for the Mediterranean basin (with a negative gradient towards the south) for mid- and end-century under RCP4.5, RCP8.5 and SSP5-8.5 and for 2°C GWL and higher (Section 11.9; MedECC, 2020).

Landslide: Rainfall periods connected to landslides are projected to increase in central Europe by up to one more period per year in flat areas in low altitudes and by up to 14 more periods per year at higher altitudes by mid-century, becoming even more evident by the end of the century (Schlögl and Matulla, 2018). An increase of landslides by up to 45.7% and 21.2% is projected for southern Italy (Calabria region) by mid-century under both RCP4.5 and RCP8.5 (Gariano and Guzzetti, 2016) and by up to 40% in central Italy (Umbria) during the winter season (Ciabatta et al., 2016). A decrease of landslides is projected in the Peloritani mountains in southern Italy (RCP4.5 and 8.5) by mid-century (Peres and Cancelliere, 2018). A slight increase for a 10-year return period landslide is projected in the eastern Carpathians, the Moldavian Subcarpathians and the northern part of the Moldavian Tableland and a higher increase in the 100-year return period event is projected in the western hilly and plateau areas of Romania (Jurchescu et al., 2017).

Aridity: The Mediterranean region shows evidence of large-scale decreasing precipitation trends over 1901–2010, which are at least partly attributable to anthropogenic forcing according to CMIP5 models (Knutson and Zeng, 2018). Nevertheless, there is low agreement among studies on observed precipitation trend in the Mediterranean region (Section 11.9.4 and Atlas.8.2).

Precipitation is projected to decrease by mid- and end-century for the RCP8.5 and SSP5-8.5 with strong agreement among CMIP5, CMIP6 and CORDEX regional climate ensemble models on the direction of change. With both temperature increase and precipitation decrease there is high confidence on increased aridity in the MED region (Sections 8.4.1.6 and 11.9.4 and Atlas.8.2; Coppola et al., 2021a). In NEU there is high confidence of decrease in aridity linked to mean precipitation increase (Section 8.4.1.6, Atlas 8.2) and meteorological drought decrease based on indicators like Standardized Precipitation Index and consecutive dry days (Section 11.9.4, Figure 12.4, Coppola et al,, 2021a).

Hydrological drought: There is high confidence that hydrological droughts have increased in the Mediterranean basin with medium confidence in anthropogenic attribution of the signal, and high confidence that they will continue to increase through the 21st century for 2°C GWL and higher and all scenarios except RCP2.6/SSP1-2.6. (Sections 8.3.1.6, 8,4.1.6, and 11.9.4). There is medium confidence in hydrological drought increase in WCE and low confidence in direction of change for EEU and NEU from mid-century onwards and for 2°C GWL and higher and all scenarios except RCP2.6/SSP1-2.6 (Section 11.9 and Figure 12.4g–i).

Streamflow droughts are projected to become more severe and persistent in the Mediterranean and western Europe (current 100-year events could occur approximately every 2–5 years by 2080; Forzieri et al., 2016).

The opposite tendency is projected in Northern, Eastern and central Europe where higher precipitation that outweighs the effects of increased evapotranspiration is expected to result in a decrease in streamflow drought frequency (Forzieri et al., 2014). For a 2°C GWL droughts will become more intense in the MED and in France and longer mainly due to less rainfall and higher evapotranspiration. A reduction of drought length and magnitude is projected for NEU and EEU (Roudier et al., 2016). In the southern Alps, both winter and summer low flows are projected to be more severe, with a 25% decrease in the 2050s (Vidal et al., 2016).

Agricultural and ecological drought: There is medium confidence that agricultural and ecologicaldroughts have increased in Western and Central Europe and in the Mediterranean region, and medium confidence that anthropogenic drivers contributed to the Mediterranean increase (Sections 8.3.1.6 and 11.9).

(Chapter 11 assesses that agricultural and ecological droughts will increase in the Mediterranean regions (high confidence) and Western and Central Europe (medium confidence) by mid-century and with high confidence by the end of the century for the MED for 2°C GWL and higher and all scenarios except RCP2.6/SSP1-2.6 (Section 11.9.4). Low confidence in direction of change is assessed for EEU and NEU under all scenarios and global warming levels (Figure 12.4k).

Recent local studies provide additional risk-relevant context to changes in European drought. Agricultural and ecologicaldrought conditions are expected to intensify in southern Europe by end-of-century based on the 12-month rainfall Drought Severity Index (a soil moisture indicator), precipitation deficit SPI and SPEI indices. There will be regions in southern Europe where this type of drought could be up to 14 times worse than the worst drought in the historical period (Guerreiro et al., 2018). One-in-10-year drought events are projected to happen every second year (Mora et al., 2018; Ruosteenoja et al., 2018). The Mediterranean region will have 100 additional stress years (years with three consecutive months of precipitation deficits greater than 25%; Giorgi et al., 2018); an increase of both drought frequency (up to two events per decade) and severity (Spinoni et al., 2014, 2020) and an increase of consecutive dry days in the southern part of the MED region (Lionello and Scarascia, 2020). In contrast, droughts are expected to decrease in winter in Northern Europe (Section 11.9; Spinoni et al., 2018a). These findings are confirmed by the EURO-CORDEX, CMIP5 and CMIP6 ensemble that show a change of frequency of drought events in the MED between 2–3 events per decade by mid-century for scenario RCP8.5 (Figure 12.SM.3; Coppola et al., 2021a).

Fire weather: Fire weather conditions have been increasing since about 1980 over a few regions in Europe including Mediterranean areas (low confidence) (Venäläinen et al., 2014; Urbieta et al., 2019; Barbero et al., 2020; Giannaros et al., 2021). However, beyond a few studies, evidence is largely missing on attribution of these trends to anthropogenic climate change (Forzieri et al., 2016). An increase in fire weather is projected for most of Europe, especially western, eastern and central regions, by 2080 (current 100-year events will occur every 5–50 years), with a progressive increase in confidence and model agreement along the 21st century (medium confidence) (Forzieri et al., 2016; Abatzoglou et al., 2019). With increased drying and heat combined, in Mediterranean areas, an increase in fire weather indices is projected under RCP4.5 and RCP8.5, or SRES A1B, as early as by mid-century (high confidence) (Bedia et al., 2014; Abatzoglou et al., 2019; Dupuy et al., 2020; Fargeon et al., 2020; Ruffault et al., 2020) and an increase in burned area of 40% and 100% for a 2°C and 3°C GWL, respectively (Turco et al., 2018).

In summary, there is high confidence that river floods will increase in central and Western Europe and medium confidence that they will decrease in Northern, Eastern and southern Europe, for mid- and end of century under RCP8.5 and with low confidence under RCP2.6. There is high confidence that aridity will increase by mid- and end-century under the RCP8.5 and SSP5-8.5, and high confidence that agricultural, ecological and hydrological droughts will increase in the Mediterranean region by mid- and far end of century under all RCPs except RCP2.6/SSP1-2.6 and also for 2°C and higher GWLs. There is high confidence in fire weather increase in the Mediterranean region.

Mean wind speed: Mean surface wind speeds have decreased in Europe as in many other areas of the Northern Hemisphere over the past four decades (medium confidence) (AR5 WGI), with a reversal to an increasing trend in the last decade (low confidence) that is, however, not fully consistent across studies (Tian et al., 2019; Zeng et al., 2019; Z. Zhang et al., 2019; Deng et al., 2021; see Section 2.3.1.4.4). Re-analyses also show declining winds in Europe (Deng et al., 2021) with large interdecadal variability (Laurila et al., 2021). The declining trend has induced a corresponding decline in wind power potential indices across Europe (low confidence) (Tian et al., 2019). However, there is low agreement and limited evidence that climate model historical trends are consistent with observed trends (Tian et al., 2019; Deng et al., 2021). Several factors have been attributed to these trends, including forest growth, urbanization, local changes in wind measurement exposure and aerosols (Bichet et al., 2012), as well as natural variability (Zeng et al., 2019).

Due to changes in mean surface wind speed patterns (C. Li et al., 2018) and the poleward shift of the North Atlantic jet stream exit, mean surface wind speeds are projected to decrease in the Mediterranean areas under RCP4.5 and RCP8.5 by the middle of the century and beyond, or for GWLs of 2°C and higher (high confidence), with a subsequent decrease in wind power potential (medium confidence) (Hueging et al., 2013; Tobin et al., 2015, 2018; Davy et al., 2018; Karnauskas et al., 2018a; Kjellström et al., 2018; Moemken et al., 2018; Figure 12.4). However, sub-regional patterns of change are shown in regional climate models, such as an increase in wind speeds in the Aegean Sea and in the northern Adriatic Sea, where a reduction of Bora events and an increase of Scirocco events are projected for mid-century and beyond under RCP4.5 and RCP8.5 (medium confidence) (Tobin et al., 2016; Davy et al., 2018; Belušić Vozila et al., 2019). Projections (as cited above) also indicate a decrease in mean wind speed in Northern Europe (medium confidence, medium agreement) (Karnauskas et al., 2018a; Tobin et al., 2018; Jung and Schindler, 2019). Daily and interannual wind variability is projected to increase under RCP8.5 only in Northern Europe (low confidence) (Moemken et al., 2018), which can influence electrical grid management and wind energy production (low confidence). Wind speeds are projected to shift towards more frequent occurrences below thresholds inhibiting wind power production (Weber et al., 2018). Wind stagnation events may become more frequent in future climate scenarios in some areas of Europe in the second half of the 21st century (Horton et al., 2014; Vautard et al., 2018), with potential consequences on air quality (low confidence).

Severe wind storm: There are large uncertainties in past evolutions of windstorms and extreme winds in Europe. Extreme near-surface winds have been decreasing in the past decades (Smits et al., 2005; Tian et al., 2019; Vautard et al., 2019) according to near-surface observations. Significant negative trends of cyclone frequency in spring and positive trends in summer have been found in the Mediterranean basin for the period 1979–2008 (Lionello et al., 2016). By contrast increasing trends have been found in Arctic Ocean areas (Wickström et al., 2020). These trends are not associated with significant trends in extratropical cyclones (Sections 8.3.2.8 and 11.7.2).

There is medium confidence that serial clustering of storms, inducing cumulated economic losses, in future climate will increase in many areas in Europe under climate projections over Europe (Karremann et al., 2014; Economou et al., 2015). Strong winds and extratropical storms are projected to have a slightly increasing frequency and amplitude in the future in northern, western and Central Europe (Outten and Esau, 2013; Feser et al., 2015; Forzieri et al., 2016; Mölter et al., 2016; Ruosteenoja et al., 2019a; Vautard et al., 2019) under RCP8.5 and SRES A1B by the end of the century (medium confidence), as well as off the European coasts (Martínez-Alvarado et al., 2018) due to the increase of intensity of extratropical storms at a 2°C GWL or above (Zappa et al., 2013) in these areas. The frequency of storms, including Medicanes, is projected to decrease in Mediterranean regions, and their intensities are projected to increase, by the middle of the century and beyond for SRES A1B, A2 and RCP8.5 (medium confidence) (Nissen et al., 2014; Feser et al., 2015; Forzieri et al., 2016; Mölter et al., 2016; Tous et al., 2016; Romera et al., 2017; González-Alemán et al., 2019; MedECC, 2020; Chapter 11).

Projections of smaller-scale hazard phenomena such as tornadoes, wind gusts, hail storms and lightning are currently not directly available partly due to the inability of climate models to simulate such phenomena. Observational networks for such phenomena usually lack homogeneity over long periods, hindering clear trends to be detected. For instance, while no robust trends have been identified (Hermida et al., 2015; Mohr et al., 2015a; Burcea et al., 2016; Ćurić and Janc, 2016), hail storm environments (favourable atmospheric configurations) have increased in frequency (low confidence, limited evidence) (Sanchez et al., 2017). In future climate periods it is more likely than not that severe convection environments will become more frequent by the end of the century under RCP8.5 (Mohr et al., 2015b; Púčik et al., 2017), and there is medium confidence that such environments will become more frequent by the 2050s in RCP4.5. There is no evidence for changes in tornado frequencies in Europe in the observations (Groenemeijer and Kühne, 2014) as well as in future climate projections. Insufficient observational record length for lightning numbers does not allow an assessment of trends.

There is high confidence that mean wind speeds will decrease in Mediterranean areas and medium confidence of such decreases in Northern Europe for global warming levels of 2°C or more and beyond the middle of the century. A slightly increased frequency and amplitude of extratropical cyclones, strong winds and extratropical storms is projected for northern, central and western Europe by the middle of the century and beyond and for global warming levels of 2°C or higher (medium confidence). The frequency of Medicanes is projected to decrease (medium confidence), but their intensity is projected to increase by mid century and beyond and for global warming levels of 2°C or more. Proxies of intense convection indicate that the large-scale conditions conducive to severe convection will tend to increase in the future climate (low confidence).

Snow: Widespread and accelerated declines in snow depth (Fontrodona Bach et al., 2018) and snow water equivalent (Marty et al., 2017a; see Figure 12.9b) have been observed in Europe. In the Pyrenees a slow snow cover decline has been observed starting from the industrial period with a sharp increase since 1955 (López-Moreno et al., 2020). Under the RCP2.6, RCP4.5 and RCP8.5 scenarios the reliability elevation for snowmaking will rise by 200–300 m in the Alps and 400–600 m in the Pyrenees by mid-century. End of century projections of natural snow conditions are highly dependent on the scenario, being stationary for the RCP2.6 and continuously decreasing under RCP8.5 to not have any more natural snow conditions at any of the locations in the French Alps and Pyrenees (Spandre et al., 2019). Similarly Norway and Austria will also see a rising of the natural snow elevation with consequences for the ski season (Scott et al., 2020; Steiger and Scott, 2020). In the Alps, recent simulations project a reduction in snow water equivalent (SWE) at 1500 m above sea level of 80–90% by 2100 under the A1B scenario and a snow season that would start 2-4 weeks later and end 5-10 weeks earlier than the 1992–2012 average (Schmucki et al., 2015), which is equivalent to a shift in elevation of about 700 m (Marty et al., 2017b). For elevations above 3000 m above sea level, a decline in SWE of at least 10% is expected by the end of the century even when assuming the largest projected precipitation increase. Similar trends are observed for the Pyrenees and Scandinavia (López-Moreno et al., 2009; Räisänen and Eklund, 2012). For the northern French Alps above 1500 m and the Ötztal locations in the Austrian alps SWE has a similar decreasing trend altitudinally dependent for RCP2.6, RCP4.5 and RCP8.5 until mid-century and with significant differentiation among them in the second half of the century up to snow-free conditions under RCP8.5 (Hanzer et al., 2018; Verfaillie et al., 2018).

Glacier: Observations and future projections of European glacier mass changes are assessed in Section 9.5.1 grouped in two main regions: Scandinavia and central Europe regions. It is virtually certain that glaciers will shrink in the future and there is medium confidence in the timing and mass change rates (Section 9.5.1). Central Europe is one of the regions where glaciers are projected to lose substantial mass even under low-emissions scenarios (Section 9.5.1.3; MedECC, 2020). GlacierMIP projections indicate that glaciers in the central Europe region will lose 63 ± 31%, 80 ± 22% and 93 ± 13% of their 2015 mass by the end of the century under RCP2.6, RCP4.5 and RCP8.5 respectively (Marzeion et al., 2020). For the same scenarios, glaciers in Scandinavia are projected to lose 55 ± 33%, 66 ± 34% and 82 ± 24% of their 2015 mass. The virtually certainshrink in glaciers is bolstered by RCM simulations from the EURO-CORDEX ensemble, with the Global Glacier Evolution Model (GloGEM) indicating a substantial reduction of glacier ice volumes in the European Alps by 2050 (47–52% with respect to 2017 for RCP2.6, RCP4.5 and RCP8.5). Under RCP2.6, about two-thirds (63 ± 11%) of the present-day (2017) ice volume is projected to be lost by 2100. In contrast, under the strong warming of RCP8.5, glaciers in the European Alps are projected to largely disappear by 2100 (94 ± 4% volume loss compared to 2017; Zekollari et al., 2019).

Permafrost: In Europe, permafrost is found in high mountains and in Scandinavia, as well as in Arctic Islands (e.g., Iceland, Novaya Zemlia or Svalbard). In recent decades permafrost has been lost (Section 9.5.2) and accelerated warming at high altitudes and latitudes has favoured an increase of permafrost temperatures of the order of 0.2 ± 0.1°C between 2007 and 2016 (Romanovsky et al., 2018; Noetzli et al., 2019). Over the 21st century, permafrost is very likely to undergo increasing thaw and degradation under all scenarios (Hock et al., 2019) and it is virtually certain that permafrost extent and volume will decrease with increase of global warming (Section 9.5.2).

Permafrost thawing is projected to affect the frequency and magnitude of high-mountain mass wasting processes (Stoffel and Huggel, 2012). The temporal frequency of periglacial debris flows in the Alps is unlikely to change significantly by the mid-21st century but is likely to decrease during the second part of the century under the A1B scenario, especially in summer (Stoffel et al., 2011, 2014). There is medium confidence that most of the Northern Europe periglacial processes will disappear by the end of the century, even in the RCP2.6 scenario (Aalto et al., 2017). The magnitude of debris flow events might increase (Lugon and Stoffel, 2010) and the debris-flow season may last longer under the A1B scenario (Stoffel and Corona, 2018). Quantitative data for the European Alps is highly site dependent (Haeberli, 2013).

Heavy snowfall, ice storms and hail: There is low confidence that climate change will affect ice and snow-related episodic hazards (limited evidence). The change in snowpack in the Alps is expected to lead to a possible reduction in overall avalanche activity by end of the century (low confidence), except possibly in winter and at high altitudes (Castebrunet et al., 2014).

For ice storms, or freezing rainstorms, there is also limited evidence due to a limited number of studies. Heavy snowfalls have decreased in frequency in the past decades and this is expected to continue in the future climate (low confidence) (Beniston et al., 2018). Freezing rain is projected to increase in western, central and southern Europe by the end of the century under RCP4.5 and RCP8.5 (low confidence) (Kämäräinen et al., 2018). Rain-on-snow events, are decreasing in northern regions (Pall et al., 2019) and by 48% on average in southern Scandinavia (Poschlod et al., 2020) due to decreases in snowfall.

In summary, future snow cover extent and seasonal duration will reduce (high confidence) and it is virtually certain that glaciers will continue to shrink. A reduction of glacier ice volume is projected in the European Alps and Scandinavia (high confidence). There is high confidence that permafrost will undergo increasing thaw and degradation over the 21st century. Most of the Northern Europe periglacial will disappear by the end of the century even for a lower emissions scenario (medium confidence) and the debris-flow season may last longer in a warming climate (medium confidence).

Relative sea level: Around Europe, over 1900–2018, a new tide gauge-based reconstruction finds a regional mean RSL change of 1.08 [0.79 to 1.38] mm yr^–1 in the subpolar North Atlantic (Frederikse et al., 2020), compared to a GMSL change of around 1.7 mm yr^–1Section 2.3.3.3 and Table 9.5). For the period 1993–2018, the RSLR rates around Europe, based on satellite altimetry, increased to 2.17 [1.66 to 2.66] mm yr^–1 (Frederikse et al., 2020), compared to a GMSL change of 3.25 mm yr^–1Section 2.3.3.3 and Table 9.5).

Relative sea level rise is extremely likely to continue in the oceans around Europe. Regional mean RSLR projections for the oceans around Europe range from 0.4–0.5 m under SSP1-2.6 to 0.7–0.8 m under SSP5-8.5 for 2081–2100 relative to 1995–2014 (median values), which means that there are locally large deviations from the projected GMSL change (Section 9.6.3.3). These RSLR projections may however be underestimated due to potential partial representation of land subsidence in their assessment (Section 9.6.3.2). The signal is strongest for the North Sea and Atlantic coasts, followed by the Black Sea. The Baltic Sea, on the contrary, shows the lowest increase due to land uplift (Vousdoukas et al., 2017). The model agreement is higher for the Mediterranean and in line with previous findings by Gualdi et al. (2013).

Coastal flood: The present-day 1-in-100-year ETWL is between 0.5 and 1.5 m in the MED basin and 2.5 and 5.0 m in the western Atlantic European coasts, around the UK and along the North Sea coast, and lower at 1.5–2.5 m along the Baltic Sea coast (Kirezci et al., 2020). Similar values are reported by Vousdoukas et al. (2018).

There is high confidence that extreme total water level (ETWL) magnitude and occurrence frequency will increase throughout Europe (see Figure 12.4p–r), except in the northern Baltic Sea. Across the region, the 5–95th percentile range of the 1-in-100-year ETWL is projected to increase (relative to 1980–2014) by 4–40 cm and by 6–47 cm by 2050 under RCP4.5 and RCP8.5, respectively. By 2100, this range is projected to be 6–88 cm and 25–186 cm under RCP4.5 and RCP8.5, respectively (Figure 12.SM.6; Vousdoukas et al., 2018; Kirezci et al., 2020). Mass addition across the Gibraltar Strait may play a role, although the extent of this contribution is currently unclear (Lionello et al., 2017). Furthermore, under RCP4.5, the present day 1-in-100-year ETWL is projected to have median return periods of between 1-in-5 and 1-in-20 years by 2050 and occur at least once per year by 2100 in the Mediterranean and Black Sea, while in the rest of Europe it is mostly projected to have median return periods of between 1-in-20-years and 1-in-50-years by 2050 and between 1-in-5-years and 1-in-20-years by 2100 (Vousdoukas et al., 2018). Under RCP8.5, occurrence of the present day 1-in-100-year ETWL is projected to increase further to median return periods of 1-in-1-year to 1-in-5-years by 2050 and occur more than once per year by 2100 in the Mediterranean and Black Sea, while in the rest of Europe it is mostly projected to have median return periods between 1-in-5-years and more than once per year by 2100.

Coastal erosion: Satellite-derived shoreline change estimates over 1984–2015 indicate shoreline retreat rates of around 0.5 m yr^–1along the sandy coasts of WCE and MED, around 4 m yr^–1 in EEU (Caspian Sea region) and more or less stable shorelines in NEU (Luijendijk et al., 2018; Mentaschi et al., 2018). Mentaschi et al. (2018) report a coastal area loss of 270 km² over a 30-year period (1984–2015) along the Atlantic coastlines of Europe.

Projections indicate that sandy coasts throughout the continent (except those bordering the northern Baltic Sea) will experience shoreline retreat through the 21st century (high confidence). Median shoreline change projections (CMIP5) relative to 2010, show that, by mid-century, shorelines will retreat by between 25 m and 60 m along sandy coasts in WCE and MED under both RCP4.5 and RCP8.5 (Athanasiou et al., 2020; Vousdoukas et al., 2020b). Mid-century median projections for NEU indicate virtually no shoreline retreat under RCP4.5, but a retreat of around 40 m under RCP8.5. By 2100, median shoreline retreats of around 50 m are projected in NEU and MED under RCP4.5, increasing to around 80 m under RCP8.5. End-century median projections for WCE are far higher at 100 m (RCP4.5) and 160 m (RCP8.5). The total length of sandy coasts in Europe that is projected to retreat by more than a median of 100 m by 2100 under RCP4.5 and RCP8.5 is about 12,000 km and 18,000 km respectively, an increase of approximately 54% (Vousdoukas et al., 2020b).

Local assessments of both long term shoreline retreat and episodic coastal erosion are given by Li et al. (2014b), Toimil et al. (2017), Bon de Sousa et al. (2018) and Le Cozannet et al. (2019). In terms of episodic coastal erosion, 31–88% of all Aegean beaches are projected to experience complete erosion, with a RCP4.5 sea level rise of 0.5 m and a surge of 0.6 m, but with substantial uncertainty (Monioudi et al., 2017).

Marine heatwave: The mean SST of the Atlantic Ocean and the Mediterranean has increased between 0.25°C and 1°C since 1982–1998. This mean ocean surface warming is correlated to longer and more frequent marine heatwaves in the region (Oliver et al., 2018). Over the period 1982–2016, the coastlines of Europe experienced on average more than 2.0 MHW yr^–1, with the eastern Mediterranean and Scandinavia experiencing 2.5–3 MHWs yr^–1. The average duration was between 10 and 15 days. Changes over the 20th century, derived from MHW proxies, show an increase in frequency of between 1.0 and 2.0 MHWs per decade in Europe, although the trend is not statistically significant; with an increase in intensity per event in the North Atlantic and the Mediterranean, and a decrease in the Atlantic off the British Isles. The total number of MHW days per decade has increased in the Mediterranean (Oliver et al., 2018).

Mean SST is projected to increase by 1°C–3°C around Europe by 2100, with a hotspot of around 4°C–5°C along the Arctic coastline of Europe under RCP4.5 and RCP8.5 scenarios (see Interactive Atlas), leading to a continued increase in MHW frequency, magnitude and duration (Oliver et al., 2018; MedECC, 2020). Projections for SSP1-2.6 and SSP5-8.5 both show an increase in MHWs around Europe by 2081–2100, relative to 1985–2014 (Box 9.2, Figure 1). Darmaraki et al. (2019) project that, by the end of the 21st century and under RCP8.5, there will be one MHW occurring every year in the northern Mediterranean sea, and that these MHWs would be three months longer, four times more intense, and 42 times more severe than present day MHWs in the region. Frölicher et al. (2018) show that, in Europe, the change in the probability for the number of days of MHWs exceeding the 99th percentile of the pre-industrial level is 4%, 15% and 30% for global warming levels of 1°C, 2°C and 3.5°C, respectively. MHW increase in the Mediterranean will impact on many species that live in shallow waters and have reduced motility, with consequences for related economic activities (Galli et al., 2017).

In general, there is high confidence that most coastal/ocean-related climatic impact-drivers in Europe will increase over the 21st century for all scenarios and time horizons. Relative sea level rise is extremely likely to continue around Europe (except in the northern Baltic Sea), contributing to increased coastal flooding in low-lying areas and shoreline retreat along most sandy coasts (high confidence). Marine heatwaves are also expected to increase around the region over the 21st century (high confidence).

Compound events: One typical compound event that is observed in the European area is compound flooding due to the combination of extreme sea level events and extreme precipitation events associated with high levels of runoff. In the present climate, the Mediterranean coasts are exposed to a higher probability of this type of compound flooding event (Bevacqua et al., 2019). Under RCP8.5, the probability of these events is projected to increase along northern European coasts (west coast of UK, northern France, the east and south coast of the North Sea, and the eastern half of the Black Sea), with the percentage of coastline now experiencing such events at least once every 6 years increasing by between 3% and 11% by the end of the 21st century (Bevacqua et al., 2019).

Under RCP8.5, regions in Russia, France and Germany are projected to experience an increase in the frequency and the length of wet and cold compound events, while Spain and Bulgaria are projected to stay longer in the hot and dry state by mid-century (Sedlmeier et al., 2016).

Compound events of dry and hot summers have increased in Europe. Manning et al. (2019) found that the probability of such compound events has increased across much of Europe between 1950–1979 and 1984–2013, notably in southern, eastern and western Europe. Compound hot and dry extremes are projected to increase in Europe by mid-century for the SRES A1B and RCP8.5 with a particularly strong signal projected in southern and eastern Germany and the Czech Republic (Sedlmeier et al., 2016).

The assessed direction of change in climatic impact-drivers for Europe and associated confidence levels are illustrated in Table 12.7, together with emergence time information (Section 12.5.2). No assessable literature could be found for sand and dust storms, although these phenomena may be relevant in parts of the region.

Table 12.7 | Summary of confidence in direction of projected change in climatic impact-drivers in Europe, representing their aggregate characteristic changes for mid-century for scenarios RCP4.5, SSP2-4.5, SRES A1B, or above within each AR6 region (defined in Chapter 1), approximately corresponding (for CIDs that are independent of sea level rise) to global warming levels between 2°C and 2.4°C (see Section 12.4 for more details of the assessment method). The table also includes the assessment of observed or projected time-of-emergence of the CID change signal from the natural interannual variability if found with at least medium confidence in Section 12.5.2.

Mean air temperature: Atlas.9.2 assessed very likely mean warming in observations across North America, with highest increases at higher latitudes and in the winter season. Atlas.9.4 assessed very likely mean warming in future decades in all North American regions, with CMIP and CORDEX models showing median increases exceeding 2°C in much of the continental interior under RCP8.5 (2041–2060 compared to 1995–2014) and higher increases towards the north. Mean temperatures at the end of century show strong scenario dependence, rising between 1°C and 2.5°C in RCP2.6 and about 4°C to 8°C in RCP8.5 (Figures Atlas.12, Atlas.26 and Atlas.27). Warming also raises stream temperatures across the continent (DOE, 2015; Trtanj et al., 2016; van Vliet et al., 2016; Chapra et al., 2017), and Hill et al. (2014) projected US stream warming by 0.6°C (±0.3°C) per 1°C increase in local air temperature. Mean warming drives shifts in the seasonal timing of temperature thresholds, including increasing growing degree days (Mu et al., 2017), longer growing seasons (Gowda et al., 2018; G. Li et al., 2018; L.A. Vincent et al., 2018), reduced chill hours (Luedeling, 2012; Lee and Sumner, 2015; Xie et al., 2015; Parker and Abatzoglou, 2019), and longer pollen and allergy seasons (Fann et al., 2016; Anenberg et al., 2017; Sapkota et al., 2019). Warmer temperatures reduce heating degree days and increase cooling degree days (high confidence) (Bartos et al., 2016; US EPA, 2016; Craig et al., 2018; X. Zhang et al., 2019; Coppola et al., 2021b)

Extreme heat: Section 11.9 assessed that extreme temperatures in North America have increased in recent decades (medium evidence, medium agreement) other than in Central and Eastern North America (low confidence), and extreme heat in all regions is projected to increase with climate change (high confidence). Observed trends in extreme heat are more positive for heat extreme indices that include temperature and humidity given historical expansion of irrigation and intensification of agriculture (Mueller et al., 2017; Grotjahn and Huynh, 2018; Thiery et al., 2020). Several studies noted statistically significant increases in intensity and particularly the frequency, duration, and seasonal length of the physiologically hazardous extreme heat conditions across North America (Grineski et al., 2015; Habeeb et al., 2015; Martínez-Austria et al., 2016; Petitti et al., 2016; L.A. Vincent et al., 2018; García-Cueto et al., 2019).

Figure 12.4b shows over a month of additional days at CMIP6 SSP5-8.5 mid-century where temperatures exceed 35°C across much of southern Mexico and regions near the US–Mexico border, and these extreme temperatures occur at least once per year up to southern Canada. Coppola et al. (2021b) found similar patterns in CMIP5 and CORDEX-Core. Using locally tailored heat thresholds, Maxwell et al. (2018) found that ‘very hot’ days in five US cities will occur a median of three to five times more often by 2036–2065 under RCP8.5 (2 to 3.5 times more often in RCP4.5), Oleson et al. (2018) projected that annual heatwave duration will exceed one month in Houston in RCP8.5 2061–2080, and Anderson et al. (2018) projected 7 to 12 times more exceedances of thresholds associated with high-mortality by 2061–2080 under RCP8.5 (6 to 7 times more exceedances in RCP4.5). Schwingshackl et al. (2021) found that Central and Eastern North America are among the regions with the strongest trend in heat stress indicators. Studies also project increasingly surpassed heat extreme thresholds for North American crops (Gourdji et al., 2013), airplane weight restrictions (Coffel et al., 2017), and peak load energy systems (Auffhammer et al., 2017).

The number of days crossing dangerous heat thresholds such as HI > 41°C will be very sensitive to the mitigation scenario at the end of the century (Wuebbles et al., 2014; Zhao et al., 2015; Dahl et al., 2019; Schwingshackl et al., 2021). At the end of the century under SSP5-8.5, a CMIP6 median increase of exceedances of 75–150 days per year is projected over much of North Central America, Central North America and the south-western USA while this increase is projected to remain limited below 60 days under SSP1-2.6 (Figure 12.4d,f and Figure 12.SM.2). Steinberg et al. (2018) also projected more frequent and longer ‘heat-health’ events in California extending into October.

Cold spell: Chapter 11 assessed high confidence in decreasing frequency and intensity of cold spells over North America (Section 11.9). The number of days with extreme wind chill hours (humidex <–30) decreased at 76% of examined Canadian stations from 1953 to 2012 (Mekis et al., 2015) and cold days and coldest nights decreased in Mexico from 1980 to 2010 (García-Cueto et al., 2019).

Cold spells are projected to decrease over North America under climate change, with the largest decreases most common in the winter season (high confidence) (Section 11.9). Minimum winter temperatures are projected to rise faster than the mean winter temperature (Underwood et al., 2017) and alter cold-hardiness zones used to determine agricultural suitability (Parker and Abatzoglou, 2016). Wuebbles et al. (2014) projections for RCP8.5 end-of-century show that the four-day cold spell that happens on average once every five years is projected to warm by more than 10°C and CMIP5 models do not project current 1-in-20-year annual minimum temperature extremes to recur over much of the continent. Multiple studies have shown that Arctic warming can alter large-scale variability and change the frequency and duration of mid-latitude cold air outbreaks, potentially leading to increasing cold hazards in some regions (low agreement) (Barcikowska et al., 2019; Cohen et al., 2020; Zhou et al., 2021).

Frost: An expansion of the frost-free season is underway and projections for North America indicate a continuation of this trend in the future (high confidence). Significant decreases in frost days, consecutive frost days, and ice days were identified in 1948–2016 station observations across Canada, along with a resulting lengthening of the frost-free season by more than a month in many regions (L.A. Vincent et al., 2018). Frost days also declined in nearly all Mexican cities from 1980 to 2010 (García-Cueto et al., 2019), and a 1917–2016 decline of about three weeks in frigid winter conditions challenges ecosystems in the north-east USA and south-east Canada (Contosta et al., 2020). Studies connect projections of a longer frost-free season in North America to a longer outdoor construction season, shifts in frost variance to orchard damages, and lower weight tolerances for runways (Daniel et al., 2018; DeGaetano, 2018; Jacobs et al., 2018). Frosts are projected to persist as an episodic hazard in many regions given natural variability and cold air outbreaks even as mean temperature rises (high confidence).

Climate change is virtually certain to shift the balance of temperature towards warming trends and away from cold extremes, with increases in the magnitude, frequency, duration and seasonal and spatial extent of heat extremes driving impacts across North America. The frequency of dangerous heat threshold exceedance (such as HI > 41°C) is particularly sensitive to scenario pathway, with 7 5–1 50 days more under SSP5-8.5 but less than 60 days more under SSP1-2.6 by the end of the century in NCA, CNA and the south-western USA.

Mean precipitation: Atlas.9.2 found that trends in annual precipitation over 1960–2015 are generally non-significant, though there are consistent positive trends over parts of ENA and CNA, together with significant decreases in precipitation in parts of the south-western USA and north-western Mexico.

Atlas.9.4 assessed very high confidence in increases in mean precipitation over most of northern and Eastern North America, with medium confidence of decrease over Northern Central America and low confidence elsewhere (see Figure Atlas.26, and Cross-Chapter Box Atlas.1, Figure 1). Changes are most dramatic in the spring and winter, when wet conditions are projected to extend from the northern portions of the continent as far south as the central Great Plains, while Mexico becomes drier; in contrast, summer changes are uncertain across most of the continent other than wetter conditions in northern Canada (Easterling et al., 2017; Bukovsky and Mearns, 2020; Almazroui et al., 2021; Teichmann et al., 2021).

River flood: There is limited evidence and low agreement on observed climate change influences for river floods in North America (Section 11.5). Trends in streamflow indices are mixed and difficult to separate from river engineering influences, with large changes but little spatial coherence across the USA, making it difficult to identify trends with confidence (Peterson et al., 2013; Mallakpour and Villarini, 2015; Archfield et al., 2016; Wehner et al., 2017; Villarini and Slater, 2018; Hodgkins et al., 2019; Neri et al., 2019). There is high confidence in historical shifts in the timing of peak streamflow towards higher winter and earlier spring flows in snowmelt-driven basins in Canada (Burn and Whitfield, 2016; Bonsal et al., 2019) and the USA (Dudley et al., 2017; Wehner et al., 2017, Neri et al., 2020). Some rivers show ice-jam floods occurring a week earlier, but changes are mixed, given localized positive and negative changes across the continent (Rokaya et al., 2018). There is medium confidence that climate change will increase river floods over the USA and Canada but low confidence for changes in Mexico. Wobus et al. (2017a) used a regional hydrologic model for 57,000 streams to project more than a doubling in the frequency of current 1-in-100-year flow events in many portions of the USA for RCP8.5 2050 with additional contributions from earlier snowmelt. CMIP6 projections for SSP5-8.5 2065–2099 show strongest peak USA runoff increases in the east (Villarini and Zhang, 2020); however, several studies applying global hydrological models disagree with regional streamflow projections, indicating a decrease in the magnitude or frequency of floods over a large portion of North America (e.g., Hirabayashi et al., 2013; Arnell and Gosling, 2016; see Figure 12.10a,c).

Figure 12.10 | Projected changes in selected climatic impact-driver indices for North America. (a) Mean change in 1-in-100-year river discharge per unit catchment area (Q100, m³s^–1km^–2), and (b) median change in the number of days with snow water equivalent (SWE) over 100 mm (from November to March), from CORDEX-North America models for 2041–2060 relative to 1995–2014 and RCP8.5. Diagonal lines indicate where less than 80% of models agree on the sign (direction) of change. (c) Bar plots for Q100 (m³s^–1km^–2) averaged over land areas for the AR6 WGI Reference Regions (defined in Chapter 1). The left-hand column within each panel (associated with the left-hand y-axis) shows the ‘recent past’ (1995–2014) Q100 absolute values in grey shades. The other columns (associated with the right-hand y-axis) show the Q100 changes relative to the recent past values for two time periods (‘mid’ 2041–2060 and ‘long’ 2081–2100) and for three global warming levels (defined relative to the pre-industrial period 1850–1900): 1.5°C (purple), 2°C (yellow) and 4°C (brown). The bars show the median (dots) and the 10–90th percentile range of model ensemble values across each model ensemble. CMIP6 is shown by the darkest colours, CMIP5 by medium, and CORDEX by light. SSP5-8.5/RCP8.5 is shown in red and SSP1-2.6/RCP2.6 in blue. (d) As for (c) but showing absolute values for number of days with SWE > 100 mm, masked to grid cells with at least 14 such days in the recent past. See Technical (Annex VI for details of indices. A Caribbean (CAR) Q100 bar plot is included here but assessed in the Small Islands section (Section 12.4.7). Further details on data sources and processing are available in the chapter data table (Table 12.SM.1).

Open figure

Heavy precipitation and pluvial flood: Section 11.4 assessed high confidence in observed increases in extreme precipitation events (including hourly totals) in Central and Eastern North America with low confidence in broad trends elsewhere in the continent despite observational increases in some portions of each region (L.A. Vincent et al., 2018; García-Cueto et al., 2019; X. Zhang et al., 2019).

(Section 11.4 found that high precipitation is projected to increase across North America (high confidence) except for portions of Western North America where projections are mixed (medium confidence of increase). Maxwell et al. (2018) identified regional ‘heavy precipitation day’ thresholds for five cities across the USA and projected that a tripling (or more) of these events is possible by RCP8.5 mid-century. Projections indicate changes to intensity-duration-frequency (IDF) curves typically used for construction design and automobile hazards, as well as increases in the 10-year recurrence level of 24-hour rainfall intensities that challenge storm water drainage systems (Hambly et al., 2013; Cheng and AghaKouchak, 2015; J.E. Neumann et al., 2015; Prein et al., 2017b; Hettiarachchi et al., 2018; Ragno et al., 2018). Precise levels of regional IDF characteristics may still depend substantially on the method and resolution of downscaling applied (DeGaetano and Castellano, 2017; L.M. Cook et al., 2020).

Landslide: There is growing yet limited evidence for unique climate-driven changes in landslide and rockfall hazards in North America, even as theory suggests decreases in slope and rockface stability due to more intense rainfall, rain-on-snow events, mean warming, permafrost thaw, glacier retreat, and coastal erosion (Cloutier et al., 2017; Coe et al., 2018; Handwerger et al., 2019; Hock et al., 2019; Patton et al., 2019) although dry trends can decelerate mass movements (Bennett et al., 2016). Landslide frequency has increased in British Columbia (Canada; Geertsema et al., 2006) and is expected to increase in North-Western North America given the combination of these factors (medium confidence) (Gariano and Guzzetti, 2016). Cloutier et al. (2017)projected an increase in landslides in western Canada due to wetter overall conditions and reduced return period for extreme rainfall. Robinson et al. (2017) used scenarios based upon projection of 50-year recurrence of 7-day precipitation periods to highlight the potential for increased landslide hazards near Seattle (USA). Broad projections for the USA are more uncertain given increases in evapotranspiration that will counteract precipitation changes over much of the country (Coe, 2016).

Aridity: Chapter 8 showed that aridity in North America generally moves opposite to mean precipitation change with an added evaporative demand from warmer temperatures (high confidence in aridity increase for Northern Central America; medium confidence for an increase in Central North America; high confidence for a decrease in North-Eastern North America; medium confidence decreases in Eastern and North-Western North America; see also Section 11.9). Projected soil moisture declines (Figure 12.4j–l) are most widespread across North America during the summer, with the largest declines in Mexico and the southern Great Plains but also extending into Canada (Section 8.4.1.6; Swain and Hayhoe, 2015; Easterling et al., 2017; Bonsal et al., 2019; J. Lu et al., 2019). Yoon et al. (2018) found net reduction in southern Great Plains groundwater storage in RCP8.5 mid-century projections despite increases in mean precipitation and both wet and dry extremes. Soil moisture drying could reach unprecedented levels by the CMIP6 RCP8.5 end-of-century, even when evaluating deeper soil columns relevant for crop rooting depth (B.I. Cook et al., 2020). Projected changes in the aridity index portray a shift the geographic range of temperate drylands northward and eastward in Central and Western North America (Schlaepfer et al., 2017; Seager et al., 2018) which also diminishes aquifer recharge rates in the southern Great Plains and in some western regions where snowpack is reduced (Meixner et al., 2016).

Hydrological drought: Section 11.9 asssessed low confidence of significant observational trends and projected future changes in the characteristics of episodic hydrological drought in North America given limited evidence and low agreement in modeled changes. C. Zhao et al. (2020) found that increases in hydrological drought frequency (particularly the 100 yr drought) were far more prevalent than for meteorological drought across 5797 watersheds in the USA and Canada, indicating a strong influence of evaporative demand. Reductions in the overall supply of meltwater from a declining snowpack also increase the potential for intermittent hydrological droughts in the western USA (Mote et al., 2018; Livneh and Badger, 2020).

Agricultural and ecological drought: Section 11.9 assessed medium confidence for an increase in agricultural and ecological drought in Western North America but otherwise found limited evidence for broadly observed changes in North American agricultural and ecological drought, even as increasing evaporative demand intensified vegetation stress and soil moisture deficits in recent events (Sections 11.6, 11.9). Section 11.9 asssessed medium confidence for more intense agricultural and ecological drought conditions over North Central America, Western North America and Central North America in a 2°C global warming level (about mid-century), with medium confidence extending to Eastern North America and high confidence for Northern Central America and Central North America under a 4°C global warming level associated with higher emissions scenarios past 2050. Figure 12.4g–i shows that the frequency of meteorological droughts (which often initiate hydrological, agricultural and ecological droughts) is largely projected to increase in North American areas where total precipitation decreases (and vice versa; see Section 11.9 and Coppola et al., 2021b), and higher evaporative demand will extend the regions where more intense ecological and agricultural droughts develop when meteorological droughts occur (Wehner et al., 2017; B.I. Cook et al., 2019, 2020). Studies utilizing a variety of drought indices and soil moisture projections consistently project increased drought extending from Mexico into the southern Canadian Plains during the summer (Swain and Hayhoe, 2015; Ahmadalipour et al., 2017; Feng et al., 2017; Bonsal et al., 2019; Tam et al., 2019; B.I. Cook et al., 2020).

Fire weather: Climatic conditions conducive to wildfire have increased in Mexico, Western and North-Western North America, primarily due to warming (high confidence). Abatzoglou and Williams (2016) found climate change led to higher values for eight fuel aridity indices over the western USA in recent decades, with 2000–2015 changes exposing 75% more forested area to high fuel aridity and adding nine more high fire-potential days each year, similar to 1979–2013 western USA and Mexico fire season expansion reported in Jolly et al. (2015). Increases in lightning-initiated fires have been distinguished from trends in man-made fire in western Canada and the USA (Balch et al., 2017; Hanes et al., 2019). Jain et al. (2017) identified a 1979–2015 expansion in fire weather season in eastern Canada and the south-western USA (with a smaller reduction in the northern Mountain West) along with regional shifts in the 99th percentile Canadian Fire Weather Index (FWI) and potential fire-spread days. Girardin and Wotton (2009) noted that 1951–2002 trends in the Monthly Drought Code fire index in eastern Canada could hardly be distinguished from decadal variability.

Climate change drives future increases in North American fire weather, particularly in the south-west (high confidence), although further studies on shifts in exposure and vulnerability are needed to understand overall fire risks (see WGII Chapter 14). A significant increase of FWI is apparent before 2050 under RCP8.5 in much of North America, including the frequency of 95th-percentile FWI days, peak seasonal FWI average, fire weather season length, and maximum fire weather index (Abatzoglou et al., 2019), and fire season across North America expands dramatically beyond 2°C global warming levels (Q. Sun et al., 2019; Jain et al., 2020). X. Wang et al. (2017b) simulated fire-spread days across Canada and found increases across most of the areas studied by 2071–2100, with median changes of –20 to +140% (RCP2.6), –20 to +250% (RCP4.5) and 40 to 360% (RCP8.5) compared to 1976–2015. Prestemon et al. (2016) found more conducive conditions for lightning-ignited fires in the south-eastern USA by mid-century, while warming conditions in Alaska increasingly push July temperatures above 13.4°C, a threshold for fire danger across Alaska’s tundra and boreal forest (Partain et al., 2016; Young et al., 2017). Longer and more intense fire seasons would also raise particulate matter and black carbon concentrations in the western USA, reducing visibility at many National Parks (Yue et al., 2013; Val Martin et al., 2015).

Changes in North American wet and dry climatic impact-drivers are largely organized by the ‘north-east (more wet) to south-west (more dry)’ pattern of mean precipitation change, although heavy precipitation increases are widespread and increasing evaporative demand expands aridity, agricultural and ecological drought, and fire weather (particularly in summer) (high confidence).

Mean wind speed: Mean wind speeds have declined in North America – as in other Northern Hemisphere areas – over the past four decades (medium confidence) (AR5 WGI) with a reversal in the last decade (low confidence) not fully consistent across studies (Tian et al., 2019; Zeng et al., 2019; Z. Zhang et al., 2019). Tian et al. (2019) found a corresponding reduction in the wind power potential across the eastern parts of North America.

Mean wind speeds are expected to decline over much of North America (Figure 12.4m–o), but the only broad signal of consistent change across model types is a reduction in wind speed in Western North America (high confidence). These declines reduce wind power endowment by 2050 and as early as the 2020–2040 near-term period in the USA Mountain West, while there is disagreement between global- and regional-model change projections in the upper and lower Great Plains, Ohio River Valley, Mexico and eastern Canada (Karnauskas et al., 2018a; Jung and Schindler, 2019; Chen, 2020).

Severe wind storm: There is limited evidence and low agreement in observed changes in North American CID indices associated with extratropical cyclones (Section 11.7), severe thunderstorms, severe wind bursts (derechos), tornadoes, or lightning strikes (Vose et al., 2014; Easterling et al., 2017; Kossin et al., 2017). Observational studies have indicated a reduction in the number of tornado days in the USA, but increases in outbreaks with 30 or more tornadoes in one day (Brooks et al., 2014), the density of tornado clusters (Elsner et al., 2015), and overall tornado power (Elsner et al., 2019).

There is medium confidence of a general decrease in the number of extratropical cyclones producing high wind speeds in North America, except over northernmost parts, for a global warming level of 2°C or by the end of the century under RCP4.5 and RCP8.5 (Kumar et al., 2015; Jeong and Sushama, 2018a; C. Li et al., 2018). GCMs cannot directly resolve tornadoes and severe thunderstorms, however projections of favourable environments for severe storms (based on convective available potential energy and wind shear) indicate medium confidence for more severe storms and a longer convective storm season in the USA, weaker increases extending north and east (Seeley and Romps, 2015; Glazer et al., 2021), and a corresponding increase in autumn and winter tornadic storms (H.E. Brooks, 2013; Diffenbaugh et al., 2013; Brooks et al., 2014; see also Section 11.7.2). Prein et al. (2017a) used a convection-permitting model to project a tripling of mesoscale convective systems over the USA for end-of-century RCP8.5.

Tropical cyclone: Section 11.7.1 identified recent reductions in tropical cyclone translation speed and higher tropical cyclone rainfall totals over the North Atlantic, as well as substantial natural variability. Projections indicate low confidence in change in North Atlantic tropical cyclone numbers, but medium confidence in Mexico and the US Gulf and Atlantic coasts for more intense storms with higher wind, precipitation, and storm surge totals when they do occur (Section 11.7.1; Diro et al., 2014; DOE, 2015; Walsh et al., 2016a; Kossin et al., 2017; Marsooli et al., 2019; Ting et al., 2019; Knutson et al., 2020). A more rapid intensification of tropical cyclone winds and destructive power also heightens the tropical cyclone hazard (Bhatia et al., 2019). Greenhouse gas forcing is projected to shift tropical cyclones poleward (Kossin et al., 2016), while also holding the potential for higher precipitation totals (Risser and Wehner, 2017; Knutson et al., 2020) particularly given evidence that storms increasingly stall near North American coastlines (Hall and Kossin, 2019).

Sand and dust storm: Land-use change has increased dust emissions in the western USA in the past 200 years (Neff et al., 2008). However, there is medium confidence for observed increases in Western North American sand and dust storm activity since 1980. In their study of Valley Fever spread, Tong et al. (2017) identified a rapid intensification of dust storm activity using PM₁₀ and PM_2.5 observations from 1980–2011 across 29 monitoring sites in the south-western USA, similar to contiguous USA observations by Brahney et al. (2013). Hand et al. (2016) attributed the earlier onset of spring dusts in the south-west in large part to the Pacific Decadal Oscillation, however. The increasing trend in dust since the 1990s in the south-western USA can be explained by precipitation deficit and surface bareness (Pu and Ginoux, 2018). Projections of future sand and dust storms over North America are based on aridity as a primary proxy for conducive conditions which lends medium confidence of an increase over Mexico and the south-western USA. Pu and Ginoux (2017) project about five more dusty days in spring and summer in the southern Great Plains under RCP8.5 at the end of the century, while dusty days decrease in northern regions where mean precipitation tends towards wetter conditions.

Tropical cyclones, severe wind and dust storms in North America are shifting towards more extreme characteristics, with a stronger signal towards heightened intensity than increased frequency, although specific regional patterns are more uncertain (medium confidence). Mean wind speed and wind power potential are projected to decrease in Western North America (medium confidence) with differences between global and regional models lending low confidence elsewhere.

Snow: The seasonal extent of snow cover has reduced over North America in recent decades (robust evidence, high agreement) (see also Sections 2.3.2.2 and 9.5.3, and Figure Atlas.25). The average snow-cover extent in North America decreased at a rate of about 8500 km²yr^–1 over the 1972–2015 period, reducing the average snow cover season by two weeks, primarily due to earlier spring melt (US EPA, 2016). Observations indicate earlier spring snowpack melting (Dudley et al., 2017) and a reduction in end-of-season snowpack metrics important to water resources over the Rocky Mountains (particularly since 1980) and Pacific Northwest (Pederson et al., 2013; Kormos et al., 2016; Kunkel et al., 2016; Fyfe et al., 2017; Mote et al., 2018). In situ measurements in Canada show more heterogenous trends in snow amount and density (Brown et al., 2019).

Climate change is expected to reduce the total snow amount and the length of the snow cover season over most of North America, with a corresponding decrease in the proportion of total precipitation falling as snow and a reduction in end-of-season snowpack (high confidence) (see Atlas.9.5). Changes include a reduction in the number of days with snowfall in across all of North America, with the exception of northern Canada (Danco et al., 2016; McCrary and Mearns, 2019), a delay of about a week in first snowfall in the western USA by 2050 under RCP8.5 (Pierce and Cayan, 2013), and more prominent reductions in Canadian snow cover in the October –December period (Mudryk et al., 2018). Reduced total snowpack and earlier snowmelt lower dry season streamflow (Kormos et al., 2016; Rhoades et al., 2018). Figure 12.10b shows a reduction in days suitable for skiing (SWE > 10 cm; Wobus et al., 2017b) across the USA and southern Canada, although some portions of northern central Canada see an increase.

Glacier: Section 9.5.1 assessed that glaciers in Alaska, western Canada and the western USA are expected to continue to lose mass and areal extent (high confidence). Compared to their 2015 state, glaciers in the western Canada and the USA region will lose 62 ± 30%, 75 ± 29% and 85 ± 23%, of their mass by the end of the century for RCP2.6, RCP4.5 and RCP8.5 scenarios, respectively (Marzeion et al., 2020). Meanwhile glaciers in Alaska will lose 26 ± 21%, 31 ± 24% and 44 ± 27%, of their 2015 mass under the same scenarios. The overall loss of glacial mass can act as a meltwater supply for freshwater resources, although this is expected to peak in the middle of the century and then fade as glaciers disappear (Fyfe et al., 2017; Derksen et al., 2018). Continued shrinkage of glaciers is projected to create further glacial lakes (medium confidence) similar to those that have led to outburst floods in Alaska and Canada (Carrivick and Tweed, 2016; Harrison et al., 2018).

Permafrost: Warmer ground temperatures are expected to extend the geographical extent and depth of permafrost thaw across northern North America (very high confidence) (Section 9.5.2). Observations across Canada show that permafrost temperature is increasing and the active layer is getting thicker (Section 2.3.2.5; Derksen et al., 2018; Biskaborn et al., 2019; Romanovsky et al., 2020). Slater and Lawrence (2013) note that the RCP8.5 end-of-century period in North America only has shallow permafrost as the most probable condition in the Canadian Archipelago. Melvin et al. (2017) noted the loss of shallow permafrost in five RCP8.5 CMIP5 models across a wide swathe of southern Alaska by 2050, along with increases of active layer thickness. There is high confidence in continued reductions in mountain near-surface permafrost area with high spatial variability given local snow and temperature changes (Section 9.5.2; Peng et al., 2018; Hock et al., 2019).

Lake, river and sea ice: Anthropogenic warming reduces the seasonal extent of lake and river ice over many North American freshwater systems, with ice-free winter conditions pushing further north with rising temperatures (high confidence). Observations in Central and Eastern North America show reduced average seasonal lake-ice cover duration (Benson et al., 2012; Mason et al., 2016; US EPA, 2016). Satellite observations show declines in lake ice (Du et al., 2017) and loss of more than 20% of winter river-ice length in much of Alaska (2008–2018 compared to 1984–1994; Yang et al., 2020a). Spring lake and river ice in Canada is projected to break up 10–25 days earlier while autumn freeze-up occurs 5–15 days later by mid-century, with larger declines in lake-ice season closer to the coasts (Dibike et al., 2012) and for rivers in the Rocky Mountains and north-eastern USA (Yang et al., 2020a), although global models have difficulty with frozen freshwater system dynamics (Derksen et al., 2018). Substantial ice loss is projected over the Laurentian Great Lakes (Hewer and Gough, 2019; Matsumoto et al., 2019). The southern extent of lakes experiencing intermittent winter ice cover moves northward with rising temperature, pushing nearly out of the continental USA at low elevations under a 4.5°C GWL (Sharma et al., 2019). Higher spring flows and the potential for winter thaws are also projected to heighten the threat of ice jams (Rokaya et al., 2018; Bonsal et al., 2019) while reducing the seasonal viability of ice roads and recreational use (Pendakur, 2016; Mullan et al., 2017; Knoll et al., 2019).

Seasonal sea ice coverage along the majority of Canadian and Alaskan coastlines is declining (robust evidence, high agreement) and there is high confidence that sea ice loss continues under climate change, as further assessed in Section 12.4.9.

Heavy snowfall: There is low agreement (limited evidence) for observed changes in heavy snowfall in North America. Kluver and Leathers (2015) noted a 1930–2008 frequency increase for all snow intensities in the northern Great Plains but declines in heavier snow events in the Pacific Northwest and declines in the south-eastern USA. Changnon (2018) found that most extreme 30-day high-snowfall periods in the 1900–2016 record over the eastern USA occurred in the 1959–1987 period, which lies between the 1930s Dust Bowl and recent warming. There is low agreement and medium evidence for broad projected changes to heavy snowfall over North America given increased heavy precipitation and warmer winter temperatures. Several recent regional studies have projected that low-intensity events decrease more rapidly than heavy snowfall events, resulting in an increase in the snowfall proportion from heavy snowfall events even as the number of such events decreases (O’Gorman, 2014; Lute et al., 2015; Zarzycki, 2016; Janoski et al., 2018; Ashley et al., 2020).

Ice storm: There is limited evidence in the literature of unique changes in ice storms observed or projected over North America. Groisman et al. (2016) examined 40 years of observations and found weak decreases in freezing rain events over the south-eastern USA in the most recent decade. Ning and Bradley (2015) project that the average snow–rain transition line, which is associated with mixed precipitation, moves 2° latitude northward over Eastern North America by the end of the 21st century under RCP4.5 (4° under RCP8.5; see also Klima and Morgan, 2015).

Hail: There is limited evidence and low agreement for observed changes in the frequency or intensity of North American hail storms. J.T. Allen et al. (2015) and Allen (2018) found that temporal inconsistencies in the US and Canadian hail records made long-term climate analysis difficult, although B.H. Tang et al. (2019) identified an increasing frequency of environmental conditions conducive for large hail (diameter ≥ 5 cm) over the central and eastern USA. There is limited evidence and medium agreement in projections of increased hail damage potential over North America. Some regional and convective-permitting model projections indicate a longer hail season with fewer events and larger hail sizes that result in higher hail damage potential (Brimelow et al., 2017; Trapp et al., 2019).

Snow avalanche: There is limited evidence of directional changes in snow avalanches over North America. Mock and Birkeland (2000) identified a 1969–1995 decrease in snow avalanches over the western United States, although they note the heavy influence of natural variability. A similar decline was observed over western Canada (Bellaire et al., 2016; Sinickas et al., 2016), but clear trends are difficult to discern given sparse observations and shifts in avalanche management. We concur with the SROCC assessment of medium confidence and high agreement that snow avalanche hazards generally decrease at low elevations given lower snowpack, even as high elevations are increasingly susceptible to wet-snow avalanches (Hock et al., 2019; see also Lazar and Williams, 2008).

Observations and projections agree that snow and ice CIDs over North America are characterized by reduction in glaciers and the seasonality of snow and ice formation, loss of shallow permafrost, and shifts in the rain/snow transition line that alters the seasonal and geographic range of snow and ice conditions in the coming decades (very high confidence).

Relative sea level: Chapter 9 found that observations indicate increasing sea levels along most North American coasts (robust evidence, high agreement), although there is substantial regional variation in relative sea level rise (robust evidence, high agreement). Around North America, over 1900–2018, a new tide gauge-based reconstruction finds a regional mean RSL change of 1.08 [0.79 to 1.38] mm yr^–1 in the subpolar North Atlantic, 2.49 [1.89 to 3.06] mm yr^–1 in the subtropical North Atlantic, and 1.20 [0.76 to 1.62] in the East Pacific (Frederikse et al., 2020), compared to a GMSL change of around 1.7 mm yr^–1Section 2.3.3.3 and Table 9.5). For the period 1993–2018, these RSLR rates, based on satellite altimetry, increased to 2.17 [1.66 to 2.66] mm yr^–1, 4.04 [2.77 to 5.24] mm yr^–1 and 2.35 [0.70 to 4.06] mm yr^–1, respectively (Frederikse et al., 2020), compared to a GMSL change of 3.25 mm yr^–1Section 2.3.3.3 and Table 9.5). Relative sea level (RSL) is falling in portions of southern Alaska (Sweet et al., 2018) and much of the northern part of north-eastern Canada and around Hudson Bay (where land is rising by >10 mm/year; Greenan et al., 2018).

Relative sea level rise is virtually certain to continue in the oceans around North America, except in the northern part of north-eastern Canada and portions of southern Alaska. Regional mean RSLR projections for the oceans around North America range from 0.4–1.0 m under SSP1-2.6 to 0.7–1.4 m under SSP5-8.5 for 2081–2100 relative to 1995–2014 (median values), which means that there are locally large deviations from the projected GMSL change (Section 9.6.3.3), including decreases in RSL in northern north-eastern Canada from land uplift (see also Sweet et al., 2017; Greenan et al., 2018; Oppenheimer et al., 2019). The RSLR projections here may however be underestimated due to potential partial representation of land subsidence in their assessment (Section 9.6.3.2).

Coastal flood: Observations indicate that episodic coastal flooding is increasing along many coastlines in North America (robust evidence, high agreement), and this episodic coastal flooding will increase in many North American regions under future climate change (high confidence) although land uplift from glacial isostatic adjustment in northern and Hudson Bay portions of North-Eastern North America leads to only medium confidence of coastal flood increases in that region. Sweet et al. (2018) found 2000–2015 observed increases of about 125% in high-tide flooding frequencies along the southern Atlantic USA coastline, with 75% increases along the USA Gulf Coast and USA northern Atlantic coastlines. That same study noted that a GMSL of 0.5 m in 2100 would increase high tide (‘nuisance’) flooding from current rates of about once a month for most coastal regions to about once every other day along the USA Atlantic and Gulf coasts and smaller increases in frequency along the Pacific coast, and Dahl et al. (2017a) found similar trends on the USA East Coast prior to mid-century. The present day 1-in-50-year ETWL is projected to occur around three times per year by 2100 with an SLR of 1 m all around North America, except in most of Eastern North America where it is expected to have return periods of 1-in-1-year to 1-in-2-years (Vitousek et al., 2017). Ghanbari et al. (2019) projected corresponding shifts towards higher frequencies of major flooding events for 20 US cities. Figure 12.4r and Figure 12.SM.6 show increases of 70 cm or more in the 100-year return period extreme total water level (ETWL) over much of the USA East Coast, British Columbia, Alaska, and the Hudson Bay under RCP8.5 by 2100 (relative to 1980–2014), with lower increases in northern Mexico, northern Canada, Labrador, and the Pacific and Gulf coasts of the USA (Vousdoukas et al., 2018). Projected increases in coastal flooding generally follow patterns of RSL change, although sea ice loss in the north also increases open water storm surge (Greenan et al., 2018).

Coastal erosion: There is limited evidence of changes in North American episodic storm erosion caused by waves and storm surges. Observations show increased extreme wave energy on the Pacific coast, but no clear trend on other USA coasts given substantial natural variability (Bromirski et al., 2013; Vose et al., 2014). In terms of long-term coastal erosion, shoreline retreat rates of around 1 m yr^–1have been observed during 1984–2015 along the sandy coasts of NWN and NCA while portions of the US Gulf Coast have seen a retreat rate approaching 2.5 m yr^–1 (Luijendijk et al., 2018; Mentaschi et al., 2018). Sandy shorelines along ENA and WNA have remained more or less stable during 1984–2014, but a shoreline progradation rate of around 0.5 m yr^–1has been observed in NEN. Mentaschi et al. (2018) report 1984–2015 coastal area land losses of 630 km² and 1260 km² along the Pacific and Atlantic coasts of the USA, respectively.

Projections indicate that sandy coasts in most of the region will experience shoreline retreat through the 21st century (high confidence). Median shoreline change projections presented by Vousdoukas et al. (2020b) show that sandy shorelines in NWN, ENA, and NCA will retreat by between 40 and 80 m by mid-century (relative to 2010) under both RCP4.5 and RCP8.5. Projections for NEN and WNA are lower at 20–30 m under the same RCPs. The highest median mid-century projection in the region is for CNA at around 125 m under both RCPs. RCP4.5 projections for 2100 show shoreline retreats of 100 m or more along the sandy coasts of NWN, CNA, and NCA, while retreats of between 40 and 80 m are projected in other regions. Under RCP8.5, retreats exceeding 100 m are projected in all regions except NEN and WNA (approximately 80 m) by 2100, with particularly high retreats in NWN (160 m) and CNA (330 m). The total length of sandy coasts in North America that are projected to retreat by more than a median of 100 m by 2100 under RCP4.5 and RCP8.5 is about 15,000 km and 25,000 km respectively, an increase of approximately 70%.

Marine heatwave: There is high confidence in observed increases in marine heatwave (MHW) frequency and future increases in marine heatwaves are very likely around North America (Box 9.2). The total number of MHW days per decade increased in the North American coastal zone, albeit somewhat more in the Pacific (Oliver et al., 2018; Smale et al., 2019). Projected increases in degree heating weeks (Heron et al., 2016) and degree heating months (Frieler et al., 2013) indicate increasing bleaching-level and mortality-level heating stress threshold events for reefs in Florida and Mexico.

Mean SST is projected to increase by 1°C (3°C) around North America by 2100, with a hotspot of around 4°C (5°C) off the North American Atlantic coastline under RCP4.5 (RCP8.5) conditions (see Interactive Atlas). Frölicher et al. (2018) projected increasing MHW frequency and spatial extent at a 2°C global warming level with the largest increases in the Gulf of Mexico and off the southern USA East Coast (>20×) as well as off the coast of the Pacific Northwest (>15×). Projections for SSP1-2.6 and SSP5-8.5 both show an increase in MHWs around North America by 2081–2100, relative to 1985–2014 (Box 9.2, Figure 1).

There is high confidence that most coastal CIDs in North America will continue to increase in the future with climate change. An observed increase in relative sea level rise is virtually certain to continue in North America (other than around the Hudson Bay and southern Alaska) contributing to more frequent and severe coastal flooding in low-lying areas (high confidence) and shoreline retreat along most sandy coasts (high confidence). Marine heatwaves are also expected to increase all around the region over the 21st century (high confidence).

The assessed direction of change in climatic impact-drivers for North America and associated confidence levels are illustrated in Table 12.8.

Table 12.8 | Summary of confidence in direction of projected change in climatic impact-drivers in North America, representing their aggregate characteristic changes for mid-century for scenarios RCP4.5, SSP2-4.5, SRES A1B, or above within each AR6 region (defined in Chapter 1), approximately corresponding (for CIDs that are independent of sea level rise) to global warming levels between 2°C and 2.4°C (see Section 12.4 for more details of the assessment method). The table also includes the assessment of observed or projected time-of-emergence of the CID change signal from the natural interannual variability if found with at least medium confidence in Section 12.5.2.

Mean air temperature: Significant warming trends are clearly evident in the small islands, such as those in the Pacific, CAR, and western Indian Ocean, particularly over the latter half of the 20th century (see Figure Atlas.11; Atlas.10.2; Cross-Chapter Box Atlas.2, Table 1). This observed warming signal in the tropical western Pacific has been attributed to anthropogenic forcing (Wang et al., 2016). There is high confidence of warming over small islands even at 1.5°C GWL (Atlas.10.4 and Figure Atlas.28; Hoegh-Guldberg et al., 2018). Mean temperature is very likely to increase by 1°C–2°C (2°C–4°C) by 2041–2060 (2081–2100) under RCP8.5 (BOM and CSIRO, 2014) and SSP3-7.0 (Atlas.10.4, Figure 4.19 and Figure Atlas.12; Almazroui et al., 2021).

Extreme heat: Observational records indicate warming trends in the temperature extremes since the 1950s in CAR and the Pacific small islands (high confidence) (Sections 11.3.2 and 11.9, and Table 11.13). A detectable anthropogenic increase in summer heat stress has been identified over a number of island regions in CAR, western tropical Pacific, and tropical Indian Ocean, based on wet bulb globe temperature (WBGT) index trends for 1973–2012 (medium confidence) (Knutson and Ploshay, 2016). An increasing trend in the maximum daytime heat index is also noted in CAR during the 1980–2014 period, as well as more extreme heat events since 1991 (Ramirez-Beltran et al., 2017).

Compared with the recent past, it is likely that the intensity and frequency of hot (cold) temperature extremes will increase (decrease) in the small islands (Section 11.9 and Table 11.13; BOM and CSIRO, 2014). Warm spell conditions will occur up to half the year in CAR at 1.5°C GWL with an additional 70 days at 2°C (Hoegh-Guldberg et al., 2018; Taylor et al., 2018), with livestock temperature–humidity tolerance thresholds increasingly surpassed (Lallo et al., 2018). In CAR, a median increase of more than a month per year where temperatures exceed 35°C is projected by end of the 21st century under SSP5-8.5 (Figure 12.4a–c and Figure 12.SM.1). Heatwaves are projected to increase in CAR by the mid- and end-century under RCP8.5 (Sections 11.3.5 and 11.9, and Table 11.13). Figure 12.4d–f and Figure 12.SM.2 also show an increase of about 30–60 days in which HI exceeds 41°C by 2041–2060 under SSP5-8.5 relative to 1995–2014 in CAR, with an additional increase of about 50–100 days by end of the 21st century for RCP8.5/SSP5-8.5, but this increase remains below 50 days for RCP2.6/SSP1-2.6. The Pacific Islands region is also among those projected to have an increase in WBGT by end-century under RCP8.5, increasing the risk of heat stress in the region (Newth and Gunasekera, 2018).

It is very likely that the significant recent warming trends observed in the small islands will continue in the 21st century, which will likely further increase heat stress in these regions.

Mean precipitation: Observational datasets have generally revealed no significant long-term trends in rainfall in the Caribbean over the 20th century when analysed at seasonal and inter-decadal timescales, except for some areas where there is evidence for decreasing trends for the period 1901–2010 but not for the period 1951–2010 (Cross-Chapter Box Atlas.2, Table 1, and Atlas.10.2; Knutson and Zeng, 2018). Although there are spatial variations, annual rainfall trends in the western Indian Ocean are mostly decreasing, with generally non-significant trends in the western tropical Pacific since the 1950s (low confidence). Significant drying trends are noted in the southern Pacific subtropics and south-western French Polynesia during the 1951–2015 period (McGree et al., 2019), and in some areas of Hawaii during the 1920–2012 period (medium confidence) (Cross-Chapter Box Atlas.2, Table 1, and Atlas.10.2).

Atlas.10.4 projects precipitation reduction over the Caribbean (high confidence) (Almazroui et al., 2021) and parts of the Atlantic and Indian oceans, particularly in June to August, by end of 21st century under SSP5-8.5. Precipitation is generally projected to increase under SSP5-8.5 and for higher GWLs in the small islands in parts of the western and equatorial Pacific, but there is low confidence in broad changes given drier conditions projected for the southern subtropical and eastern Pacific Ocean (limited agreement given spatial and seasonal variability) (Atlas.10.4 and Figure Atlas.28).

River flood: There is limited evidence on observed changes in river flooding in the small islands. Long-term records in Hawaii indicate no clear trends in peak flow, except for the significant decrease in peak streamflow in Hawaii Island over the period 1967–2016 (Bassiouni and Oki, 2013; Clilverd et al., 2019). Similarly, there is no significant trend in the frequency and height (after adjusting for average sea level rise) of river flood in Fiji over the period 1892–2013 (McAneney et al., 2017). There is low confidence on the direction of future change of river flooding in the small islands due to the limited literature. In Oahu, Hawaii, extreme peak flow events with high return periods are projected to increase by end of the 21st century under RCP8.5, but there is also high uncertainty in these projections (Leta et al., 2018).

Heavy precipitation and pluvial flood: Heavy precipitation days in CAR have increased in magnitude, and have been more frequent in the northern part during the latter part of the 20th century (low confidence) (Section 11.4.2 and Table 11.14). The direction of change in extreme precipitation varies across the Pacific and depends on the season (low confidence) (Section 11.4.2 and Cross-Chapter Box Atlas.2, Table 1). Although pluvial flooding events have been observed in some islands, there is limited evidence for an assessment on past changes in pluvial flooding, unlike in other regions. There is low confidence in the projected change in magnitude of very heavy precipitation days in CAR across different GWLs (Table 11.14). On the other hand, there is high confidence in the increase in frequency and intensity of extreme rainfall events (i.e., 1-in-20-year rainfall events) in the western tropical Pacific in the 21st century, even for RCP2.6 scenario, based on model agreement and mechanistic understanding but low confidence in the magnitude of change in extreme rainfall due to model bias (BOM and CSIRO, 2014).

Landslide: Heavy rainfall, such as from tropical cyclones, can trigger landslides over steep terrain in the small islands (Bessette-Kirton et al., 2019). There is limited evidence to determine long-term trends in rainfall-induced landslides in the small islands (Kirschbaum et al., 2015; Sepúlveda and Petley, 2015; Froude and Petley, 2018; Bessette-Kirton et al., 2019). There is low confidence in future changes in landslides in the small islands. The direction of change may depend on future changes in precipitation, tropical cyclones, climate modes (e.g., El Niño–Southern Oscillation, ENSO), as well as human disturbance, but more data and understanding of the complexity of these relationships are needed, especially in these vulnerable areas (Sepúlveda and Petley, 2015; Gariano and Guzzetti, 2016; Froude and Petley, 2018).

Aridity: Current estimates identify many small islands as being under water stress and thus particularly sensitive to variations in rainfall and groundwater, population growth and demand, and land-use change, among others (Cross-Chapter Box Atlas.2; Holding et al., 2016). From 1950 to 2016, a heterogeneous but prevalent drying trend is found in CAR (low confidence), where drought variability is modulated by the tropical Pacific and North Atlantic oceans (Table 11.15 and Cross-Chapter Box Atlas.2, Table 1; Herrera and Ault, 2017). In the future, increased aridity and decreased freshwater availability are projected in many small islands due to higher evapotranspiration in a warmer climate that partially offsets increases or exacerbates reductions in precipitation (Karnauskas et al., 2016, 2018b; Hoegh-Guldberg et al., 2018). Increased aridity is projected for the majority of the small islands, such as in CAR, southern Pacific and western Indian Ocean, by 2041–2059 relative to 1981–1999 under RCP8.5 or at 1.5°C and 2°C GWLs, which will further intensify by 2081–2099 (medium confidence) (Karnauskas et al., 2016, 2018b). Groundwater recharge is projected to increase in Maui, Hawaii except on the leeward side of the island, which underscores the importance of topography and elevation on freshwater availability in different island microclimates (Brewington et al., 2019; Mair et al., 2019).

Hydrological drought: There is low confidence of widespread changes to hydrological drought in CAR or Pacific small islands in recent decades, although an increasing number of studies document local changes. Records in Hawaii indicate downward trends in low streamflow and base flow from 1913 to 2008 (Bassiouni and Oki, 2013). Decadal variability of Hawaiian streamflow coincides with rainfall fluctuations associated with the Pacific Decadal Variability although significant average declines in surface and baseflow runoff of about 8% and 11% per decade, respectively, have been noted during the 1987–2016 period (Clilverd et al., 2019).

There is low confidence in hydrological drought change projections, given low signal-to-noise ratios and the challenge in representing island scales in global analyses. Prudhomme et al. (2014) recognized CAR as one of the regions with the highest increase in regional deficit index (RDI; a measure of the fraction of area in hydrological drought conditions) by the end of the 21st century under RCP8.5. Daily streamflow and extreme low flows in two watersheds in Oahu, Hawaii are projected to decline by mid- and end of the 21st century under RCP4.5 and RCP8.5, which would result in more frequent hydrological droughts in this area (Leta et al., 2018).

Agricultural and ecological drought: Recent trends toward more frequent and severe droughts have been noted in the small islands but only with low confidence in broad trend patterns, given high spatial variability including heightened drought on the leeward side of islands (e.g., Frazier and Giambelluca, 2017; Herrera and Ault, 2017; McGree et al., 2019; see Table 11.15, Cross-Chapter Box Atlas.2, Table 1). Agricultural and ecological droughts are projected to increase in frequency, duration, magnitude, and extent in small islands, such as in CAR (medium confidence) and parts of the Pacific (low confidence), particularly where future declines in precipitation are compounded by higher evapotranspiration, under increasing levels of warming (Naumann et al., 2018; Taylor et al., 2018; Vichot-Llano et al., 2021). Relative to the period 1985–2014, decreases in annual surface and total column soil moisture become more robust in more areas in CAR by 2071–2100 under SSP3-7.0 and SSP5-8.5 scenarios (B.I. Cook et al., 2020), but reliably representing drought features in small island domains with global simulations is challenging (see also Section 11.9).

Fire weather: There is limited evidence on trends in wildfire in CAR and the Pacific. Records of wildfire in Hawaii from 2005 to 2011 indicate a peak in area burned during the hot and dry summer months, but Trauernicht et al. (2015) note the difficulty in establishing the link between past climate and wildfire trends due to human activities and vegetation changes. Availability of literature limits assessment on future fire weather in the small islands. Drying and warming trends tend to increase fire probability aside from the climate impact on fuel loading, for example, grassland fires in Hawaii (Trauernicht, 2019), and wildfires in Puerto Rico (Van Beusekom et al., 2018).

Observed and projected rainfall trends vary spatially across the small islands. Higher evapotranspiration under a warming climate are projected to partially offset future increases or amplify future reductions in rainfall, resulting in drier conditions and increased water stress in the small islands (medium confidence).

Mean wind speed: Scarcity of observations limits assessment of long-term changes in winds over the small islands in the Pacific and CAR. Records indicate that average daily wind speeds have slowly declined in Hawaii, but have remained constant across western and southern Pacific sites since the mid-20th century (Marra and Kruk, 2017). Recent studies of reanalyses and hindcast simulations indicate an intensification of the Pacific trade winds during the 1992–2011 period, which contributed to the ocean cooling in the tropical central and eastern Pacific (England et al., 2014; Takahashi and Watanabe, 2016). Projections estimate up to 0.4 m s^–1 (8%) increase in annual winds in CAR under RCP8.5, which is associated with changes in the extension of the North Atlantic Subtropical High that enhances the Caribbean low-level jet during the wet season, and stronger local easterlies due to enhanced land–ocean temperature differences in the dry season (Costoya et al., 2019) (low confidence).

Tropical cyclone: Tropical cyclones have devastating impacts on the small islands due to intense winds, storm surge and rainfall, although the associated rainfall can also be beneficial for freshwater resources. It is likely that tropical cyclone intensity and intensification rates at a global scale have increased in the past 40 years but it is not clear if regional-scale changes are basin-wide or due to shifts in tropical cyclone tracks (Section 11.7.1.2). Other, less data-sensitive tropical cyclone features, such as the poleward migration of where tropical cyclones reach peak intensity in the western North Pacific since the 1940s (medium confidence) and the slowdown in tropical cyclone translational speed over contiguous USA since 1900 (medium confidence), can affect rainfall and flooding over small islands in CAR and the Pacific (Section 11.7.1.2).

Projections of global changes in tropical cyclones indicate more frequent Category 4–5 storms (high confidence) and increased rain rates (high confidence) (Knutson et al., 2020), with relative sea level rise exacerbating storm surge potential, but with large regional differences (see Section 11.7.1.5). By the late 21st century, tropical cyclones are projected to be less frequent in the basins of the western and eastern North Pacific, Bay of Bengal, Caribbean Sea and in the Southern Hemisphere, but will be more frequent in the subtropical central Pacific (Murakami et al., 2014; Yoshida et al., 2017; Bell et al., 2019; Knutson et al., 2020). Over CAR, tropical cyclone intensity is expected to increase by the end of the century under RCP8.5 due to higher sea surface temperatures but can be inhibited by increases in vertical wind shear in the region (medium confidence) (Kossin, 2017; Ting et al., 2019). The poleward movement of the area in which tropical cyclones reach peak intensity in the western North Pacific is likely to continue, which affects the tropical cyclone frequency over the small islands in the area (Section 11.7.1.5; Kossin et al., 2016). Projections also indicate an increase (decrease) in the tropical cyclone frequency during El Niño (La Niña) events in the Pacific at the end of the 21st century (Chand et al., 2017). RCP8.5 2080–2099 projections indicate a 2% increase in the number of tropical cyclones in the north-central Pacific relative to 1980–1999, with tracks shifting northward towards Hawaii (N. Li et al., 2018). Given projected reductions to the overall number of tropical cyclones but increases in storm intensity, total rainfall and storm surge potential, we assess medium confidence of overall changes to tropical cyclones affecting the Caribbean and Pacific small islands.

Projections indicate that small islands will generally face fewer but more intense tropical cyclones (medium confidence) although there is substantial variability across small island regions given projected regional shifts in storm tracks.

Relative sea level: Relative sea level rise (RSLR) continues to be a major threat to small islands and atolls, since it can exacerbate the impacts of other climate hazards on low-lying coastal communities and infrastructures, ecosystems, and freshwater resources (Nurse et al., 2014; Hoegh-Guldberg et al., 2018). In the Indian Ocean–South Pacific region, a new tide gauge-based reconstruction finds a regional mean RSL change of 1.33 [0.80 to 1.86] mm yr^–1 over 1900–2018 (Frederikse et al., 2020) compared to a GMSL change of around 1.7 mm yr^–1Section 2.3.3.3 and Table 9.5). RSLR rates based on satellite altimetry for the period 1993–2018 in the region increased to 3.65 [3.23 to 4.08] mm yr^–1 (Frederikse et al., 2020), compared to a GMSL change of 3.25 mm yr^–1Section 2.3.3.3 and Table 9.5).

Relative sea-level rise is very likely to continue surrounding the oceans in the Small Island States. Around the small islands, regional mean RSLR projections vary widely, from 0.4–0.6 m under ­SSP1-2.6 to 0.7–1.6 m under SSP5-8.5 for 2081–2100 relative to 1995–2014 (median values), but in general they are situated in areas with RSL changes ranging from the mean projected GMSL change to above-average values (Section 9.6.3.3). These RSLR projections may however be underestimated due to potential partial representation of land subsidence in their assessment (Section 9.6.3.2).

Coastal flood: Relative sea level rise, storm surges and swells contribute to coastal inundation in the small islands, where studies on historical trends in coastal flooding are currently limited. For example, a swell event due to distant extratropical cyclones in December 2008 raised extreme water levels leading to flooding affecting five Pacific island nations: Marshall Islands, Micronesia, Papua New Guinea, Kiribati and Solomon Islands (Hoeke et al., 2013; Merrifield et al., 2014). Over low-lying atoll islands in the north-west tropical Pacific, potential increases in the frequency and areal extent of coastal flooding, especially at higher SLR scenarios, are expected to have negative consequences for freshwater resources and island habitability (Storlazzi et al., 2015, 2018). Select tide gauges across the Pacific also indicate increasing trends in the frequency of minor flooding since the 1960s (Marra and Kruk, 2017).

As relative sea levels increase, the potential for coastal flooding increases in the small islands (high confidence). Across the Pacific and CAR small islands, the 5–95th percentile range of the 1-in-100-year ETWL is projected to increase (relative to 1980–2014) by 10–35 cm and by 14–41 cm by 2050 under RCP4.5 and RCP8.5, respectively (Figure 12.4q). By 2100, this range is projected to be 27–81 cm and 44–188 cm under RCP4.5 and RCP8.5, respectively (Figure 12.4p,r; Vousdoukas et al., 2018; Kirezci et al., 2020). Furthermore, by 2050, the present-day 1-in-100-year ETWL is projected to have median return periods of between 1-in-1-year and 1-in-50-year in both the Pacific and CAR small islands, with some Pacific islands projected to experience the present-day 1-in-100-year ETWL more than once a year (Vousdoukas et al., 2018; Oppenheimer et al., 2019). By 2100, the present-day 1-in-50-year ETWL is projected to occur around three times a year by 2100 with an SLR of 1 m at Pacific and CAR small islands (Vitousek et al., 2017). In the western tropical Pacific, the magnitude and frequency of coastal flooding due to SLR can be modulated by changes in the wave climate (Shope et al., 2016).

Coastal erosion: Recent studies have indicated variable and dynamic changes in shorelines of reef islands (medium confidence), including both erosion and accretion, which suggest factors other than SLR affecting shoreline changes, such as in the central and western Pacific within the past 50-to-60-year timeframe (Webb and Kench, 2010; Le Cozannet et al., 2014; Ford and Kench, 2015; Duvat and Pillet, 2017). For example, islands on atolls in the central and western Pacific have not substantially eroded or reduced in size in the past decades while sea level has been rising, but their position and morphology have changed due to anthropogenic factors (e.g., seawalls, reclamation) and climate–ocean processes (Biribo and Woodroffe, 2013; McLean and Kench, 2015). Analysis of aerial and satellite imagery revealed severe shoreline retreat in six islands and the disappearance of five vegetated reef islands in Solomon Islands in the western Pacific between 1947 and 2014, which may be due to the interaction between SLR and waves (Albert et al., 2016). In French Polynesia, changes in shoreline and island area have been observed since the 1960s, partly due to the effect of TCs on sediment changes and human activities (Duvat and Pillet, 2017; Duvat et al., 2017). Coastal erosion has also been noted over the small, low-lying, sandy islands, such as in French Polynesia and Solomon Islands, among others (Luijendijk et al., 2018; Mentaschi et al., 2018). Average shoreline retreat rates between 1 and 2 m yr^–1 are estimated for the islands in the equatorial Pacific and in CAR, while a retreat rate of 0.5 m yr^–1is estimated for islands in the South Pacific, based on satellite observations from 1984–2016 (Luijendijk et al., 2018; Mentaschi et al., 2018). There was also a loss of 610 km² compared with a gain of 520 km² in coastal area in Oceania during the 1984–2015 period (Mentaschi et al., 2018).

Projections indicate that shoreline retreat will occur over most of the small islands in the Pacific and CAR throughout the 21st century with spatial variability (high confidence). Median shoreline change projections (CMIP5) relative to 2010, presented by Vousdoukas et al. (2020b), show that, by mid-century, shorelines in the islands in the equatorial Pacific and South Pacific will retreat by around 40 m, under both RCP4.5 and RCP8.5. In CAR islands, sandy shorelines are projected to retreat by about 80 m by mid-century under both RCPs. By 2100, more than 100 m of median shoreline retreat is projected for all small islands under both RCPs; notably in CAR where retreats approaching 200 m (relative to 2010) are projected under both RCPs. The total length of sandy coasts in CAR and Pacific small islands that is projected to retreat by more than a median of 100 m by 2100 under RCP4.5 and RCP8.5 is about 1100 km and 1200 km respectively, an increase of approximately 14%.

Marine heatwave: Ocean temperatures from satellite observations noted a moderate increase of 1–4 annual marine heat wave (MHW) events between 1982–1988 and 2000–2016 over some areas in the Indian Ocean, subtropical parts of the North and South Atlantic, and central and western parts of the North and South Pacific, but a decrease in frequency (two annual events) over the eastern Pacific Ocean (Box 9.2; Oliver et al., 2018). The intensity of MHWs has also increased between 0.2°C and 0.5°C over the equatorial portions of the North Atlantic and the South Pacific. Over the eastern tropical Pacific, the decrease in intensity and duration of MHW is between 0.5°C and 1.0°C and between 30 and 75 days, respectively (Box 9.2; Oliver et al., 2018). There is high confidence that MHWs will increase around all small island nations. Marine heatwaves are projected to be more intense and prolonged where the largest changes are noted in the equatorial region with maximum annual intensities up to 1.2°C (1.8°C) and annual mean duration reaching 100 days (200 days) at 1.5°C (2.0°C) warming levels (Frölicher et al., 2018). Projections for SSP1-2.6 and SSP5-8.5 both show an increase in MHWs around all small island nations by 2081–2100, relative to 1985–2014 (Box 9.2, Figure 1).

In summary, relative sea level rise is very likely in the oceans around small islands, and along with storm surges and waves will exacerbate coastal inundation in small islands. Shoreline retreat is projected along sandy coasts of most small islands (high confidence). There is high confidence that MHWs will increase around all small island nations.

The assessed direction of change in climatic impact-drivers for CAR and Pacific small islands and associated confidence levels are illustrated in Table 12.9. Cold, snow, and ice-related climatic impact-drivers, and sand and dust storms are not broadly relevant in the small islands that were assessed.

Table 12.9 | Summary of confidence in direction of projected change in climatic impact-drivers in the small islands, representing their aggregate characteristic changes for mid-century for scenarios RCP4.5, SSP2-4.5, SRES A1B, or above within each AR6 region (defined in Chapter 1), approximately corresponding (for CIDs that are independent of sea level rise) to global warming levels between 2°C and 2.4°C (see Section 12.4 for more details of the assessment method). The table also includes the assessment of observed or projected time-of-emergence of the CID change signal from the natural interannual variability if found with at least medium confidence in Section 12.5.2.

Mean air temperature: Atlas.11.2 shows high confidence in warming of the Arctic in observations and projections, measuring among the fastest-warming places at more than twice the global mean, with substantially higher temperature increases in the cold season (see also AMAP, 2017; Meredith et al., 2019). Atlas.11.1 assessed very likely warming in observations of West Antarctica from 1957 to 2016, but limited evidence of mean air temperature change across East Antarctica even as there is high confidence in future warming across the continent (Figure Atlas.29; Meredith et al., 2019).

Extreme heat, cold spell and frost: Ecosystem and societal temperature thresholds in polar regions often reflect lower tolerance to heat and higher tolerance to cold. Extreme heat events have increased around the Arctic and Iceland since 1979, including increases in cold season warm days and nights, melt days, and Arctic winter warm events (T > –10°C) as well as decreases in cold days and nights (Mernild et al., 2014; Matthes et al., 2015; Vikhamar-Schuler et al., 2016; Graham et al., 2017; Sui et al., 2017; Dobricic et al., 2020; Peña-Angulo et al., 2020). Heatwaves causing high temperature records have been recently documented in West and East Antarctica (Wille et al., 2019; Robinson et al., 2020). There is high confidence that polar amplification will drive increases in Arctic heat extremes as well as continuing declines in the magnitude and frequency of cold extremes (Matthes et al., 2015; Kharin et al., 2018), although dynamical effects will still bring substantial cold air anomalies over the Arctic (Wu and Francis, 2019). There is medium confidence for equivalent changes in extreme heat in Antarctica based primarily on higher mean temperatures, with J.R. Lee et al. (2017) projecting more than 50 additional degree days above freezing (2098 RCP8.5 compared with 2014) over parts of the Antarctic Peninsula but smaller changes over mainland Antarctica.

Mean precipitation: Atlas.11.2 indicated medium confidence in observed increases in Arctic precipitation, with the largest rises in the cold season. Antarctic precipitation showed no significant overall trend since the 1970s, with a positive trend over the 20th century (Sections 9.4.2.1 and Atlas.11.1). Increases in Arctic and Antarctic precipitation during the 21st century are very likely , with projected percentage increases that are much higher than most subpolar regions of the world (Figure Atlas.29).

Floods and heavy precipitation: Observations and model projections indicate high confidence in increasing Arctic river runoff in response to increasing total precipitation (Box et al., 2019; Durocher et al., 2019; Meredith et al., 2019) with a shift towards earlier meltwater flooding (AMAP, 2017). Higher Arctic precipitable water totals are also connected with observed increases in heavy precipitation and convective activity (high confidence) (Ye et al., 2015; Kharin et al., 2018; Chernokulsky et al., 2019). Higher flood magnitudes are also driven by future increases in rain-on-snow event days, amounts, and runoff, which are more significant in the Arctic than in mid-latitudes (where seasonal snow cover is often further reduced; AMAP, 2017; Jeong and Sushama, 2018b).

Landslide and snow avalanche: There is a growing number of studies on mass movements in polar regions. Although there is low confidence in widespread observational trends for landslides or snow avalanches, a rise in the number of future landslides is supported by strong links to increases in heavy precipitation, glacier retreat, and thawing of ice-rich permafrost that can lead to retrogressive thaw slumps in Arctic regions (Section 2.3.2.5; Kokelj et al., 2015; Derksen et al., 2018; Lewkowicz and Way, 2019; Patton et al., 2019; Ward Jones et al., 2019).

Aridity and drought: Recent decades have seen a general decrease in Arctic aridity, with projections indicating a continuing trend towards reduced aridity (high confidence) as increased moisture transport leads to higher precipitation, humidity and streamflow (Meredith et al., 2019) and a corresponding decrease in dry days (Khlebnikova et al., 2019a). There is low confidence overall of recent or projected drought changes in polar regions (Section 11.9) even as increasing evidence shows that drainage from permafrost thaw, higher potential evapotranspiration, and changing seasonal patterns of melt have caused lake reduction and soil moisture deficits in several areas that match with projections of future drought increase despite overall precipitation increases (Andresen and Lougheed, 2015; Bring et al., 2016; Spinoni et al., 2018a; Feng et al., 2019; Finger Higgens et al., 2019).

Fire weather: Fire season lengthened from 1979 to 2015 over Arctic portions of North America (Jain et al., 2017), corresponding also to a 1975–2015 increase in lightning-ignited fires in Arctic North-Western North America (Girardin et al., 2013; Veraverbeke et al., 2017). Abatzoglou et al. (2019) climate model simulations project significant fire weather index increases in boreal forests of Arctic Europe, Arctic Russia and Arctic North-Eastern North America (medium confidence). Trends towards more frequent fires in tundra regions are expected to continue, driven in particular by increasing potential evapotranspiration and changes in vegetation (high confidence) (Hu et al., 2015; AMAP, 2017; Young et al., 2017).

Mean wind speed and severe storm: There is medium confidence of mean wind decrease over the Russian Arctic, Greenland and Iceland, and Arctic North-Eastern North America (Karnauskas et al., 2018a; Jung and Schindler, 2019), but low confidence of changes in the other Arctic regions and Antarctica. Bintanja et al. (2014) projected that a strengthening of the Southern Annular Mode would decrease easterlies along Antarctica’s coasts with only small changes in katabatic winds (although this effect may diminish with stratospheric ozone recovery). In contrast, Gorter et al. (2014) regional climate model projections indicated a reduction in mean winds over the interior of Greenland by RCP4.5 2100 while coastal winds increase. Reanalysis data and climate models indicate few coherent regional trends of polar cyclone frequency or relationships with cyclone depth and size (Akperov et al., 2018, 2019; Day and Hodges, 2018; Zahn et al., 2018).

Snow: Atlas.11.1 identified likely increases in surface mass balance (driven by snowfall) across Antarctica in the 20th century (medium confidence). In the Arctic, overall snow extent and seasonal duration are projected to continue recent declines (high confidence), although mid-winter snowpack increases in some of the coldest and high-elevation locations given higher precipitation totals (medium confidence) (Section 9.5.3, Atlas.9 and Atlas.11.2; Bring et al., 2016; Danco et al., 2016; AMAP, 2017; Meredith et al., 2019). Higher temperatures result in a higher percentage of Arctic precipitation falling as rain (particularly in autumn and spring) (high confidence), with most land regions (outside of Greenland and Antarctica) becoming dominated by rainfall (more than 50% of total precipitation) by RCP8.5 2100 (Bintanja and Andry, 2017; Irannezhad et al., 2017).

Glacier and ice sheet: Section 9.5.1 and Section 2.3.2.3 found that glaciers have lost mass in all polar regions since 2000 (high confidence), and Section 9.4 assessed high confidence in Greenland Ice Sheet mass losses since 1980 and Antarctic Ice Sheet losses since 1992 (dominated by West Antarctica, with losses in parts of East Antarctica in the past two decades). New simulations from GlacierMIP (Marzeion et al., 2020) indicate glaciers in Iceland will lose 31 ± 35%, 41 ± 46% and 53 ± 45% of their mass in 2015 by the end of the century for RCP2.6, RCP4.5 and RCP8.5 scenarios, respectively. Marzeion et al. (2020) projected mass losses (high confidence) for those same scenarios in the Greenland Periphery: 22 ± 23%, 29 ± 26%, and 42 ± 28%; Svalbard: 35 ± 34%, 50 ± 36%, and 66 ± 35%; Russian Arctic: 26 ± 26%, 38 ± 28%, and 52 ± 30%; Northern Arctic Canada: 12 ± 13%, 18 ± 12%, and 27 ± 18%; Southern Arctic Canada: 23 ± 27%, 33 ± 29%, and 48 ± 32%; and Antarctic Periphery: 7 ± 12%, 13 ± 10%, and 16 ± 19%. Areas with receding glaciers are also potentially vulnerable to glacial lake outburst floods (Harrison et al., 2018).

Permafrost: Observations from recent decades (assessed in Section 9.5.2 and Section 2.3.2.5) show increases in permafrost temperature (very high confidence) and active layer thickness (medium confidence) across the Arctic (AMAP, 2017; Derksen et al., 2018; Markon et al., 2018; Biskaborn et al., 2019; Farquharson et al., 2019; Meredith et al., 2019; Romanovsky et al., 2020). Section 9.5.2 noted that observations of active layer thickness in Antarctica are too limited to assess long-term trends (see also Hrbáček et al., 2018; Biskaborn et al., 2019). Future projections indicate continuing increases in permafrost temperature and active layer thickness with loss of permafrost across the Arctic (Section 9.5.2). Streletskiy et al. (2019) noted that changes to Russian permafrost temperature and active layer thickness are most pronounced in areas where permafrost is continuous (underlying >90% of landmass). CMIP5 analyses by Slater and Lawrence (2013) projected that, by RCP8.5 2100, shallow (<3 m) permafrost would be most probable only in portions of the Canadian Arctic Archipelago and the Russian Arctic coastal and eastern upland regions.

Sea ice: Consistent with SROCC (Meredith et al., 2019), Section 9.3.1 and Section 2.3.2.1.1 assess very high confidence that Arctic sea ice thickness, extent, and average age have significantly decreased over the past four decades, with largest declines in September (when sea ice is at an annual minimum). Declines in landfast ice are most rapid in the Laptev Sea (Selyuzhenok et al., 2015), with warming also breaking perennial landfast ice blocking ocean channels (‘ice plugs’) in the Canadian Archipelago (Pope et al., 2017), and landfast ice declining in the cold season by 7% per decade across the Arctic (1976–2007; Yu et al., 2014). Observed trends and projections suggest that perennial sea ice is being replaced by thin, seasonal ice, although multi-year ice will persist above the Canadian Archipelago and drift into sea transportation lanes (Howell et al., 2016; Derksen et al., 2018). Trends from 1979 to 2013 show slightly earlier spring melt for Arctic sea ice, but substantially delayed autumn freeze-up and a melt season lengthened by more than 3 days per decade off northern Alaska and Canada with the exception of portions of the Bering Sea (Parkinson, 2014; Stroeve et al., 2014). Section 9.3.2 assessed low confidence in long-term trends in sea ice extent or thickness near Antarctica.

Future declines in Arctic sea ice are virtually certain, although there is low confidence in declines of Antarctic sea ice given dynamical processes in the Southern Ocean and the recovery of stratospheric ozone (Section 9.3; Meredith et al., 2019). Projections of an ‘ice-free’ Arctic vary, depending on definitions representing transportation needs, however Laliberté et al. (2016) noted that the median of 42 CMIP5 models projected <5% sea ice for the month of September by 2050, with equivalent conditions for the entirety of the August–October period by 2090. Section 9.3.1 assessed high confidence that practically ice-free conditions (<1 million km² in the September mean) would likely first appear before 2050 even under strong mitigation scenarios (Sigmond et al., 2018; Stroeve and Notz, 2018; Notz and SIMIP Community, 2020).

Lake and river ice: There is high confidence in observations of significant declines in seasonal ice cover thickness and duration over most Arctic lakes, with many lakes projected to lose more than one month of ice cover by mid-century (medium confidence) (Meredith et al., 2019; Sharma et al., 2019). Some lakes that previously froze to the bottom (‘bedfast’) now maintain liquid bottom water year round, and others shift from perennial to seasonal ice cover (Surdu et al., 2016; Engram et al., 2018). Yang et al. (2020a) identified a decline in Arctic cold-season river ice extent in satellite observations (particularly in Alaska) and projected reductions in average Northern Hemisphere seasonal river ice duration of 6.10 ± 0.08 days per degree global surface air temperature.

Heavy snowfall and ice storm: There is limited evidence of changes in heavy snowfall due to competing influences of shortened snowfall seasonality with more intense (and larger overall) precipitation in the Arctic. Episodic heavy snowfall trends in Antarctica are difficult to separate from large interannual variability (limited evidence) (Gorodetskaya et al., 2014, Turner et al., 2020). Limited evidence also hinders clear signals in ice storms, although warming shifts the freezing line (around which ice storms occur) poleward and upslope (Bintanja and Andry, 2017). Groisman et al. (2016) used 40 years of observations to identify an increase of freezing rain events in Norway, North America, and eastern and western Russia. Increases in winter rainfall have led to more frequent development of difficult wildlife and livestock grazing conditions as basal ice conditions coat the ground below snowpack (Peeters et al., 2019).

Table 12.11 | Summary of confidence in direction of projected change in climatic impact-drivers in the polar regions, representing their aggregate characteristic changes for mid-century for scenarios RCP4.5, SSP2-4.5, SRES A1B, or above within each AR6 region (defined in Chapter 1), approximately corresponding (for CIDs that are independent of sea level rise) to global warming levels between 2°C and 2.4°C (see Section 12.4 for more details of the assessment method). The table also includes the assessment of observed or projected time-of-emergence of the CID change signal from the natural interannual variability if found with at least medium confidence in Section 12.5.2. Note that the Arctic portions of the NEU, NEN and NWN differ from the full AR6 regions assessed in the Europe and North America sections above (see also Figure 1.18c).

Relative sea level: Satellite altimetry and tide data show that relative sea levels (with glacial isostatic adjustment) are rising in Arctic Europe and Arctic North-Western North America, declining in portions of southern Alaska and Arctic North-Eastern North America and no clear trend in Greenland and the Russian Arctic (Sweet et al., 2018; Rose et al., 2019), which is broadly consistent with findings in Oppenheimer et al. (2019). Areas with low or negative change have substantial land uplift counteracting the global mean sea level trend (Greenan et al., 2018; Sweet et al., 2018; Madsen et al., 2019). SROCC projections indicate high confidence in future rises in relative sea level for all Arctic regions other than areas of substantial land uplift in north-eastern Canada, the west coast of Greenland, and narrow portions of West Antarctica (Oppenheimer et al., 2019).

Coastal flooding and erosion: Higher sea levels and reduced coastal sea ice protection will increase future extreme sea levels in the Arctic (high confidence for Arctic Northern Europe, the Russian Arctic, and Arctic North-Western North America (medium confidence) for Greenland and Iceland and Arctic North-Eastern North America given glacial isostatic adjustment). Vousdoukas et al. (2018) project that the current 1-in-100-year extreme total water level would have median return periods of 1-in-20-years to 1-in-50-years by 2050, increasing to 1-in-5-years to 1-in-20-years by 2100 under RCP4.5 along nearly the entire Arctic coastline by 2100 (excluding GIC for which projections are not available). Projections for RCP8.5 indicate that the present-day 1-in-100-year ETWL would have median return periods of 1-in-10-years to 1-in-50-years by 2050 and would occur once every five years (or more frequently) by 2100. Arctic coastal erosion is also expected to increase with climate change (medium confidence; high agreement but limited evidence of projections), accelerated in some regions by subsurface permafrost thaw and increased wave energy (Gibbs and Richmond, 2015; Fritz et al., 2017; Oppenheimer et al., 2019; Casas-Prat and Wang, 2020). A longer ice-free season for the RCP8.5 2080s is projected to help drive more than 100 m of shoreline retreat in North-Western North America Arctic coastal communities (Melvin et al., 2017; Greenan et al., 2018; Magnan et al., 2019). Assessment of coastal flooding and erosion changes in Antarctica are limited by a lack of studies.

Marine heatwave: Recent years have seen marine heatwaves (MHWs) and increasing extreme coastal SSTs in Arctic systems (Lima and Wethey, 2012; Collins et al., 2019; Frölicher, 2019). Projections show increases in MHW intensity, frequency and duration will be larger over the Arctic Ocean than mid-latitude oceans due in part to low interannual variability under current sea ice (high confidence). Frölicher et al. (2018) used 12 CMIP5 models to project median MHW days increasing about 25-fold and 50-fold at the 2°C and 3.5°C GWLs, respectively, in response to mean ocean warming and sea ice loss, and the smallest global changes still leading to increases in the Southern Ocean around Antarctica (see also Cross-chapter Box 9.1).

Climate change has caused and will continue to induce an enhanced warming trend, increasing heat-related extremes and decreasing cold spells and frosts in the Arctic (high confidence), with similar changes in Antarctica but medium confidence for extreme heat increases and West Antarctic frost change decreases and low confidence for cold spell changes and East Antarctica frost. The water cycle is projected to intensify in polar regions, leading to more rainfall, higher river flood potential and more intense precipitation (high confidence). Projections indicate reductions in glaciers at both poles, with sea ice loss, enhanced permafrost warming, decreasing permafrost extent, and decreasing seasonal duration and extent of snow cover in the Arctic (high confidence) even as some of the coldest regions will see higher total snowfall given increased precipitation (medium confidence). Projections indicate relative sea level rises in polar regions (high confidence) , with the exception of regions with substantial land uplift including North-Eastern North America (high confidence), western Greenland, the northern Baltic Sea, and portions of West Antarctica. Higher sea levels also contribute to high confidence for projected increases of Arctic coastal flooding and higher coastal erosion (aided by sea ice loss) (medium confidence) with lower confidence for those CIDs in regions with substantial land uplift.

Hotspots of biodiversity are defined by the AR6 WGII as ‘geographic areas with exceptionally high richness of species, including rare (endemic) species’ (WGII Cross-Chapter Paper 1). The AR6 assessment is based on 238 distinctive regions often called the ‘Global 200 ecoregions’ (Olson and Dinerstein, 2002).

Mean temperature increase is a major climatic impact-driver for biodiversity hotspots, and it is very likely that it will affect all hotspot areas identified in the literature, at various rates in all climate scenarios, except those located in the North Atlantic where warming is uncertain (see Chapter 4). Terrestrial ecosystems will experience an enhanced warming compared to ocean ecosystems, because land temperatures are warming faster than ocean temperatures (Chapter 4). Marine ecoregions will experience ocean acidification and temperatures that increase faster in high latitudes (high confidence), but critical temperature and oxygen thresholds are projected to be crossed earlier (by mid-century RCP8.5) in tropical areas (Hughes et al., 2017a; Bruno et al., 2018). A warming trend is also expected for freshwater ecosystems, with different local magnitudes due to combined effects of groundwater system inertia as well as hydrology changes (Knouft and Ficklin, 2017). In tropical land areas, because interannual temperature variability is weak compared to mean changes, the temperature distribution range is likely to be shifted to a very different range in all projection scenarios, with unprecedented values relative to pre-industrial conditions. High climate velocities are particularly noteworthy for biodiversity hotspots given complex ecosystem dynamics and niche climates not easily replicated under shifted geographies (Burrows et al., 2014; Halpern et al., 2015; Dobrowski and Parks, 2016). In some regions (e.g., Central Africa, Amazon, South East Asia) the mean temperature change is already beyond the normal range of variations as it has reached levels higher than three (and up to six) times larger than the standard deviation of the interannual variations (Hawkins et al., 2020). Together with global warming, land and marine heatwaves are very likely to increase in the future climate in biodiversity hotspots (Sections 12.4.1–12.4.7).

There is low confidence in broad patterns of future drying or wet trends across the land and freshwater biodiversity hotspots in the humid tropics, although drying trends have been detected and predicted in parts of the Amazon (Fu et al., 2013; Boisier et al., 2015). There is medium confidence (limited evidence, high agreement) that in several regions the length of the dry season has already increased and is projected to further increase in some parts of the Mediterranean, Amazonia and sub-Saharan Africa (Debortoli et al., 2015; Dunning et al., 2018; Hochman et al., 2018; Saeed et al., 2018). Longer dry seasons also extend the seasonal length and geographical extent of fire weather in all future scenarios (medium confidence) (Jolly et al., 2015; Abatzoglou et al., 2019).

In conclusion, biodiversity hotspots around the world will each face unique challenges as climatic impact-drivers change. However, heat, drought and length of dry season, fire weather, sea surface temperature and deoxygenation are relevant drivers to terrestrial, freshwater and marine ecosystems, and have marked increasing trends.

Cities and settlements by the sea are exposed to specific climate and climate change patterns and to compound coastal hazard risks (AR6 WGII Cross-Chapter Paper 2). The AR5 WGII found that, in general, ‘urban climate change-related risks are increasing (including rising sea levels and storm surges, heat stress, extreme precipitation, inland and coastal flooding, landslides, drought, increased aridity, water scarcity, and air pollution)’ (Revi et al., 2014). Since AR5 a number of studies have been carried out to understand urban climate and its change. Box 10.3 identified a continuing strong role of the urban heat island in amplifying heat extremes in cities, although changes in the urban heat island are an order of magnitude smaller than projected localized warming trends (very high confidence).

Coastal cities’ proximity to the sea somewhat mitigates the effect of urban heat islands (high confidence) (Salvati et al., 2017; Santamouris et al., 2017; Y. Wang et al., 2018; Martinelli et al., 2020). Cities and settlements by the sea typically experience higher humidity levels than inland regions, combining with heat to enhance heat stress and induce exceedance of critical heat stress thresholds for outdoor activities, with potential enhanced exposure to heat for informal settlements (J. Wang et al., 2019). Such threshold exceedances are projected to increase for many coastal areas (high confidence), including the Persian Gulf where heat stress is projected to be extreme (Pal and Eltahir, 2016; Ahmadalipour and Moradkhani, 2018), and some low-lying areas in Europe such as the Po Valley and coastal Mediterranean areas (Coppola et al., 2021a; Schwingshackl et al., 2021; see also the heat stress index shown in Figure 12.4d–f).

Climate change-related variations in oceanic drivers (e.g., relative sea level, storm surge, ocean waves), combined with tropical cyclones, extreme precipitation and river flooding, are expected to lead to more frequent and more intense coastal flooding and erosion (very high confidence) impacting cities and settlements located especially in low-elevation coastal zones and mega-deltas (Chan et al., 2012, 2018; Karymbalis et al., 2012; Hemer et al., 2013; Aerts et al., 2014; B. Neumann et al., 2015; Hauer et al., 2016; Ranasinghe, 2016; Hinkel et al., 2018; Mavromatidi et al., 2018; Marcos et al., 2019; see also Sections 12.3, 12.4.1–12.4.7 and 12.4.9). Coastal erosion and flooding also pose challenges to critical infrastructure such as roads, subway tunnels, electricity and phone networks, wastewater management plants and buildings (Grahn and Nyberg, 2017; Pregnolato et al., 2017). Compound flooding due to simultaneous storm surges and high river flows have been found to be increasingly frequent in several cities and/or low-lying areas in Europe and the USA (high confidence) (Wahl et al., 2015; Bevacqua et al., 2019; Ganguli and Merz, 2019). Chapter 11 found that the frequency of such compound flood events is projected to increase (high confidence). In addition to changes induced by sea level change, many cities and settlements by the sea are in regions where tropical cyclones are projected to become more intense and severe tropical cyclones more frequent (high confidence) (Section 11.7).

The SROCC highlighted coastal settlements in the Arctic as being particularly exposed to several CID changes (Magnan et al., 2019). Enhanced waves, due to extended season of sea ice retreat, are projected to foster coastal flooding and erosion (Section 12.4.9; Gudmestad, 2018; Casas-Prat and Wang, 2020). Climate change is also affecting sea ice quality and season length along coasts of the Arctic Ocean where populations depend on sea ice for hunting or transportation (Section 12.4.9; Pearce et al., 2015).

In summary, coastal cities and settlements are particularly affected by a number of climatic impact-drivers that have already changed and will continue to change whatever the emissions scenario. These include increases in extreme heat, pluvial floods, coastal erosion and coastal flood (high confidence). Increasing relative sea level, compounding with increasing tropical cyclone storm surge and rainfall intensity, will increase the probability of coastal city flooding (high confidence). Arctic coastal settlements are particularly exposed to climate change due to sea ice retreat (high confidence).

Drylands, which include hyper-arid, arid, semi-arid and dry sub-humid areas (IPCC, 2019c), lie on all continents and cover 46% of the global land area and host more than one-third of the current population (Olsson et al., 2019). Huang et al. (2016b) found that aridity changes have helped expand dryland area by about 4% from 1948 to 2004, with the largest expansion of drylands occurring in semi-arid regions since the early 1960s. Section 4.5.1 assessed high confidence of a future poleward expansion of the Hadley cell, leading to a poleward shift of dryland areas in all scenarios considered. There is no evidence of a future global trend in aridification of drylands (IPCC, 2019a), but high confidence of aridification in some areas (e.g., Mediterranean, Central America, Southern Africa; IPCC, 2019a; see also Figure 12.4j–l). However, drivers of desertification largely include land-cover changes and land-use management, along with climate change (IPCC, 2019a).

Warming temperatures and extreme heat are major climatic impact-drivers with multiple potential impacts on societies, health, and habitability in semi-arid and arid regions that are already near physiological limits for outdoor activities. Semi-arid regions will very likely undergo a warming in all future scenarios (Chapter 4 and Atlas) and likely undergo an increase in duration, magnitude and frequency of heatwaves (Chapter 11) (Figure 12.4a–c). It is likely that heat stress will be much more intense by the end of the century in many areas under all scenarios, such as deserts and semi-arid zones in Asia (Murari et al., 2015; Mishra et al., 2017), Australia and Africa (Zhao et al., 2015; Xia et al., 2016; Guo et al., 2017; Dosio et al., 2018; Schwingshackl et al., 2021), with consequences for labour productivity with respect to high heat-humidity conditions (Figure 12.4d–f).

Drought is another major climatic impact-driver for semi-arid areas, imposing major challenges on agriculture given existing water availability constraints (Kusunose and Lybbert, 2014; Barlow et al., 2016; Otto et al., 2018). Over the period 1961–2013, the annual area of drylands in drought has increased, on average by slightly more than 1% per year, with large interannual variability (Olsson et al., 2019). In general, droughts have increased in several arid and semi-arid areas over the last decades (medium confidence), and are likely to increase in the future as indicated by a number of indices calculated from climate (Liu et al., 2018b; Zkhiri et al., 2019; Coppola et al., 2021b; Driouech et al., 2021; see also Figure 12.4j–l).

Deserts and semi-arid areas are prone to dust storms, which can drive impacts on health and several other sectors (X. Zhang et al., 2016; Tong et al., 2017). The SRCCL indicated that the evolution of dust under climate change is uncertain (Mirzabaev et al., 2019), and there is a lack of evidence and agreement of a change in their frequency or intensity so far in most regions (Sections 12.4.1–12.4.9). Model projections of future changes in dust are hindered by the uncertainties in future regional wind and precipitation as the climate warms (Evan et al., 2016); in the effect of CO₂ fertilization on source extent (Huang et al., 2017); and in the impact of human activities upon the land surface (Ginoux et al., 2012; see Chapter 10). Projected trends in dust storms and dust loads in deserts and semi-arid areas vary from region to region. Dust loadings are expected to decrease over most of the Sahara and Sahel (low confidence) (Section 12.4.1), increase over Mexico and the south-west USA (medium confidence) (Section 12.4.6), and there is low confidence of a future trend due to climate change in other continents (Sections 12.4.2–12.4.5).

In conclusion, desert and semi-arid areas are strongly affected by climatic impact-drivers such as extreme heat, drought and dust storms. Heat hazards are very likely increasing in all future climate scenarios, but uncertainty remains regarding any broadly consistent future changes in other climatic impact-drivers for deserts and semi-arid regions.

Mountains cover about 30% of the land areas on Earth (not counting Antarctica) and deliver a number of vital services to humanity (WGII Cross-Chapter Paper 5; IPCC, 2019b). Climate change in high mountains was addressed in SROCC, which emphasized changes in several climatic impact-drivers. These included an observed general decline in low-elevation snow cover, glaciers and permafrost (high confidence), which induced changes in natural hazards such as decrease in slope stability (high confidence), changes to the frequency of glacial lake outbursts (limited evidence), and climate effects on other climatic impact-drivers (avalanche, rain-on-snow floods) with various degrees of confidence (Hock et al., 2019).

There is a growing body of literature indicating elevation-dependent warming (EDW; different rates of warming by altitude although not necessarily increasing with altitude) in several mountain regions but not globally (Hock et al., 2019; Pepin et al., 2019; Ahmed et al., 2020; B. Li et al., 2020; Williamson et al., 2020; You et al., 2020; Micu et al., 2021). Statistically significant elevational enhancement to long-term trends in maximum near-surface air temperatures and diurnal temperature range were observed in southern central Himalaya and in the Swiss Alps (Rottler et al., 2019; Thakuri et al., 2019). Aguilar-Lome et al. (2019) reported that winter daytime land surface temperatures in the Andean region between 7°S and 20°S show the strongest trends at higher elevations: +1.7°C per decade above 5000 m above sea level. Palazzi et al. (2019) identified changes in albedo and downward thermal radiation as key drivers of EDW according to the simulation outputs of a high-spatial-resolution model in three important mountainous areas: the Colorado Rocky Mountains, the Greater Alpine Region and the Himalayas–Tibetan Plateau, but mechanisms for EDW remain complex (Hock et al., 2019). Warming is also affecting mountain lake surface temperatures, increasing probabilities of ice-free winters and the frequency and duration of ‘lake heatwaves’ (high confidence) (O’Reilly et al., 2015; Woolway et al., 2020, 2021) with a high variability from lake to lake.

Elevation-dependent warming could speed up the observed, rapid upward shifts of the freezing level height (FLH) in several mountainous regions of the world and lead to faster changes in the snowline, the glacier equilibrium-line altitude and the snow/rain transition height (high confidence). In the Indus, Ganges and Brahmaputra basins in Asia, the FLH is projected to rise at a rate of 4.4 to 10.0 m yr^–1under RCP8.5 (Viste and Sorteberg, 2015). In the Argentinian Andes, FLH is projected under RCP8.5 to move up more than twice as much by 2070 as during the entire Holocene under the worst case scenario (Drewes et al., 2018). On the western slope of the subtropical Andes (30°S–38°S) in central Chile, the mean value of the free tropospheric height of the 0°C isotherm under wet conditions is projected to be close to or higher than the upper quartile of the distribution in the current climate, towards the end of the century and under RCP8.5 (Mardones and Garreaud, 2020). In the Alps and the Pyrenees, Spandre et al. (2019) projected a rise in the natural snow elevation of between 200–300 m and 400–600 m by mid-century under RCP2.6 and RCP8.5, respectively. In the same region, the environmental equilibrium-line altitude is projected to exceed the maximum elevation of 69%, 81% and 92% of the glaciers by the end of the century under RCPs 2.6, 4.5 and 8.5, respectively (Žebre et al., 2021).

Orographic effects enhance convection and stratiform heavy precipitation (due to uplift) and make mountains prone to extreme precipitation events. These events are projected to increase in major mountainous regions (Alps, parts of the Andes, British Columbia, North-Western North America, Calabria, Carpathian, Hindu-Kush-Himalaya, Rocky Mountains, Umbria; medium to high confidence depending on location), with potential cascading consequences of floods, landslides and lake outbursts in mountainous areas in all scenarios (medium confidence) (Chapter 11 and Sections 12.4.1–12.4.9; Geertsema et al., 2006; Gaire et al., 2015; Kim et al., 2015; Ciabatta et al., 2016; Gariano and Guzzetti, 2016; Kharuk et al., 2016; Syed and Al Amin, 2016; Cloutier et al., 2017; Gądek et al., 2017; Jurchescu et al., 2017; Rajczak and Schär, 2017; Alvioli et al., 2018; Coe et al., 2018; Schlögl and Matulla, 2018; C.-W. Chen et al., 2019; Handwerger et al., 2019; Hock et al., 2019; Patton et al., 2019; Vaidya et al., 2019; Kirschbaum et al., 2020; Coppola et al., 2021b).

Declines in low-elevation snow depth and seasonal extent are projected for all SSP-RCPs (see Sections 12.4.1–12.4.6), along with reductions in mountain glacier surface area, increases in permafrost temperature, decreases in permafrost thickness, changes in lake and river ice, changes in the amount and seasonality of streamflows and hydrologic droughts in snow-dominated and glacier-fed river basins (e.g., in Central Asia; Sorg et al., 2014; Reyer et al., 2017b) (medium confidence), and decreases in the stability of mountain slopes and snowfields. Glacier recession could lead to the creation of new glacial lakes in places like the Himalaya-Karakoram region (Linsbauer et al., 2016) and in Alaska and Canada (Carrivick and Tweed, 2016; Harrison et al., 2018) (medium confidence). With increasing temperature and precipitation these can increase the occurrence of glacier lake outburst floods and landslides over moraine-dammed lakes (high confidence) (Carey et al., 2012; Rojas et al., 2014; Iribarren Anacona et al., 2015; Cook et al., 2016; Haeberli et al., 2017; Kapitsa et al., 2017; Narama et al., 2018; Wilson et al., 2018; Drenkhan et al., 2019; S. Wang et al., 2020).

In conclusion, mountains face complex challenges from specific climatic impact-drivers drastically influenced by climate change: regional elevation-dependent warming (high confidence), low-to-mid-altitude snow cover and sno w-sea son decrease even as some high elevations see more snow (high confidence), glacier mass reduction and permafrost thawing (high confidence), and increases in extreme precipitation and floods in most parts of major mountain ranges (medium confidence).

Tropical forests, which are among the world’s most biologically diverse ecosystems, are essentially located in Central and South America, Africa and South East Asia (AR6 WGII Cross-Chapter Paper 7). The AR5 and SR1.5 indicated several specific climatic impact-driver changes that are particularly important to tropical forests: mean temperature increase, long-term drying trends (including shifts in the length of the dry season), prolonged drought, wildfires and surface CO₂ increase for inland forests (IPCC, 2013, 2018). The SRCCL assessed an enhanced risk and severity of wildfires in tropical rainforests (high confidence), but fires are not only a natural process but are also affected by deforestation and other human influences (IPCC, 2019a).

Temperature is rising in all tropical regions covered with forests and will very likely continue to rise, reaching levels unprecendented in recent decades as the temperature trends rapidly emerge from weak historical interannual variability (Sections 12.4.1–12.4.4 and 12.5.2; see also Chapter 4 and Atlas).

Regional patterns of increasing drought or unusual wet and dry periods are predicted with agreement over many climate models such as over the Amazon basin (Boisier et al., 2015; Duffy et al., 2015; Zulkafli et al., 2016; Coppola et al., 2021b). There is medium confidence (limited evidence, high agreement) that in several tropical-forest regions (e.g., Amazonia, West Africa) the dry season length has increased (Fu et al., 2013; Debortoli et al., 2015; Saeed et al., 2017; Dunning et al., 2018; Wadsworth et al., 2019), and there is low confidence (limited evidence) that deforestation influences the shift in the onset of the wet season in south Amazonia (Leite-Filho et al., 2019). In contrast, the wet season is increasing in northern Australia tropical forests (Catto et al., 2012).

Tropical forests typically reach peak fire weather conditions in the dry season (Taufik et al., 2017), in particular during long-lived droughts (Brando et al., 2014; Marengo et al., 2018), with consequences for tree mortality, forest and carbon sink loss (Brando et al., 2019), and on the hydrological cycle in South America (Martinez and Dominguez, 2014; Espinoza et al., 2020). Observations and reanalyses over the past three to four decades, combined into fire risk indices, show that the fire weather season length has been increasing by about 20% globally (Jolly et al., 2015), and this index exhibits particularly high trend values over tropical forest areas of South and Central America and Africa. There is generally low confidence in future projections of general fire weather risk evolution in tropical forests and evolutions depend on the region (Abatzoglou et al., 2019). Over the Amazon basin the fire risk increase is projected to emerge well before 2050 while for other equatorial forests no significant evolution is found. In Savanna areas the risk increase is found to be more general.

In conclusion, most tropical forests are challenged by a mix of emerging warming trends that are particularly large in comparison to historical variability (medium confidence). Water cycle changes bring prolonged drought, longer dry seasons, and increased fire weather to many tropical forests, with plants also responding to CO₂ increases (medium confidence).