Water is vital to all life on Earth. Seventy-one percent of the Earth is covered by water, with saline ocean water accounting for around 97% of total water availability (Figure 8.1). Terrestrial freshwater represents less than 2% of all water on Earth, and the remainder (around 1–2%) is primarily made up of saline groundwater and saline lakes (Durack, 2015; Abbott et al., 2019). Ice sheets, glaciers and snow pack account for approximately 96% of all freshwater, with less than 4% of freshwater considered easily accessible and available for essential ecosystem functioning and human society’s water resource needs (Durack, 2015; Abbott et al., 2019). This very small fraction of freshwater represents a total volume of about 835,000 km³, mostly contained in groundwater (630,000 km³), the remaining 205,000 km³being stored in lakes, rivers, wetlands and soils (Abbott et al., 2019). Although the natural cycling rate of this amount is theoretically enough to meet global human and ecosystem needs, there are large geographical and seasonal differences that influence the availability of freshwater to meet regional demands.

Freshwater is the most essential natural resource on the planet (Mekonnen and Hoekstra, 2016; Djehdian et al., 2019) and underpins almost all Sustainable Development Goals (SDGs), which require access to adequate and safe resources for drinking and sanitation (SDG 6) and many other purposes. Freshwater supports a range of human activities from irrigation to industrial processes including the generation of hydro-electricity and the cooling of thermoelectric power plants (Bates et al., 2008; Schewe et al., 2014). These activities require sufficient quantities of freshwater that can be drawn from rivers, lakes, groundwater stores, and in some cases, desalinated sea water (Schewe et al., 2014). Recent estimates of global water pools and fluxes suggest that half of global river discharge is redistributed each year by human water use (Abbott et al., 2019). This emphasizes the need to consider both anthropogenic climate change and direct human influences, such as population increase or migration, economic development, urbanization, and land use change, when planning water-related mitigation or adaptation strategies (Jiménez Cisneros et al., 2014).

Water scarcity occurs when there are insufficient freshwater resources to meet water demands, although water problems may also arise from water quality issues or from economic and institutional barriers (AR6 WGII Chapter 4). This affects the preservation of environmental flows that ultimately influence ecosystem functioning and services (Schewe et al., 2014; Mekonnen and Hoekstra, 2016; Djehdian et al., 2019). As such, water availability is a major constraint on human society’s ability to meet the future food and energy needs of a growing population (D’Odorico et al., 2018). Water plays a key role in the production of energy, including hydro-electricity, bioenergy, and the extraction of unconventional fossil fuels (Schewe et al., 2014; D’Odorico et al., 2018; Djehdian et al., 2019). These dependencies have resulted in increasing competition for water between the food and energy sectors. Pressures on this ‘food-energy-water nexus’ are further compounded by increasing globalization, which can transfer large-scale water demands to other regions of the world, raising serious concerns about local food and water security in regions that are highly dependent on agricultural exports or imports (D’Odorico et al., 2018).

The consequences of climate change on terrestrial ecosystems and human societies are primarily experienced through changes to the global water cycle (JiménezCisneros et al., 2014). Changes in the quantity and seasonality of water due to climate change have long been recognized by IPCC and global development agencies as heavily influencing the food security and economic prosperity of many countries, particularly in the arid and semi-arid areas of the world including Asia, Africa, Australia, Latin America, the Mediterranean, and small island developing states (Bates et al., 2008; Schewe et al., 2014; Mekonnen and Hoekstra, 2016). Having too much or too little water increases the likelihood of flooding and drought, as precipitation variability increases in a warming climate (Stockeret al., 2013; Hoegh-Guldberg et al., 2018). Climate change poses a threat to both regional water availability and global water security. Changes in precipitation and glacier runoff and snowmelt influence other hydroclimate variables like surface and subsurface runoff, and groundwater recharge, which are critical to the water, food and energy security of many regions (Oki andKanae, 2006; Jiménez Cisneros et al., 2014; Schewe et al., 2014; Mekonnen and Hoekstra, 2016).

Currently, around four billion people live under conditions of severe freshwater scarcity for at least one month of the year, with half a billion people in the world facing severe water scarcity all year round (Mekonnen and Hoekstra, 2016). The AR5 WGII reported that approximately 80% of the world’s population already suffers from high levels of threat to water security (Jiménez Cisneros et al., 2014). Given the vulnerability of the planet’s freshwater resources and the role of climate change in intensifying adverse impacts on human societies and ecosystems (Hoegh-Guldberg et al., 2018; IPCC, 2018), this chapter evaluates advances in the theoretical, observational and model based understanding of the global water cycle made since AR5 (IPCC, 2013) and AR6 Special Reports.

As shown in Figure 8.1, the global water cycle is the continuous, naturally occurring movement of water through the climate system from its liquid, solid and gaseous forms among reservoirs of the ocean, atmosphere, cryosphere and land (Stocker et al., 2013). In the atmosphere, water primarily occurs as a gas (water vapour), but it is also present as ice and liquid water within clouds where it substantially affects Earth’s energy balance (Sections 7.4.2.2 and 7.4.2.4). The water cycle primarily involves the evaporation¹ and precipitation of moisture at the Earth’s surface including transpiration associated with biological processes. Water that falls on land as precipitation, supplying soil moisture, groundwater recharge, and river flows, was once evaporated from the ocean or sublimated from ice-covered regions before being transported through the atmosphere as water vapour, or in some areas was generated over land through evapotranspiration (Gimeno et al., 2010; van der Ent and Savenije, 2013). In addition, the net flux of atmospheric and continental freshwater is a key driver of sea surface salinity, which in turn influences the density and circulation of the ocean (Chapter 9).

Figure 8.1 | Depiction of the present-day water cycle based on previous assessments (Trenberth et al., 2011; Rodell et al., 2015; Abbott et al., 2019) with adjustments for groundwater flows (Zhou et al., 2019c; Luijendijk et al., 2020), seasonal snow (Pulliainen et al., 2020) and ocean precipitation and evaporation (Stephens et al., 2012; Allan et al., 2020; Gutenstein et al., 2021). The net loss of frozen and liquid water from land to ocean is estimated from Chapter 9, Table 9.5. In the atmosphere, which accounts for only 0.001% of all water on Earth, water primarily occurs as a gas (water vapour), but it is also present as ice and liquid water within clouds. The ocean is the primary water reservoir on Earth: it comprises mostly liquid water across much of the globe but also includes areas covered by ice in polar regions. Liquid freshwater on land forms surface water (lakes, rivers) and, together with soil moisture and mostly unusable groundwater stores, accounts for less than 2% of global water (Stocker et al., 2013). Solid terrestrial water that occurs as ice sheets, glaciers, snow and ice on the surface, and permafrost currently represents nearly 2% of the planet’s water (Stocker et al., 2013). Water that falls as snow in winter provides soil moisture and streamflow after melting, which are essential for human activities and ecosystem functioning. Note that these best estimates do not lead to a perfectly closed global water budget and that this budget has no reason to be closed given the ongoing human influence through both climate change (e.g., melting of ice sheets and glaciers, see Chapter 9) and water use (e.g., groundwater depletion through pumping into fossil aquifers, see Figure 8.10).

Open figure

Understanding the interactions between the water and energy cycles is one of the four core projects of the World Climate Research Programme (WCRP). Latent heat fluxes, released by condensation of atmospheric water vapour and absorbed by evaporative processes, are critical to driving the circulation of the atmosphere on scales ranging from individual thunderstorm cells to the global circulation of the atmosphere (Stocker et al., 2013; Miralles et al., 2019). Water vapour is the most important gaseous absorber in the Earth’s atmosphere, playing a key role in the Earth’s radiative budget (Schneider et al., 2010). As atmospheric water vapour content increases with temperature, it has a considerable influence on climate change (Section 7.4.2.2). Additionally, a small fraction of the atmospheric water content is liquid or solid and has a major effect on both solar and longwave radiative fluxes, from the Earth’s surface to the top of the atmosphere. The cloud response to anthropogenic radiative forcings, both in the tropics and in the extratropics (Zelinka et al., 2020), is therefore also crucial for understanding climate change (Section 7.4.2.4).

The terrestrial water and carbon cycles are also strongly coupled (Cross-Chapter Box 5.1). As atmospheric carbon dioxide (CO₂) concentration increases, the physical environment in which plants grow is altered, including the availability of soil moisture necessary for plants’ CO₂ uptake and, potentially, the effectiveness of CO₂ removal techniques to mitigate climate change (Section 5.6.2.1.2). Rising surface CO₂ concentrations also modify stomatal (small pores at the leaf surface) regulation as well as the plants’ biomass, thus affecting ecosystem photosynthesis and transpiration rates and leading generally to a net increase in water use efficiency (Lemordant et al., 2018). These coupled changes have profound implications for the simulation of the carbon and water cycles (Gentine et al. , 2019; see also Section 5.4.1), which can be better assessed with the new generation Earth system models, although both the carbon concentration and carbon-climate feedbacks remain highly uncertain over land Section 5.4.5; Arora et al., 2020). The water constraints on the terrestrial carbon sinks are a matter of debate regarding the feasibility or efficiency of some land-based CO₂ removal and sequestration techniques requested to comply with the Paris Agreement (Section5.6.2.2.1; Fuss et al. , 2018; Belyazid and Giuliana, 2019).

Based on long-term observational evidence (Hartmann et al., 2013), AR5 concluded it was likely that anthropogenic influence has affected the water cycle since the 1960s (IPCC, 2018). Detectable human influ­ence on changes to the water cycle were found in atmospheric moisture content (medium confidence), global-scale changes of precipitation over land (medium confidence), intensification of heavy precipitation events over land regions where sufficient data networks exist (medium confidence), and very likely changes to ocean salinity through its connection with evaporation minus precipitation change patterns (Sections 2.5, 2.6, 3.3, 7.6, 10.3 and 10.4; Stocker et al., 2013). The AR5 also reported that it is very likely that global surface air specific humidity increased since the 1970s. There was low confidence in the observations of global-scale cloud variability and trends, medium confidence in reductions of pan-evaporation, and medium confidence in the non-monotonic changes of global evapotranspiration since the 1980s. In terms of streamflow and runoff, AR5 identified that there is low confidence in the observed increasing trends of global river discharge during the 20th century. Similarly, AR5 concluded that there is low confidence in any global-scale observed trend in drought or dryness (lack of rainfall) since the mid-20th century. Yet, the frequency and intensity of drought likely increased in the Mediterranean and West Africa, while theylikely decreased in central North America and north-western Australia since 1950.

Water cycle projections in AR5 (Collins et al., 2013) were considered primarily in terms of water vapour, precipitation, surface evaporation, runoff, and snowpack. Globally-averaged precipitation was projected to increase with global warming with virtual certainty (Chapter 12 Executive Summary and Section 12.4.1.1). Regionally, precipitation in some areas of the tropics and polar regions could increase by more than 50% by the end of the 21st century under the RCP8.5 emissions scenario, while precipitation in large areas of the subtropics could decrease by 30% or more (AR5 FAQ 12.2, Figure 12.22). Overall, the contrast of annual mean precipitation between dry and wet regions and between dry and wet seasons (‘wet get wetter, dry get drier’) was projected to increase over most of the globe with high confidence (Chapter 12 Executive Summary and Section 12.4.5.2). Globally, the frequency of intense precipitation events was projected to increase while the frequency of all precipitation events was projected to decrease, leading to the contradictory-seeming projection of a simultaneous increase in both droughts and floods (12.2 and Section 12.4.5.5 in AR5 WGI). Surface evaporation change was projected to be positive over most of the ocean and to generally follow the pattern of precipitation change over land Chapter 12 Executive Summary, and Section 12.4.5.4). Near-surface relative humidity reductions over many land areas were projected to be likely , with medium confidence (Section 12.4.5.1). General decreases in soil moisture in present-day dry regions were considered likely , and projected with medium confidence under the RCP8.5 scenario (Section 12.4.5.3). Soil moisture drying in the Mediterranean, south-west USA and southern African regions was considered likely , with high confidence by the end of this century under the RCP8.5 scenario (Section 12.4.5.3). Projections for annual runoff included both decreases and increases. Decreases in Northern Hemisphere snow cover were assessed as very likely with continued global warming (Section 12.4.6.2). As temperatures increase, snow accumulation was projected to begin later in the year and melting to start earlier, with related changes in snowmelt-driven river flows (FAQ 12.2 and Section 12.4.6.2 in AR5 WGI). In terms of the potential for abrupt change in components of the water cycle, long-term droughts and monsoonal circulation were identified as potentially undergoing rapid changes, but the assessment was reported with low confidence (Sections 12.5.5.8.1 and 12.5.5.8.2, and Table 12.4).

The SR1.5 assessed the impacts of global warming of 1.5°C above pre-industrial levels. The dominant human influence on observed global warming and related water cycle changes was confirmed. Further evidence that anthropogenic global warming has caused an increase in the frequency, intensity and/or amount of heavy precipitation events at the global scale (medium confidence), as well as in drought occurrence in the Mediterranean region (medium confidence) was also reported. Chapter 3 of SR1.5 (Hoegh-Guldberg et al., 2018) highlights that each half degree of additional global warming influences the climate response. Heavy precipitation shows a global tendency to increase more at 2°C compared to 1.5°C, though there is low confidence in projected regional differences in heavy precipitation at 1.5°C compared to 2°C global warming, except at high latitudes or at high altitude where there is medium confidence. A key finding is that ‘limiting global warming to 1.5°C compared to 2°C would approximately halve the proportion of the world population expected to suffer water scarcity, although there is considerable variability between regions (medium confidence)’ (SR1.5). This is consistent with greater adverse impacts found at 2°C compared to 1.5°C for a number of dryness or drought indices (Schleussner et al., 2016; Lehner et al., 2017; Greve et al., 2018). There is also medium confidence that land areas with increased runoff and exposure to flood hazards will increase more at 2°C compared to 1.5°C of global warming.

The Special Report on the Ocean and Cryosphere in a Changing Climate (SROCC) provides a comprehensive assessment of recent and projected changes, specifically in snow and ice-covered areas that form a key component of the water cycle in high-elevation and high-latitude areas. High mountain regions have experienced significant warming since the early 20th century, resulting in reduced snowpack on average (Marty et al., 2017), with glaciers retreating globally since the mid-20th century (Marzeion et al., 2018; Zemp et al., 2019). Glacier shrinkage and snow cover changes have led to changes (both increases and decreases) in streamflow in many mountain regions in recent decades (Milner et al., 2017). Permafrost regions have undergone degradation and ground-ice loss due to recent warming (Lu et al., 2017). Glacier mass loss is projected to continue through the 21st century under all scenarios. In high mountain areas, low-elevation snow cover is also projected to decrease, regardless of emissions scenario. Widespread permafrost thaw is projected to continue through this century and beyond. River runoff in snow- or glacier-fed basins is projected to increase in winter and to decrease in summer (and in the annual mean) by 2100. In the oceans, the Atlantic Meridional Overturning Circulation (AMOC) will very likely weaken over the 21st century under all emissions scenarios (SROCC), with potential effects on atmospheric circulation and the water cycle at the regional scale (see also Section 8.6).

The Special Report on climate change, desertification, land degradation, sustainable management, food security, and greenhouse gas (GHG) fluxes in terrestrial ecosystems (SRCCL) has clear connections with the water cycle. This Report indicates that since 1850–1900, land surface temperature has risen nearly twice as much as global surface temperature (high confidence), with an increase in dry climates (high confidence). Land surface processes modulate the likelihood, intensity and duration of many extreme events including droughts (medium confidence) and heavy precipitation (medium confidence). The direction and magnitude of hydrological changes induced by land use change and land surface feedbacks vary with location and season (high confidence). Desertification exacerbates climate change through feedbacks involving vegetation cover, greenhouse gases and mineral dust aerosol (high confidence). Urbanization increases extreme rainfall events over or downwind of cities (medium confidence). Intensification of rainy events increase their consequences on land degradation.

A warming climate drives increases in atmospheric moisture and horizontal moisture transport from the divergent to the convergent portions of the atmospheric circulation (including storm systems, the tropical rain belt and monsoons) that on average amplifies existing precipitation minus evaporation (P–E) patterns (Held and Soden, 2006). Increased latent heat transports in high latitudes also contribute to polar amplification of warming (Section 7.4.4.1). Although convergent parts of the atmospheric circulation are expected to become wetter (in terms of increasing P–E) and net evaporative regions drier (increasing E–P) these regions are not geographically and seasonally fixed and their location and timing are expected to alter (Section 8.2.2.2). Atmospheric and ocean circulation changes overall decrease the amplification of P–E and salinity patterns. Paleoclimate evidence confirms that during the LGM zonal mean changes were roughly in agreement with thermodynamic expectations (G. Li et al., 2013). However regional changes can be dominated by dynamics, including responses to the large Northern Hemisphere (NH) ice sheets (DiNezio and Tierney, 2013; T. Bhattacharya et al., 2017; Scheff et al., 2017; D’Agostino et al., 2019; Lowry and Morrill, 2019) such that altered P–E patterns are not well described by thermodynamic drivers (Osteret al., 2015; Lora, 2018; Morrill et al., 2018).

There is robust evidence and high agreement across thermodynamics, detailed modelling and observations that amplification of P–E patterns occurs over the oceans (Figure 8.5a) with an associated ‘fresh gets fresher, salty gets saltier’ signature in ocean salinity (Sections 2.3.3.2 and 3.5.2). This amplification is moderated by proportionally larger increases in subtropical ocean evaporation and weakening of the tropical circulation (Section 8.2.2.2), an expectation supported by observations (Skliriset al., 2016) and process understanding (Yang andRoderick, 2019). Thermodynamics explain a smaller low latitude evaporation increase (1% °C^–1) than in high latitudes (5% °C^–1) with changes in surface radiation, boundary layer adjustments and ocean heat uptake playing a secondary role, based on idealized modelling (Siler et al., 2019). Increased evaporation from warmer oceans and lakes is exacerbated by the loss of surface ice in some regions (Bintanja and Selten, 2014; Laîné et al., 2014; W. Wang et al., 2018; Sharma et al., 2019; Woolway et al., 2020). This can generate a more local moisture source for precipitation, for example in north-west Greenland during non-summer months since the 1980s (Nusbaumer et al., 2019), though moisture transport changes can counteract this effect (Nygård et al., 2020). Ocean stratification due to heating of the upper layers through radiative forcing has been identified as a mechanism that further amplifies surface salinity patterns beyond the responses driven by water cycle changes alone (Zika et al., 2018).

Figure 8.5 | Zonally-averaged annual mean changes in precipitation minus evaporation (P–E) over (a) ocean and (b) land between the historical (1995–2014) andSSP2-4.5 (2081–2100) CMIP6 simulations (blue lines, an average of the CanESM5 and MRI-ESM2-0 models). Dashed lines show estimated P–E changes using a simple thermodynamic scaling (Held and Soden, 2006); dotted lines show estimates using an extended scaling (Byrne and O’Gorman, 2016). All curves have been smoothed in latitude using a three grid-point moving-average filter. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

Since AR5, numerous studies have confirmed that changes in P–E with warming over land cannot be interpreted simply as a ‘wet regions get wetter, dry regions get drier’ response (Chadwick et al., 2013; Greve et al., 2014; Roderick et al., 2014; Byrne and O’Gorman, 2015; Scheff and Frierson, 2015). Firstly, P–E is a simplistic diagnostic of the water cycle that inadequately describes ‘dryness’ or aridity (Fu and Feng, 2014; Roderick et al., 2014; Greve and Seneviratne, 2015; Scheff and Frierson, 2015; Greve et al., 2019; Vicente-Serrano et al., 2020). Secondly, terrestrial P–E is generally positive and balanced by surface runoff and percolation into subsurface soils and aquifers (Figure 8.1). As a result, the simple thermodynamic scaling (Figure 8.5b) predicts that P–E over land will become more positive (wetter) with warming (Greve et al., 2014; Roderick et al., 2014; Byrne and O’Gorman, 2015). This is not necessarily true, however, in the dry seasons and regions where terrestrial water is lost to the atmosphere and exported (Sheffield et al., 2013; Kumar et al., 2015; Keune and Miralles, 2019). Thirdly, regional P–E patterns over land are affected by changes in atmospheric circulation, oceanic moisture supply and land surface feedbacks. As the land warms more than oceans, spatial gradients in temperature and relative humidity influence moisture supply and reduce P–E over some land regions, such as southern Chile and Argentina around 30°S–50°S as captured by an extended thermodynamic scaling (Figure 8.5b). Drying of soils can be amplified by vegetation responses (Berg et al., 2016; Byrne and O’Gorman, 2016; Lambert et al., 2017) but limited by atmospheric circulation feedbacks (Zhou et al., 2021). Changes in soil moisture and rainfall intensity (Sections 8.2.3.2 and 8.2.3.3) can alter the partitioning of precipitation between evaporation and runoff, further complicating terrestrial P–E responses (Short Gianotti et al., 2020).

The strong physical basis for regionally and seasonally dependent responses of P–E and the expectation for an increasing contrast between wet and dry seasons and weather regimes is supported byhigh agreement across multiple observational and CMIP5/CMIP6 modelling studies (Liu and Allan, 2013; Kumar et al., 2015; Polson and Hegerl, 2017; Ficklin et al., 2019; Deng et al., 2020; Schurer et al., 2020). Increased moisture transports into storm systems, monsoons and high latitudes increase the intensity of wet events (Section 8.2.3.2), while stronger atmospheric evaporative demand with warming (Scheff and Frierson, 2014; Vicente-Serrano et al., 2018; Cook et al., 2019) is an important mechanism for intensifying dry events (Section 8.2.3.3) and decreasing soil moisture over many subtropical land regions. However, aridification is modulated regionally by poleward migration of the subtropical dry zones and an increasing land –ocean temperature contrast that drives declining relative humidity (Section 8.2.2.2).

To summarize, increased moisture transport from evaporative oceans to high precipitation regions of the atmospheric circulation will drive amplified P–E and salinity patterns over the ocean (high confidence) while more complex regional changes are expected over land. Greater warming over land than ocean alters atmospheric circulation patterns and on average reduces continental near-surface relative humidity which along with vegetation feedbacks can contribute to regional decreases in precipitation (high confidence). Based on an improved understanding of thermodynamic drivers since AR5 and multiple lines of evidence, there is high confidence that very wet or dry seasons and weather patterns will intensify in a warming climate such that wet spells become wetter and dry spells drier.

Responses of the large-scale atmospheric circulation to a warming climate are not as well understood as thermodynamic drivers (Shepherd, 2014). The AR5 identified robust features including a weakening and broadening of tropical circulation with poleward movement of tropical dry zones and mid-latitude jets (Collins et al., 2013). These can dominate regional water cycle changes, affecting the availability of freshwater and the occurrence of climate extremes. Atmospheric circulation changes generally dominate the spatial pattern of rapid precipitation adjustments (Section 8.2.1) to different forcing agents in the tropics (Bony et al., 2013; He and Soden, 2015; T.B. Richardson et al., 2016, 2018a; Tian et al., 2017; X. Li et al., 2018). Radiative forcing with heterogeneous spatial patterns such as ozone and aerosols (including cloud interactions; Section 6.4.1 and Box 8.1) drive substantial responses in regional atmospheric circulation through uneven heating and cooling effects(L. Liu et al., 2018; Dagan et al., 2019b; Wilcox et al., 2019). Changes in atmospheric circulation are also driven by slower, evolving patterns of warming and associated changes in temperature and moisture gradients (Bony et al., 2013; Samset et al., 2016, 2018a; Ceppi et al., 2018; Ma et al., 2018). There is strong evidence that large regional water cycle changes arise from the atmospheric circulation response to radiative forcings and associated SST pattern evolution butlow agreement in the sign and magnitude (Chadwick et al., 2016b). The role of prolonged weather regimes in determining wet and dry extremes is also better understood since AR5 (Kingston and McMecking, 2015; Schubert et al., 2016; D. Richardson et al., 2018; Barlow et al., 2019). Advances in knowledge of expected large-scale dynamical responses of the water cycle are further assessed in this section (see also Figure 8.21).

Long-term weakening of the tropical atmospheric overturning circulation is expected as climate warms in response to elevated CO₂ (Collins et al., 2013). A weaker circulation is required to reconcile global mean low-level water vapour increases (around 7% °C^–1) with the smaller global precipitation responses of about 1–3% °C^–1 (Section 8.2.1). The slowdown can occur in both the Hadley and Walker circulations, but occurs preferentially in the Walker circulation in most climate models (Vecchi and Soden, 2007) but this response has been questioned on the basis of model bias in east Pacific SST (Seager et al., 2019a). Weakening is expected to drive P–E decreases over the western Pacific and increases over the eastern Pacific. However, the driving mechanisms for Walker circulation weakening differ to those involved in determining ENSO variability, so it is too simplistic to interpret changes as an El Niño pattern of regional hydrological cycle extremes (Sohn et al., 2019). Internal variability is also capable of temporarily strengthening the Walker circulation (Section 2.3.1.4.1; L’Heureux et al., 2013; Chung et al., 2019) while regional responses depend on the pattern of warming (Sandeep et al., 2014).

Model simulations show a stronger Pacific Walker circulation during the LGM in response to a cooler climate (consistent with an expected weakening in a warmer climate), but a weaker Indian Ocean east– west circulation in response to the exposure of the Sunda and Sahul shelves due to lowered sea level (DiNezio et al., 2011). The latter effect is detectable in proxies for hydroclimate, as well as salinity and sea surface temperature (DiNezio and Tierney, 2013; DiNezio et al., 2018). More relevant to future warming is the mid-Pliocene period (3 million years ago), the last time the Earth experienced CO₂ levels comparable to present (see Cross-Chapter Box 2.4). Sea surface temperature (SST) reconstructions show a weakening of the Pacific zonal gradient and a pattern of warmth consistent with a weaker Walker cycle response (Corvec and Fletcher, 2017; Tierney et al., 2019; McClymont et al., 2020). Although the Pliocene SST pattern and wet subtropics contrast with present conditions (Burls and Fedorov, 2017), the paleoclimate record strengthens evidence that a warmer climate is associated with a weaker Walker circulation (Cross-Chapter Box 2.4; Section 3.3.3).

Since AR5, weakening of the tropical circulation has been explained as a rapid response to increasing CO₂ concentrations and slower response to warming and evolving SST patterns (He and Soden, 2017; Xia and Huang, 2017; Shaw and Tan, 2018; Chemke and Polvani, 2020). Large-scale tropical circulation weakens by 3–4% in a rapid response to a quadrupling of CO₂ concentrations (Plesca et al., 2018), which suppresses tropospheric radiative cooling, particularly in subtropical ocean subsidence regions (Bony et al., 2013; Merlis, 2015; Richardson et al., 2016). The resulting increased atmospheric stability explains the rapid weakening of the Walker circulation (Wills et al., 2017) and Northern Hemisphere Hadley Cell (Chemke and Polvani, 2020). Subsequent surface warming contributes up to a 12% slowing of circulation for a uniform 4°C SST increase, driven by thermodynamic decreases in temperature lapse rate (Plesca et al., 2018).

The regional Inter-tropical Convergence Zone (ITCZ) position, width and strength determine the location and seasonality of the tropical rain belt. Since AR5, multiple studies have linked cross-equatorial energy transport to the mean ITCZ position (Donohoe et al., 2013; Frierson et al., 2013; Bischoff and Schneider, 2014; Boos and Korty, 2016; Loeb et al., 2016; Adam et al., 2018; Biasutti and Voigt, 2019). Multi-model studies agree that aerosol cooling in the NH led to a southward shift in the ITCZ and tropical precipitation after the 1950s up to the 1980s that is linked with the 1980s Sahel drought (Box 8.1; Section 8.3.2.4 and 10.4.2.1). In particular, aerosol-cloud interaction was identified as a potentially important driver of this shift (Chung and Soden, 2017) but this is uncertain since observations suggest that models may overestimate (Malavelle et al. , 2017; Toll et al. , 2017) or underestimate (Rosenfeld et al. , 2019) the aerosol cloud-mediated cooling effects. In addition, greenhouse gas forcing has been invoked in explaining much of the increase in Sahel precipitation since the 1980s through enhanced meridional temperature gradient, with only a secondary role for aerosol (Dong and Sutton, 2015).

Understanding of how ITCZ width and strength respond to a warming climate has improved since AR5 (Byrne and Schneider, 2016; Harrop and Hartmann, 2016; Popp and Silvers, 2017; Dixit et al., 2018; Zhou et al., 2020). Studies suggest that convection gets stronger and more focused within the core of the ITCZ (Lau and Kim, 2015; Byrne et al., 2018). This leads to drying on the equatorward edges of the ITCZ and a moistening tendency in the ITCZ core (Byrne and Schneider, 2016). Feedbacks involving clouds have been identified as an important mechanism leading to tightening and strengthening of the ITCZ (Popp and Silvers, 2017; Su et al., 2017, 2019, 2020; Talib et al., 2018). Stronger ascent in the core amplifies the ‘wet get wetter’ response while reduced moisture inflow near the ITCZ edges reduces this response below the 7% °C^–1 Thermodynamic increase in moisture transport. Thus, there is a range of evidence and medium agreement for strengthening and contraction of the ITCZ with warming that sharpens contrasts between wet and dry regimes. However, understanding of how the regional ITCZ location responds in a warming climate is not robust (Section 8.4.2.1) with limited evidence of distinct regional responses to GHG forcing including a northward shift over eastern Africa and the Indian Ocean and a southward shift in the eastern Pacific and Atlantic oceans (Mamalakis et al., 2021). Paleoclimate evidence highlights the distinct regional ITCZ responses to hemispheric asymmetry in volcanic and orbital forcing (McGee et al., 2014; Boos and Korty, 2016; Colose et al., 2016; Denniston et al., 2016; PAGES Hydro2K Consortium, 2017; Singarayer et al., 2017; Atwood et al., 2020) and rapid (>1° latitude over decades) shifts in the ITCZ and regional monsoons in response to AMOC collapse cannot be ruled out (Sections 8.6.1.1 and 5.1.3).

Monsoons are key components of the tropical overturning circulation that can be understood as a balance between net energy input (e.g., radiative and turbulent fluxes) and the export of moist static energy. This is determined by contrasting surface heat capacity between ocean and land and modified through changes in atmospheric dynamics, tropical tropospheric stability and land surface properties (Jalihal et al., 2019). Thermodynamic increases in moisture transport are expected to increase monsoon strength and area (Christensen et al., 2013). Since AR5, evidence continues to demonstrate that monsoon circulation is sensitive to spatially varying radiative forcing by anthropogenic aerosols (Hwang et al. , 2013; R.J. Allen et al. , 2015; Z. Li et al. , 2016b) and GHGs (Dong andSutton, 2015). Changes in SST patterns also play a role (Guo et al., 2016; W. Zhou et al., 2019; Cao et al., 2020) by altering cross-equatorial energy transports and land–ocean temperature contrasts. This evidence continues to support a thermodynamic strengthening of monsoon precipitation that is partly offset by slowing of the tropical circulation but with weak evidence and low agreement for regional aspects of circulation changes. Disagreement between paleoclimate and modern observations, physical theory and numerical simulations of global monsoons have been partly reconciled (Section 3.3.3.2) through improved understanding of regional processes (Harrison et al. , 2015; R. Bhattacharya et al. , 2017; Bhattacharya et al. , 2018; Biasutti et al. , 2018; D’Agostino et al. , 2019; Jalihal et al. , 2019; Seth et al. , 2019), although interpreting past changes in the context of future projections requires careful account of differing forcings and feedbacks (D’Agostinoet al., 2019). Assessment of past changes and future projections in regional monsoons are provided in Sections 2.3.1.4.2, 8.3.2.4 and 8.4.2.4.

Since AR5, understanding of poleward expansion of the Hadley Cells has improved (Section 2.3.1.4.1) but its role in subtropical drying is limited to the zonal mean and dominated by ocean regions (Byrne and O’Gorman, 2015; Grise and Polvani, 2016; He and Soden, 2017; Schmidt and Grise, 2017; Siler et al., 2018; Chemke and Polvani, 2019; Grise and Davis, 2020). Over subtropical land, evolving SST patterns and land–ocean warming contrasts, that are partly explained by rapid responses to CO₂ increases, can dominate aspects of the atmospheric circulation response (Byrne and O’Gorman, 2015; He and Soden, 2015; Chadwick et al., 2017; H. Yang et al., 2020) and resultant regional water cycle changes, particularly for projected drying in semi-arid, winter-rainfall dominated subtropical climates (Deitch et al. , 2017; Brogli et al. , 2019; Seager et al. , 2019b; Zappa et al. , 2020). Poleward expansion of the tropical belt is expected to drive a corresponding shift in mid-latitude storm tracks, but the controlling mechanisms differ between hemispheres. Southern Hemisphere expansion is driven by GHG forcing and amplified by stratospheric ozone depletion, while weaker Northern Hemisphere expansion in response to GHG forcing is modulated by tropospheric ozone and aerosol forcing, particularly black carbon (Davis et al. , 2016; Grise et al. , 2019; Watt-Meyer et al. , 2019; Zhao et al. , 2020). However, internal variability is found to dominate observed responses in the NH, precluding attribution to radiative forcing (D’Agostino et al., 2020a). Paleoclimate evidence of poleward expansion and weakening of westerly winds in both hemispheres in the warmer Pliocene is linked to reduced equator-to-pole thermal gradients and ice volume(Abell et al., 2021).

The influence of amplified Arctic warming on mid-latitude regional water cycles is not well understood based on simple physical grounds due to the large number of competing physical processes (Cross-Chapter Box 10.1). The thermal gradient between polar and lower latitude regions decreases at lowlevels due to Arctic warming amplification. However, at higher altitudes, the corresponding thermal gradient increases with warming due to cooling of the Arctic stratosphere and this is consistent with a strengthening of the winter jet stream in both hemispheres, yet there is low agreement on the precise mechanisms (Vallis et al., 2015; Vihma et al., 2016). Changes in the strength of the polar stratospheric vortex can also alter the mid-latitude circulation in winter, but responses are not consistent across models (Oudar et al., 2020a). Nevertheless, thermodynamic strengthening of moisture convergence into weather systems and polar regions is robust (Section 8.2.2.1) and remains valid despite weak understanding of atmospheric circulation change.

In summary, there is high confidence that altered atmospheric wind patterns in response to radiative forcing and evolving surface temperature patterns will affect the regional water cycle in most regions. Mean tropical circulation is expected to slow with global warming (high confidence) but temporary multi-decadal strengthening is possible due to internal variability (medium confidence). Slowing of the tropical circulation reduces the meridional P–E gradient over the Pacific and can partly offset thermodynamic amplification of P–E patterns and strengthening of monsoons (high confidence) but regional characteristics of tropical rain belt changes are not well understood. There is medium confidence in processes driving strengthening and tightening of the ITCZ that increase the contrasts between wet and dry tropical weather regimes and seasons. There is high confidence in understanding of how radiative forcing and global warming drive a poleward expansion of the subtropics and mid-latitude storm tracks but onlylow confidence in how poleward expansion influences drying of subtropical and mid-latitude climates. There is low confidence in understanding how Arctic warming amplification affects mid-latitude regional water cycles but high confidence that thermodynamic strengthening of precipitation within weather systems and in monsoons and polar regions is robust to large-scale circulation changes.

Declining ice-sheet mass, glacier extent and Northern Hemisphere (NH) sea ice, snow cover and permafrost (Collins et al., 2013; Vaughan et al., 2013) is an expected consequence of a warming climate (Sections 2.3.2, 3.4, 4.3.2.1 and 9.3–9.5). A decline in mountain snow cover and increased snow and glacier melt will alter the amount and timing of seasonal runoff in mountain regions (Sections 3.4.2, 3.4.3 and 9.5). Earlier and more extensive winter and spring snowmelt (X. Zeng et al., 2018) can reduce summer and autumn runoff in snow-dominated river basins of mid–high latitudes of the NH (Rhoades et al., 2018; Blöschl et al., 2019). Since AR5, an earlier but less rapid snowmelt has been explained by reduced winter snowfall and less intense solar radiation earlier in the season (Musselman et al. , 2017; Wu et al. , 2018; Grogan et al. , 2020). Reduced snow cover also increases energy available for evaporation, which can dominate declining river discharge based on modelling of the Colorado River (Milly and Dunne, 2020). An increase in the fraction of precipitation falling as rain compared with snow can lead to declines in both streamflow and groundwater storage in regions where snowmelt is the primary source of recharge (Earman and Dettinger, 2011; Berghuijs et al., 2014). Such regions include western South America and western North America, semi-arid regions which rely on snowmelt from high mountain chains (Ragettli et al., 2016; Milly and Dunne, 2020). Rain-on-snow melt events reduce at lower altitudes due to declining snow cover but increase at higher altitudes where snow tends to be replaced by rain based on observations and modelling (Musselman et al., 2018; Pall et al., 2019), thereby altering seasonal and regional characteristics of flooding (Section 11.5).

Seasonal melt water from high mountain glaciers in Asia (see Cross-Chapter Box 10.4) supply the basic needs of 221 ± 97 million people (Pritchard, 2019; Immerzeel et al., 2020). Glacier-melt in response to warming can initially lead to increased runoff volumes, especially in peak summer flows, but they will eventually decline as most glaciers continue to shrink. SROCC concluded there is high confidence that the peak runoff has already been passed for some smaller glaciers (Hock et al., 2019a). Increased precipitation and glacier-melt can also contribute to rising lake levels and flood hazards in regions such as the inner Tibetan Plateau, Patagonia, Peru, Alaska and Greenland (Lei et al. , 2017; Shugar et al. , 2020; Stuart-Smith et al. , 2020). Since AR5, evidence from multiple locations (New Zealand, Greenland, Antarctica) shows that intrusions of warm, moist air are important in controlling glacier mass balance, the likelihood of extreme ablation or snowfall events depending on air temperature (Gorodetskaya et al. , 2014; Mackintosh et al. , 2017; Mattingly et al. , 2018; Little et al. , 2019; Oltmanns et al. , 2019; Wille et al. , 2019; Adusumilli et al. , 2021). Sensible heating from warm air and increased longwave radiation from atmospheric moisture and low clouds drive melt events (Stuecker et al., 2018).

Reductions in snow, freshwater ice and permafrost affect terrestrial hydrology. Permafrost degradation reduces soil ice and alters the extent of thermokarst lake coverage (Section 9.5.2; M. Meredith et al., 2019). A lag between current climate change and permafrost degradation is expected, given the slow response rates in frozen ground and the fact that snow cover insulates soil from sensible heat exchanges with the air above (Hoegh-Guldberg et al. , 2018; García-García et al. , 2019; Soong et al. , 2020). Post‐wildfire areas are also linked with permafrost degradation in the Arctic based on satellite observations (Yanagiya and Furuya, 2020). An increase in spring rainfall can increase heat advection by infiltration, exacerbating permafrost thaw and leading to increased methane emissions (Section 5.4.7; Neumann et al., 2019). Increased heat transport by Arctic rivers can also contribute to earlier sea ice melt (Park et al., 2020).

In summary, it is virtually certain that warming will cause a loss of frozen water stores, except in areas where temperatures remain below 0°C for most of the year. There is high confidence that warming and reduced snow volume drives an earlier snowmelt, leading to seasonally dependent changes in streamflow. There is medium confidence that weaker sunlight earlier in the season can reduce the rate of snowmelt. Melting of snowpack or glaciers can increase streamflow in high-latitude and high-altitude catchments until frozen water reserves are depleted (high confidence). There is high confidence that warm, moist airflows and associated precipitation dominate glacier mass balance in some regions (New Zealand, Greenland, Antarctica).

Evidence that heavy precipitation events (from sub-daily up to seasonal time scales) intensify as the planet warms has strengthened since AR5 (Section 11.4, Box 11.1 and Cross-Chapter Box 3.2) based on improved physical understanding, extensive modelling and increasing observational corroboration (O’Gorman, 2015; Fischer and Knutti, 2016; Neelin et al., 2017). There is robust evidence, with medium agreement across a range of modelling and observational studies, of thermodynamic intensification of wet seasons (Chou et al., 2013; Liu and Allan, 2013; Dunning et al., 2018; Lan et al., 2019; Zhang and Fueglistaler, 2019). Extreme daily precipitation is expected to increase at close to the 7% °C^–1 increase in the near-surface atmospheric moisture-holding capacity determined by the Clausius–Clapeyron equation (Section 11.4, Figure 8.4), with limited evidence that higher rates apply for shorter duration precipitation events (Formayer and Fritz, 2017; Lenderink et al. , 2017; Ali et al. , 2018; Guerreiro et al. , 2018; Burdanowitz et al. , 2019; W. Zhang et al. , 2019a). However, observed estimates sample multiple synoptic weather states, mixing thermodynamic and dynamic factors, so are not directly relatable to climate change responses (Bao et al., 2017; Drobinski et al., 2018). The contrasting spatial scales sampled by the observations and models (from global to cloud resolving) explain the large range of daily and sub-daily precipitation scaling with temperature assessed in Figure 8.4.

Since AR5, advances in understanding the expected changes in intense rainfall at the sub-daily time scale (Section 11.4, Figure 8.4) are provided by idealized or high resolution model experiments and observations (Westra et al., 2014; Fowler et al., 2021). There is robust evidence from simplified calculations, convection resolving models and observations that thermodynamics drives an increase in convective available potential energy (CAPE) with warming and therefore the intensity of convective storms (Singh and O’Gorman, 2013; Romps, 2016; Barbero et al., 2019). Also, declining relative humidity over land (Sections 2.3.1.3.2 and 8.2.2.1) increases lifting condensation level, thereby delaying but intensifying convective systems (Louf et al., 2019; J. Chen et al., 2020a). Larger systems are linked with increasing tropopause height (Lenderink et al., 2017) that can also amplify storm precipitation (Prein et al., 2017). However, the heaviest rainfall is not necessarily associated with the most intense (deepest) storms based on satellite data (Hamada et al., 2015; Hamada and Takayabu, 2018). Precipitation intensification can exceed thermodynamic expectations where and when additional latent heating invigorates individual storms (Section 11.4.1) as implied by medium agreement across modelling and observational studies (Berg et al. , 2013; Molnar et al. , 2015; Scoccimarro et al. , 2015; Prein et al. , 2017; Zhou and Wang, 2017; Nie et al. , 2018; Kendon et al. , 2019; Z. Zhang et al. , 2019). This intensification depends on time of day, based on convection-permitting simulations (E.P. Meredith et al., 2019).

Intensification of sub-daily rainfall is inhibited in regions and seasons where available moisture is limited (Prein et al., 2017). However, a fixed threshold temperature above which precipitation is limited by moisture availability is not supported by modelling evidence (Neelin et al., 2017; Prein et al., 2017). Enhanced latent heating within storms can also suppress convection at larger scales due to atmospheric stabilization as demonstrated with high resolution, idealized and large ensemble modelling studies (Loriaux et al. , 2017; Chan et al. , 2018; Nie et al. , 2018; Tandon et al. , 2018; Kendon et al. , 2019). Stability is also increased by the direct radiative heating effect of higher CO₂ concentrations (Baker et al., 2018) and influenced by aerosol effects on the atmospheric energy budget and cloud development (Box 8.1). Since AR5, modelling evidence shows increases in convective precipitation extremes are limited by droplet/ice fall speeds (Singh and O’Gorman, 2014; Sandvik et al., 2018) but these processes are only crudely represented (Tapiador et al., 2019a). Idealized regional and coupled global models combined with limitedobservational evidence shows that instantaneous precipitation extremes are sensitive to microphysical processes, while daily extremes are determined more by the degree of convective aggregation (Bao and Sherwood, 2019; Pendergrass, 2020a).

Dynamical changes modify and can dominate thermodynamic drivers of local rainfall and flood hazard change (Box 11.1). For example, increased land –ocean temperature gradients (Section 8.2.2.2) explain more intense rain from convective systems over the Sahel based on satellite data since the 1980s (Taylor et al., 2017) and dynamical feedbacks can invigorate active to break phase transition over India (Karmakar et al., 2017; Roxy et al., 2017). Satellite data shows long-lived, organized mesoscale convective systems contribute disproportionally to extreme tropical precipitation (Roca and Fiolleau, 2020). Since AR5, the spatial variability in soil moisture has been linked with the timing and location of convective rainfall by altering the partitioning between latent and sensible heating. This was demonstrated for the Sahel, Europe and India in observations (C.M. Taylor et al., 2013; Taylor, 2015; Petrova et al., 2018; Barton et al., 2020; Klein and Taylor, 2020) but depends on the moisture-convergence regime (Welty et al., 2020). Only high-resolution convection-permitting models can capture the sub-grid scale mechanisms for convective initiation (C.M. Taylor et al. , 2013; H. Moon et al. , 2019). There is medium evidence that greater tropical cyclone rainfall totals can be caused by dynamical feedbacks (Chauvin et al., 2017) and slower propagation speed as tropical circulation weakens (Kossin, 2018). These processes amplify the thermodynamic intensification of rainfall (Section 11.7.1.2), yet observational support is weak (Chan, 2019; Lanzante, 2019; I.J. Moon et al. , 2019; Knutson et al. , 2020). Slower decay following landfall, explained by larger stores of heat and moisture at higher SSTs, can also amplify rainfall amount based on observations and modelling (Li and Chakraborty, 2020). Rainfall intensity from the outer rain bands of tropical cyclones is also increased by aerosol–cloud interactions (Box 8.1).

The amount and intensity of rainfall within extratropical storms is expected to increase with atmospheric moisture. This is particularly evident for atmospheric rivers (see Glossary) and research since AR5 has confirmed their link with flooding and terrestrial water storage (Froidevaux and Martius, 2016; Paltan et al. , 2017; Waliser and Guan, 2017; Adusumilli et al. , 2019; Ionita et al. , 2020; Payne et al. , 2020). There is robust evidence based on simple physics and detailed modelling that extratropical cyclone rainfall, including atmospheric river events, will intensify through increased atmospheric moisture flux (Lavers et al., 2013; Ramos et al., 2016; Yettella and Kay, 2017; V. Espinoza et al., 2018; Algarra et al., 2020; Xu et al., 2020; Zavadoff and Kirtman, 2020; Zhao, 2020), although changes in dynamical aspects will modify responses regionally (Section 8.4.2.8). For example, stronger latitudinal temperature gradients in the high-latitude upper troposphere drive increased extratropical storm speed around 30°N–70°N based on CMIP5 simulations (Dwyer and O’Gorman, 2017), causing reduced precipitation accumulation.

The response of flood hazard to changing rainfall characteristics depends on time and space scale and the nature of the land surface (Section 11.5.1 and FAQ 8.2). Sustained and heavy rainfall can lead to widespread flooding and landslides while intensification of short-duration intense rainfall can increase the severity and frequency of flash flooding (Marengo et al., 2013; Chan et al., 2016; Gariano and Guzzetti, 2016; Sandvik et al., 2018). Flooding events in many tropical regions (e.g., north-western South America, southern Africa and Australasia) are associated with ENSO variability (Emerton et al., 2017; Takahashi and Martínez, 2019; Pabón-Caicedo et al., 2020) and amplified by thermodynamic increases in water vapour. Flood hazard from heavy rainfall is modulated by snowmelt (Section 8.2.3.1), vegetation characteristics (Page et al., 2020; Murphy et al., 2021) and direct human intervention (Sections 8.2.3.4 and FAQ 8.2) but also can be compounded by sea level rise (Sections 4.3.2.2 and 9.6.4) in coastal and delta regions (Bevacqua et al., 2019; Ganguli and Merz, 2019; Eilander et al., 2020). Antecedent soil moisture conditions are an important modulator of flooding (Section 11.5.1) but become less important for smaller catchments and for more severe floods (Wasko and Nathan, 2019). Depleted soil moisture after more intense dry seasons (Section 8.2.2.1) can allow greater uptake of wet season rainfall before soils saturate. Since AR5, evidence confirms that more intense rainfall increases the proportion of runoff and reservoir recharge relative to infiltration into the soil (Eekhout et al., 2018; Yin et al., 2018). More intense but less frequent storms (Kendon et al., 2019) favour focused groundwater recharge through leakage from surface waters (R.G. Taylor et al., 2013a; Cuthbert et al., 2019a) and runoff and flash flooding where the percolation capacity of the soil is exceeded (Yin et al., 2018).

Increased severity of flooding on larger, more slowly-responding rivers is expected as precipitation accumulations increase during persistent wet events over a season. This can occur where atmospheric blocking patterns repeatedly steer extratropical cyclones across large river catchments, as identified for NH mid-latitudes and Asia (Takahashi et al., 2015; Pfleiderer et al., 2018; Zhou et al., 2018; Blöschl et al., 2019; Lenggenhager et al., 2019; Nikumbh et al., 2019; Zanardo et al., 2019), although groundwater flooding and antecedent conditions including soil moisture and snowmelt also play a role (Muchan et al., 2015; Berghuijs et al., 2019). Increased atmospheric moisture amplifies the severity of these events when they occur in a warmer climate, yet drivers of change in the occurrence of blocking patterns, stationary waves and jet stream position are not well understood (Section 8.2.2.2 and Cross-Chapter Box 10.1).

In summary, there is very high confidence that heavy precipitation events will become more intense in a warming climate. There is high confidence that increased moisture and its convergence within extratropical and tropical cyclones and storms will increase rainfall totals during wet events at close to the 7% °C^–1 Thermodynamic response, with low confidence of higher rates for sub-daily intensities. There is medium confidence that more intense but less frequent rainfall increases the proportion of rainfall leading to surface runoff and focused groundwater recharge from temporary water bodies. There is low confidence in how the frequency of flooding will change regionally as it is strongly dependent on catchment characteristics, antecedent conditions and how atmospheric circulation systems respond to climate change, which is less certain than thermodynamic drivers (Section 11.5). However, there is high confidence that increases in precipitation intensity and amount during very wet events (from sub-daily up to seasonal time scales) will intensify severe flooding when these extremes occur.

Regional changes in aridity – broadly defined as a deficit of moisture – are expected to occur in response to anthropogenic forcings as a consequence of shifting precipitation patterns, warmer temperatures, changes in cloudiness (affecting solar radiation), declining snowpack, changes in winds and humidity, and vegetation cover (Figure 8.6). Evapotranspiration (see Annex VII: Glossary) is a key component of aridity, and is composed of two main processes: evaporation from soil, water and vegetation surfaces; and transpiration, the exchange of moisture between plants and atmosphere through plant stomata. On a global level, warmer temperatures increase evaporative demand in the atmosphere, and thus (assuming sufficient soil moisture is available) increase moisture loss from evapotranspiration (high confidence) (Dai et al., 2018; Vicente-Serrano et al., 2020). On a regional level, aridity is further modulated by seasonal rainfall patterns, runoff, water storage, and interactions with vegetation.

Figure 8.6 | Climatic drivers of drought, effects on water availability, and impacts. Plus and minus signs denote the direction of change that drivers have on factors such as snowpack, evapotranspiration, soil moisture, and water storage. The three main types of drought are listed, along with some possible environmental and socio-economic impacts of drought (bottom).

Open figure

Vegetation is a crucial interface between subsurface water storage (in soil moisture and groundwater) and the atmosphere. Plants alter evapotranspiration and the surface energy balance, and thus can have a large influence on regional aridity (Lemordant et al., 2018). SRCCL concluded there is high confidence that higher atmospheric CO₂ increases the ratio of plant CO₂ uptake to water loss (water-use efficiency; WUE) through the combined enhancement of photosynthesis and stomatal regulation (Section 5.4.1; DeKauwe et al. , 2013; C.D. Jones et al. , 2013; Deryng et al. , 2016; Swann et al. , 2016; Cheng et al. , 2017; Knauer et al. , 2017; Peters et al. , 2018; Guerrieri et al. , 2019). Modelling studies suggest that increasing WUE can partly counteract water losses from increased evaporative demand in a warmer atmosphere, potentially mitigating aridification (Milly and Dunne, 2016; Bonfils et al. , 2017; Cook et al. , 2018; Y. Yang et al. , 2018). However, observational studies suggest that this effect may be counter-balanced by the increase in plant growth in response to elevated CO₂, which results in increased water consumption (De Kauwe et al. , 2013; Donohue et al. , 2013; Ukkola et al. , 2016b; Yang et al. , 2016; Guerrieri et al. , 2019; Mankin et al. , 2019; A. Singh et al. , 2020). In semi-arid regions, increased plant water consumption can reduce streamflow and exacerbate aridification (Ukkola et al. , 2016b; Mankin et al. , 2019; A. Singh et al. , 2020). Thus, there is low confidence that increased WUE in plants can counterbalance increased evaporative demand (Cross-Chapter Box 5.1).

A drought is a period of abnormally dry weather that persists for long enough to cause a serious hydrological imbalance (Glossary; Wilhite and Glantz, 1985; Wilhite, 2000; Cook et al., 2018). Most droughts begin as persistent precipitation deficits (‘meteorological drought’) that propagate over time into deficits in soil moisture, streamflow, and water storage (Figure 8.6), leading to a reduction in water supply (‘hydrological drought’). Increased atmospheric evaporative demand increases plant water stress, leading to ‘agricultural and ecological drought’ (Williams et al. , 2013; C.D. Allen et al. , 2015; Anderegg et al. , 2016; McDowell et al. , 2016; Grossiord et al. , 2020). Evaporative demand affects plants in two ways. It increases evapotranspiration, depleting soil moisture and stressing plants through lack of water (Teuling et al., 2013; Sperry et al., 2016), and also directly affects plant physiology, causing a decline in hydraulic conductance and carbon metabolism, leading to mortality (Figure 8.6; Breshears et al., 2013; Hartmann, 2015; McDowell and Allen, 2015; Fontes et al., 2018). While droughts are traditionally viewed as ‘slow moving’ disasters that typically take months or years to develop, rapidly evolving and often unpredictable flash droughtscan also occur (Otkin et al., 2016, 2018). Flash droughtscan develop within a few weeks, causing substantial disruption to agriculture and water resources (Pendergrass et al., 2020). Conversely, droughts that persist for a long time (usually a decade or more) are called megadroughts. Droughts span a large range of spatial and temporal scales, arise through a variety of climate system dynamics (e.g., internal atmospheric variability, ocean teleconnections), and can be amplified or alleviated by a variety of physical and biological processes. As such, droughts occupy a unique space within the framework of extreme climate and weather events, possessing no singular definition.

While the role of precipitation in droughts is obvious, other climatic drivers are also important, such as temperature, radiation, wind, and humidity (Figure 8.6). These factors have a strong influence on atmospheric evaporative demand, which affects evapotranspiration and soil moisture (Figure 8.6). In snow-dominated regions, high temperatures increase the fraction of precipitation falling as rain instead of snow and advance the timing of spring snowmelt (high confidence) (Vincent et al. , 2015; Mote et al. , 2016, 2018; Berg and Hall, 2017; Solander et al. , 2018). This can result in lower than normal snowpack levels (a ‘snow drought’), and thus reduced streamflow, even if total precipitation is at or above normal for the cold season (Harpold et al., 2017). Plants also affect the severity of droughts by modulating evapotranspiration (Figure 8.6). As discussed above, the effect of elevated CO₂ on plants has the potential to both increase and reduce water loss through evapotranspiration via enhanced WUE and plant growth, respectively (Figure 8.6), but there is low confidence in whether one process dominates over another at the global scale.

Drought severity also depends on human activities and decision-making (AghaKouchak et al. , 2015; Van Loon et al. , 2016; Pendergrass et al. , 2020). Societies have developed a variety of strategies to manipulate the water cycle to increase resiliency in the face of water scarcity, including irrigation, creation of artificial reservoirs, and groundwater pumping. While potentially buffering water resource capacity, in some cases these interventions may unexpectedly increase vulnerability (medium confidence). For example, while increased irrigation efficiency may ensure more water is available to crops, the corresponding reduction in runoff and subsurface recharge may exacerbate hydrologic drought (Grafton et al., 2018). Furthermore, while building dams and increasing surface reservoir capacity can boost water resources, they may actually increase drought vulnerability if demands rise to take advantage of the increased supply or if over-reliance on these surface reservoirs is encouraged (Di Baldassarre et al., 2018). Interactions between adaptation, vulnerability, and drought impacts are discussed further in WGII (Chapters 2 and 4).

In summary, there is high confidence that a warming climate drives an increase in atmospheric evaporative demand, decreasing available soil moisture. There is high confidence that higher atmospheric CO₂ increases plant water-use efficiency, but low confidence that this physiological effect can counterbalance water losses. Since drought can be defined in a number of ways, there are potentially different responses under a warming climate depending on drought type. Beyond a lack of precipitation, changes in evapotranspiration are critical components of drought, because these can lead to soil moisture declines (high confidence). Under very dry soil conditions, evapotranspiration becomes restricted and plants experience water stress in response to increased atmospheric demand (medium confidence). Human activities and decision-making have a critical impact on drought severity (high confidence).

Human activities influence the regional water cycle directly through modifying and exploiting stores and flows from rivers, lakes and groundwater and by altering land cover characteristics. These actions alter surface energy and water balances through changes in permeability, surface albedo, evapotranspiration, surface roughness and leaf area. Direct redistribution of water by human activities for domestic, agricultural and industrial use of about 24,000 km³yr^–1 (Figure 8.1) is equivalent to half the global river discharge or double the global groundwater recharge each year (Abbott et al., 2019). Since AR5, both modelling studies and observations have demonstrated that land use change can drive local and remote responses in precipitation and river flow by altering the surface energy balance, moisture advection and recycling, land –sea thermal contrast and associated wind patterns (Alter et al. , 2015; Wey et al. , 2015; De Vrese et al. , 2016; Pei et al. , 2016; Wang-Erlandsson et al. , 2018; Vicente-Serrano et al. , 2019). There is robust evidence that a warming climate combined with direct human demand for groundwater will deplete groundwater resources in already dry regions (Wada and Bierkens, 2014; D’Odorico et al., 2018; Jia et al., 2019).

The SRCCL presented evidence that extraction of water from the ground or river systems and intensive irrigation increases evaporation and atmospheric water vapour locally (Jia et al., 2019; Mishra et al., 2020). Irrigation can explain declining groundwater storage in some regions, including north-western India and North America (Asoka et al., 2017; G. Ferguson et al., 2018). Simulations spanning 1960–2010 indicate that approximately 30% of the present human water consumption is supplied from non-sustainable water resources (Wada and Bierkens, 2014). However, there is onlylimited evidence that groundwater extraction is lowering streamflow (Mukherjee et al., 2018; de Graaf et al., 2019). Model experiments show that irrigation can either aggravate or alleviate climate‐induced changes of surface or subsurface water (Lenget al., 2015). Widespread extraction of water from rivers can reduce flows and decrease the level and area of inland seas and lakes (Wurtsbaugh et al. , 2017; Torres-Batlló et al. , 2020; X. Wang et al. , 2020). Between 1985 and 2015, about 139,000 km² of inland water areas have become land, while creation of dams has converted about 95,000 km² of land to water, particularly in the Amazon and Tibetan Plateau (Donchyts et al., 2016). Direct management of river flow is comparable in magnitude to climate change effects for snow-fed rivers at a continental scale based on a global analysis and a study of 96 Canadian catchments (Tan and Gan, 2015; Arheimer et al., 2017).

The SRCCL assessed with medium confidence that mean and extreme precipitation is increased over and downwind of urban areas (Jia et al., 2019). There is medium confidence that altered thermodynamic and aerodynamic properties of the land surface from urbanization affects evaporation and increases precipitation over or downwind of cities (Box 10.3) due to altered stability and turbulence (Han et al. , 2014; Pathirana et al. , 2014; Jiang et al. , 2016; D’Odorico et al. , 2018; Sarangi et al. , 2018; Boyaj et al. , 2020). However, reduced biogenic aerosol, but increased anthropogenic aerosol emissions modify cloud microphysics and precipitation processes (Box 8.1; Schmidand Niyogi, 2017; D’Odorico et al. , 2018; Fan et al. , 2020; Zheng et al. , 2020). Urbanization also decreases permeability of the surface, leading to increased surface runoff (Chen et al., 2017; Jia et al., 2019). Large-scale infrastructure, such as the construction and operation of dikes, weirs, and hydropower plants, also alters surface energy and moisture fluxes, potentially influencing the regional water cycle. Limited modellingevidence suggests that large-scale solar and wind farms can increase precipitation locally (over the Sahel and North America) when dynamic vegetation responses are represented (Y. Li et al., 2018; Pryor et al., 2020), with remote effects also possible (Lu et al., 2021).

Changes in land use from forest to agriculture can exert profound regional effects on the water cycle (FAQ 8.1) by modifying the surface energy balance and moisture recycling (Krishnan et al. , 2016; Paul et al. , 2016; Llopart et al. , 2018; Singh et al. , 2019). There is medium evidence from modelling and observations over the Amazon and East Africa that deforestation drives increased streamflow (Dos Santos et al., 2018; Guzha et al., 2018; Levy et al., 2018) butlimited evidence that increases in global runoff due to deforestation are counterbalanced by decreases resulting from irrigation (Hoegh-Guldberg et al., 2018). Total Amazon deforestation drives reductions in precipitation but with a large 90% confidence range (–38 to +5 %) based on 44 primarily pre-AR5 climate model simulations (Spracklen and Garcia-Carreras, 2015) with smaller reductions (–2.3 to–1.3 %) attributed to observed Amazon deforestation up to 2010. Climate model development has reduced this uncertainty range but has not altered the median change (Lejeune et al., 2015). Large-scale global deforestation (20 million km²) simulated by 9 CMIP6 models confirms a large range in precipitation amount reduction of–37 ± 54 mm yr^–1 over the deforested regions (Boysen et al., 2020). However, small-scale deforestation can increase precipitation locally (Lawrence and Vandecar, 2015). A 50–60% deforestation rate corresponded to a wet season delay of about one week and greater chance of dry spells of eight days or longer based on correlation analysis of rain gauge and land-use data for South America (Leite-Filhoet al., 2019). Forest and grassland fires can also modify hydrological response at the watershed scale (Havel et al., 2018). Afforestation or reforestation aimed at removing CO₂ from the atmosphere can also alter the water cycle at the regional scale (Section 8.4.3 and Cross-Chapter Box 5.1).

In summary, there is high confidence that land-use change and water extraction for irrigation drive local, regional and remote responses in the water cycle. Large-scale deforestation is likely to decrease precipitation over the deforested regions but there is low confidence in the effects of limited deforestation. There is medium confidence that deforestation drives increased streamflow relative to the responses caused by climate change. Urbanization can increase local precipitation (medium confidence) and resulting runoff intensity (high confidence). A warming climate combined with direct human demand for water is expected to deplete groundwater resources in dry regions (high confidence).

The human influence on the global water cycle is often summarized as an intensification (Huntington, 2006; DeAngelis et al., 2015; W. Zhang et al., 2019b) or an overall strengthening which has been observed since at least 1980 (high confidence) (see Chapter 2). There is, however, no unique definition of the global water cycle intensity (Trenberth, 2011; Ficklin et al., 2019; Sprenger et al., 2019). One simple metric is the global and annual mean amount of precipitation. Although an increase in global precipitation is consistent with physical expectations (Section 8.2.1), it has not yet been detected and attributed to human activities given large observational uncertainties and low signal-to-noise ratio (Section 3.3.2.2). Other metrics are more suitable to detect and attribute changes in the global water cycle, including the likely increase in global land precipitation since 1950 (Section 2.3.1.4) which is likely due to a human influence (Section 3.3.2.3).

The flux of freshwater between the ocean and atmosphere is determined by the difference between precipitation and evaporation (P–E). Evaporation is measured in very few locations across the global ocean, so that directly assessing P–E over the ocean is very challenging and relies on indirect reanalysis estimates (Robertson et al., 2020). The AR5 presented robust evidence ofan amplifiedoceanic pattern in P–E since the 1960s from both regional and global surface and subsurface salinity measurements and reanalyses. This pattern is consistent with our theoretical understanding of human-induced changes in the water cycle, leading to the conclusion that these changes are very likely the result of anthropogenic forcings (Section 9.2.2.2).

In contrast, AR5 did not provide a conclusive assessment of observed changes in P–E over land. Continental P–E estimated from reanalyses and data-driven land surface models indicate that interannual variations are linked to ENSO (Robertson et al., 2014, 2020). Increasing trends in P–E since 1979 based on land models are not statistically significant. Observations and models show evidence that P–E increases in the wet parts and decreases in the dry parts of tropical circulation systems, which shift in location seasonally and from year to year, with increases in seasonality since 1979 (see Box 8.2; Chou et al., 2013; Liu and Allan, 2013; Fu and Feng, 2014).

In summary, a low signal-to-noise ratio, observational uncertainties and current data assimilation techniques limit the assessment of recent global trends in P–E over both land and ocean. It is likely that the global land P–E variations observed since the late 1970s were dominated by internal variability, mostly linked to ENSO teleconnections (medium confidence). In contrast, the attribution of changes in sea surface salinity (Section 3.5.2.2) suggests that it is extremely likely that human influence has contributed to the regional changes in P–E observed over the global ocean since the mid-20th century.

The AR5 presented evidence of increases in global near-surface and tropospheric specific humidity since the 1970s but with medium confidence of a slowing of near-surface moistening trends over land associated with reduced relative humidity since the late 1990s. According to AR5, radiosonde, Global Positioning System (GPS) and satellite observations of tropospheric water vapour indicate very likely increases at near global scales since the 1970s occurring at a rate that is generally consistent with the Clausius–Clapeyron relation (about 7% °C^–1 at low altitudes) and the observed atmospheric warming (Hartmann et al., 2013).

Since AR5, it is very likely that increases in global atmospheric water vapour were observed based on in situ, satellite and reanalysis data (with medium confidence in the magnitude; Section 2.3.1.3). Satellite records show increases in upper tropospheric water vapour (constant relative humidity while temperatures have increased) since 1979 (E.-S. Chung et al. , 2014; Blunden and Arndt, 2020), to which human influence has likely contributed (Section 3.3.2.2). Combined satellite and reanalysis estimates and CMIP6 atmosphere-only simulations (1988–2014) show global mean precipitable water vapour increases of 6.7 ± 0.3 % °C^–1, very close to the Clausius–Clapeyron rate (Allan et al., 2020). Satellite-based products show increases close to the Clausius–Clapeyron rate over the ice-free oceans (about 7 to 9 % °C^–1; 1998–2008), but reanalysis estimates outside this range (Schröder et al., 2019) are an expected consequence of their changing observing systems (Allan et al., 2014; Parracho et al., 2018). Increases in precipitable water vapour are found over the central and sub-Arctic based on multiple reanalyses with some corroboration from sparse, in situ data (Vihma et al., 2016; Rinke et al., 2019; Nygård et al., 2020).

Declining near-surface relative humidity over land areas (e.g., the USA, Mediterranean, South Asia, South America and southern Africa) is evident in surface observations (Willett et al., 2014, 2020; Dunn et al., 2017). This is consistent with a faster rate of warming over land than ocean (Sections 2.3.1.3 and 8.2.2.1; Byrne and O’Gorman, 2018). CMIP5 simulations underestimate the observed decreases in relative humidity over much of global land during 1979–2015 (Douville and Plazzotta, 2017; Dunn et al., 2017)even when observed SSTs are prescribed (–0.05 to –0.25% per decade compared with an observed rate of –0.4 to –0.8% per decade). It is not yet clear if this discrepancy is related to internal variability or can be explained by deficiencies in models (Vannière et al., 2019; Douville et al., 2020)or observations (Willett et al., 2014). Over the NH mid-latitude continents, there is medium confidence that human influence has contributed to a decrease in near-surface relative humidity in summer(Sections 2.3.1.3 and 3.3.2.3).

Water vapour transport (or convergence) estimates from observations have substantial uncertainties even in regions of high quality radiosonde data. Consequently many studies use reanalyses for water transport estimates instead of instrumental observations. For example, increases in low-level (800–1000 hPa) moisture convergence into the tropical wet regime with a smaller outflow increase in the mid-troposphere (400–800 hPa) with warming was detected in one reanalysis (ERA-Interim; Allan et al., 2014). Modelling evidence combined with statistical analysis demonstrate consistency between reanalysis moisture convergence and P–E over land (Robertson et al., 2016). Advances in reanalysis representation of atmospheric moisture and winds in addition to new observational isotope analysis have improved the ability to identify the main sources of water vapour for key continental regions and quantify the relative contributions from moisture advection and recycling (Gimeno et al. , 2012; van Der Ent et al. , 2014; Joseph et al., 2016).

Observed changes in moisture transport can also arise from changes in atmospheric circulation as well as thermodynamics. For instance, moisture transport into the Arctic region estimated from reanalyses datasets is consistent with radiosonde data (Dufour et al., 2016), with increases since 1979 linked to atmospheric circulation (Nygård et al., 2020). Moisture transport into the Eurasian Arctic was identified to increase by 2.6% per decade during 1948–2008 based on a reanalysis estimate (X. Zhang et al., 2013). More intense moist intrusions associated with atmospheric rivers affecting the Arctic and Europe have been documented since 1979, but with a substantial influence from decadal internal variability (Ummenhofer et al., 2017; Mattingly et al., 2018). A recent strengthening of tropical circulation and associated moisture convergence has been identified since around 2000 for the Amazon region (Arias et al. , 2015; Barichivich et al. , 2018; J.C. Espinoza et al. , 2018; X.Y. Wang et al. , 2018). This was also strenghtened by increased moisture transport from the North Atlantic, driving more abundant latent heat release (Segura et al., 2020) and leading to an increased frequency of extreme floods in the northern Amazon (Barichivich et al., 2018; Heerspink et al., 2020). Overall, increased moisture transport has been linked to increased precipitation over wet tropical land areas (Gimeno et al., 2020) and to more extreme and persistent wet and dry weather events (Konapala et al., 2020) in many regions worldwide.

In summary, there is high confidence that human-caused global warming has led to an overall increase in water vapour and moisture transport throughout the troposphere, at least since the mid-1990s. In particular, there is high confidence that moisture transport into the Arctic has increased but only medium confidence in the attribution of such a trend to a human influence. There is medium confidence that human influence has contributed to a decrease in near-surface relative humidity over the Northern Hemisphere mid-latitude continents during summer(see also Sections 2.3.1.3 and 3.3.2.3).

This section assesses observed changes in precipitation at global and regional scales. Note that changes in precipitation seasonality are assessed in Box 8.2 and that changes in regional monsoons are assessed in section 8.3.2.4 where observed changes in both circulation and rainfall are considered. Further assessment of regional changes in precipitation is presented in Chapters 10, 12 and Atlas, while extreme precipitation is presented in Chapter 11.

The AR5 concluded that it is likely there has been an overall increase in annual mean precipitation amount over mid-latitude land areas in the NH, with low confidence since 1901, butmedium confidence after 1951. There is further evidence of a faster increase since the 1980s (medium confidence) (Sections 2.3.1.3.4 and 3.3.2.2). Precipitation has increased from 1950 to 2018 over mid-high latitude Eurasia, most of North America, south-eastern South America, and north-western Australia, while it has decreased over most of Africa, eastern Australia, the Mediterranean region, the Middle East, and parts of East Asia, central South America, and the Pacific coasts of Canada, as simulated by the CMIP5 multi-ensemble mean (Dai, 2021). Since AR5, there have been updates of several precipitation datasets, including satellite estimates, reanalysis and merged products (Adler et al., 2017; Roca, 2019). However, observational uncertainties remain an issue for assessing regional trends in seasonal or annual mean precipitation amount (Hegerl et al. , 2015; Maidment et al. , 2015; Sarojini et al. , 2016; Beck et al. , 2017), as well as the convective and stratiform types of precipitation (e.g., Ye et al., 2017). Precipitation trends at regional scales are dominated by internal variability across much of the world (Knutson and Zeng, 2018). Regional changes in precipitation amounts can also be obscured by contrasting responses to GHG compared with aerosol forcings (Wu et al., 2013; Hegerl et al., 2015; Xie et al., 2016; Zhao and Suzuki, 2019; Zhao et al., 2020) and changes in precipitation intensity versus frequency (Shang et al., 2019).

Global and regional changes in precipitation frequency and intensity have been observed over recent decades. An analysis of 1875 rain gauge records worldwide over the period 1961–2018 indicates that there has been a general increase in the probability of precipitation exceeding 50 mm day^–1, mostly due to an overall boost in rain intensity (Benestad et al., 2019). Such changes in precipitation intensity and frequency have not been formally attributed to human activities, but are consistent with the heating effect of increasing CO₂ levels on the distribution of daily precipitation rates (Section 8.2.3.2) and with a distinct overall intensification of heavy precipitation events found in both observations and CMIP5 models, though with an underestimated magnitude (Fischer and Knutti, 2014). Beyond amplified precipitation extremes (Section 11.4.2), CMIP5 models also indicate that anthropogenic forcings have increased temporal variability of annual precipitation amount over land from 1950 to 2005, which is most pronounced in annual mean daily precipitation intensity (Konapala et al., 2017).

Anthropogenic aerosols can alter precipitation intensities both through radiative and microphysical effects (Box 8.1 and Section 8.5.1.1.2). Precipitation suppression through aerosol microphysical effects has been observed in shallow cloud regimes over South America and the south-eastern Atlantic, associated with local biomass burning (Andreae et al., 2004; Costantino and Bréon, 2010), and in industrial regions in Australia (Rosenfeld, 2000; Hewson et al., 2013; Heinzeller et al., 2016). In contrast, precipitation intensification through aerosol microphysical effects in deep convective clouds is seen in many regions such as the Amazon, southern USA, India, and Korea. This is associated with anthropogenic aerosols from cities (Hewson et al., 2013; Fan et al., 2018; S.S. Lee et al., 2018; Sarangi et al., 2018).

In the tropics, increases in precipitation amount are observed in convergence zones and decreases in the descending branches of the atmospheric circulation since 1979 (Chou et al., 2013; Liu and Allan, 2013; Gu et al., 2016; Polson et al., 2016; Polson and Hegerl, 2017), consistent with increased moisture transports with warming (Gimeno et al., 2020). Over tropical land areas, there is substantial variability in the ‘wet convergent regimes get wetter’ and ‘dry divergent regimes get drier’ pattern of trends observed since 1950 that are modulated by decadal changes in ENSO (Liu and Allan, 2013; Gu and Adler, 2018). CMIP6 models indicate an increased contrast between wet and dry regions in the tropics and subtropics (Figure 8.7; Schurer et al., 2020). This provides further evidence that rainfall has increased in wet regimes, and slightly decreased in dry regimes over the period 1988–2019 (Figure 3.14). This greater contrast is primarily attributable to greenhouse gas forcings, although the observed trends are statistically larger than the model responses (Section 3.3.2.3).

Over the African continent, there are distinct precipitation trends observed in multiple datasets since the 1980s (Figure 8.7; Maidment et al., 2015; P. Nguyen et al., 2018). Increases in intense convective storms affecting the Sahel have been attributed to increased land –ocean temperature gradients (Taylor et al., 2017), enhanced by intense heating of the Sahara (Dong and Sutton, 2015) rather than thermodynamics (Section 8.2.2). Changes in Sahel rainfall, with reduced precipitation amounts from the 1960s to the 1980s and a subsequent recovery, are assessed in Sections 8.3.2.4.3 and 10.4.2.1. In eastern Africa, decreasing precipitation amount (−2 to −7 % per decade for 1983–2010) was reported for the March to May ‘long rains’ season (Lyon and Dewitt, 2012; Viste et al., 2013; Liebmann et al., 2014; Maidment et al., 2015; Rowell et al., 2015) and evidence of a recovery since, with internal variability playing a large role in these decadal changes (Wainwright et al., 2019). In contrast, the second ‘short rains’ season in eastern Africa (October to December) does not exhibit significant precipitation trends (Rowell et al., 2015). Increases in annual southern African rainfall of 6–7% per decade during 1983–2010 are linked with the Pacific Decadal Oscillation (PDO; Maidment et al., 2015).

Figure 8.7 | Linear trends in annual mean precipitation (mm day^–1per decade) for 1901–1984 (left) and 1985–2014 (right): (a, e) observational dataset, and the CMIP6 multi-model ensemble mean historical simulations driven by: (b, f) all radiative forcings; (c, g) GHG-only radiative forcings; (d, h) aerosol-only radiative forcings experiment. Colour shades without grey cross correspond to the regions exceeding 10% significant level. Grey crosses correspond to the regions not reaching the 10% statistically significant level. Nine CMIP6-DAMIP models have been used having at least three members. The ensemble mean is weighted per each model on the available and used members. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

(Section 8.3.1.6 assesses changes in precipitation over the Mediterranean region and its connection with drought and aridity.

Rainfall increases have been observed over northern Australia since the 1950s, with most of the increases occurring in the north-west (Dey et al., 2019a, b; Dai, 2021) and decreases observed in the north-east (J. Li et al., 2012) since the 1970s. In contrast, there has been a decline in rainfall over southern Australia related to changes in the intensification and position of the subtropical ridge (CSIRO and BoM, 2015) and anthropogenic effects (Knutson and Zeng, 2018). The drying trend over south-west Australia is most pronounced during May to July, where rainfall has declined by 20% below the 1900–1969 average since 1970 and by about 28% since 2000 (BoM and CSIRO, 2020).

Over South America, there is observational and paleoclimate evidence of declining precipitation amount during the past 50 years over the Altiplano and central Chile, primarily explained by the PDO but with at least 25% of the decline attributed to anthropogenic influence (Morales et al., 2012; Neukom et al., 2015; Boisier et al., 2016; Seager et al., 2019b; Garreaud et al., 2020). In contrast, a significant rainfall increase has been detected over the Peruvian–Bolivian Altiplano (from observational data and satellite-based estimations) since the 1980s (Figure 8.7; Imfeld et al., 2020; Segura et al., 2020). Long-term (1902–2005) precipitation data indicate positive trends over south-eastern South America and negative trends over the southern Andes, with at least a partial contribution from anthropogenic forcing (Gonzalez et al., 2014; Vera and Díaz, 2015; Díaz and Vera, 2017; Boisier et al., 2018; Knutson and Zeng, 2018; see further assessment in Section 10.4.2.2 and Atlas.7.2.2). The Peruvian Amazon has exhibited significant rainfall decreases during the dry season since 1980 (Lavado et al., 2013; Ronchail et al., 2018). Increases in wet season rainfall in the northern and central Amazon since the 1980s and decreases during the dry season in the southern Amazon (Barreiro et al., 2014; Gloor et al., 2015; Martín-Gómez and Barreiro, 2016; J.C. Espinoza et al., 2018; X.Y. Wang et al., 2018; Haghtalab et al., 2020) are not explained by radiative forcing based on CMIP6 experiments (Figure 8.7) and trends are insignificant over longer periods since 1930 (Kumar et al., 2013)or more recently, since 1973 (Almeida et al., 2017). See (Section 8.3.2.4.5 for monsoon-related changes. For the tropical Andes region, trends in annual precipitation show heterogenous patterns, ranging between –4% per decade and +4% per decade in the northern and southern tropical Andes for a 30-year period at the end of the 20th century, although increases during 1965–1984 and decreases since 1984 have been registered in Bolivia (Carmona and Poveda, 2014; Pabón-Caicedo et al., 2020).

Over China, annual precipitation totals changed little from 1973 to 2016, but precipitation intensity significantly increased at a rate of 0.12 mm day^–1 per decade, while the number of days with precipitation exceeding 0.1 mm day^–1significantly decreased at a rate of 0.9 days per decade (Shang et al., 2019). There is consistency in trend estimates during 1998–2015 over mainland China among satellite-based products and station data, which show increased precipitation amounts in autumn and winter and decreases in summer (Chen and Gao, 2018), consistent with a decreased intensity of East Asian monsoon precipitation (Lin et al., 2014; Deng et al., 2018). Further assessment of precipitation changes over the South and South East Asian and the East Asian monsoon regions is presented in Section 8.3.2.4. An increasing trend in the frequency of heavy rainfall occurrences at the expense of low and moderate rainfall occurrences is found over central India (Krishnan et al., 2016; Roxy et al., 2017) and over eastern China with the latter due to increasing high aerosol levels (Qian et al., 2009; J. Guo et al., 2017; Xu et al., 2017; Day et al., 2018), consistent with the effects of absorbing aerosol on stability and convective inhibition (Box 8.1).

Observed precipitation records since the early 1900s show increases in precipitation totals over central and north-eastern North America that are attributable to anthropogenic warming but larger in magnitude than found in CMIP5 simulations (Knutson and Zeng, 2018; Guo et al., 2019). Decreases in precipitation amount over the central and south-western USA and increases over the north-central USA during 1983–2015 (Cui et al., 2017; P. Nguyen et al., 2018), are not clearly associated with forced responses in CMIP6 simulations (Figure 8.7; see also Section 10.4.2.3). Over Europe, precipitation trends since 1979 do not show coherence across datasets (Zolina et al., 2014; P. Nguyen et al., 2018). Longer records since 1910 show increases for much of Scandinavia, north-western Russia, and parts of north-western Europe/United Kingdom and Iceland (Knutson and Zeng, 2018). Records since 1930 show increases of annual preciptation amount over western Russia (see also Atlas.8.2). Widespread increases in daily precipitation intensity appear clearly over regions with a high density of rain gauges, such as Europe and North America over the 1951–2014 period (Alexander, 2016). Observations during 1966–2016 over northern Eurasia show increases in the contribution of heavy convective showers to total precipitation by 1–2% on average (with local trends of up to 5%) for all seasons except for winter (Chernokulsky et al., 2019). Increases in convective precipitation intensity have been identified, particularly on sub-daily time scales, using a range of modelling and observational data (Berg et al., 2013; Kanemaru et al., 2017; Pfahl et al., 2017).

Snowfall is an important component of precipitation in high-latitude and mountain watersheds. Reanalysis data indicate significant reductions in annual mean potential snowfall areas over NH land by 0.52 million km² per decade, with the largest decline over the Alps, with snow water equivalent reductions of about 20 mm per decade (Tamang et al., 2020). In the Tibetan Plateau, region-wide winter snowfall has increased but summer snowfall has decreased during the 1960–2014 period (Deng et al., 2017). State-of-the-art model simulations indicate reduced mean annual snowfall in the Arctic, despite the strong precipitation increase, mainly in summer and autum, when temperatures are close to the melting point (Bintanja and Andry, 2017).

In summary, regional changes in precipitation amounts can be obscured by the contrasting responses to GHG and aerosol forcings across much of the 20th century and can thus be dominated by internal variability at decadal to multi-decadal time scales (high confidence). There is, however, a detectable increase in northern high-latitude annual precipitation over land which has been primarily driven by human-induced global warming (high confidence) (Section 3.3.2). Human influence has strengthened the zonal mean precipitation contrast between the wet tropics and dry subtropics since the 1980s (medium confidence), although regional studies suggest a more complex precipitation response to evolving anthropogenic forcings. There is high confidence that daily mean precipitation intensities have increased since the mid-20th century in a majority of land regions with available observations and it is likely that such an increase is mainly due to GHG forcing (see Section 11.4). Section 8.3.2.4 assesses monsoon precipitation changes in detail.

The AR5 assessed that there was medium confidence that pan evaporation declined in most regions over the last 50 years, yetmedium confidence that evapotranspiration increased from the early 1980s to the late 1990s. Since AR5, these conflicting observations have been attributed to internal variability and by the fact that evapotranspiration is less sensitive to trends in wind speed and is partly controlled by vegetation greening (K. Zhang et al., 2015; Y. Zhang et al., 2016; Z. Zeng et al., 2018b). Observation-based estimates show a robust positive trend in global terrestrial evapotranspiration between the early 1980s and the early 2010s (Miralles et al., 2014b; Z. Zeng et al., 2014, 2018b; K. Zhang et al., 2015; Y. Zhang et al., 2016). The rate of increase varies among datasets, with an ensemble mean terrestrial average rate of 7.6 ± 1.3 mm yr^–1 per decade for 1882–2011 (Z. Zeng et al., 2018a). In addition, a decreasing trend in pan evaporation plateaued or reversed after the mid-1990s (C.M. Stephens et al., 2018) has been reported as due to a shift from a dominant influence of wind speed to a dominant effect of water vapour pressure deficit, which has increased sharply since the 1990s (Yuan et al., 2019). The absence of a trend in evapotranspiration in the decade following 1998 was shown to be at least partly an episodic phenomenon associated with ENSO variability (Miralles et al. , 2014b; K. Zhang et al. , 2015; Martens et al. , 2018). Thus, there is medium confidence that the apparent pause in the increase in global evapotranspiration from 1998 to 2008 is mostly due to internal variability. In contrast to AR5, there are now consistent trends in pan evaporation and evapotranspiration at the global scale, given the recent increase in both variables since the mid-1990s (medium confidence). Given the growing number of quantitative studies, there is high confidence that global terrestrial annual evapotranspiration has increased since the early 1980s.

Since AR5, the predominant contribution of transpiration to the observed trends in terrestrial evapotranspiration has been revisited and confirmed (Good et al., 2015; Wei et al., 2017). Using satellite and ecosystem models, Zhu et al. (2016) found a positive trend in leaf area index during 1982–2009, indicating that greening could contribute to the observed positive trend of evapotranspiration, in line withsimilar studies that focused on the 1981–2012(Y. Zhanget al. , 2016) and1982–2013(K. Zhanget al. , 2015) periods. Zeng et al. (2018) determined that the 8% global increase in satellite-observed leaf area index between the 1980s and the 2010s may explain an increase in evapotranspiration of 12.0 ± 2.4 mm yr^–1 (about 55 ± 25% of the total observed increase). Forzieri et al. (2020) estimated that the recent increase in leaf area index led to 3.66 ± 0.45 W m^–2 in latent heat flux (about 51 ± 6 mm yr^–1) and that the sensitivity of energy fluxes to leaf area index increased by about 20% over the 1982–2016 period. Overall, there is medium confidence that greening has contributed to the global increase in evapotranspiration since the 1980s.

Plant water use efficiency(WUE)is expected to rise with CO₂levels (high confidence) (Section 8.2.3.3 and Box 5.2), andcan in theory counteract rising evapotranspiration in a warmer atmosphere (Section 8.2.3.3). However, observational studies suggest that this may not be the case in some ecosystems. For example, Frank et al. (2015) found that while the WUE increased in European forests across the 20th century, transpiration also increased due to more plant growth, a lengthened growing season, and increased evaporative demand. Likewise Guerrieri et al. (2019) observed that while WUE and photosynthesis increased in North American forests, stomatal conductance experienced only modest declines that were restricted to moisture-limited forests. Other studies further suggest that in many ecosystems increased WUE will not compensate for increased plant growth, amplifying declines in surface water availability(De Kauweet al. , 2013; Ukkolaet al. , 2016b; A. Singhet al. , 2020), while drought conditions can also offset the CO₂ fertilization effect and lead to a decline in WUE (N. Liu et al., 2020). There is low confidence regarding the impact of plant physiological effects on observed trends in evapotranspiration.

An increasing number of studies have identified signals of attribution in the recent observed trends in evapotranspiration. Douville et al. (2013) found that the post-1960 rise in evapotranspiration in both the mid-latitudes and northern high latitudes was related to anthropogenic radiative forcing. An analysis of CMIP5simulations suggests that anthropogenic forcing accounts for a large fraction of the global mean evapotranspiration trend from 1982 to 2010 (Dong and Dai, 2017). Padrón et al. (2020) determined that increases in evapotranspiration were responsible for the majority of the anthropogenic pattern in dry-season water availability that dominates global trends since 1984. These findings are further supported by CMIP6 model results (Figure 8.8) that show that the recent summer increase in evapotranspiration in the northern mid- and high latitudes is due to GHG forcing and decreasing anthropogenic aerosol emissions over Europe.

Figure 8.8 | Linear trends in annual mean evapotranspiration (mm day^–1 per decade) for1901–1984 (left) and 1985–2014 (right): (a, e) Land Model Intercomparison Project (LMIP) and observational dataset, and the CMIP6 multi-model ensemble mean historical simulations driven by (b, f) all radiative forcings, (c, g) GHG-only radiative forcings, (d, h) aerosol-only radiative forcings experiment. Colour shade without grey cross correspond to the regions exceeding 10% significant level. Grey crosses correspond to the regions not reaching the 10% statistically significant level. Nine CMIP6-DAMIP models have been used having at least three members. The ensemble mean is weighted per each model on the available and used members. The Global Land Data Assimilation System (GLDAS) was not available over the early 20th century so was replaced by a multi-model off-line reconstruction, LMIP, which is consistent with GLDAS over the recent period but may be less reliable over the early 20th century given larger uncertainties in the atmospheric forcings. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

In summary, there is high confidence that terrestrial evapotranspiration has increased since the 1980s. There is medium confidence that this trend is driven by both increasing atmospheric water demand and vegetation greening, and high confidence that it can be partly attributed to anthropogenic forcing. There is low confidence about the extent to which increases in plant water use efficiency have influenced observed changes in evapotranspiration.

The AR5 reported low confidence in the assessment of trends in global river discharge during the 20th century. This is because many streamflow observations have been impacted by land use and dam construction, and the largest river basins worldwide differ in many characteristics, including geography and morphology. In regions with seasonal snow storage, AR5 WGII assessed that there is robust evidence and high agreement that warming has led to earlier spring discharge maxima and robust evidence of earlier breakup of Arctic river ice, as well as indications that warming has led to increased winter flows and decreased summer flows where streamflows are lower and that the observed increases in extreme precipitation led to greater probability of flooding at regional scales with medium confidence. The SROCC found robust evidence and high agreement that discharge due to melting glaciers has already reached its maximum point and has begun declining with smaller glaciers, but onlylow confidence that anthropogenic climate change has already affected the frequency and magnitude of floods at the global scale.

Significant trends in streamflow and continental runoff were observed in 55 out of 200 large river basins during 1948–2012, with an even distribution of increasing and decreasing trends (Section 2.3.1.3.6; Dai, 2016). A global detection and attribution study shows that the simulation of spatially heterogeneous historical trends in streamflow is consistent with observed trends only if anthropogenic forcings are considered (Gudmundsson et al., 2019). Section 3.3.2.4 assesses with medium confidence that anthropogenic climate change has altered regional and local streamflows, although a significant trend has not been observed in the global average (Sections 2.3.1.3.6 and 3.3.2.3). Multiple human-induced and natural drivers have been shown to play an important but variable role in observed regional trends of streamflow for several different areas (Fenta et al. , 2017; Ficklin et al. , 2018; Glas et al. , 2019; Vicente-Serrano et al. , 2019). For instance, decreasing runoff during the dry season has been observed over the Peruvian Amazon since the 1980s (Lavado et al., 2013; Ronchail et al., 2018). Up to 30–50% of the recent multi-decadal decline in streamflow across the Colorado River Basin can be attributed to anthropogenic warming and its impacts on snow and evapotranspiration (Woodhouse et al., 2016; McCabe et al., 2017; Udall and Overpeck, 2017; Xiao et al., 2018; Milly and Dunne, 2020). In the Upper Missouri River basin, Martin et al. (2020) found that warming temperatures have contributed to streamflow reductions since at least the late 20th century. Cold regions in the NH have experienced an earlier occurrence of snowmelt floods, an overall increase in water availability and streamflow during winter, and a decrease in water availability and streamflow during the warm season (Aygün et al., 2019).

Some studies have suggested that dam construction and water withdrawals can be the dominant drivers in observed trends in streamflow amount (Wada et al., 2013). Regionally, land-use and land cover changes have been identified as important factors for streamflow (H. Chen et al., 2020). The impact of surface dimming from aerosol emissions on evaporation was identified as a discernible influence in NH streamflows (Gedney et al., 2014). While changes in annual mean streamflow present a complicated picture, recent studies of changes in the timing of streamflow in snow-influenced basins continue to support a prominent influence from warming (Kang et al., 2016; Dudley et al., 2017; Kam et al., 2018). Global land runoff variations correlate significantly with ENSO variability (Miralles et al., 2014b; Schubert et al., 2016).

Observed changes in flooding are assessed in Section 11.5.2 and are summarized as follows. For changes in the magnitude of peak flow, recent studies show strong spatial heterogeneity in the sign, size and significance of trends. For changes in timing of peak flows, recent studies further support observed changes in snowmelt-driven rivers. Observed changes in runoff and flood magnitude cannot be explained by precipitation changes alone given the possible season- and region-dependent decreases in antecedent soil moisture and snowmelt, which can partly offset the increase in precipitation intensity (Sharma et al., 2018), or the expected effect of urbanization and deforestation which can, on the contrary, amplify the runoff response (Chen et al., 2017; Abbott et al., 2019; Cavalcante et al., 2019). Simulations of mean and extreme river flows are consistent with the observations only when anthropogenic radiative forcing is considered (Gudmundsson et al., 2021).

In summary, the assessment of observed trends in the magnitude of runoff, streamflow, and flooding remains challenging, due to the spatial heterogeneity of the signal and to multiple drivers. There is, however, high confidence that the amount and seasonality of peak flows have changed in snowmelt-driven rivers due to warming. There is also high confidence that land-use change, water management and water withdrawals have altered the amount, seasonality, and variability of river discharge, especially in small and human-dominated catchments.

The AR5 reported low confidence that changes in drought since the mid-20th century could be attributed to human influence, owing to observational uncertainties and difficulties in distinguishing decadal-scale variability from long-term trends. Changes in soil moisture, a metric of aridity, were not assessed thoroughly in AR5. Since AR5, new satellite products, land surface reanalyses, and land surface models have been used to document recent changes in soil moisture at the global scale. The science of detection and attribution has also progressed considerably (Trenberth et al., 2015; Easterling et al., 2016; Stott et al., 2016). Attribution efforts have further benefited from the increased use of paleoclimate information, which provides an important constraint on natural variability that is insufficiently sampled by short observational record (Cook et al., 2018; Kageyama et al., 2018).

Several studies have identified a persistent ‘fingerprint’ of anthropogenic forcing in global trends in aridity spanning the last 120 years. Using a combination of tree ring data, CMIP5 model simulations, and reanalysis products, Marvel et al. (2019) determined that the dominant trend in aridity since 1900, characterized by drying in North and Central America and the Mediterranean, is detectable and attributable to external forcing from 1900 to 1949. This trend weakens from 1950 to 1975, possibly due to aerosol forcing (Marvel et al., 2019), but then emerges again from 1981 to present, although it is not detectable in the GLEAM nor MERRA-2 soil moisture reanalysis products. Likewise, Bonfils et al. (2020) investigated changes in precipitation, temperature and continental aridity in CMIP5 historical simulations and found that the dominant multivariate fingerprint, an amplification of wet–dry latitudinal patterns and progressive continental aridification, was associated with greenhouse gas emissions (Figure 8.9a, d), and the second leading fingerprint was associated with anthropogenic aerosols (Figure 8.9e, h). This study found that the anthropogenic greenhouse gas signal is statistically detectable in reanalyses over the 1950–2014 period (signal-to-noise ratio above 1.96). Gu et al. (2019) found that a global trend in declining soil moisture is detectable in the GLDAS-2 reanalysis product and is attributable to greenhouse gas forcing. Padrón et al. (2020) reconstructed the global patterns of dry season water availability from 1902–2014, and found itextremely likely (99% range) that trends in the last three decades of the analysis period could be attributed to anthropogenic forcing, mainly due to increases in evapotranspiration. It is very likely (>90% range) that anthropogenic forcing has affected global patterns of soil moisture over the 20th century.

Figure 8.9 | Spatial expressions (a–c, e–g) of the leading multivariate fingerprints of temperature (°C), precipitation (mm day^–1), and aridity (CMI; the Climate Moisture Index) in CMIP5 historical simulations and the corresponding temporal evolution in both CMIP5 and reanalysis products (d, h).

The first leading fingerprint is associated with greenhouse gas forcing (a–d) and the second leading fingerprint is associated with aerosol forcing (e–h). CMI is a dimensionless aridity indicator that combines precipitation and atmospheric evaporative demand. Figure after Bonfils et al. (2020). Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

On a regional scale, the robustness of trend attribution for drought and aridity varies widely. Key trends and their attributions are summarized here, while a complete regional assessment of observed trends in drought and aridity is in Chapter 11 (Sections 11.6.2, 12.3.2 and 12.4).

Several studies have analyzed CMIP5 and land surface models and detected a significant summer drying trend in the NH across the late 20th century that is attributable to anthropogenic forcings (Mueller and Zhang, 2016; Douville and Plazzotta, 2017). This trend is mainly driven by dryland areas such as the western USA and west-central Asia, where both reanalysis products and satellite data confirm there has been a persistent decline in soil moisture since 1990 (Y. Liu et al., 2019a). In the western USA, snow deficits have very likely contributed to recent drying (Mote et al., 2018). Spring snow water equivalent across the Sierra Nevada Mountains reached a record low in 2015 (Margulis et al., 2016; Mote et al., 2016), possibly the lowest of the last five hundred years (Belmecheri et al., 2016). Over the longer California drought (2011–2015) anthropogenic warming alone reduced snowpack levels in the Sierras by 25% (Berg and Hall, 2017). The north-western USA also experienced snow drought in 2015, despite near-normal levels of total cold season precipitation (Mote et al., 2016; Marlier et al., 2017). There is high confidence that anthropogenic warming contributed to these recent snow droughts (Belmecheri et al., 2016; Mote et al., 2016).

In the western USA, anthropogenic warming is amplifying drought and aridity by increasing evaporative demand and water loss to the atmosphere (Weiss et al., 2009; Overpeck, 2013; Cook et al., 2014; Griffin and Anchukaitis, 2014; Williams et al., 2020). For the California drought between 2012­–2014, Griffin and Anchukaitis (2014) used paleoclimate reconstructions to determine that while rainfall deficits were not unprecedented, record-high temperatures drove an exceptional decline in soil moisture relative to the last millennium. Williams et al. (2015) concluded that anthropogenic warming accounted for 8–27% of these soil moisture deficits. Robeson (2015) estimated that the California drought was a 1-in-10,000 year event. Tree ring reconstructions indicate that prolonged megadroughts have occurred in the western USA throughout the last 1200 years (Cook et al. , 2004, 2010; B.I. Cook et al. , 2015), forced by internal variability (Coats et al., 2016; Cook et al., 2016b). However, Williams et al. (2020) determined that 2000–2018 drought across the south-western USA was the second driest 19-year period since 800 CE, and attributed nearly half the magnitude of this event to anthropogenic forcing (see also Section 10.4.2.3). Evidence for human signals in drought can also be found in western North American streamflow records, as noted above in Section 8.3.1.5. There is high confidence that anthropogenic forcing has contributed to recent droughts and drying trends in western North America.

Large areas of east-central Asia experienced drying in the early 2000s as a result of warmer temperatures, lower humidity, and declining soil moisture(Wei and Wang, 2013; Z. Liet al. , 2017; Hesslet al. , 2018). Paleoclimate data from the Mongolian plateau suggest that this recent central Asian drought exceeds the 900-year return interval, but is not unprecedented in the last 2060 years (Hessl et al., 2018). There is low confidence due to limited evidence that recent droughts in central Asia can be attributed to anthropogenic forcing.

The Mediterranean region has experienced notable changes in drought and aridity. A number of studies have identified a decline in precipitation since 1960 and attributed this to anthropogenic forcing (Hoerling et al., 2012; Gudmundsson and Seneviratne, 2016; Knutson and Zeng, 2018; Seager et al., 2019b). Kelley et al. (2015) showed that climate change caused a three-fold increase in the likelihood of the 2007–2010 meteorological drought in the eastern Mediterranean. However, historical trends in precipitation across the Mediterranean are spatially variable and contain substantial decadal variability, such that an anthropogenic influence may not be detectable in all areas (Zittis, 2018; Vicente-Serrano et al., 2021). Records of soil moisture provide a clearer signal, indicating that higher temperatures and increased atmospheric demand have played a strong role in driving Mediterranean aridity (Vicente-Serrano et al., 2014). Hydrological modeling suggests that the recent decline in soil moisture in the Mediterranean is unprecedented in the last 250 years (Hanel et al., 2018). Paleoclimate evidence extends this view, additionally indicating that dryness in the Mediterranean is approaching an extreme condition compared to the last millennium (Markonis et al., 2018) and that the 15-year drought in the Levant (1998–2012) has an 89% likelihood of being the driest of the last 900 years (Cook et al., 2016a). Marvel et al. (2019) found that the Mediterranean region contributes strongly to the anthropogenic warming component of the global trend in aridity. There is high confidence that anthropogenic forcings are causing increased aridity and drought severity in the Mediterranean region.

Both central and north-eastern Africa have experienced a decline in rainfall since about 1980 (high confidence) (Lyon and Dewitt, 2012; Lyon, 2014; Hua et al., 2016; Nicholson, 2017). In Central Africa, the decline has been attributed to atmospheric responses to Indo-Pacific sea surface temperature variability (Hua et al., 2018). In north-eastern Africa, droughts have become longer and more intense in recent decades, continuing across rainy seasons (Hoell et al., 2017b; Nicholson, 2017), and this trend appears to be unusual in the context of the last 1500 years (Tierney et al., 2015). Knutson and Zeng (2018) attribute decreased annual precipitation over the Sudan to anthropogenic forcing, but other studies argue that the recent trend cannot yet be distinguished from natural variability, at least over parts of this region (Hoell et al., 2017b; Philip et al., 2018). There remains low confidence due to limited evidence that drying the north-eastern Africa is attributable to human influence. In the Western Cape region of South Africa, human influence increased the likelihood of the severe 2015–2017 drought by a factor of 3–6, depending on the analysis (Otto et al., 2018; Pascale et al., 2020). Anthropogenic forcing also contributed to the 2018 drought, mainly by increasing evapotranspiration (Nangombe et al., 2020). While some analysis of instrumental precipitation data in this region detect a slight long-term drying trend consistent with the simulated anthropogenic response (Seager et al., 2019b), there is strong multi-decadal variability in the data (Wolski et al., 2021). However, a study of streamflow in southern Africa detected a significant decline (Gudmundsson et al., 2019; see also Section 10.6.2). There is medium confidence in the long-term drying trend in this region and its attribution to anthropogenic forcing, and medium confidence that anthropogenic warming has contributed to recent severe drought events.

Several subtropical, semi-arid regions in the Southern Hemisphere have experienced long-term drying trends in the late 20th century. South-western South America (central Chile) experienced a multi-decadal decline in precipitation and streamflow culminating in a post-2010 megadrought that has been partly attributed to anthropogenic GHG emissions and ozone depletion (Boisier et al. , 2016, 2018; Saurral et al. , 2017; Knutson and Zeng, 2018; Seager et al. , 2019b; Garreaud et al. , 2020). There is medium confidence that drying in central Chile can be attributed to human influence. The tree-ring paleoclimate record demonstrates that the mid-century increase in exteme drought events in southern South America is unusual in the context of the last 600 years, suggesting an emerging influence of anthropogenic forcing (Morales et al., 2020).

There has been a 20% decrease in winter (May to July) rainfall in south-western Australia since 1970, with the decline increasing to around 28% since 2000 (Delworth and Zeng, 2014; BoM and CSIRO, 2020). There has also been a significant increase in the average intensity of seasonal droughts in the region since 1911in response to both lower precipitation and increased atmospheric evaporative demand (Gallant et al., 2013). Several studies attribute the precipitation declines in south-western Australia to anthropogenic changes in GHG and ozone (Delworth and Zeng, 2014; Knutson and Zeng, 2018; Seager et al., 2019b). There is high confidence that the observed drying in south-western Australia can be attributed to anthropogenic forcing.

In south-eastern Australia, the average length of droughts have increased significantly, lasting between 10 and 69% longer than droughts during the first half of the 20th century (Gallant et al., 2013). Paleoclimate reconstructions indicate a 97.1% probability that the decadal rainfall anomaly recorded during the 1997–2009 Millennium drought in south-eastern Australia was the worst experienced since 1783 (Gergis et al., 2012), and that the spatial extent and duration of cool season (April to September) rainfall anomalies were either very much below average or unprecedented over at least the last 400 years (Freund et al., 2017). Other paleoclimate studies suggest that the Millennium drought in eastern Australia was not unusual in the context of natural variability reconstructed over the past millennium (Palmer et al. , 2015; Cook et al. , 2016c; Kiem et al. , 2020). While there is currentlylow confidence that recent droughts in eastern Australia can be clearly attributed to human influence (Cai et al., 2014; Delworth and Zeng, 2014; Rauniyar and Power, 2020), there is emerging evidence that declines in April to October rainfall in south-eastern Australia since the 1990s would not have been as large without the influence of increasing levels of atmospheric GHGs (Rauniyar and Power, 2020).

In summary, it is very likely that anthropogenic factors have influenced global trends in aridity, mainly through competing changes in evapotranspiration and/or atmospheric evaporative demand due to anthropogenic emissions of GHG and aerosols. There is high confidence that the frequency and the severity of droughts has increased over the last decades in the Mediterranean, western North America, and south-western Australia and that this can be attributed to anthropogenic warming. There is medium confidence that recent drying and severe droughts in southern Africa and south-western South America can be attributed to human influence. In some regions of western North America and the Mediterranean, paleoclimate evidence suggests that recent warming has resulted in droughts that are of similar or greater intensity than those reconstructed over the last millennium (medium confidence).

The AR5 concluded it is likely that the tropical belt, as delimited by the Hadley circulation, has widened since the 1970s. Observations in the satellite era indicate precipitation increases in the core of the Pacific Inter-tropical Convergence Zone (ITCZ) and decreases on the ITCZ margins (Gu et al., 2016; Su et al., 2017). As the satellite period has lengthened, observations have increasingly been used to assess trends in the ITCZ and tropical rain belt. Since AR5, significant narrowing and strengthening of the Pacific ITCZ after 1979 have been identified in atmospheric reanalyses (Wodzicki and Rapp, 2016), but no change in the ITCZ location (Byrne et al., 2018). Atmospheric model simulations suggest that with a narrower ITCZ, the subtropical jet becomes baroclinically unstable at a lower latitude and allows mid-latitude eddies to propagate farther equatorward (Watt-Meyer and Frierson, 2019). Observational analyses also show that the ITCZ narrowing (Zhou et al., 2020) is associated with increased precipitation in the ITCZ core region that is strongly coupled to increasing Outgoing Longwave Radiation (OLR) in the expanding dry zones, particularly over land regions in the subtropics and mid-latitudes (Lau and Tao, 2020). In addition, an eastward movement of the South Pacific Convergence Zone (SPCZ) between 1977 and 1999 has been reported, with associated significant precipitation trends in the South Pacific regions (Salinger et al., 2014).

ITCZ trends seen in satellites, precipitation measurements and reanalysis data are further supported by ocean surface-salinity observations. Long-term salinity observations show a freshening in the cores of the Atlantic and Pacific ITCZs and increased salinity on the ITCZ margins (Durack and Wijffels, 2010; Durack et al., 2012; Terray et al., 2012; Skliris et al., 2014). By investigating simultaneous changes in precipitation, temperature and continental aridity in CMIP5 historical simulations, Bonfils et al. (2020) found a secondary signal (Figure 8.9, right column) characterized by a robust inter-hemispheric temperature contrast (Section 3.3.1.1), a latitudinal shift in the ITCZ (in accordance with the theory of cross-equatorial energy transport; Section 8.2.2.2), and changes in aridity in the Sahel (Section 8.3.1.6). These forced changes are statistically detectable in reanalyses datasets over the 1950–2014 period at the 95% confidence level.

Reconstructions in the Sahel (Carré et al., 2019) and Belize (Ridley et al., 2015) support the southward displacement of the tropical rain belt since 1850 and the narrowing trend of the tropical rainbelt detected in observations (Rotstayn et al., 2002; Hwang et al., 2013). Decreasing precipitation trends in the NH during the 1950s to 1980s have been attributed to anthropogenic aerosol emissions from North America and Europe, which peaked during the late1970s and declined thereafter following improved air quality regulations, causing dimming (brightening) through reduced (increased) surface solar radiation (Box 8.1 Figure 1), in agreement with model simulations (Chiang et al., 2013; Hwang et al., 2013). This is consistent with energetic constraints where tropical precipitation shifts are anti-correlated with cross-equatorial energy transport (Section6.3.3, Box 8.1). It also provides a physical mechanism for the severe drought in the Sahel that peaked in the mid-1980s (Sections 8.3.2.4.3 and 10.4.2.1) and the southward shift of the NH tropical edge from the 1950s to the 1980s (Allen et al., 2014; Brönnimann et al., 2015). However, CMIP5 and CMIP6 models still exhibit strong biases in representing the ITCZ, such as the simulation of a double ITCZ (Oueslati and Bellon, 2015; Adam et al., 2018; Tian and Dong, 2020). The impacts of aerosols and volcanic activity on the position of the ITCZ have been investigated but changes are difficult to characterize from observations (Section 6.3.3.2; Friedman et al. , 2013; J.M. Haywood et al. , 2013; Iles and Hegerl, 2014; Colose et al. , 2016; Chung and Soden, 2017). Such systematic shifts of the ITCZ can have important regional impacts like changes in precipitation (Figure 8.9).

In summary, there is medium confidence that the tropical rain belts over the oceans have been narrowing and strengthening in recent decades, leading to increased precipitation in the ITCZ core region (Section 8.2.2.2). Decreasing precipitation trends in the NH during the 1950s–1980s have been attributed to anthropogenic aerosol emissions from North America and Europe (high confidence).

The AR5 reported low confidence in trends in the strength of the Hadley circulation (HC) due to uncertainties in reanalyses but high confidence on the widening of the tropical belt since 1979. In AR6, Chapter 2 (Section 2.3.1.4.1) states that the HC has very likely widened and strengthened since at least the 1980s, mostly in the NH (medium confidence).

The poleward shift of the HC is closely related to migration of the location of tropical cyclone trajectories in bothhemispheres (Sharmila and Walsh, 2018; Studholme and Gulev, 2018), with avery likely poleward shift over the western North Pacific Oceans since the 1940s (Section 11.7.1.2). Moreover, the Western North Pacific Subtropical High has extended westward since the 1970s, resulting in a monsoon rain band shift over China, with excessive rainfall along the middle and lower reaches of the Yangtze River valley along about 30°N over eastern China. At the same time, the effect of anthropogenic aerosols dominated the response to GHG increases over East Asia, resulting in a weakening of the East Asian summer monsoon and causing a drying trend in north-eastern China (Hu, 2003; Yu and Zhou, 2007; T. Wang et al., 2013; Z. Li et al., 2016b; Lau and Kim, 2017) and northern parts of South Asia (Section 8.3.2.4.2; Preethi et al., 2017). During 1977–2007, the precipitation variability over the eastern USA increased due to changes in the intensity and position of the western ridge of the North Atlantic Subtropical High (Li et al., 2011; Diem, 2013).

In the Southern Hemisphere (SH), the HC expansion has been associated with both the intensification and poleward shift of the subtropical high pressure belt (Nguyen et al., 2015), with consequences for precipitation amount over Africa, Australia, South America, and subtropical Pacific islands (Cai et al. , 2012; Grose et al. , 2015; Nguyen et al. , 2015; Sharmila and Walsh, 2018; McGree et al. , 2019). The subtropical ridge in Australia has intensified significantly since 1970, with marked declines observed in April to October rainfall across south-eastern and south-western Australia (Timbal and Drosdowsky, 2013).

The local tropical edges of the meridional overturning cells (as diagnosed from the horizontally divergent wind) are more closely associated with hydroclimate variations than the subtropical ridge (Staten et al., 2019). Poleward expansion of the tropical belt strongly contributes to precipitation decline in the poleward edge of the subtropics (Cai et al., 2012; Scheff and Frierson, 2012; Timbal and Drosdowsky, 2013; He and Soden, 2017; H. Nguyen et al., 2018; Tang et al., 2018), although recent modelling evidence suggests that subtropical precipitation declines are a response to direct CO₂ radiative forcing mainly over ocean, irrespective of the HC expansion (He and Soden, 2017). Both reanalyses datasets and climate model simulations suggest that the HC expansion is not associated with widespread, zonally symmetric subtropical drying over land (Schmidt and Grise, 2017).

Since AR5, an improved understanding of the key drivers of the recent HC expansion has been achieved, identifying the role of both internal variability and anthropogenic climate change. Part of the recent expansion (1979–2005) of the HC has been driven by a swing from warm to cold phase of the Pacific Decadal Variability (PDV; Meehl et al., 2016; Grise et al., 2019). The presence of large multi-decadal variability in 20th-century reanalyses means there is limited evidence on the human influence on the recent HC strengthening, yet the southward shift of the southern edge and widening of the SH HC appeared as robust features in all reanalysis datasets, and their trends have accelerated during 1979–2010 (D’Agostino and Lionello, 2017). As assessed in Section 3.3.3.1, GHG increases and stratospheric ozone depletion have contributed to the expansion of the zonal mean HC in the SH since around 1980, and the expansion of the NH HC has not exceeded the range of internal variability (medium confidence). Moreover, Antarctic ozone depletion can cause a poleward shift in the SH mid-latitude jet and HC (Sections 3.3.3 and 6.3.3.2). Further assessment of the attribution of recently observed changes in the HC extent and intensity is found in Section 3.3.3.1.

In summary, it is very likely that the recent HC expansion was associated with poleward shifts of tropical cyclone tracks over the western North Pacific Ocean since the 1940s, and of extratropical storm tracks in the SH since the 1970s. Changes to the HC in the NH may have contributed to subtropical drying and a poleward expansion of aridity during the boreal summer, but there is low confidence due to limited evidence. GHG increases and stratospheric ozone depletion have contributed to expansion of the zonal mean HC in the SH since around 1970, while the expansion of the NH HC has not exceeded the range of internal variability (medium confidence).

The AR5 concluded that the long-term weakening of the Pacific Walker circulation (WC) from the late 19th century to the 1990s has been largely offset by a recent strengthening (high confidence), though with low confidence in trends of the WC strength due to reanalysis uncertainties and large natural variability. The observed trends in the WC since 1980 are consistent with avery likely WC strengthening in the Pacific, similar to a La Niña pattern, with medium confidence in the magnitude of these changes due to differences between satellite observations and reanalyses.

The causes of the observed strengthening of the WC during 1980–2014 are not well understood due to competing influences from individual external forcings and since this strengthening is outside the range of variability simulated in coupled models (medium confidence), as assessed in Chapter 3 (Section 3.3.3.1). Recent strengthening in the WC has been linked with internal variability (Chung et al., 2019), although one study argues that it could be a response forced by GHG that models do not capture because of common sea surface temperature (SST) biases in the equatorial Pacific (Seager et al., 2019a). It could be also related to an interbasin thermostat mechanism whereby the human-induced Indian Ocean warming emerged earlier than in the tropical Pacific (L. Zhang et al., 2018) and induced a transient strengthening of the zonal sea level pressure gradient and easterly trades in the tropical Pacific (L. Zhang et al., 2019).

The weakening of the PWC observed during most of the 20th century is associated with reductions in land rainfall over the Maritime Continent during 1950–1999 (Tokinaga et al., 2012; Yoden et al., 2017). In contrast, the recent strengthening of the WC has been associated with an intensification of extreme flooding (Barichivich et al., 2018) and an increased frequency of wet days (J.C. Espinoza et al., 2016, 2018) over the north-western Amazon, increased precipitation in South America (Yim et al., 2017), reduced precipitation over eastern Africa (Williams and Funk, 2011; Lyon and Dewitt, 2012), and increased rainfall in southern Africa (Maidment et al., 2015). Internal variability has been shown to have a dominant role in the recent strengthening of the WC (Chung et al., 2019).

In summary, there is high confidence that changes in the WC are associated with changes in the water cycle over regions like the Maritime Continent, South America and Africa. It is very likely that the WC has strengthened in the Pacific since the 1980s, with medium confidence that this strengthening is within the range of internal variability.

The AR5 reported low confidence in the attribution of changes in monsoons to human influence, although a detailed attribution assessment of the observed changes in the regional monsoons was not presented.

Large human populations in the monsoon regions of the world heavily depend on freshwater supply for agriculture, water resources, industry, transport and various socio-economic activities. The effects of GHG forcing combined with water vapour feedback (R.J. Allen et al., 2015; Dong and Sutton, 2015; Evan et al., 2015; Dunning et al., 2018) and cloud feedbacks (Stephens et al., 2015; Potter et al., 2017) are fundamental to monsoon precipitation changes in a warming world. Since AR5 there has been improved understanding of precipitation changes associated with regional monsoons. Sections 2.3.1.4.2 and 3.3.3.2 provide an assessment of observed changes and attribution for the global monsoon. Here we provide an assessment of the observed changes in regional monsoons (see Annex V and Figure 8.11) and underlying causes. In AR6, the definition of regional monsoons slightly differs from AR5 and the rationale for it is provided in Annex V (see Glossary). Specific examples of regional monsoons are discussed further in Section 10.4.2, from the perspective of climate change attribution and in Section 10.6.3, from the viewpoint of constructing regional climate messages.

Figure 8.11 | Regional monsoon precipitation changes from observations and model attribution. Precipitation changes during 1951–2014 are shown as least-square linear trends in box-whisker plots (first and fourth rows) over the six regional monsoons, for example, North American monsoon (NAmerM, July–August–September, JAS), West African monsoon (WAfriM, June–July–August–September, JJAS), South and South East Asian monsoon (SAsiaM, June–July–August–September, JJAS), East Asian monsoon (EAsiaM, June–July–August), South American monsoon (SAmerM, December–January–February, DJF), Australian and Maritime Continent monsoon (AusMCM, December–January–February, DJF), and over the two land domains, for example, equatorial America (EqAmer, June–July–August, JJA) and South Africa (SAfri, December–January–February, DJF), as identified in the map shown in the middle and as described in Annex V. Precipitation changes are computed from observations and from Detection and Attribution Model Intercomparison Project (DAMIP) CMIP6 experiments over the historical period with all-forcing (ALL), GHG-only forcing (labelled GHG), Aerosol-only (AER) and Natural (NAT) forcings prescribed. Observations are based on the CRU (light green) and GPCC (light blue) datasets and the APHRODITE (light orange) dataset for SAsiaM and EAsiaM. CMIP6 simulations are taken from nine CMIP6 models contributing to DAMIP, with at least three members. Ensembles are weight-averaged for the respective model ensemble size. Observed trends are shown as coloured circles and the simulated trends from the CMIP6 multi-model experiments are shown as box-whisker plots. Precipitation anomaly time-series are shown in the second and third row. The thick black line is the multi-model ensemble-mean precipitation anomaly time-series from the ALL experiment and the grey shading shows the spread across the multi-model ensembles. An 11-year running mean has been applied on the precipitation anomaly time-series prior to calculating the multi-model ensemble mean. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

The AR5 assessed low confidence in centennial changes in tropical cyclone (TC) activity globally, and in the attribution of observed changes in TCs to anthropogenic forcing. Since AR5, there has been considerable progress in understanding the observed changes of TCs and an overall improved knowledge of the sensitivity of TCs to both GHG and aerosol forcing (Knutson et al., 2019; Sobel et al., 2019).

Although observational data limitations (Lau and Zhou, 2012) tend to limit detection of anthropogenic forced increases in TC precipitation (Knutson et al., 2019), there is medium confidence that anthropogenic forcing has contributed to observed heavy rainfall events over the USA associated with TCs (Kunkel et al., 2012) and other regions with sufficient data coverage (Section 11.7.1.2; Bindoff et al., 2013). There has been increased frequency of TC heavy rainfall events over several areas in the USA since the late 19th century that is greater than what would be expected solely from changes in US landfall frequency, suggesting the increasing role of TCs have in causing heavy rainfall events (Kunkel et al., 2010). For example, there is evidence for an anthropogenic contribution to the extreme rainfall of Hurricane Harvey in 2017 (Emanuel, 2017; Risser and Wehner, 2017; van Oldenborgh et al., 2017; Trenberth et al., 2018; S.-Y.S. Wang et al., 2018).

While TCs cause extreme local rainfall and flooding, they can be also an important contributor to annual precipitation and regional fresh water resources (Hristova-Veleva et al., 2020). Transport of moisture by TCs is an important contributor for precipitation over the coastal areas of East Asia mostly from July through October, with the TC rainfall accounting for nearly 10% to 30% of the total rainfall in the region (L. Guo et al., 2017). Local TC rainfall totals depend on rain-rate and translation speed (the speed of TC movement along the storm track) with slow TCs such as Hurricane Harvey (2017), providing a clear example of the effect of slow translation speed on local rainfall accumulation, with urbanization exacerbating the storm total rainfall and flooding (Section 11.7.1; W. Zhang et al., 2018).

In addition to evidence that rain-rates have increased, there is evidence that TC translation speed has slowed globally (Kossin, 2018) thus amplifying thermodynamic intensification of rainfall and may be linked to anthropogenic forcing (Gutmann et al., 2018). This is limited evidence however, so there is medium confidence of a detectable change in TC translation speed over the US. Since the 1900s, and there is low confidence for a global signal because of limited agreement among models and due to data heterogeneity. However, the slowdown is consistent with theoretical and modelling studies that indicate a general weakening of the tropical circulation with warming that reduces the speed of the TC system (Chauvin et al., 2017), though there is limitedobservational evidence (Sections 8.2.3.5 and 11.7.1).

In summary, there is medium confidence of an observed increase in TC precipitation intensity in regions with sufficient data coverage Robust physical understanding (Section 8.2.3.2) and detailed singular event attribution studies provide evidence that tropical cyclone rainfall has increased with a warming climate (high confidence, Section 11.7.1.4).

Stationary waves are planetary-scale waves that are approximately stable (stationary) in terms of geographic position, as opposed to propogating planetary waves, and are important both as part of the climatological general circulation and seasonal and shorter-term anomalies. They are related to surface features including land –ocean contrasts and major mountain ranges, as well as atmospheric features including the jet stream, storm tracks, and blocking, which are considered separately in the following sections. While zonal mean changes in P–E (precipitation minus evaporation) are dominated by thermodynamic effects (Section 8.2.2.1), changes in stationary waves are of key importance in understanding zonal asymmetries in the water cycle response to global warming (Wills and Schneider, 2015; Wills et al., 2019). The AR5 did not explicitly assess stationary waves, but noted changes in related circulation features such as a likely poleward shift of the Northern Hemisphere (NH) storm tracks and an increase in frequency and eastward shift in North Atlantic blocking anticyclones, although there was low confidence in the global assessment of blocking.

Since AR5, several studies have demonstrated a link between stationary wave amplitude and wet and dry extremes in several different regions of the NH (Liu et al., 2012; Coumou et al., 2014; Screen and Simmonds, 2014; Yuan et al., 2015) with changes in moisture transport playing an important role (Yuan et al., 2015). A ‘resonance mechanism’ has been proposed for an increasing amplitude of stationary waves (Petoukhov et al., 2013, 2016; Coumou et al., 2014; Kornhuber et al., 2017)and several studies have linked increasing amplitude of stationary waves to Arctic warming (Francis and Vavrus, 2012, 2015; Liuet al. , 2012; Tanget al. , 2014) as well as to global warming (Mannet al. , 2017). However, other studies have not identified an increase in stationary wave amplitude (Barnes, 2013; Screen and Simmonds, 2013a, b).

There has been considerable work on linkages (teleconnections) between Arctic warming and the mid-latitude circulation (see also Cross-Chapter Box 10.1). The limited amount of research on Southern Hemisphere (SH) stationary waves suggests changes in high-latitude, mid-tropospheric stationary waves which influence Antarctic precipitation (Turner et al., 2017) and changes in stratospheric stationary waves that are associated with ozone depletion rather than increases in GHGs (L. Wang et al., 2013). The observed climatology of NH winter stationary waves is well-represented in the CMIP5 multi-model mean (Wills et al., 2019) but individual models have important deficiencies in reproducing stationary wave variability (Lee and Black, 2013). In the SH, the observed climatology of stationary waves in CMIP5 models has considerable bias in both phase and amplitude (Garfinkel et al., 2020). A comprehensive assessment is not yet available for CMIP6 models.

In summary, there is low confidence in strengthened winter stationary wave activity over the North Atlantic, associated with increased poleward moisture fluxes east of North America There is medium confidence in a recent amplification of the NH stationary waves in summer, but no formal attribution to anthropogenic climate change.

Atmospheric blocking refers to persistent, semi-stationary weather patterns characterized by a high-pressure (anticyclonic) anomaly that interrupts the westerly flow in the mid-latitudes of both hemispheres. By redirecting the pathways of mid-latitude cyclones, blocking can affect the water cycle and lead to negative precipitation anomalies in the region of the blocking anticyclone and positive anomalies in the surrounding areas (Sousa et al., 2017). In this way, blocking can also be associated with extreme events such as heavy precipitation (Lenggenhager et al., 2019), drought (Schubert et al., 2014) and heatwaves (Miralles et al., 2014a). The AR5 reported low confidence in global-scale changes in blocking, due to methodological differences between studies.

Currently no consensus exists on observed trends in blocking during 1979–2013. (Horton et al., 2015) identified increasing trends in anticyclonic circulation regimes based on geopotential height fields in the mid-troposphere, which may be partly related to the tropospheric warming itself and thus not represent real changes in the statistics of weather (Horton et al., 2015; Woollings et al., 2018). Hanna et al. (2018) and (Davini and D’Andrea, 2020) reported a significant increase in the frequency of summer blocking over Greenland. A weakening of the zonal wind, eddy kinetic energy and amplitude of Rossby waves in summer in the NH (Coumou et al., 2015, Kornhuber et al., 2019) and an increased ‘waviness’ of the jet stream associated with Arctic warming (Francis and Vavrus, 2015; Pfahl et al., 2015; Luo et al., 2019) have also been identified, which may be linked to increased blocking.

In contrast, it has been shown that observed trends in blocking are sensitive the choice of the blocking index, and that there is a large internal variability that complicates the detection of forced trends (Barnes et al., 2014; Cattiaux et al., 2016; Woollings et al., 2018), compromising the attribution of any observed changes in blocking. Many climate models still underestimate the occurrence of blocking, at least in winter over north-eastern Atlantic and Europe (Dunn-Sigouin and Son, 2013), which leads to caution in the interpretation of their results for these regions. However, over the Pacific Ocean there have been large improvements in the simulation of blocking for the last 20 years(Davini and D’Andrea, 2016; Patterson et al. , 2019). In the SH, increases in blocking frequency have occurred in the South Atlantic in austral summer (Dennison et al., 2016) and in the southern Indian Ocean in austral spring (Schemm, 2018). A reduced blocking frequency has been found over the south-western Pacific in austral spring (Sections 2.3.1.4.3 and 3.4.1.3.3; Schemm, 2018).

In summary, no robust trend in atmospheric blocking has been detected in modern reanalyses and in CMIP6 historical simulations (medium confidence). The lack of trend is explained by strong internal variability and/or the competing effects of low-level Arctic amplification and upper-level tropical amplification of the equator-to-pole temperature gradient (medium confidence).

Following on from the assessment in Chapters 2 and 3, this section considers changes in modes of variability at seasonal to interannual time scales in terms of their implications on recent water cycle changes. These modes are described in details in Technical Annex IV.

As discussed in 8.3.1.1, the definition of global water cycle intensity varies from the simple metric of increases in global mean precipitation to broader joint considerations of water vapour and its transport, precipitation minus evaporation (P–E) rates and continental runoff (Figure 8.1). The AR5 determined that globally averaged precipitation is virtually certain to increase with temperature and that there is high confidence that the contrast of annual mean precipitation between dry and wet regions and seasons will increase over most of the globe as temperatures and moisture transports increase (Collins et al., 2013). The AR5 also highlighted that continued ocean warming for a few decades after GHG forcing stabilizes or begins to decrease will also lead to further increases in global mean precipitation and evaporation.

In this Report, Chapter 4 provides an updated assessment of global annual precipitation (Section 4.3.1), finding that it is very likely that annual precipitation averaged over all land regions continuously increases as global surface temperatures increase in the 21st century (high confidence). CMIP6 projections for long-term changes in P–E (Figure 8.13) show that, for all scenarios, P–E increases over the tropics and high latitudes and decreases over the subtropics, resulting from a thermodynamically driven amplification of P–E patterns (Section 8.2.2.1). Both the intensity of changes and the spread among the models is larger for the higher emissions scenarios. A less coherent latitudinal pattern and smaller magnitude of P–E changes over land reflect the complex influence of land–ocean warming contrast, atmospheric circulation change and vegetation feedbacks (Section 8.2.2.1). However, stronger atmospheric moisture transport, increases in precipitation and evaporation over global land and ocean and larger continental runoff that is in part fed by melting of glaciers characterizes a more intense water cycle with global warming.

Figure 8.13 | Zonal and annual-mean projected long-term changes in the atmospheric water budget. Zonal and annual mean projected changes (mm day^–1) in P (precipitation, left column), E (evaporation, middle column), and P–E (right column) over both land and ocean areas (coloured lines) and over land only (black lines) averaged across available CMIP6 models (number provided at the top left of each panel) in the SSP1-2.6 (top row), SSP2-4.5 (middle row) and SSP5-8.5 (bottom row) scenario, respectively. Shading denotes confidence intervals estimated from the CMIP6 ensemble under a normal distribution hypothesis. Colour shading denotes changes over both land and ocean. Grey shading represents internal variability derived from the pre-industrial control simulations. All changes are estimated for 2081–2100 relative to the 1995–2014 base period. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

Global and global land mean water cycle changes from CMIP6 projections are shown in Table 8.1. Increases in global and continental precipitation, P–E and runoff in both the mid-term and long-term illustrate the future intensification of the water cycle, with the magnitude of change increasing with emissions scenarios. Consistent with AR5, CMIP6 simulations of global mean precipitation show a systematic multi-model mean increase of 1.6 to 2.9 % °C^–1 warming (apparent hydrological sensitivity; Section 8.2.1) by 2081–2100 relative to present day across the new SSP scenarios (using global surface air temperature change from Table 4.1). It is well understood that rising concentrations of CO₂ drive a long-term increase in global precipitation with warming, but with the increase partly offset by rapid atmospheric adjustments to the direct atmospheric heating from radiative forcing agents (Section 8.2.1). The largest apparent hydrological sensitivity is found for SSP1-1.9, where the suppressing effects on precipitation from atmospheric heating by greenhouse gases (GHGs) rapidly reduce as their concentration falls. Additional warming due to reduced aerosol loadings under the SSP scenarios (Lund et al. , 2019) further increases global precipitation (Rotstayn et al., 2013; Wu et al., 2013; Salzmann, 2016; T.B. Richardson et al., 2018b; Samset et al., 2018b; Westervelt et al., 2018), with particularly strong contributions from increased monsoon rainfall over East and South Asia (Levy et al., 2013; Westervelt et al., 2015; Dwyer and O’Gorman, 2017).

Over global land there is a small range in global mean multi-model mean precipitation increase across scenarios in the mid-term (2.6–4.0 %), which widens (to 2.6–8.8 %) in the long-term (Table 8.1). The long-term projections are consistent with the Chapter 4 assessment that global annual precipitation over land is projected to increase on average by 2.4 [–0.2 to +4.7] % (likely range) in the SSP1-1.9 low-emissions scenario and by 8.3 [0.9 to 12.9] % in the SSP5-8.5 high emissions scenario by 2081–2100 relative to 1995–2014. Small differences in assessed model mean changes in Chapter 4, Table 4.2 result from a slightly different set of models considered for Table 8.1. Over land, P–E increases by around 2–3% in the mid-term (apart from SSP5-8.5 where increases are almost 5%) and around 1–12% in the long-term, determined by increased moisture transport from the ocean to land Section 8.4.1.2). Runoff increases are larger and less certain due to additional inputs from glacier melt and changes in groundwater storage (Section 8.4.1.7). Overall, precipitation and runoff are very likely to increase over the global land in all scenarios in the mid- and long term. P–E is likely to increase over global land in the mid- and long term and very likely in SSP1-1.9, SSP3-7.0 and SSP5-8.5 pathways. The mid-term consistency in projections across scenarios is not apparent for precipitable water vapour, which increases over land by around 6–15% in the mid-term and 5–36% in the long-term across all scenarios. This implies that increases in extreme precipitation (closely related to atmospheric water vapour content; Section 8.2.3.2) are dependent on mitigation pathway, even in the mid-term (Section 11.4.5). Water vapour residence time (computed as the ratio of precipitable water vapour to precipitation from values in Table 8.1) increases from eight days in the present to nine days in mid-term and up to about ten days in the long-term over land in SSP3-7.0, indicating a longer time to moisten the atmosphere between precipitation events. The CMIP6 projections are therefore consistent with anintensification but not acceleration of the global water cycle.

In summary, it is virtually certain that global water cycle intensity, considered in terms of global and continental mean precipitation, evaporation and runoff, will increase with continued global warming. Global annual precipitation over land is projected to increase on average by 2.4 [–0.2 to +4.7] % (likely range) in the SSP1-1.9 low-emissions scenario and by 8.3 [0.9 to 12.9] % in the SSP5-8.5 high emissions scenario by 2081–2100 relative to 1995–2014.

Globally, AR5 assessed that by the end of the 21st century, the average quantity of water vapour in the atmosphere could increase by 5–25%, depending on emissions. The AR5 assessed that increases in near-surface specific humidity over land are very likely , but that it was also likely that near-surface relative humidity would decrease over many land areas, although with only medium confidence. In terms of moisture transport, AR5 assessed that it was likely that moisture transport into the high latitudes would increase and that there was high confidence that, over the ocean, atmospheric moisture transport from the evaporative regions to the wet regions would increase.

CMIP6 climate models continue to project a steady increase in global mean column-integrated water vapour by around 6–13% by 2041–2060 and 5–32% by 2081–2100, depending on scenario (Table 8.1). This is consistent with projected atmospheric warming (Section 4.5.1.2) and the Clausius–Clapeyron relationship (Section 8.2.1) where every degree Celsius of warming is associated with an approximate 7% increase in atmospheric moisture in the lower atmospheric layers where most of the water vapour is concentrated. This increase sustains a positive feedback on anthropogenic global warming (Section 7.4.2.2). In contrast, the response of clouds is much more spatially heterogeneous, microphysically complex, and model-dependent so that the projected cloud feedbacks remain a key uncertainty for constraining climate sensitivity (Section 7.4.2.4).

CMIP6 models project an overall decrease in near-surface relative humidity over land, although with some regional and seasonal variations in their response (Figure 4.26). Regional changes in near-surface humidity over land are dominated by thermodynamic processes and are primarily controlled by moisture transport from the warming ocean (Chadwick et al., 2016a). Increases in specific humidity lower than the thermodynamic rate are explained by greater warming over land than ocean and modulated by land –atmosphere feedbacks such as soil moisture and plant stomatal changes (Section 8.2.2.1; Berg et al., 2017; Douville et al., 2020). This explains why climate models continue to project a contrasting response of near-surface relative humidity, with a slight and possibly overestimated increase over the oceans and a consistent but possibly underestimated decrease over land (Byrne and O’Gorman, 2016; Douville and Plazzotta, 2017; R. Zhang et al., 2018).

While projections of water vapour are well understood due to the constraints of the Clausius–Clapeyron relationship, projections of water vapour transport are complicated regionally by the role of changes in the wind field, which is influenced by a wide variety of factors. Additionally, there has been relatively little general evaluation of moisture transport in models. In CMIP5 models, both the mean and variability of the vertically-integrated moisture transport is projected to increase, largely due to increases in water vapour (Lavers et al., 2015), with substantial regional differences (Levang and Schmitt, 2015). Single-model studies have illustrated projected increases in low-altitude moisture transport into convergence regions (Allan et al., 2014) and from ocean to land (Zahn and Allan, 2013) that are consistent with present day trends. Increases in moisture transport have been linked to increases in large precipitation accumulations over land (Norris et al., 2019). Based on robust physics and supported by modelling studies, it is well understood that moisture transport increases into convergent parts of the atmospheric circulation such as storm systems, the tropical rain belt and high latitudes (Section 8.2.2.1), but changes in atmospheric circulation that are less well understood alter moisture transport regionally (Section 8.2.2.2). Therefore, given the limited examination of moisture transport in models, regional projections should be considered with caution. Changes in moisture transport specifically associated with monsoons, atmospheric rivers, and other specific circulation features are discussed further in the following sections.

In summary, there is high confidence in continued increases in global mean column integrated water vapour and near-surface specific humidity over land. There is medium confidence in region and season-dependent decreases in near-surface relative humidity over land, due to the complex physical processes involved . In general, there will be increases in moisture transport into storm systems, monsoons and high latitudes (medium confidence).

This section assesses projected changes in precipitation at regional scales. Note that changes in precipitation seasonality are assessed in Box 8.2 and that changes in regional monsoons are assessed in Section 8.4.2.4, where both circulation and rainfall are considered. Further assessments of regional projections of precipitation are presented in Chapters 10, 12 and the Atlas, while a comprehensive assessment of changes in precipitation extremes is provided in Chapter 11.

The AR5 assessed that the contrast of mean precipitation amount between dry and wet regions and seasons is expected to increase over most of the globe as temperatures increase (high confidence), but with large regional variations. Precipitation over the high latitudes, equatorial Pacific Ocean, mid-latitude wet regions, and monsoon regions were assessed as likely to increase under the RCP8.5 scenario, and in many mid-latitude and subtropical dry regions as likely to decrease (AR5 Chapters 7, 12, and 14). Extreme precipitation over most mid-latitude land areas and wet tropical regions was assessed as very likely to become more intense and more frequent.

Geographical patterns of projected precipitation changes show substantial seasonal contrasts and regional differences, including over land (Figure 8.14 and Figure 4.27). Projections for 2081–2100under the SSP2-4.5scenariosuggest increased precipitation over the tropical oceans, north-eastern Africa, the Arabian Peninsula, India, south-eastern Asia and the Polar regions while decreased precipitation is projected mainly over the subtropical regions (Section 4.5.1.4). Precipitation changes contrast regionally in the tropics with wetter wet seasons over South Asia, central Sahel and eastern Africa, but less precipitation over Amazonia and coastal West Africa (Section 8.4.2.4). These large-scale responses are associated with stronger moisture transports in a warmer climate that are modulated by the greater warming over land than ocean, atmospheric circulation responses and land surface feedbacks (Section 8.2.2). There is agreement across CMIP5 and CMIP6 modelling studies that precipitation increases in wet parts of the atmospheric circulation and decreases in dry parts (Liu and Allan, 2013; Kumar et al., 2015; Deng et al., 2020; Schurer et al., 2020) although these regions shift with atmospheric circulation changes. The overall pattern is robust across different model scenarios and time horizons (Tebaldi and Knutti, 2018), but some deviations from the mean pattern cannot be excluded due to the multiple time scales and non-linear atmospheric or land surface processes involved (Section 8.5.3). Near-term regional changes in precipitation are more uncertain because of a stronger sensitivity to natural variability (Section 8.5.2) and non-GHG anthropogenic forcings (Section 4.4.1.3 and 8.4.3.1).

Figure 8.14 | Projected long-term relative changes in seasonal mean precipitation. Global maps of projected relative changes (%) in seasonal mean of precipitation averaged across available CMIP6 models (number provided at the top right of each panel) in the SSP2-4.5 scenario. All changes are estimated for 2081–2100 relative to the 1995–2014 base period. Uncertainty is represented using the simple approach. No overlay indicates regions with high model agreement, where ≥80% of models agree on sign of change; diagonal lines indicate regions with low model agreement, where <80% of models agree on sign of change. For more information on the simple approach, please refer to the Cross-Chapter Box Atlas.1. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

Projected changes in regional precipitation also arise as a response to changes in large-scale atmospheric circulation (Section 8.2.2.2 and 8.4.2), both in the tropics (Chadwick et al. , 2016b; Byrne et al. , 2018) and extratropics (Shaw, 2019; Oudar et al., 2020b). Despite variability in simulated changes, CMIP5 climate models consistently project large rainfall changes (of varying sign) over considerable proportions of tropical land during the 21st century (Chadwick et al., 2016b). Since AR5, some robust responses in large-scale circulation patterns have been identified. For example, and as further assessed in Section 8.4.2, CMIP6 models project a northward shift in the tropical rain belt over eastern Africa and the Indian Ocean and a southward shift in the eastern Pacific and Atlantic oceans (Mamalakis et al., 2021). A projected strengthening and tightening of the tropical rain belt increases the contrasts between wet and dry tropical weather regimes and seasons. It is less clear how the well understood poleward expansion of the subtropics and mid-latitude storm tracks influences precipitation over subtropical and mid-latitude continents (Section 8.2.2.2).

An ensemble of 31 CMIP6 models under the SSP5-8.5 scenario projects increases precipitation by 10–30% over much of the USA and decreases by 10–40% over Central America and the Caribbean by 2080–2099 (Almazroui et al., 2021). This CMIP6 ensemble also projects an increase in annual precipitation over the southern Arabian Peninsula and a decrease over the northern Arabian Peninsula, as also projected by CMIP3 and CMIP5 models (Almazroui et al., 2020a). Annual mean precipitation is projected to increase over South Asia during the 21st century under all scenarios, although the rate of change varies within the region based on 27 CMIP6 models (Almazroui et al., 2020c). CMIP6 projections also display a reduction in annual mean precipitation over northern and southern Africa while increases are projected over Central Africa, under the SSP1-2.6, SSP2-4.5 and SSP5-8.5 scenarios (Almazroui et al., 2020b). The AR6 (Atlas assesses that regions where annual mean rainfall is likely to increase include the Ethiopian Highlands, East, South and North Asia, south-eastern South America, northern Europe, northern and eastern North America, and the Polar Regions. In contrast, regions where annual mean rainfall is likely to decrease include southern Africa, coastal West Africa, Amazonia, south-western Australia, Central America, south-western South America, and the Mediterranean.

The AR5 identified that high-latitude precipitation increase may lead to an increase in snowfall in the coldest regions and a decrease of snowfall in warmer regions due to a decreased number of freezing days. The fraction of precipitation falling as snow and the duration of snow cover was projected to decrease. Heavy snowfall events globally are not expected to decrease significantly with warming as they occur close to the water freezing point, which will migrate poleward and in altitude (O’Gorman, 2014; Turner et al., 2019). There are only a small number of studies evaluating the implications of this mechanism in specific regions. A study for the north-eastern USA indicates smaller reductions for major snowfall events against the broader decline in snowfall expected from thermodynamic effects (Bintanja and Andry, 2017). Arctic snowfall is projected to decrease as rainfall makes up more of the precipitation (Zarzycki, 2018).

Beyond annual or seasonal mean precipitation amounts, an implication of the parallel intensification of the global water cycle and of the increased residence time of atmospheric water vapour (Section 8.2.1) is that the distribution of daily and sub-daily precipitation intensities will experience significant changes (Pendergrass and Hartmann, 2014b; Pendergrass et al., 2015; Bador et al., 2018; Douville and John, 2021), with fewer but potentially stronger events (high confidence) (Section 4.3.3). CMIP6 projections show that in the long-term more drier days but more intense single events of precipitation are expected, regardless of scenario (Figure 8.15). Over almost all land regions, it is very likely that extreme precipitation will intensify at a rate close to the 7% °C^–1 of global warming, but with large spatial differences (Sections 11.4 and 8.2.3.2). The projected increase in precipitable water is expected to lead to an increase in the highest possible precipitation intensities and an increase in the probability of occurrence of extreme precipitation events on the global scale (Neelin et al., 2017), regardless of how annual-mean precipitation changes (O’Gorman and Schneider, 2009; O’Gorman, 2015). The projected increase in heavy precipitation intensity is also found for daily mean precipitation intensity though at a lower rate (Pendergrass and Hartmann, 2014a).

Figure 8.15 | Projected long-term relative changes in daily precipitation statistics. Global maps of projected seasonal mean relative changes (%) in the number of dry days (i.e., days with less than 1 mm of rain) and daily precipitation intensity (in mm day–1, estimated as the mean daily precipitation amount at wet days – for example, days with intensity above 1 mm day–1) averaged across available CMIP6 models (number provided at the top right of each panel) in the SSP1-2.6 (a, b), SSP2-4.5 (c, d) and SSP5-8.5 (e, f) scenario respectively. Uncertainty is represented using the simple approach. No overlay indicates regions with high model agreement, where ≥80% of models agree on sign of change; diagonal lines indicate regions with low model agreement, where <80% of models agree on sign of change. For more information on the simple approach, please refer to the Cross-Chapter Box Atlas.1. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

An increase in the number of dry days is also projected in several regions of the world (Polade et al., 2014; Berthou et al., 2019a), which can dominate the annual precipitation change at least in the subtropics (Polade et al., 2014; Douville and John, 2021). These findings are supported by CMIP6 projections showing a widespread increase in daily mean precipitation intensity over land (Figure 8.15b,d,f) as well as an increase in the number of dry days in the subtropics and over Amazonia and Central America (Figure 8.15a,b,c). Such changes in precipitation regimes, as well as the general increase in the frequency and intensity of precipitation extremes (Section 11.4.5), contribute to an overall increase in precipitation variability (Polade et al., 2014; Pendergrass et al., 2017; Douville and John, 2021). This is also found in CMIP6 models, which show a stronger increase of interannual variability than in seasonal mean precipitation changes, apart from in the winter extratropics where both quantities increase at the same rate with increasing global warming levels (Figure 8.16).

Figure 8.16 | Rate of change in components of water cycle mean and variability across increasing global warming levels. Relative change (%) in seasonal mean total precipitable water (grey line), precipitation (red solid lines), runoff (blue solid lines), as well as in standard deviation of precipitation (red dashed lines) and runoff (blue dashed lines) averaged over extratropical land in (c) summer and (d) winter, and tropical land in (a) June–July–August (JJA) and (b) December–January–February (DJF) as a function of global mean surface temperature for the CMIP6 multi-model mean across the SSP5-8.5 scenario. Extratropical winter refers to DJF for Northern Hemisphere and JJA for Southern Hemisphere (and the reverse for extratropical summer). Each marker indicates a 21-year period centred on consecutive decades between 2015 and 2085 relative to the 1995–2014 base period. Precipitation and runoff variability are estimated by their standard deviation after removing linear trends from each time series. Error bars show the 5–95% confidence interval for the warmest 5°C global warming level. Figure adapted from Pendergrass et al. (2017) and updated with CMIP6 models. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

In summary, it is virtually certain that global precipitation will increase with warming due to increases in GHG concentrations and decreases in air pollution. There is high confidence that total precipitation will increase in the high latitudes, with a shift from snowfall to rainfall except in the coldest regions and seasons. There is also high confidence that precipitation will decrease over the Mediterranean, southern Africa, Amazonia, Central America, south-western South America, south-western Australia and coastal West Africa and that monsoon precipitation will increase over South Asia, East Asia and central-eastern Sahel. See (Section 8.4.2.4 for a more detailed assessment of changes in regional monsoons. Daily mean precipitation intensities, including extremes, are projected to increase over most regions (high confidence). The number of dry days is projected to increase over the subtropics, Amazonia, and Central America (medium confidence). There is high confidence in an overall increase in precipitation variability over most land areas.

CMIP5 projections show no consistent shift in the zonal mean position of the ITCZ (Donohoe et al. , 2013; Donohoe and Voigt, 2017; Byrne et al. , 2018). The ITCZ position is strongly connected to cross-equatorial energy transport (Kang et al. , 2008; Bischoff and Schneider, 2014), which also shows no consistent change in future projections (Donohoe et al., 2013). Since AR5 it has been reported that most CMIP5 models project a narrowing of the ITCZ in response to surface warming together with intensified ascent in the core region and weakened ascent on the ITCZ edges (Lau and Kim, 2015; Byrne et al., 2018), implying a narrowing of precipitation regions influenced by the ITCZ. Modelled changes in the width and intensity of the zonal mean ITCZ are strongly anti-correlated, for example, narrowing is associated with increased intensity while broadening with decreased intensity. Such changes are associated with changes in tropical high cloud fraction and outgoing longwave radiation (Su et al., 2017; Byrne et al., 2018).

Regional shifts in tropical convergence zones are much larger than their zonal mean, and associated regional changes in precipitation (Chadwick et al., 2013; Mamalakis et al., 2021) are characterized by considerable uncertainties across models (Kent et al., 2015; Oueslati et al., 2016). Over the tropical oceans, shifts in rain bands are strongly coupled with changes in SSTs (Xie et al. , 2010; Huang et al. , 2013). Over tropical land, factors including remote SST increases (Giannini, 2010), the direct CO₂ effect (Biasutti, 2013) and land–atmosphere interactions (Chadwick et al., 2017; Kooperman et al., 2018) influence projections. CMIP6 models project a clear northward ITCZ shift over eastern Africa and the Indian Ocean as well as a southward shift over the eastern Pacific and Atlantic oceans, as a result of regionally-contrasting inter-hemispheric energy flows (Mamalakis et al., 2021). The northward movement of the ITCZ over Africa has been linked to an intensification of the Saharan heat low associated with greenhouse gas (GHG) warming (Dong and Sutton, 2015), causing the tropical rain belt to seasonally migrate farther northward and reside there longer (Cook and Vizy, 2012; Dunning et al., 2018). In southern Africa, the projected delay in the wet season onset (Dunning et al. , 2018) is also associated with a circulation-based northward shift in the tropical rain band (Lazenby et al., 2018).

In summary, consistent with the AR5, the overall weakening of the tropical circulation is projected in CMIP5 and CMIP6 simulations with high confidence. It is likely that the zonal mean of the ITCZ will narrow and strengthen in the core region with projected surface warming (high confidence). Distinct regional shifts in the ITCZ will be associated with regional changes in precipitation amount and seasonality (medium confidence).

The AR5 found that the Hadley cells are likely to slow down and expand in response to radiative forcing, but with considerable internal variability. Given the complexities in forcing mechanisms, AR5 assigned low confidence to near-term changes in the structure of the Hadley circulation. The widening Hadley cells were expected to result in a poleward expansion of subtropical dry zones.

Model simulations since AR5 project a more noticeable and consistent weakening of the Northern Hemisphere (NH) winter Hadley cell than the Southern Hemisphere (SH) winter cell (Seo et al., 2014; Zhou et al., 2016), related to changes in meridional temperature gradient, static stability, and tropopause height (Seo et al., 2014; D’Agostino et al., 2017). Changes in SST patterns reduces the magnitude of Hadley cell weakening (Gastineau et al., 2009; Ma et al., 2012). There is considerable structure in Hadley circulation strength changes with longitude, associated with cloud-circulation interactions (Su et al., 2014). Subtropical anticyclones are projected to intensify over the north Atlantic and south Pacific but to weaken elsewhere (He et al., 2017).

A consistent poleward expansion of the edges of the Hadley cells is projected (Nguyen et al., 2015; Grise and Davis, 2020), particularly in the SH, consistent with observed trends (Figure 8.21 and Section 8.3.2.2; Nguyen et al. , 2015). The main driver of future expansion appears to be greenhouse gas forcing (Grise et al., 2019), with uncertainty in magnitude due to internal variability (Kang et al., 2013). Proposed mechanisms for poleward expansion include increased dry static stability (Frierson et al. , 2007; Lu et al. , 2007), increased tropopause height (Chen and Held, 2007; Chen et al. , 2008), stratospheric influences (Kidston et al., 2015) and radiative effects of clouds and water vapour (Shaw and Voigt, 2016; see also Section 4.5.1.5). Hadley cell expansion is thought to be associated with the precipitation declines projected in many subtropical regions (Shaw and Voigt, 2016), but more recent work suggests that these reductions are mainly due to the direct radiative effect of CO₂ forcing (He and Soden, 2015), land –sea contrasts in the response to forcing (Shaw and Voigt, 2016; Brogli et al. , 2019) and SST changes (Sniderman et al., 2019). In semi-arid, winter rainfall-dominated regions (such as the Mediterranean), thermodynamic processes associated with the land –sea thermal contrast and lapse rate changes dominate the projected precipitation decline in summer, whereas circulation changes are of greater importance in winter (Brogli et al. , 2019). The hydroclimates in these regions are projected to evolve with time due to changing contributions from rapid atmospheric circulation changes and their associated SST responses, as well as slower SST responses to anthropogenic forcing (Zappa et al., 2020).

In summary, CMIP5 and CMIP6 models project a weakening of the Hadley cells, with high confidence for the NH in boreal winter and low confidence for the SH in austral winter. The Hadley cells are projected to expand polewards with global warming, most notably in the SH (high confidence). There is currentlylow confidence in the impacts on regional precipitation in subtropical regions.

The AR5 determined that the Pacific Walker circulation was likely to slow down over the 21st century, which would lead to decreased precipitation over the western tropical Pacific and increases over the central and eastern Pacific. Recent studies show consistency with AR5 conclusions but also show an eastward shift over the Pacific, mostly due to a shift towards more ‘El Niño-like’ conditions under global warming (Bayr et al., 2014). Other studies suggest that the weakening of the Walker circulation is related to the response of the western North Pacific monsoon and to changing land–sea temperature contrasts, while a positive ocean–atmosphere feedback amplifies the weakening of both east–west SST gradient and trade winds in the tropical Pacific (Zhang and Li, 2017).

Since AR5, the paradox between the projected weakening and the observed strengthening of the Walker circulation since the 1990s (Section 8.3.2.2) has triggered debate about the drivers of these changes (England et al. , 2014; McGregor et al. , 2014; Kociuba and Power, 2015; Vilasa et al. , 2017; Chung et al. , 2019). Projected changes in equatorial SST gradients are not entirely consistent with observed trends (Coats and Karnauskas, 2017; Seager et al., 2019a), and one CMIP5 model that projects a future strengthening of the Walker circulation is more consistent with observations than other models (Kohyama et al., 2017). Other studies suggest that these differences arise from the dominant influence of internal climate variability to the observed trends (Chung et al., 2019), or as a consequence of a systematic cold bias of most CMIP5 models in their Equatorial Pacific cold tongues (Seager et al., 2019a). However, the latter hypothesis is based on a simplified model of tropical Pacific dynamics and is not consistent with the current physical understanding of the tropical circulation response to increasing CO₂ levels (Section 8.2.2.2) or with independent paleoclimate evidence suggesting a weaker Walker circulation under warmer climates (Tierney et al., 2019; McClymont et al., 2020). Different time scales of the tropical Pacific responses to global warming have been highlighted by numerical experiments with both comprehensive and simplified models. Results suggest a transient strengthening of the Walker circulation related to Indian Ocean warming (L. Zhang et al., 2018), followed by a slower weakening linked to a strengthened eastern Pacific cold tongue warming emerging after 50–100 years (Section 7.4.4.2.1; Heede et al., 2020).

CMIP6 projections provide further evidence of a significant long-term weakening of the Walker circulation (Figure 8.21). For instance, a pronounced weakening of the upper-level tropical easterly jet is projected both over the Indian Ocean and tropical eastern Pacific, where declines are projected to exceed 70% by 2100 in the high-emissions SSP5-8.5 scenario (S. Huang et al., 2020). CMIP6 models agree on a future decrease of the equatorial zonal temperature gradient (Fredriksen et al., 2020), which can lead to weaker trade winds over the tropical Pacific. However, CMIP6 models show a diversity of SST warming patterns in the tropical Pacific (Freund et al., 2020), which contributes to uncertainties in the response of both Walker circulation and ENSO to continued warming.

Figure 8.21 | Schematic depicting large-scale circulation changes and impacts on the regional water cycle. The central figures show precipitation minus evaporation (P–E) changes at 3°C or global warming relative to an 1850–1900 base period (mean of 23 CMIP6 SSP5-8.5 simulations). Annual mean changes (large map) include contours (ocean only) depicting control climate P–E = 0 mm day–1lines with the solid contour enclosing the tropical rain belt region and dashed lines representing the edges of subtropical regions. Confidence levels assess understanding of how large-scale circulation change affect the regional water.

Open figure

In summary, there is high confidence that the Pacific Walker circulation will weaken by the end of the 21st century, and will be associated with decreased precipitation over the western tropical Pacific and increases farther east. Discrepancies between observed and simulated changes in SSTs in the tropics indicate that a temporary strengthening of the Walker Circulation can arise from a transient response to GHG radiative forcing (low confidence) and from internal variability (medium confidence).

In AR5, monsoon precipitation over land was projected to intensify by the end of the 21st century, due to thermodynamic increases in moisture convergence despite weakening of the tropical circulation (see Section 8.2.1.3). Following the definition of regional monsoons in Annex V and Figure 8.11, and the assessment of the observed changes (Section 8.3.2.4), here we provide an assessment of projected changes in regional monsoons. Assessment is provided either in terms of SSP and RCP scenarios and global warming levels available since AR5, or from the newly available CMIP6 projections (Figure 8.22 and Table 8.2). Table 8.2 provides projected changes across the five SSPs used in this Report for precipitation (mm day^–1), P–E (mm day^–1) and runoff (mm day^–1) over the regional monsoons for the mid (2041–2060) and long term (2081–2100).

Figure 8.22 | Projected regional monsoons precipitation changes. Percentage change in projected seasonal mean precipitation over regional monsoon domains (as defined in Figure 8.11, Section 8.3.2.4 and Annex V) for near term (2021–2040), mid-term (2041–2060), and long term (2081–2100) periods based on 24 CMIP6 models and three SSP scenarios (SSP1-2.6, SSP2-4.5 and SSP5-8.5). Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

Tropical cyclones (TCs) projections are primarily assessed in Section 11.7.1.5. Here, we extend this analysis by assessing the implications of projected changes in tropical cyclones on the water cycle.

The AR5 concluded that TC rainfall rate was likely to increase through the 21st century. Section 11.7.1.5 assesses that the average tropical cyclone rain-rate is projected to increase with warming (high confidence), and peak rain rates are projected to increase at greater than the Clausius–Clapeyron scaling rate of 7%°C^–1 warming in some regions due to increased low-level moisture convergence (medium confidence). The increase in TC rainfall rate is explained by increased TC intensity resulting from increasing SSTs, and increased environmental water vapour (Chauvin et al., 2017; M. Liu et al., 2019).

Consistent with the observed poleward migration of tropical cyclone activity (Kossin et al., 2014), in the SH a larger proportion of storms are projected to decay south of 25°S at the end of the 21st century but with negligible changes in genesis latitude and storm duration for the Australian region (CSIRO and BoM, 2015; Sharmila and Walsh, 2018). An analysis of projections for North Pacific islands indicate that the maximum intensity of storms will increase but the number of tropical cyclones will decrease in some places, such as Guam and Kwajalein Atoll in the tropical north-western Pacific, or remain the same in other regions like near Okinawa (Japan) or Oahu (Hawaii) (Widlansky et al., 2019). TC-induced storm tides affecting landfall in the Pearl River delta over South China are projected to increase by the end of the 21st century (J. Chen et al., 2020b)

In summary, there is high confidence that heavy precipitation associated with tropical cyclones is projected to increase, in response to well-understood processes related to increased low-level moisture convergence and environmental water vapour.

The AR5 did not provide an assessment of stationary wave projections as distinct from other related aspects of circulation, such as blocking, modes of variability, and storm tracks. Here we provide a brief assessment of stationary wave projections from the water cycle perspective, with the related circulation aspects considered separately in the following sections.

Several studies based on CMIP5 projections show changes in NH winter stationary waves that increase precipitation over the west coast of North America and decrease it over the eastern Mediterranean and parts of south-western North America (Neelin et al. , 2013; Seager et al. , 2014a, b, 2019b; Simpson et al. , 2016; Wills et al. , 2019), although the underlying dynamics are not yet fully understood (Seager et al., 2019b; Wills et al., 2019). For the NH winter global teleconnection pattern, the majority of the models analyzed in (Sandler and Harnik, 2020) project the development of a preferred longitudinal phasing for the pattern, but with strong disagreement among models over the details of the phasing and therefore the associated regional hydrologic impacts.

While the potential role of increasing hydrologic extremes with quasi-resonant stationary waves during NH summer has received considerable attention (see Section 8.3.2.6), as yet there is no clear evidence in model projections that this variability will increase (Teng and Branstator, 2019). The influence of the Arctic on mid-latitude circulation is assessed in Cross-Chapter Box 10.1, which reports that there is low confidence in the dominant contribution of Arctic warming compared to other drivers in future projections. Potential changes to the stratospheric polar vortex in CMIP5 models have a substantial influence on tropospheric stationary waves and associated hydrologic impacts in both the NH (Zappa and Shepherd, 2017) and SH (Mindlin et al., 2020). CMIP5 models have some important limitations in their representation of stationary waves (Lee and Black, 2013; Simpson et al., 2016; Garfinkel et al., 2020) and this aspect of CMIP6 models has not yet been comprehensively evaluated.

In summary, future changes in stationary waves may have an important influence on both the mean state and variability of the water cycle. Limitations in model representation, dynamical understanding, and the number of targeted studies on the topic currently constrain the assessment of future changes in stationary waves. Based on current knowledge, there is low confidence that projected changes in stationary wave activity will contribute to decreases of cold season precipitation over the eastern Mediterranean and increases over the west coast of North America.

In AR5, the increased ability of models to simulate blocking and higher agreement on projections led to an assessment with medium confidence that the frequency of NH and SH blocking will not increase, but future changes in blocking intensity and persistence were deemed uncertain (AR5 Chapter 14, ES and Box 14.2). Blocking influences precipitation (e.g., Trigo et al., 2004), flooding (e.g., Yamada et al. , 2016), drought (e.g., Dong et al. , 2018b), snow (e.g., García-Herrera and Barriopedro, 2006), and glacier melt (e.g., Hanna et al., 2013), and so is of broad importance to the water cycle in areas of blocking activity.

Blocking projections are assessed in this Report in Chapter 4 (Section 4.5.1.6), and model performance in simulating blocking is also discussed in Chapter 3 (Section 3.3.3.3). CMIP5 projections suggest a complex response in blocking frequencies with an eastward shift in NH winter blocking, mid-latitude decreases in boreal summer except in eastern Europe– western Russia, and SH decreases in the Pacific sector during austral spring and summer. CMIP6 projections (Figure. 4.28) show a notable decrease in blocking activity over Greenland and the North Pacific for the SSP3-7.0 and SSP5-8.5 scenarios. However, the continued large differences among current models as well as the sensitivity to blocking detection methods limits confidence in projected regional changes in blocking (see also Section 10.3.3.3.1). The influence of blocking on multiple elements of the water cycle means that the uncertainty in blocking projections adds a corresponding layer of uncertainty to water cycle projections.

In summary, and despite recent improvements in the simulation of blocking, there is limited evidence in model projections of future changes, except for boreal winter over Greenland and the North Pacific where there is high confidence that blocking events are not expected to increase in the SSP3-7.0 and SSP5-8.5 scenarios. As with stationary waves, this adds uncertainty to mid-latitude water cycle projections at the regional scale.

Following on from the assessment of projected changes in modes of climate variability (MoVs) and regional teleconnections (Section 4.5.3), here we assess their consequences for projected water cycle changes.

The AR5 (Chapter 7 recognized that the simulation of clouds and precipitation remains challenging for state-of-the-art GCMs. Model development and evaluation have continued since AR5, with a particular emphasis on the representation of new model components, like interactive vegetation, aerosols and biogeochemical cycles. For example, the comparison of simulated tropical precipitation across three successive generations of CMIP models (including CMIP6) indicates overall little improvement for the summer monsoons, the double-ITCZ bias, the diurnal cycle and the frequency of precipitation (Fiedler et al., 2020). Some of these issues are related to inherent model limitations in three specific areas: atmospheric convection, cloud–aerosol interactions and land surface processes (ocean and cryosphere-related processes are addressed in Chapter 9). These limitations do not weaken the overall progress made in the large-scale simulation of present-day climate (FAQ 3.3 and Section 3.3.2.3), even though the improvement of CMIP6 compared with CMIP5 models is limited (Figure 3.12) and is generally less systematic or obvious at the regional scale (e.g., Gusain et al. , 2020; Monerie et al. , 2020; Oudar et al. , 2020a). Instead, they call for a careful interpretation of hydrological projections with the full range of plausible outcomes, rather than only considering the most likely scenarios (Sutton, 2018, 2019).

Coarse spatial resolution of climate models has often been considered a key limitation in global climate projections (Di Luca et al., 2015; Roberts et al., 2018). Proposed and tested solutions include a uniform or regional increase in the resolution of GCMs, or the use of regional climate models (RCMs). The increase in computing resources has also led to the development of convection-permitting models (Prein et al., 2015), which have been integrated over larger domains, but are still unsuitable for CMIP simulations. Statistical downscaling tools are also widely used to generate fine-scale regional climate information necessary for climate impacts and adaptation studies. A comprehensive assessment of the added value of increased spatial resolution and of the benefits and shortcomings of statistical downscaling tools are addressed in Chapter 10 (Section 10.3.3).

Estimating internal variability is an important challenge in the assessment of human-induced changes in the water cycle since its magnitude and range of variability can exceed the anthropogenic signal, at least at the regional scale and for near-term projections or low-emissions scenarios (Sections 4.4.1.4 and 8.4.2.9; Deser et al. , 2012; Shepherd, 2014; Xie et al. , 2015; Sarojini et al. , 2016; Dai and Bloecker, 2019; Lehner et al. , 2020). Underestimating internal variability in models may result in the overestimation of anthropogenic climate change because the ‘noise’ in the signal-to-noise ratio is underestimated (Knutson and Zeng, 2018). There is medium confidence that this underestimation affects global water cycle projections, for instance, in terms of drought persistence and severity in the south-western USA, eastern Australia, southern Africa, the Mediterranean, the southern Amazonian basin and China (Ault et al., 2014; Cook et al., 2018; Gu et al., 2018). In CMIP6 models, the uncertainty in future projections of 20-year mean precipitation changes attributable to internal variability ranges from 41% in the near term (2021–2040) to 5% in the long term (2081–2100) (Figure 8.23). For decadal-mean precipitation changes, the relative contribution of internal variability is even larger when using large ensembles (Lehner et al., 2020).

Over the 20th century, CMIP5 models show a realistic magnitude of decadal precipitation variability, if not a slight overestimation in some regions (Knutson and Zeng, 2018). However, the relatively short and human-influenced instrumental record limits our ability to quantify the magnitude of internal variability in the water cycle, particularly over long time scales (decadal and beyond). Global extended reanalyses (Section 1.5.2) have been used to derive long-term variability in the regional water cycle components (Caillouet et al., 2017), merged with historical meteorological and hydrological local observations (Bonnet et al., 2017; Devers et al., 2020). Specific assessment of these types of methodology and related uncertainties is provided in Chapter 10 (Sections 10.2 and 10.3). Paleoclimate archives (tree rings, corals, ice core, speleothems, lake and ocean sediments) provide extended reconstructions of key water cycle metrics and large-scale circulation features. Some studies have suggested that CMIP5 models underestimate internal variability at decadal and longer time scales, and therefore may be missing important processes in the climate system (Ault et al. , 2012, 2013; Bunde et al. , 2013; Franke et al. , 2013; Cheung et al. , 2017; Hope et al. , 2017; Kravtsov, 2017; Cassou et al. , 2018). However, recent assessments using paleoclimate records have found that CMIP5 models are able to reproduce decadal-to-centennial variability, including the severity, persistence and spatial extent of megadroughts (Coats et al., 2015; Stevenson et al., 2015; PAGES Hydro2K Consortium, 2017), once signal reddening (autocorrelation) in proxy archives is accounted for (Deeet al., 2017; PAGES Hydro2K Consortium, 2017). Implementation of proxy system models, that is, functions that transform model variables into proxy units, has reduced model–proxy disagreement, although some differences in the magnitude of internal variability remain, particularly at centennial time scales (Deeet al., 2017; Parsons et al., 2017). It is unclear whether remaining discrepancies represent limitations of the climate models, or limitations of the proxy system models. Therefore, there is medium to high confidence (i.e., depending on the region) that climate models do not underestimate water cycle internal variability.

The mechanisms driving internal variability in the water cycle in climate model simulations varies. While models indicate that cool SSTs in the eastern tropical Pacific (La Niña or the cool phase of the PDO) are associated with drought in south-western North America, they also show that atmospheric internal variability may be a more prominent driver (Coats et al., 2015, 2016; Stevenson et al., 2015; Parsons et al., 2018). Simulations of the last millennium from CMIP5–PMIP3 reproduce the observed negative correlation between eastern Australian rainfall and the central equatorial Pacific SSTs with varying skill, and also display periods when the ENSO teleconnection weakens substantially for several decades (Brownet al., 2016a). Differences in simulated internal variability have been found to be responsible for the inter-model spread in predicted shifts in subtropical dry zones for a given shift in the Hadley cell (Seviour et al., 2018). CMIP5 models show that both internal variability and anthropogenic forcings are responsible for the drying over the South Atlantic Convergence Zone region, though with large uncertainties (Zilli and Carvalho, 2021). Moreover, the detection of the anthropogenic forcing on the South Atlantic Convergence Zone is strongly dependent on the characterization of model internal variability (Talento and Barreiro, 2012).

Beyond the tropics, North Pacific decadal variability (Annex IV.2.6, 2.4.5, 3.7.6) exerts a strong modulation of extratropical ENSO teleconnections, but also influences low-frequency variability of the Walker circulation, which is underestimated by most CMIP5 models (England et al., 2014). Atlantic Multi-decadal Variability (Annex IV.2.7, 2.4.6, 3.7.7) teleconnections show a high model spread among CMIP5 models, both in terms of persistence and spatial coherence (Qasmiet al., 2017), which has potential consequences for the water cycle variability simulated over Europe. For example, internal variability will continue to play an important role in the variability of river flows over France in coming decades (medium confidence) (Giuntoli et al., 2013; Boé and Habets, 2014; Bonnet et al., 2017).

Ensembles of atmosphere-only simulations driven by observed or reconstructed SST are useful for evaluating the ability of models to capture the circulation and/or precipitation variability observed over the historical period (Zhou et al., 2016; Deng et al., 2018; Douville et al., 2019). However, limitations of such AGCM-based attribution methods, that is, related to the lack of air–sea interactions in the response, may lead to erroneous attribution conclusions in some regions for local circulation and mean and extreme precipitation (Dong et al., 2017). Other methods to measure the portion of precipitation variability include the partitioning into dynamical as opposed to thermodynamical components (Saffioti et al., 2016; Fereday et al., 2018; Lehner et al., 2018), the analysis of variance (Dong et al., 2018a) and direct characterization of stochastic weather-noise (Short Gianotti et al., 2014).

Single-model initial condition large ensembles (SMILEs) are a powerful tool for estimating the magnitude of internal variability in historical and future climates (Section 1.4.4). Using SMILEs, it has been shown, for example, that internal NAO variability imparts substantial uncertainty to future changes in European precipitation (Figure 8.24; Deser et al., 2017). For the South Asian summer monsoon, internal variability can overshadow the forced monsoon rainfall trend, thereby increasing near-term projection uncertainties (X. Huang et al., 2020a). Specific regional applications of the use of large ensembles are further assessed in Sections 10.3.4.3 and 10.3.4.4.

Figure 8.24 | Impact of the North Atlantic Oscillation (NAO) on2016–2045climate trends. (a) Regressions of winter sea level pressure (SLP) and precipitation trends upon the normalized leading principal component (PC) of winter SLP trends in the CESM1 Large Ensemble, multiplied by two to correspond to a two standard deviation anomaly of the PC (as internal climate variability component); (b) CESM1 ensemble-mean winter SLP and precipitation trends (as forced climate variability component); (c) b – a (forced minus internal climate variability component); (d) b + a (forced plus internal climate variability component). Precipitation in colour shading (mm day^–1 per 30 years) and SLP in contours (interval = 1 hPa per 30 years with negative values dashed). Figure adapted from Deser et al. (2017), https://creativecommons.org/licenses/by/4.0/; further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

Since AR5, SMILEs have helped quantify the time of emergence of climate change signals (see Sections 1.4.2.2 and 10.4.3). Results from SMILEs indicate that by 2000–2009 (compared to 1950–1999), simulated anthropogenic shifts in mean annual precipitation already emerged over 36–41% of the globe including high latitudes (Frankcombe et al., 2018; Kumar and Ganguly, 2018), the eastern subtropical oceans, and the tropics (Zhang andDelworth, 2018). By 2050 (2100), more than 60% (85%) of the globe is projected to show detectable anthropogenic shifts in mean annual precipitation (Zhang andDelworth, 2018). Other SMILE results for the 1950–2100 period (Kay et al., 2015; Sigmond and Fyfe, 2016) indicate that internal variability can obscure the detection of the anthropogenic hydroclimatic signal until the middle to late 21st century in many parts of the world for both mean and extreme precipitation (Martel et al., 2018; Dai and Bloecker, 2019). A common finding is that changes in the characteristics of wet extreme events will emerge earlier than changes in average conditions (Gaetani et al., 2020; Hawkins et al., 2020; Kusunoki et al., 2020). An assessment of the methods used to estimate time of emergence is presented in Chapter 10 (Section 10.3.4.3). For specific regional examples of climate change attribution and emergence of anthropogenic signal, see Section 10.4.2.

In summary, there is medium confidence that climate models reproduce the general magnitude and character of internal variability that influences water cycle variables. There is high confidence that internal variability will continue to be a major source of uncertainty, at least for near-term water cycle projections at the regional scale. There is low confidence in the region-dependent time of emergence of water cycle changes (see also Section 10.4.3), but there is medium confidence that changes in wet extreme events will emerge earlier than changes in average conditions.

Adapting water resource management in the face of climate change will greatly benefit from improved prediction of land surface hydrology at the decadal time scale. Climate predictions (Section 1.4.4) differ from climate projections by constraining the initial state of the slow components of the climate system (i.e., the ocean, the cryosphere and the terrestrial hydrology) as well as volcanic aerosols and ozone depleting substances with observations. Anthropogenic and natural radiative forcing and low-frequency modes of variability (e.g., AMV and PDV, Annex IV.2.7 and IV.2.6) suggest the possible predictability of climate in the first decade or so of the 21st century, in addition to the projected response to the anthropogenic forcing (Sections 4.2.3 and 4.4.1.3).

In AR5, decadal prediction of precipitation over some land areas showed improved skill due to specified radiative forcing, with almost no added value from ocean initialization. Since AR5, more studies have been devoted to understanding the potential or effective water cycle predictability related to ocean multi-decadal variability. Decadal hindcast experiments based on large ensembles highlight increasing skill scores in annual mean precipitation three to seven years ahead, at least over the Sahel and Europe (Yeager et al., 2018). There is relatively high predictability of the AMV impacts over the Mediterranean basin, Central Asia and the Americas (from the USA to northern South America) during boreal summer, but in boreal winter the signal-to-noise ratio shows only weak predictability over land (Yamamoto and Palter, 2016; Ruprich-Robert et al., 2017). The link between South Asian summer monsoon changes and the AMOC and the decadal variability in the Pacific Ocean open the possibility of increased predictability for the near future (Kushnir et al. , 2017; X. Huang et al. , 2020b; Sandeep et al., 2020).

The additional skill associated with the initialization of the cryosphere and the land surface has received limited attention. However, there is observational evidence that oceanic decadal variations can propagate into the atmosphere and, consequently accumulate into terrestrial land surface reservoirs(e.g., Bonnet et al., 2017) and vegetation (e.g., Zeng et al., 1999). This land surface memory, like in soil moisture (Alessandri and Navarra, 2008; Catalano et al., 2016) or snow(Loranty et al., 2014), may also contribute to the decadal predictability of the terrestrial component of the water cycle, but remains difficult to assess given the limitations of observational records. Vegetation initialization seems to generate as much noise as signal and does not necessarily translate into improved skill in early decadal predictions based on ESMs (Weiss et al., 2014).

Decadal hydrological predictability in an idealized setting has also been investigated through offline land surface hindcast experiments, driven by observed atmospheric forcing and/or initial conditions, suggesting the potential for skilful predictions for terrestrial water storage, deep soil moisture, and groundwater (Yuan and Zhu, 2018). Yet, a real-world assessment is hampered by the lack of observations and is only feasible when multi-decadal records of satellite estimates of terrestrial water storage, snow mass or soil moisture are available.

In summary, there is high confidence that the water cycle changes that have already emerged from internal variability will become more pronounced in near-term (2021–2040) projections. However, there is low confidence in decadal predictions of precipitation changes, particularly over most land areas, because internal variability remains difficult to predict and can offset or amplify the forced water cycle response.

Volcanic eruptions can affect climate projections in the near term (2021–2040; Section 4.4.4 and Cross-Chapter Box 4.1). In this chapter, they are of interest because they can trigger a transient departure from the water cycle response to anthropogenic radiative forcing. Major volcanic eruptions temporarily reduce total global and wet tropical region precipitation (high confidence) (Iles and Hegerl, 2014), can weaken or shift the ITCZ (Iles and Hegerl, 2014; Colose et al., 2016; Liu et al., 2016), and reduce summer monsoon rainfall (medium confidence) (Pausata et al. , 2015b; Zambri and Robock, 2016; Zambri et al. , 2017; Zuo et al. , 2019; M. Singh et al. , 2020). Monsoon precipitation in one hemisphere can be enhanced by the remote volcanic forcing occurring in the other hemisphere (medium confidence) (Pausata et al. , 2015a; Liu et al. , 2016; Zuo et al. , 2019). Over the Sahel, the sign of hydrological changes depend on the hemisphere where the volcanic eruptions occur (J.M. Haywood et al., 2013). Out of phase changes in the Sahel and the Amazonian basin are expected from the effect of volcanic aerosols on tropical Atlantic SST and the ITCZ (Hua et al., 2019). Over the last millennium, uncertainties remain in the symmetry/asymmetry of the monsoon response because it is difficult to estimate the exact latitude and season of past volcanic eruptions further back in time (Colose et al., 2016; Fasullo et al., 2019).

Data for six major eruptions over the last century along with CMIP5 historical experiments indicate that volcanic eruptions cause a detectable decrease in streamflow in northern South America, Central Africa, high-latitude Asia and in wet tropical–subtropical regions, and a detectable increase in south-western North America and southern South America (Iles and Hegerl, 2015). Attempts to include volcanic forcing in future projections show enhanced precipitation variability on annual to decadal time scales with small reductions in Asian monsoon rainfall (Bethke et al., 2017). The occurrence of volcanic eruptions in the coming century, either as single large events or clustered smaller ones, can alter the water cycle (see also Cross-Chapter Box 4.1), and regional drought events may be enhanced by co-occurring volcanic (Liu et al., 2016; Gao and Gao, 2017; Zambri et al., 2017) and GHG (e.g., Cook et al., 2018) forcing (low confidence). Volcanic eruptions may also lead to widespread precipitation anomalies up to several years following an eruption through their potential influence on the El Niño Southern Oscillation (low confidence) (Stevenson et al., 2016; Dee et al., 2020; McGregor et al., 2020).

In summary, large volcanic eruptions reduce global mean precipitation, as well as precipitation in tropical wet regions (high confidence). There is low confidence in specific regional and seasonal responses, primarily due to the limitations of the observational record.

Since AR5, there is further evidence that the pattern-scaling technique has limitations (Lopez et al., 2014; Wartenburger et al., 2017; Tachiiri et al., 2019), and that alternative approaches, such as multiple regressions using the land–sea warming contrast as an additional predictor, offer added value (Joshi et al., 2013). The simplest traditional pattern-scaling approach approximates future changes by the product of a time-evolving global surface temperature change and a pattern that varies spatially but is constant across time, scenarios, and models. This technique was shown to be more robust across scenarios rather than across models, with better results for temperature compared with precipitation (Tebaldi and Arblaster, 2014; see also Section 4.2.4). One approach which avoids scaling is to consider a period in a different scenario with the same global surface temperature change (Herger et al., 2015). It is attractive as it provides patterns of any temporal resolution that are consistent across variables. Nonetheless, this technique is still only based on global surface temperature and is not necessarily suitable for precipitation changes projected in stabilized versus transient scenarios (at the same global warming level) given the fast-atmospheric adjustment to GHG radiative forcing (Sections 8.2.1 and 8.4.1.1).

Even in a theoretical climate system governed by linear processes, pattern-scaling assumptions can fail because the different forcing time response of different parts of the Earth system cause evolving spatial warming patterns (Good et al., 2016a). This occurs primarily because different feedbacks occur at different time scales (Armour et al., 2013; Andrews et al., 2015), which in turn implies that the atmospheric circulation and water cycle is dependent both on the level of warming and the rate of change (Ceppi et al., 2018). The usual distinction between the fast adjustment to increased GHG concentrations and the slower response to SST warming (Section 8.2.2.2) may, however, not be sufficient to explain the time evolution of the hydroclimatic response at the regional scale, especially in subtropical land areas where this response critically depends on shifts in atmospheric circulation associated with distinct ‘fast’ (typically five to ten years, that is however much slower than the atmospheric adjustment assessed in Section 8.2.1) and slow SST warming patterns (Zappa et al., 2020). The changing balance between the water cycle response to anthropogenic GHG and aerosol forcings is another source of non-linearity across time and global warming levels (Ishizaki et al. , 2013; Rowell et al. , 2015; Y. Liu et al. , 2019b; Wilcox et al., 2020).

Non-linearities in the climate response are thought to arise from multiple factors. These include state-dependent ice-albedo feedback and its potential influence on Northern Hemisphere (NH) storm tracks (Peings and Magnusdottir, 2014; Semenov and Latif, 2015; see also Cross-Chapter Box 10.1 and Section 8.6.1.2); a state-dependent sensitivity of tropical precipitation to increased SST (Schewe and Levermann, 2017; He et al., 2018); a complex response of the Atlantic meridional overturning circulation (AMOC; Sections 9.2.4.1 and 8.6.1.1) and its model- and magnitude dependent teleconnections with regional temperature and precipitation (Kageyama et al., 2013; Jackson et al., 2015; Qasmi et al., 2017, 2020); and other atmospheric and terrestrial (Section 8.5.3.2) processes such as cloud and land surface feedbacks (Ceppi and Gregory, 2017; King, 2019). The response of convective precipitation may exhibit non-linearities because it is itself modulated by both dynamics and atmospheric water content, each responding independently to warming (Chadwick and Good, 2013; Neupane and Cook, 2013).

Based on a simple model, it was also suggested that the Indian summer monsoon may exhibit a moisture-advection feedback which allows multiple stable states as boundary conditions change (Zickfeld et al., 2005). However, limitations of this theory and comprehensive GCMs suggest a near-linear monsoon response to a broad range of radiative forcings (Boos and Storelvmo, 2016). Non-linear precipitation responses to global warming have been reported in the Indo-Pacific, where a linear increase in SSTs can trigger non-linear changes in precipitation and a shift in the ITCZ depending on the relative amplitudes of uniform and structured SST anomalies (C.T.Y. Chung et al., 2014; Toda and Watanabe, 2018).

Compared to atmospheric circulation and seasonal mean precipitation, extreme precipitation has been found to scale more accurately with local and global mean temperature (Chou et al., 2012; Pendergrass et al., 2015). The projected increase in the magnitude of extreme precipitation is generally proportional to the global warming level, with an increase of around 7% per 1°C warming (Section 11.4.5) although this rate shows seasonal and geographical variations and is slightly less for five-day than for one-day precipitation maxima. Projected changes in extreme precipitation are the result of both thermodynamical and more model-dependent and potentially less linear dynamical contributions (Pfahl et al., 2017). Projected changes in precipitation extremes are also potentially sensitive to a non-linear response of spatial convective organization (Pendergrass et al., 2016), and can exhibit a quadratic rather than linear response to global warming (Pendergrass et al., 2019).

Within CMIP6, the linearity to CO₂ forcing can be assessed through the comparison of the model response to abrupt doubling compared with abrupt quadrupling of atmospheric CO₂ (Webb et al., 2017). Preliminary analyses based on CMIP5 models showed that annual precipitation changes following a doubling step change in CO₂ from pre-industrial levels are not necessarily consistent with the response to the step from doubling to quadrupling despite a similar change in radiative forcings (Good et al., 2016a; Ceppi and Shepherd, 2017). Beyond the visual comparison of the climate response at various global warming levels (e.g., Figure 4.35), the linearity across global warming levels can be assessed by using the highest emissions scenario and comparing seasonal mean relative precipitation changes at +2°C versus +4°C above pre-industrial (1850–1900) temperatures (Figure 8.25). The results support the previous finding (Good et al., 2016b) that a second 2°C warming does not necessarily lead to the same precipitation anomaly pattern as the first 2°C, especially in the tropics where regional differences can be large but not necessarily consistent among different models. They are also consistent with a recent analysis of CMIP5 models showing that the projected drying in the Mediterranean and in Chile is substantially faster than the increase in GSAT, and therefore does not scale linearly with global warming (Zappa et al., 2020).

In summary, there is high confidence that continued global warming will further amplify GHG-induced changes in large-scale atmospheric circulation and precipitation. Nonetheless, there are cases where regional water cycle changes are not linearly related to global warming due to the interaction of multiple forcings, feedbacks and time scales (medium confidence, see also Sections 4.2.4, 7.4.3 and 8.2.1). Aridity in subtropical regions is highly sensitive to fast shifts in large-scale atmospheric circulation so are particularly susceptible to such non-linearities.

Land surface responses and feedbacks represent a potential source of non-linearity for the water cycle response, at least at regional and local scales. The forced response of soil moisture and freshwater resources not only depends on precipitation, but also on evaporation (Laîné et al., 2014), snowmelt (Thackeray et al., 2016), and runoff(X. Zhang et al., 2018) which are intrinsically non-linear processes depending on soil moisture or temperature thresholds. Bare ground evaporation is, for instance, usually estimated as a non-linear function of surface soil moisture (Jefferson and Maxwell, 2015). Plant transpiration requires more complex formulations with non-linear dependencies on multiple environmental factors including root-zone soil moisture and atmospheric CO₂ concentration (Franks et al., 2017). Globally, land surface evaporation is both energy and soil-moisture limited, but one of these limitations can become dominant depending on regions and seasons. Non-linearities may be particularly strong in transitional regimes where and when soil moisture limitation plays a major role (Berg and Sheffield, 2018b).

Snowmelt is a nonl-inear process and projected changes in snowfall are also a non-linear combination of changes in total precipitation and in the fraction of solid precipitation. In cold regions, snowfall may first increase because of the increased water capacity of a warmer atmosphere and then decrease because snow falls as rain in an even warmer atmosphere. Such non-linearities can contribute to elevation, latitudinal and seasonal contrasts in the observed and projected retreat of the Northern Hemisphere (NH) snow cover (Shi and Wang, 2015; Thackeray et al., 2016). Mountain glaciers also represent source of non-linear runoff responses since the annual runoff can first increase due to additional melting and then decrease as the glaciers shrink (Kraaijenbrink et al., 2017; Shannon et al., 2019). Section 9.5.1.3 concludes with high confidence that the average annual runoff from glaciers will generally reach a peak at the latest by the end of the 21st century, and decline thereafter. This peak may have already occurred for small catchments with little ice cover, but tends to occur later in basins with large glaciers. Permafrost thawing is another mechanism which can trigger a non-linear hydrological response in the high latitudes of the NH(Walvoord and Kurylyk, 2016), whose magnitude and potential abruptness is assessed in Section 5.4.3.3.

Land surface runoff and groundwater recharge are highly non-linear process, depending for instance on rainfall intensity, soil infiltration capacity, vertical profile of soil moisture and water table depth. A non-linear relationship between rainfall and groundwater recharge was observed in the tropics where intense seasonal rainfalls associated with internal climate variability contribute disproportionately to recharge (R.G. Taylor et al. , 2013a; Cuthbert et al. , 2019a). Groundwater fluxes in arid regions are generally less responsive to climate variability than in humid regions, which can temporarily buffer climate change impacts on water resources or lead to a long, initially hidden, hydrological responses to global warming (Cuthbert et al., 2019a). Hydrological model simulations driven by individual and combined forcing show that decreased precipitation can cause larger deficits in soil moisture, streamflow and water table depth than other forcings, but also that these factors are not linearly cumulative when applied in combination (Hein et al., 2019). Surface runoff was found to scale only approximately with global warming (Tanaka et al., 2017). Significant non-linearities were found in the projected annual mean runoff response to global warming in CMIP5 projections, which could not be entirely explained by precipitation changes(X. Zhang et al., 2018). Similar non-linear behaviours are found in CMIP6 models over the Amazon, Yangtze, Niger, Euphrates and Mississippi river basins (Figure 8.26), highlighting the need to reassess the assumption of linearity when estimating regional water cycle changes.

Figure 8.26 | Rate of change in basin-scale annual mean runoff with increasing global warming levels. Relative changes (%) in basin-averaged annual mean runoff estimated as multi-model ensemble median from a variable subset of CMIP6 models for each SSP over nine major river basins: (a) Mississippi, (b) Danube, (c) Lena, (d) Amazon, (e) Euphrates, (f) Yangtze, (g) Niger, (h) Indus, and (i) Murray. The basin averages have been estimated after a first-order conservative remapping of the model outputs on the 0.5° by 0.5° river network of Decharme et al. (2019). The shaded area indicates the 5–95% confidence interval of the ensemble values across all SSPs. Note that the y-axis range differs across basins and is particularly large for Niger and Murray (panels g and i). The number of models considered is specified for each scenario in the legend located inside panel b. Further details on data sources and processing are available in the chapter data table (Table 8.SM.1).

Open figure

Beyond changes in land surface water fluxes, non-linearities in the response of soil moisture and freshwater reservoirs have not been well documented in global climate projections but deserve further attention given the complex interactions between the water, energy and carbon cycles (Berg and Sheffield, 2018a), the growing direct human influence on rivers and groundwater (Abbott et al., 2019), and a possible offset between the linear components of changes in precipitation and evapotranspiration. Significant non-linearities were found in water scarcity projections, as seen by the stronger sensitivity to the first 2°C increase in global warming (Gosling and Arnell, 2016).

In summary, there is both numerical and process-based evidence that terrestrial water cycle changes can be non-linear at the regional scale (high confidence). Non-linear regional responses of runoff, groundwater recharge and water scarcity have been documented based on both CMIP5 and CMIP6 models, and highlight the limitations of simple pattern-scaling techniques (medium confidence). Water resources fed by melting glaciers are particularly exposed to such non-linearities (high confidence).

The Amazon forest plays an active role in driving atmospheric moisture transport and generating precipitation in the South American region (SRCCL; Drumond et al. , 2014; Poveda et al. , 2014; Yin et al. , 2014; Staal et al. , 2018, 2020; Agudelo et al. , 2019; Espinoza et al. , 2019). This close association between the land surface and the water cycle makes the Amazon a potential hotspot for abrupt change (Torres and Marengo, 2014). Both deforestation and drying are projected to increase by 2100, resulting in a worst-case scenario of up to a 50% loss in forest cover by 2050 (Soares-Filho et al. , 2006; Boisier et al. , 2015; ter Steege et al. , 2015; Gomes et al. , 2019). Deforestation in the Amazon also raises the probability of catastrophic fires (Brando et al., 2014). The combination of deforestation, drier conditions, and increased fire can push the rainforest ecosystem past a tipping point, beyond which there is rapid land surface degradation, a sharp reduction in atmospheric moisture recycling, an increase in the fraction of precipitation that runs off, and a further shift towards a drier climate (Staal et al. , 2015; Boers et al. , 2017; Zemp et al. , 2017; Ruiz-Vásquez et al. , 2020). A rapid drop in precipitation has a direct impact on river flows, driving basin-scale shifts from a regulated to unregulated state (Salazar et al., 2018). Regional climate modeling experiments confirm that increased deforestation leads to a drier climate, although not all models show a true tipping point, at least under present-day climatic conditions (Lejeune et al., 2015; Spracklen and Garcia-Carreras, 2015).

In AR5, some simulations using a coupled climate–carbon cycle model exhibited an abrupt dieback of the Amazon forest in future climate scenarios (Oyama and Nobre, 2003; Cox et al., 2004; Malhi et al., 2008).However, subsequent work demonstrated that abrupt Amazon dieback does not occur consistently across, or even within, Earth system models (Lambert et al., 2013; Boulton et al., 2017). The occurrence of dieback is highly dependent on both how dry the simulated climate is in the present day (Malhi et al., 2009) as well as the representation of forest structure and competitive dynamics (Levine et al., 2016). Models with a low diversity of plant characteristics and types have a higher tendency for abrupt change (Sakschewski et al., 2016). Abrupt shifts and ecosystem disruptions can occur on the sub-regional level (Pires and Costa, 2013), highlighting the need for higher-resolution modelling studies. Since AR5, CMIP6 projections suggest that a tipping point in the Amazon system may be crossed on a local or regional scale (Staal et al., 2020) but continue to be highly dependent on model biases in precipitation and the simulation of the land surface. Consequently, the timing, and probability, of an abrupt shift remains difficult to ascertain.

In summary, while there is a strong theoretical expectation that Amazon drying and deforestation can cause a rapid change in the regional water cycle, currently there is limited model evidence to verify this response, hence there is low confidence that such a change will occur by 2100.

Greening of the Sahara and Sahel regions in North Africa, in response to an increase in precipitation, has long been considered an amplifying mechanism that can lead to abrupt change. Although the high surface albedo of the desert stabilizes the energy balance of the system (Charney, 1975), greening can induce strong, positive feedbacks between the land surface and precipitation that can shift the region into a ‘Green Sahara’ state. The fact that the transition phase between a Desert Sahara and Green Sahara is not theoretically stable (Brovkin et al., 1998) creates a tipping point and allows for the possibility of an abrupt shift between dry and wet climate regimes. Paleoclimate reconstructions provide evidence of past Green Sahara states (DeMenocal and Tierney, 2012), under which rainfall rates increased by an order of magnitude (Tierney et al., 2017), leading to a vegetated landscape (Jolly et al., 1998) with large lake basins (Gasse, 2000; Drake and Bristow, 2006). The underlying driver of the Green Sahara is the periodic increase in summer insolation associated with the orbital precession cycle (Kutzbach, 1981). In this sense, Green Saharas are not direct analogues for a response to anthropogenic greenhouse gas emissions (GHGs), as these past states were forced by natural, seasonal changes in solar radiation. However, the climate dynamics of Green Sahara periods (which have global impacts, Pausata et al., 2020), and the speed of the transitions between Desert Saharas and Green Saharas, are relevant for future projections.

Since AR5, paleoclimatic studies have improved our view of the timing, spatial extent, and speed of transitions associated with the early Holocene (11,000–5,000 years ago) Green Sahara. Observed transitions into and out of Green Sahara states are always faster than the underlying forcing, in agreement with theoretical considerations (high confidence) (Tierney and DeMenocal, 2013; Shanahan et al., 2015; Tierney et al., 2017). However, there is low confidence in the duration of the transition because sedimentary records cannot typically resolve changes on decadal to multi-decadal time scales (Tierney and DeMenocal, 2013). Both paleoclimate data and modelling experiments suggest that the timing and speed of the transition was spatially heterogeneous (high confidence), with northern Saharan locations becoming drier thousands of years before more equatorial locations (Shanahan et al., 2015; Tierney et al., 2017; Dallmeyer et al., 2020). These observations are consistent with theoretical studies suggesting that spatial heterogeneity and diversity in ecosystems can mitigate the probability of catastrophic change (Van Nes and Scheffer, 2005; Bathiany et al., 2013). Conversely, low ecosystem diversity can produce local or regional ‘hot spots’ of abrupt change such as those seen in some paleoclimate records (Claussen et al., 2013).

CMIP5 and CMIP6 models, some of which include dynamic vegetation schemes, cannot simulate the magnitude, nor the spatial extent, of greening and precipitation change associated with the last Green Sahara under standard mid-Holocene (6,000 years ago) boundary conditions (high confidence) (Figure 3.11; Harrison et al. , 2014; Tierney et al. , 2017; Brierley et al. , 2020). This result remains unchanged since AR4 (Jansen et al., 2007). This may be due to climatological biases in the models (Harrison et al., 2015) or could imply that the strength of the feedbacks between vegetation and the water cycle in the models is too weak (Hopcroft et al., 2017). To date, climate models still only produce the amount and spatial extent of rainfall that is needed to sustain a Green Sahara if they are given prescribed changes in the land surface, such as albedo, soil moisture, vegetation cover and/or dust emissions (Pausata et al., 2016; Skinner and Poulsen, 2016; Tierney et al., 2017).

Some climate model simulations suggest that under future high-emissions scenarios, CO₂ radiative forcing causes rapid greening in the Sahel and Sahara regions via precipitation change (Claussen et al., 2003; Drijfhout et al., 2015). For example, in the BNU-ESM RCP8.5 simulation, the change is abrupt with the percentage of bare soil dropping from 45% to 15%, and percentage of tree cover rising from 50% to 75%, within 10 years (2050–2060; Drijfhout et al., 2015). However, other modelling results suggest that this may be a short-lived response to CO₂ fertilization (Bathiany et al., 2014).

In summary, given outstanding uncertainties in how well the current generation of climate models capture land surface feedbacks in the Sahel and Sahara, there is low confidence that an abrupt change to a greener state will occur in these regions before 2100 or 2300.

Mineral dust aerosols in the climate system originate from both semi-permanent and transient sources (Prospero et al., 2002; Ginoux et al., 2012). The former are typically arid regions where significant alluvial sediments have accumulated over time, while the latter are often associated with natural (e.g., droughts, wildfires) and anthropogenic (e.g., land use change, desertification) disturbances. Modern-day dust emissions are dominated by natural sources (Ginoux et al., 2012), although human emissions may contribute 10–60% of the global atmospheric dust load (Webb and Pierre, 2018). Paleo-dust records suggest that human factors (land use change and landscape disturbance) may have doubled global dust emissions between 1750 and the last quarter of the 20th century (Section 2.2.6; Hooper and Marx, 2018).

Dust aerosols influence the climate system and hydrologic cycle through both direct impacts on radiation (absorbing and scattering longwave and shortwave) and via indirect effects on cloud and precipitation processes (Box 8.1; Choobari et al., 2014; Kok et al., 2018; Schepanski, 2018). The capacity of dust aerosols to suppress precipitation by reducing humidity and energy availability, and increasing stability in the atmosphere (Cook et al., 2013; Huang et al., 2014) can drive positive feedbacks (see also Section 6.3.6). Thus there is strong potential for dust to contribute to abrupt changes in the water cycle, especially in semi-arid regions where wind erosion is highly sensitive to vegetation cover and drought variability (Yu et al., 2015). One such event occurred over the Central USA during the 1930s: the Dust Bowl drought, an iconic event characterized by widespread land degradation and historically unprecedented levels of dust storm activity (Hansen and Libecap, 2004; Lee and Gill, 2015). While initialized by warm sea surface temperatures in the North Atlantic, modeling work indicates that land cover changes and resulting dust emissions contributed to the severity and spatial extent of the drought by further suppressing precipitation (Cook et al., 2009; Hu et al., 2018; Cowan et al., 2020). There is also increasing evidence that dust aerosol feedbacks are necessary to explain the magnitude of rainfall increase during the mid-Holocene Green Sahara (Pausata et al., 2016; Tierney et al., 2017).

The importance of dust aerosol feedbacks in future abrupt climate events, like droughts or rapid aridification, is unclear. In part, this is because the response of dust aerosol emissions and loading levels in the atmosphere to climate change is highly uncertain (Tegen and Schepanski, 2018; Webb and Pierre, 2018). This difficulty in predicting future dust responses is rooted in the fact that emissions depend on both changes to the land surface (e.g., land use/land cover change, aridification, ecological responses to climate change) and the state of the atmosphere (Tegen and Schepanski, 2018). While there is some evidence that global dust aerosol concentrations in the future will increase (Allen et al., 2016; Tegen and Schepanski, 2018), it is highly dependent on changes in precipitation patterns and atmospheric circulation (see the SRCCL, Section 2.4.1), and it is not clear what the radiative impact will be (Allen et al., 2016; Kok et al., 2018).

In summary, due to limited evidence, there is low confidence regarding the role of dust in abrupt climate change events over the next century.