Extremes, Abrupt Changes and Managing Risks

In discussing concepts such as abrupt changes, irreversibility, tipping points and extreme events it is important to define precisely what is meant by those terms. The following definitions are therefore adopted (based on either AR5, Special Report on Global Warming of 1.5°C (SR15) or Special Report on Climate Change and Land (SRCCL) Glossaries):

Abrupt climate change: A large-scale change in the climate system that takes place over a few decades or less, persists (or is anticipated to persist) for at least a few decades, and causes substantial disruptions in human and natural systems.

Extreme weather/climate event: An extreme event is an event that is rare at a particular place and time of year. Definitions of ‘rare’ vary, but an extreme event would normally be as rare as or rarer than the 10th or 90th percentile of a probability density function estimated from observations. By definition, the characteristics of what is called an extreme event may vary from place to place in an absolute sense. When a pattern of extreme weather persists for some time, such as a season, it may be classed as an extreme climate event, especially if it yields an average or total that is itself extreme (e.g., high temperature, drought, or total rainfall over a season).

Irreversibility: A perturbed state of a dynamical system is defined as irreversible on a given timescale, if the recovery timescale from this state due to natural processes is significantly longer than the time it takes for the system to reach this perturbed state. In the context of this report, the recovery time scale of interest is hundreds to thousands of years. 

Tipping point: A level of change in system properties beyond which a system reorganises, often in a nonlinear manner, and does not return to the initial state even if the drivers of the change are abated. For the climate system, the term refers to a critical threshold when global or regional climate changes from one stable state to another stable state. Tipping points are also used when referring to impact; the term can imply that an impact tipping point is (about to be) reached in a natural or human system.

These above four terms generally refer to aspects of the physical climate system. Here we extend their definitions to natural and human systems. For example, there may be gradual physical climate change which causes an irreversible change in an ecosystem. An adaptation tipping point could be reached when an adaptation option no longer remains effective. There may be a tipping point within a governance structure. 

We also introduce two new key terms relevant for discussing risk-related concepts:

Compound events refer to the combination of multiple drivers and/or hazards that contribute to societal or environmental risks.

Cascading impacts from extreme weather/climate events occur when an extreme hazard generates a sequence of secondary events in natural and human systems that result in physical, natural, social or economic disruption, whereby the resulting impact is significantly larger than the initial impact. Cascading impacts are complex and multi-dimensional, and are associated more with the magnitude of vulnerability than with that of the hazard. 

Some potentially abrupt or irreversible events are assessed in other chapters, hence Table 6.1 presents a cross-chapter summary of those. Subsection numbers indicate where detailed information may be found.

The attribution of changes in the observed statistics of extremes are generally addressed using well-established detection-attribution methods. In contrast, record-breaking weather and climate events are by definition unique, and can be expected to occur with or without climate change as the observed record lengthens. Therefore, event attribution begins with the premise that the climate is changing, the goal being to determine statistically how much climate change has contributed to the severity of the event (Trenberth et al., 2015⁵; Shepherd, 2016⁶). Annual reports dedicated to extreme event attribution (Peterson et al., 2012⁷; Peterson et al., 2013⁸; Herring et al., 2014⁹; Herring et al., 2015¹⁰; Herring et al., 2018¹¹) have helped stimulate studies that adopt recognised methods for extreme event attribution. The increasing pool of studies allows different approaches to be contrasted and builds consensus on the role of climate change when individual climate events are studied by multiple teams using different methods. A number of these events are summarised in Table 6.2 and Figure 6.2. Collectively, these studies show that the role of climate change in the ocean and cryosphere extreme events is increasingly driving extreme climate and weather events across the globe including compound events (high confidence). Some regions including Africa and the Pacific have had relatively fewer event attribution studies undertaken, possibly reflecting the lack of capacity by regional and national technical institutions. A caveat of this approach is that there is a potential for ‘null results’, that is, cases where attribution is not possible, to be reported. Nevertheless, there is no evidence that this is the case, and the number of recent studies and wide range of phenomena addressed suggests increasing influence of climate change on extreme events.

As shown in previous assessments, increasing exposure is a major driver of increased cyclone risk (wind damages), as well as flood risk associated with cyclone rainfall and surge, besides possible changes in hazard intensities from anthropogenic climate change (Handmer et al., 2012²⁷³; Arent et al., 2014²⁷⁴). Changes in TC trajectories are potentially a major source of increased risk, as the degree of vulnerability is typically much higher in locations that were previously not exposed to the hazard (Noy, 2016). Typhoon Haiyan’s move to the south of the usual trajectories of TCs in the western North Pacific basin (Yonson et al., 2018²⁷⁵) made the evacuation more difficult as people were less willing to heed storm surge warnings they received. 

Abrupt changes in impacts therefore are not only determined by changes in cyclone hazard, but also by the sensitivity or tipping points that are crossed in terms of flooding for instance, that can be driven by SLR but also by changes in local exposure. The frequency of nuisance flooding along the US east coast is expected to accelerate further in the future (Sweet and Park, 2014²⁷⁶). The loss of coral reef cover and mangrove forests have also been shown to increase damages from storm surge events (e.g., Beck et al., 2018). Cyclones also affect marine life, habitats and fishing. There is some evidence that fish may evacuate storm areas or be redistributed by storm waves and currents (FAO, 2018; Sainsbury et al., 2018²⁷⁷). Other examples of damage to fisheries from cyclones and storm surges can be found in FAO (2018: Chapter V, Table 1).

With regard to property losses, according to most projections, increasing losses from more intense cyclones are not offset by a possible reduction in frequency (Handmer et al., 2012²⁷⁸). While the relation between aggregate damages and frequency may be linear, the relationship between intensity and damages is most probably highly nonlinear; with research suggesting a 10% increase in wind speed associated with a 30–40% increase in damages (e.g., Strobl, 2012). Although it is clear that direct damages from cyclones could increase, investigations into the economic impact of past cyclone events is less common, as these are much more difficult to identify. Examples of such work include Strobl (2012)²⁷⁹ on hurricane impacts in the Caribbean, Haque and Jahan (2016) on TC Sidr in Bangladesh, Jakobsen (2012) on Hurricane Mitch in Nicaragua, and Taupo and Noy (2017) on TC Pam in Tuvalu. The relation between changes in TCs and property losses is complex, and there are indications that wind shear changes may have larger impact than changes in global temperatures (Wang and Toumi, 2016²⁸⁰). With regard to loss of life, total fatalities and mortality from cyclone-related coastal flooding is globally declining, probably as a result of improved forecasting and evacuation, although in some low-income countries mortality is still high (Paul, 2009²⁸¹; Lumbroso et al., 2017²⁸²; Bouwer and Jonkman, 2018²⁸³). A global analysis finds that despite adaptation efforts, further SLR could increase storm surge mortality in many parts of the developing world (Lloyd et al., 2016²⁸⁴). 

An assessment of future changes in coastal impacts based on direct downscaling of indicators of flooding such as total water level and number of hours per year with breakwater overtopping over a given threshold for port operability is provided by Camus et al. (2017)²⁸⁵. These indicators are multivariable and include the combined effect of SLR, storm surge, astronomical tide and waves. Regional projected wave climate is downscaled from global multi-model projections from 30 CMIP5 model realisations. For example, projections by 2100 under the RCP8.5 scenario show a spatial variability along the coast of Chile with port operability loss between 600–800 h yr–1 and around 200 h yr–1 relative to present (1979–2005) conditions. Although wave changes are included in projected overtopping distributions, future changes of operability are mainly due to the SLR contribution.

The most effective risk management strategy in the last few decades has been the development of early warning systems for cyclones (Hallegatte, 2013²⁸⁶). Generally, however, a lack of familiarity with the changed nature of storms prevails. Powerful storms often generate record storm surges (Needham et al., 2015²⁸⁷), such as in the cases of Cyclone Nargis and Typhoon Haiyan but surge warnings had been less well understood and followed because they had tended to be new or rare to the locality (Lagmay et al., 2015²⁸⁸). A US study on storm surge warnings highlights the issue of the right timing to warn, as well as the difficulty in delivering accurate surge maps (Morrow et al., 2015²⁸⁹). Previous experience with warnings that were not followed by hazard events show the ‘crying wolf’ problem leading many to ignore future warnings (Bostrom et al., 2018²⁹⁰).

There is scant literature on the management of storms that follow less common trajectories. The most recent and relatively well-studied ones are Superstorm Sandy in 2012 in the USA and Typhoon Haiyan in 2013 in the Philippines. These two storms were unexpected and having underestimated the levels of impact, people ignored warnings and evacuation directives. In the case of Typhoon Haiyan, the dissemination of warnings via scripted text messages were ineffective without an explanation of the difference between Haiyan’s accompanying storm surge and that of other ‘normal’ storms to which people were used to (Lejano et al., 2016²⁹¹). Negative experiences of previous evacuations also lead to the reluctance of authorities to issue mandatory evacuation orders, for example, during Superstorm Sandy (Kulkarni et al., 2017²⁹²), and contributes to a preventable high number of casualties (Dalisay and De Guzman, 2016²⁹³). These examples also show that saving lives and assets through warning and evacuation is limited. Providing biophysical protection measures as well as improving self-reliance during such events can complement warning and evacuation. 

After the storms, retreat or rebuild options exist. Rebuilding options can depend on whether insurance is still affordable after the event. Buyout programs, a form of ‘managed retreat’ whereby government agencies pay people affected by extreme weather events to relocate to safer areas, gained traction in recent years as a potential solution to reduce exposure to changing storm surge and flood risk. The decision to retreat or rebuild in situ depends, at least partially, on how communities have recovered in the past and therefore on the perceived success of a future recovery (Binder, 2014²⁹⁴). However, political and jurisdictional conflicts between local, regional, and national government over land management responsibilities, lack of coordinated nation-wide adaptation plans, and clashes between individual and community needs have led to some unpopular buyout programs after Hurricane Sandy (Boet-Whitaker, 2017²⁹⁵). Relocation (i.e., managed retreat) is often very controversial, can incur significant political risk even when it is in principle voluntary (Gibbs et al., 2016²⁹⁶), and is rarely implemented with much success at larger scales (Beine and Parsons, 2015²⁹⁷; Hino et al., 2017²⁹⁸). In addition, managed retreats are often fraught with legal, distributional and human rights issues, as seen in the case of resettlements after Typhoon Haiyan (Thomas, 2015²⁹⁹; see also Cross-Chapter Box 5 in Chapter 1), and extend to loss of cultural heritage and indigenous qualities in the case of small island states. 

If rebuilding in situ is pursued after catastrophic events and without decreased exposure, it is often accompanied by actions that aim to reduce vulnerability in order to adapt to the increasing risk (Harman et al., 2013³⁰⁰). In many cases, resilient designs and sustainable urban plans integrating climate change concerns, that are inclusive of vegetation barriers as coastal defences and hybrid designs, are considered (Cheong et al., 2013³⁰¹; Saleh and Weinstein, 2016³⁰²). However, often more physical structures that are known to be less sustainable in the long-term, but potentially more protective in the short-term, are constructed (Knowlton and Rotkin-Ellman, 2014³⁰³; Rosenzweig and Solecki, 2014³⁰⁴). Anticipatory planning approaches are under way to warn and enable decision making in time (Bloemen et al., 2018³⁰⁵; Lawrence et al., 2018³⁰⁶).

Risk management strategies to respond to MHWs include early warning systems as well as seasonal (weeks to several months) and multi-annual predictions systems. Since 1997, the National Oceanic and Atmospheric Administration’s (NOAA) Coral Reef Watch has used satellite SST data to provide near real-time warning of coral bleaching (Liu et al., 2014a⁴⁶¹). These satellite-based products, along with NOAA Coral Reef Watch’s four month coral bleaching outlook based on operational climate forecast models (Liu et al., 2018⁴⁶²), and coral disease outbreak risk (Heron et al., 2010⁴⁶³) provide critical guidance to coral reef managers, scientists, and other stakeholders (Tommasi et al., 2017b⁴⁶⁴; Eakin et al., 2018⁴⁶⁵). These products are also used to implement proactive bleaching response plans (Rosinski et al., 2017⁴⁶⁶), brief stakeholders, and allocate monitoring resources in advance of bleaching events, such as the 2014–2017 global coral bleaching event (Eakin et al., 2017⁴⁶⁷). For example, Thailand closed ten reefs for diving in advance of the bleaching peak in 2016, while Hawaii immediately began preparation of resources both to monitor the 2015 bleaching and to place specimens of rare corals in climate controlled, onshore nurseries in response to these forecast systems (Tommasi et al., 2017b⁴⁶⁸). New measurement techniques, such as Argo and deep Argo floats, may help to further develop prediction systems for subsurface MHWs, but such systems are not yet in place.

SST forecasts ranging from seasonal to decadal (5–10 years) have also been used or are planned to be used as early warning systems for multiple other ecosystems and fisheries in addition to coral reefs, including aquaculture, lobster, sardine, and tuna fisheries (Hobday et al., 2016b⁴⁶⁹; Payne et al., 2017⁴⁷⁰; Tommasi et al., 2017b⁴⁷¹). For example, seasonal forecasts of SST around Tasmania may help farm managers of salmon aquaculture to prepare and respond to upcoming MHWs by changing stocking densities, varying feed mixes, transferring fish to different locations in the farming region and implementing disease management (Spillman and Hobday, 2014⁴⁷²; Hobday et al., 2016b⁴⁷³). Skilful multi-annual to decadal SST predictions may also inform and improve decisions about spatial and industrial planning, as well as the management of various extractive sectors such as the adjustments to quotas for internationally shared fish stocks (Tommasi et al., 2017a⁴⁷⁴). It has been shown that global climate forecasts have significant skill in predicting the occurrence of above average warm or cold SST events at decadal timescales in coastal areas (Tommasi et al., 2017a⁴⁷⁵), but barriers to their widespread usage in fishery and aquaculture industry still exist (Tommasi et al., 2017b⁴⁷⁶). 

Even with a monitoring and prediction system in place, MHWs have developed without warning and had catastrophic effects (Payne et al., 2017⁴⁷⁷). For example, governmental agencies, socioeconomic sectors, public health officials and citizens were not forewarned of the Coastal Peruvian 2017 MHW, despite a basin-wide monitoring system across the Pacific. The reason was partly due to a coastal El Niño definition problem and a new government (in Nicaragua) that may have hindered actions (Ramírez and Briones, 2017⁴⁷⁸). Therefore, early warning systems should not only provide predictions of physical changes, but should also connect different institutions to assist decision makers in performing time-adaptive measures (Chang et al., 2013⁴⁷⁹). 

Monitoring and prediction systems are important and can be advanced by the use of common metrics to describe MHWs. So far, MHWs are often defined differently in the literature, and it is only recently that a categorising scheme (Categories I to IV; based on the degree to which temperatures exceed the local climatology), similar to what is used for hurricanes, has been developed (Hobday et al., 2018⁴⁸⁰). Such a categorising scheme, can easily be applied to real data and predictions, and may facilitate comparison, public communication and familiarity with MHWs. Similar metrics (e.g., DHW) have been successfully developed and used to identify ocean regions where conditions conducive to coral bleaching are developing. 

Increasing frequency of extreme ENSO and IOD events have the potential to have widespread impacts on natural and human systems in many parts of the globe. Though the occurrence of the extreme 2015–2016 El Niño has produced a large body of literature, it is still not clear how climate change may have altered such an impact, nor how such impacts might change in the future with increasing frequency of extreme ENSO events. We highlight here some studies that have attempted to assess the joint impact of mean change and variability. In addition to observed high variability of rainfall, severe weather events and impacts on TCs activity (Yonekura and Hall, 2014⁵⁴¹; Zhang and Guan, 2014⁵⁴²; Wang and Liu, 2016⁵⁴³; Zhan, 2017⁵⁴⁴), extreme El Nino events have substantial impacts on natural systems which include those on marine ecosystems (Sanseverino et al., 2016⁵⁴⁵; Mogollon and Calil, 2017⁵⁴⁶; Ohman et al., 2017⁵³⁷), such as severe and repeated bleaching of corals (Hughes et al., 2017a⁵⁴⁸; Hughes et al., 2017b⁵⁴⁹; Eakin et al., 2018⁵⁵⁰), and glacial growth and retreat (Thompson et al., 2017⁵⁵¹). On the other hand, impacts on human, including managed systems are: increased incidences of forest fires (Christidis et al., 2018b⁵⁵²; Tett et al., 2018⁵⁵³), degraded air quality (Koplitz et al., 2015⁵⁵⁴; Chang et al., 2016⁵⁵⁵; Zhai et al., 2016⁵⁵⁶) such as the dense haze over most parts of Indonesia and the neighbouring countries in Southeast Asia as a result of prolonged Indonesian wildfires, thus imposing adverse impacts on public health in the affected areas (Koplitz et al., 2015⁵⁵⁷; WMO, 2016), decreased agricultural yields in many parts of the globe (e.g., in most of the Pacific Islands countries, Thailand, eastern and southern Africa and others which had resulted food insecurity, particularly in eastern and southern Africa (UNSCAP, 2015; WMO, 2016; Christidis et al., 2018b⁵⁵⁸; Funk et al., 2018⁵⁵⁹), and regional uptick in the number of reported cases of plague and hantavirus in Colorado and New Mexico, cholera in Tanzania, dengue in Brazil and Southeast Asia (Anyamba et al., 2019⁵⁶⁰) and Zika virus in South America (Caminade et al., 2017⁵⁶¹), including increases in heat stroke cases (Christidis et al., 2018b⁵⁶²). Substantial economic losses had resulted from droughts and floods across various parts of the globe due to teleconnections. For instance, direct losses of 10 billion USD (Sun and Miao, 2018⁵⁶³; Yuan et al., 2018⁵⁶⁴) and 6.5 billion USD (Christidis et al., 2018b⁵⁶⁵) were estimated to have been incurred from severe urban inundation in cities along the Yangtze River in China and the extreme drought in Thailand, respectively. 

ENSO events affect TCs activity through variations in the low-level wind anomalies, vertical wind shear, mid-level relative humidity, steering flow, the monsoon trough and the western Pacific subtropical high in Asia (Yonekura and Hall, 2014⁵⁶⁶; Zhang and Guan, 2014⁵⁶⁷). The subsurface heat discharge due to El Niño can intensify TCs in the eastern Pacific (Jin et al., 2014⁵⁶⁸; Moon et al., 2015b⁵⁶⁹). TCs are projected to become more frequent (~20–40%) during future-climate El Niño events compared with present climate El Niño events (medium confidence), and less frequent during future-climate La Niña events, around a group of small island nations (for example, Fiji, Vanuatu, Marshall Islands and Hawaii) in the Pacific (Chand et al., 2017⁵⁷⁰). The Indian Ocean basin-wide warming has led to an increase in TC heat potential in the Indian Ocean over the last 30 years, however the link to the changes in the frequency of TCs is not robust (Rajeevan et al., 2013⁵⁷¹).

During the early stages of an extreme El Niño event (2015–2016 El Niño), there is an initial decrease in atmospheric CO2 concentrations over the tropical Pacific Ocean, due to suppression of equatorial upwelling, reducing the supply of CO2 to the surface (Chatterjee et al., 2017⁵⁷²), followed by a rise in atmospheric CO2 concentrations due reduced terrestrial CO2 uptake and increased fire emissions (Bastos et al., 2018⁵⁷³). It is not clear how a future increase in the frequency extreme events would modulate the carbon cycle on longer decadal time scales.

Studies on projections of changes in ENSO impacts or teleconnections are rather limited. Nevertheless, Power and Delage (2018) provide a multi-model assessment of CMIP5 models and their simulated changes in the precipitation response to El Niño in the future (Figure 6.6). They identify different combinations of changes that might further impact natural and human systems. El Niño causes either positive or negative precipitation anomalies in diverse regions of the globe. Dry El Niño teleconnection anomalies may be further strengthened by, either mean climate drying in the region (Amazon, Central America and Australia in June to August (JJA)), or a strengthening of the El Niño dry teleconnection, or both. Conversely, wet El Niño teleconnections can be further strengthened by either increases in mean precipitation (East Africa and southeastern South America in December to February (DJF)) or a strengthening of the El Niño wet teleconnection (southeastern South America in JJA), or both (Tibetan Plateau, DJF). However, a present day dry El Niño response may be dampened by a wet mean response (South, East and Southeast Asia in JJA) or a wet present day El Niño response may be weakened by a dry mean change (Southern Europe/Mediterranean and West Coast South America in JJA). Finally, changes in the mean and El Niño response may be in the opposite direction (Southeast Asia, JJA and Central North America, DJF). Such changes could have an impact on phenomena such as wildfires (Fasullo et al., 2018⁵⁷⁵). However, in many other regions that are currently impacted by El Niño, e.g., regions of South America, studies have found no significant changes in the ENSO-precipitation relationship (Tedeschi and Collins, 2017⁵⁷⁶) and agreement between models for many regions suggests low confidence in projections of teleconnection changes (Yeh et al., 2018⁵⁷⁷).

Along with extreme El Niño events, abrupt warming in the Indian Ocean and extreme IOD events have largely altered the Asian and African monsoon, impacting the food and water security over these regions. As a response to rising global SSTs and partially due to extreme El Niño events, the NH summer monsoon showed substantial intensification during 1979–2011, with an increase in rainfall by 9.5% per degree Celsius of global warming (Wang et al., 2013⁵⁷⁸). However, the Indian summer monsoon circulation and rainfall exhibits a statistically significant weakening since the 1950s. This weakening has been hypothesised to be a response to the Indian Ocean basin-wide warming (Mishra et al., 2012⁵⁷⁹; Roxy et al., 2015⁵⁸⁰) and also to increased aerosol emissions (Guo et al., 2016⁵⁸¹) and changes in land use (Paul et al., 2016⁵⁸²). Warming in the north Indian Ocean has resulted in increasing fluctuations in the southwest monsoon winds and a three-fold increase in extreme rainfall events across central India (Roxy et al., 2017⁵⁸³). The frequency and duration of heatwaves have increased over the Indian subcontinent, and these events are associated with the Indian Ocean basin-wide warming and frequent El Niños (Rohini et al., 2016⁵⁸⁴). In April 2016, as a response to the extreme El Niño, Southeast Asia experienced surface air temperatures that surpassed national records, increased energy consumption, disrupted agriculture and resulted in severe human discomfort (Thirumalai et al., 2017⁵⁸⁵). A strong negative IOD event in 2016 led to large climate impact on East African rainfall, with some regions recording below 50% of normal rainfall, leading to devastating drought, food insecurity and unsafe drinking water for over 15 million people in Somalia, Ethiopia and Kenya.

Risk management of ENSO events has focussed on two main aspects: better prediction and early warning systems, and better mechanisms for reducing risks to agriculture, infrastructure, fisheries and aquaculture, wildfire and flood management. Extreme ENSO events are rare, with three such events since 1950 and they are difficult to predict due to the different drivers influencing them (Puy et al., 2017⁵⁸⁶). The impacts of ENSO events also vary between events and between the different regions affected (Murphy et al., 2014⁵⁸⁷; Fasullo et al., 2018⁵⁸⁸; Power and Delage, 2018⁵⁸⁹) however, there is limited literature on the change in the impacts of extreme ENSO compared to other ENSO events. In addition, there are also no specific risk management and adaptation strategies for human and natural systems for more extreme events other than what is in place for ENSO events (see also Chapter 4, Section 4.4 for the response to sea level change, an observed impact of ENSO). A first step in risk management and adaptation is thus to better understand the impacts these events have and to identify conditions that herald such extreme events that could be used to better predict extreme ENSO events. 

Monitoring and forecasting are the most developed ways to manage extreme ENSOs. Several systems are already in place for monitoring and predicting seasonal climate variability and ENSO occurrence. However, the sustainability of the observing system is challenging and currently the Tropical Pacific Observing System 2020 (TPOS 2020) has the task of redesigning such a system, with ENSO prediction as one of its main objectives. These systems could be further elaborated to include extreme ENSO events. Westerly wind events in the Western Tropical Pacific, (Lengaigne et al., 2004⁵⁹⁰; Chen et al., 2015a⁵⁹¹; Fedorov et al., 2015⁵⁹²) strong easterly wind events in the tropical Pacific (Hu and Fedorov, 2016⁵⁹³; Puy et al., 2017⁵⁹⁴), nonlinear interaction between air-sea fluxes and atmospheric deep convection (Bellenger et al., 2014⁵⁹⁵; Takahashi and Dewitte, 2016⁵⁹⁶) and advection of mean temperature by anomalous eastward zonal currents (Kim and Cai, 2014⁵⁹⁷) are some of the factors that play an important role in the evolution of extreme ENSO events, which can be considered while improving the monitoring and forecasting system. 

Despite the specificity of each extreme El Niño event, their forecasting is expected to improve through monitoring of recently identified precursory signals that peak in a window of two years before the event (Varotsos et al., 2016⁵⁹⁸). An early warning system for coral bleaching associated, among other stressors, with extreme ENSO heat stress is provided by the NOAA Coral Reef Watch service with a 5 km resolution (Liu et al., 2018⁵⁹⁹). The impacts of ENSO-associated extreme heat stress are heterogeneous, indicating the influence of other factors either biotic such as coral species composition, local adaptation by coral taxa reef depth or abiotic such as local upwelling or thermal anomalies (Claar et al., 2018⁶⁰⁰). When identified and quantified, these factors can be used for risk analysis and risk management for these ecosystems.

In principle, it is easier to transfer the financial risk associated with extreme ENSO events through, for example, insurance products or other risk transfer instruments such as Catastrophe Bonds, than for more moderate events. An accurate prediction system is not required, but the measurement of these events, and quantification of likely impacts is required. As in other types of insurance systems, this can be done through, for example, calculations of average annual losses associated with extreme ENSO, and the design of appropriate financial instruments. Examples of research that can support the design of risk transfer instruments include Anderson et al. (2018) and Gelcer et al. (2018) for specific crops yields, and Aguilera et al. (2018) and Broad et al. (2002) for specific fisheries. Several risk transfer instruments have been implemented to deal with ENSO impacts, including parametric insurance based on SSTs for heavy rainfall damages, and another scheme for agricultural damages, both in Peru. Other examples include forecast-based financial aid (Red Cross Climate Centre, 2016). More broadly, other forms of risk management and governance can be designed with better information about the likely impacts of extreme ENSO events (e.g., Vignola et al., 2018).

In the last two decades, total water transport from the Pacific to the Indian Ocean and the Indian Ocean to the Atlantic Ocean has increased (high confidence). Increased ITF has been attributed to Pacific cooling and basin-wide warming in the Indian Ocean. The ITF annual average is 15 x 106 m3 s–1 (Susanto et al., 2012). ITF varies from intraseasonal to decadal time scales. On seasonal time scale, South China Sea Throughflow controls freshwater flux and modulates the main ITF (Fang et al., 2010⁶⁰⁶; Susanto et al., 2013⁶⁰⁷; Lee et al., 2019⁶⁰⁸; Wang et al., 2019⁶⁰⁹; Wei et al., 2019⁶¹⁰). During the extreme El Niño of 1997–1998, the ITF transport was reduced to 9.2 x 106 m3 s–1. Based on observations and proxy records from satellite altimetry and gravimetry, in the last two decades, 1992–2012, ITF has been stronger (Sprintall and Revelard, 2014⁶¹¹; Liu et al., 2015a⁶¹²; Susanto and Song, 2015⁶¹³), which translates to an increase in ocean heat-flux into the Indian Ocean (Lee et al., 2015b⁶¹⁴). Exchanges of heat and fresh water between ocean basins are important at the global scale (Flato et al., 2013⁶¹⁵). ITF may have played a key role in the slowdown of the Pacific SST warming during 1998–2013, and the rapid warming in the surface and subsurface Indian Ocean during this period (Section 6.5.1.2; Makarim et al., 2019⁶¹⁶), by transferring warm water from the western Pacific into the Indian Ocean (Lee et al., 2015b⁶¹⁷; Dong and McPhaden, 2018⁶¹⁸). 

Under 1.5°C warming both El Niño and La Niña frequencies may increase (see Section 6.5) and hence ITF variability may also increase. ITF is also influenced by the IOD events, with an increase in transport during a positive IOD and vice-versa during a negative IOD event (Potemra and Schneider, 2007⁶¹⁹; Pujiana et al., 2019⁶²⁰). Positive IODs are projected to increase threefold in the 21st century as a response to changes in the mean state rather than changes in the El Niño frequency (Section 6.5.1.2; Cai et al., 2014b⁶²¹) and this may have an impact on the ITF, additional to the changes due to increasing extreme ENSO events. In response to greenhouse warming, climate models predict that on interannual time scale, it is likely that the mean ITF may decrease due to wind variability (Sen Gupta et al., 2016), but recent observation trend tends to strengthen which has led to speculations about the fidelity of the current climate models (Chung et al., 2019⁶²²). On multidecadal and centennial timescales, it is likely that mean ITF decreases which is not associated with wind variability but due to reduction of net deep ocean upwelling in the tropical South Pacific (Sen Gupta et al., 2016; Feng et al., 2017⁶²³; Feng et al., 2018⁶²⁴). Due to a lack of long-term sustained ITF observations, their impacts on Indo-Pacific climate varibility, biogeochemisty, ecosystem as well as society are not fully understood.

Pacific SST cooling trends and strengthened the equatorial Pacific trade winds have been linked to anomalously warm tropical Indian and Atlantic oceans. The period following the mid-1990s saw a marked strengthening of both the easterly trade winds in the central equatorial Pacific (Figure 6.7) and the Walker circulation (L’Heureux et al., 2013; England et al., 2014⁶²⁵). Both the magnitude and duration of this trend are large when compared with past variability reconstructed using atmosphere reanalyses. (The 1886–1905 extreme weakening trend is poorly constrained by observations and we note the disparity between reanalysis products going back in time.) Moreover, it is very unusual when model simulations are used as an estimate of internal climate variability (Figure 6.7; England et al., 2014⁶²⁹; Kociuba and Power, 2015⁶³⁰). The slowdown in global surface warming is dominated by the cooling in the Pacific SSTs, which is associated with a strengthening of the Pacific trade winds (Kosaka and Xie, 2013⁶³¹). This pattern leads to cooling over land and possibly to additional heat uptake by the ocean, although recent studies suggest that ocean heat uptake may even slow down during surface warming slowdown periods (Xie et al., 2016⁶³²; von Känel et al., 2017). The intensification of the Pacific trade winds has been related to inter-ocean basin SST trends, with rapid warming in the Indian (see section 6.5.1.2) and Atlantic Oceans both hypothesised as drivers (Kucharski et al., 2011⁶³³; Luo et al., 2012⁶³⁴; McGregor et al., 2014⁶³⁵; Zhang and Karnauskas, 2017⁶³⁶). While the extreme event of strengthening trade winds are potentially a result of natural internal variability, a role of anthropogenic contribution has not been ruled out. Nevertheless, the CMIP5 models indicate no general change in trends into the future (Figure 6.7), giving more weight to natural internal variability as an explanation. 

Among the number of potential causes of this decadal variability in surface global temperature, a prolonged negative phase of the Pacific Decadal Oscillation/Interdecadal Pacific Oscillation (PDO/IPO) was suggested as a contributor. Because of the magnitude and duration of this Pacific-centred variability (Figure 6.7), it is identified as an extreme decadal climate event. One line of research has explored the role of the warm tropical Atlantic decadal variability in forcing the trade wind trends and associated cooling Pacific SST trends (Kucharski et al., 2011⁶³⁷; McGregor et al., 2014⁶³⁸; Li et al., 2016b⁶³⁹). It appears that climate models may misrepresent this link due to tropical Atlantic biases (Kajtar et al., 2018⁶⁴⁰; McGregor et al., 2018⁶⁴¹) and thus potentially underestimate global mean temperature decadal variability. Nevertheless, there is no indication that such an underestimation of global temperature variability is evident in the models (Flato et al., 2013⁶⁴²; Marotzke and Forster, 2015⁶⁴³). The impact of modes of natural variability on global mean temperature decadal variability remains an active area of research.

In the Indian Ocean, water exits the Indonesian Seas mostly flowing westward along with the South Equatorial Current, and some supplying the Leeuwin Current. The South Equatorial Current feeds the heat and biogeochemical signatures from the Indian Ocean into the Agulhas Current, which transports it further into the Atlantic Ocean. Observations of Mozambique Channel inflow from 2003–2012 measured a mean transport of 16.7 x 106 m3 s–1 with a maximum in austral winter, and IOD related interannual variability of 8.9 x 106 m3 s–1 (Ridderinkhof et al., 2010⁶⁴⁴). A multidecadal proxy, from three years of mooring data and satellite altimetry, suggests that the Agulhas Current has been broadening since the early 1990s due to an increase in eddy kinetic energy (Beal and Elipot, 2016⁶⁴⁵). Numerical model experiments suggest an intensification of the Agulhas leakage since the 1960s, which has contributed to the warming in the upper 300 m of the tropical Atlantic Ocean (Lübbecke et al., 2015⁶⁴⁶). Agulhas leakage is found to covary with the AMOC on decadal and multi-decadal timescales and has likely contributed to the AMOC slowdown (Biastoch et al., 2015⁶⁴⁷; Kelly et al., 2016⁶⁴⁸). Meanwhile, climate projections indicate that Agulhas leakage is likely to strengthen and may partially compensate the AMOC slowdown projected by coarse-resolution climate models (Loveday et al., 2015⁶⁴⁹).

Interannual to decadal variability of Indo-Pacific SST variability is likely to affect extreme hydroclimate in East Africa (Ummenhofer et al., 2018⁶⁵⁰). The Pacific cooling pattern is often synonymous with predominance of La Niña events in 1998 and 2012 is linked to megadroughts in the USA (Baek et al., 2019⁶⁵¹). On decadal to multidecadal time scales, PDO/IPO and Atlantic variability may have impacts on megadroughts in North America (Coats et al., 2016⁶⁵²; Diodato et al., 2019⁶⁵³) and Australia (Vance et al., 2015⁶⁵⁴) as well as Indian subcontinent (Bao et al., 2015⁶⁵⁵; Joshi and Rai, 2015⁶⁵⁶). It is likely that occurrence of megadroughts in North America and Australia increased (Kiem et al., 2016⁶⁵⁷; Baek et al., 2019⁶⁵⁸). PDO and North Pacific Gyre Oscillation may also influence the decadal variability of North Pacific nutrient, chlorophyll and zooplankton taxa (Di Lorenzo et al., 2013).

The Pacific cooling pattern may have significant impacts on terrestrial carbon uptake via teleconnections. The reduced ecosystem respiration due to the smaller warming over land has significantly accelerated the net biome productivity and therefore increased the terrestrial carbon sink (Ballantyne et al., 2017⁶⁵⁹) and paused the growth rate of atmospheric CO2 despite increasing anthropogenic carbon emissions (Keenan et al., 2016⁶⁶⁰). During the 2000s, the global ocean carbon sink has also strengthened (Fay and McKinley, 2013⁶⁶¹; Landschützer et al., 2014⁶⁶²; Majkut et al., 2014⁶⁶³; Landschützer et al., 2015⁶⁶⁴; Munro et al., 2015⁶⁶⁵), reversing a trend of stagnant or declining carbon uptake during the 1990s. It has been suggested that the upper ocean overturning circulation has weakened during the 2000s thereby decreasing the outgassing of natural CO2, especially in the Southern Ocean (Landschützer et al., 2015⁶⁶⁶), and enhanced the global ocean CO2 sink (DeVries et al., 2017⁶⁶⁷). How this is connected to the global warming slowdown is currently unclear.

Even though the AMOC is very unlikely to collapse over the 21st century, its weakening may be substantial, which may therefore induce strong and large-scale climatic impacts with potential far-reaching impacts on natural and human systems (e.g., Good et al., 2018). Furthermore, the SPG subsystem has been shown to potentially shift, in the future, into a cold state over a decadal time scale, with significant climatic implications for the North Atlantic bordering regions (Sgubin et al., 2017). There have been far more studies analysing impacts on climate of an AMOC weakening than SPG collapse. We will thus in the following mainly depict impacts of an AMOC substantial weakening. 

The AR5 report concludes that based on palaeoclimate data, large changes in the Atlantic Ocean circulation can cause worldwide climatic impacts (Masson-Delmotte et al., 2013), with notably, for an AMOC weakening, a cooling of the North Atlantic, a warming of the South Atlantic, less evaporation and therefore precipitation over the North Atlantic, and a shift of the ITCZ. Impacts of AMOC or SPG changes and their teleconnections in the atmosphere and ocean are supported by a large amount of palaeo-evidence (Lynch-Stieglitz, 2017). Such impacts and teleconnections have been further evaluated over the last few years both using new palaeo-data and higher resolution models. Furthermore, multi-decadal variations in SST observed over the last century, the so-called Atlantic Multidecadal Variability (AMV) or Atlantic Multidecadal Oscillation (AMO), also provide observational evidence of potential impacts of changes in ocean circulation. Nevertheless, due to a lack of long-term direct measurements of the Atlantic Ocean circulation, the exact link between SST and circulation remains controversial (Clement et al., 2015; Zhang, 2017). 

The different potential impacts of large changes in the Atlantic Ocean circulation are summarised in Figure 6.10. Based on variability analysis, it has been shown that a decrease in the AMOC strength has impacts on storm track position and intensity in the North Atlantic (Gastineau et al., 2016), with a potential increase in the number of winter storms hitting Europe (Woollings et al., 2012; Jackson et al., 2015), although some uncertainty remains with respect to the models considered (Peings et al., 2016). The influence on the Arctic sea ice cover has also been evidenced at the decadal scale, with a lower AMOC limiting the retreat of Arctic sea ice (Yeager et al., 2015; Delworth and Zeng, 2016). The climatic impacts could be substantial over Europe (Jackson et al., 2015), where an AMOC weakening can lead to high pressure over the British Isles in summer (Haarsma et al., 2015), reminiscent of a negative summer NAO, inducing an increase in precipitation in Northern Europe and a decrease in Southern Europe. In winter, the response of atmospheric circulation may help to reduce the cooling signature over Europe (Yamamoto and Palter, 2016), notably through an enhancement of warming maritime effect due to a stronger storm track (Jackson et al., 2015), driving more powerful storms in the North Atlantic (Hansen et al., 2016). The observed extreme low AMOC in 2009–2010, which was followed by a reduction in ocean heat content to the north (Cunningham et al., 2013), has been possibly implicated in cold European weather events in winter 2009–2010 and December 2010 (Buchan et al., 2014) although a robust attribution is missing. In summer, cold anomalies in the SPG, like the one occurring during the so-called cold blob (Josey et al., 2018), have been suspected to potentially enhance the probability of heatwaves over Europe in summer (Duchez et al., 2016). Nevertheless, considerable uncertainties remain with regard to this aspect due to the lack of historical observations before 2004 and due to poor model resolution of small-scale processes related to frontal dynamics around the Gulf Stream region (Vanniere et al., 2017). In addition, oceanic changes in the Gulf Stream region may occur in line with AMOC weakening (Saba et al., 2016) with potential rapid warming due to a northward shift of the Gulf Stream. However, these changes are largely underestimated in coarse resolution models (Saba et al., 2016) . In North America, a negative phase of the AMV, reminiscent of a weakening of the AMOC, lowers agricultural production in a few Mexican coastal states (Azuz-Adeath et al., 2019).

Changes in ocean circulation can also strongly impact sea level in the regions bordering the North Atlantic (McCarthy et al., 2015a; Palter et al., 2018). A collapse of the AMOC or of the SPG could induce substantial increase of sea level up to a few tens of centimetres along the western boundary of the North Atlantic (Ezer et al., 2013; Little et al., 2017; cf. Chapter 5). For instance, such a link may explain 30% of the extreme observed SLR event (a short-lived increase of 12 mm during 2 years) in northeast America in 2009–2010 (Ezer, 2015; Goddard et al., 2015). This illustrates that monitoring changes in AMOC may have practical implications for coastal protection.

The AMOC teleconnections are widespread and notably strongly affect the tropical area, as evidenced in palaeo-data for the Sahel region (Collins et al., 2017; Mulitza et al., 2017) and in model simulations (Jackson et al., 2015; Delworth and Zeng, 2016). These teleconnections may affect vulnerable populations. For instance, Defrance et al. (2017) found that a substantial decrease in the AMOC, at the very upper end of potential changes, may strongly diminish precipitation in the Sahelian region, decreasing the millet and sorghum emblematic crop production, which may impact subsistence of tens of millions of people, increasing their potential for migration. Smaller amplitude variations in Sahelian rainfall, driven by North Atlantic SST, has been found to be predictable up to a decade ahead (Gaetani and Mohino, 2013; Mohino et al., 2016; Sheen et al., 2017), potentially providing mitigation and adaptation opportunities. The number of tropical storms in the North Atlantic has been found to be very sensitive to the AMOC (Delworth and Zeng, 2016; Yan et al., 2017) as well as to the SPG (Hermanson et al., 2014) variations, so that a large weakening of the AMOC or cooling of the SPG may decrease the number of Atlantic tropical storms. The Asian monsoon may also potentially weaken in the case of large changes in the AMOC (Marzin et al., 2013; Jackson et al., 2015; Zhou et al., 2016; Monerie et al., 2019) implying substantial adverse impacts on populations. The interactions of the Atlantic basin with the Pacific has also been largely discussed over the last few years, with the supposed influence of a cool North Atlantic inducing a warm tropical Pacific (McGregor et al., 2014; Chafik et al., 2016; Li et al., 2016b), although not found in all models (Swingedouw et al., 2017), which may induce stronger amplitudes of El Niño (Dekker et al., 2018).

The AMOC plays an important function in transporting excess heat and anthropogenic carbon from the surface to the deep ocean (Kostov et al., 2014; Romanou et al., 2017), and therefore in setting the pace of global warming (Marshall et al., 2014). A large potential decline in the AMOC strength reduces global surface warming. This is due to changes in the location of ocean heat uptake and associated expansion of the cryosphere around the North Atlantic, which increases surface albedo (Rugenstein et al., 2013; Winton et al., 2013), as well as cloud cover variations and modifications in water vapour content (Trossman et al., 2016). As the uptake of excess heat occurs preferentially in regions with delayed warming (Winton et al., 2013; Frölicher et al., 2015; Armour et al., 2016), a potential large reduction of the AMOC may shift the uptake of excess heat from the low to the high latitudes (Rugenstein et al., 2013; Winton et al., 2013), where the atmosphere is more sensitive to external forcing (Winton et al., 2010; Rose et al., 2014; Rose and Rayborn, 2016; Rugenstein et al., 2016). A decrease in AMOC may also decrease the subduction of anthropogenic carbon to deeper waters (Zickfeld et al., 2008; Winton et al., 2013; Randerson et al., 2015; Rhein et al., 2017). A potential impact of methane emissions has also been highlighted for past Heinrich events during which massive icebergs discharge in the North Atlantic may have led to large AMOC disruptions. Large increases (>100 ppb) in methane production have been associated with these events (Rhodes et al., 2015) potentially due to increased wetland production in the SH, related to teleconnections of the North Atlantic with tropical area (Ringeval et al., 2013; Zurcher et al., 2013). All these different effects indicate a potentially positive feedback of the AMOC on the carbon cycle (Parsons et al., 2014), although other elements from the terrestrial biosphere may limit its strength or even reverse its sign (Bozbiyik et al., 2011). 

Changes in Atlantic Ocean circulation can also strongly impact marine life and can be seen at all levels of different ecosystems. For instance, changes in the abundance and distribution of species in response to circulation changes in the SPG have been documented amongst plankton (Hátún et al., 2009), fish (Payne et al., 2012; Miesner and Payne, 2018), seabirds (Descamps et al., 2013) and top predators such as tuna, billfish and pilot whales (Hátún et al., 2009; MacKenzie et al., 2014). Nutrient concentrations in the northeast Atlantic have also been shown to be limited by the recent weakening of the SPG, with concomitant ecosystem impacts (Johnson et al., 2013; Hátún et al., 2016). The influence of SPG circulation also extends to ecosystems beyond from the immediate area, and has a clear impact on the productivity of cod (Gadus morhua) in the Barents Sea, for example (Årthun et al., 2017; Årthun et al., 2018). On a broader scale, changes in the AMOC are an important driver of AMV, which has also been linked to substantial changes in marine ecosystems on both sides of the North Atlantic (Alheit et al., 2014; Nye et al., 2014). Recent AMOC weakening is also suspected to explain large marine deoxygenation in the northwest coastal Atlantic (Claret et al., 2018). In addition, a recent study using a marine productivity proxy from Greenland ice cores suggest that net primary productivity has decreased by 10 ± 7% in the subarctic Atlantic over the past two centuries possibly related to changes in AMOC (Osman et al., 2019). Finally, a model study investigated the impact of mitigation by reversing the forcing from a RCP8.5 scenario from 2100 and found that global marine net productivity may recover very rapidly and even overshoot contemporary values at the end of the reversal, highlighting the potential benefit of mitigation (John et al., 2015).

Following all these potential impacts, it has been suggested that a collapse of the AMOC may have the potential to induce a cascade of abrupt events, related to the crossing of thresholds from different tipping points, itself potentially driven by GIS rapid melting. For example, a collapse of the AMOC may induce causal interactions like changes in ENSO characteristics (Rocha et al., 2018), dieback of the Amazon rainforest and shrinking of the WAIS due to seesaw effect, ITCZ southern migration and large warming of the Southern Ocean (Cai et al., 2016). However, such a worst case scenario remains very poorly constrained quantitatively due to the large uncertainty in GIS and AMOC response to global warming. 

The potential impacts of such rapid changes in ocean circulation on agriculture, economy and human health remain poorly evaluated up to now with very few studies on the topic (Kopits et al., 2014). The available impact literature on AMOC weakening has focussed on impacts from temperature change only (reduced warming), globally leading to economic benefits (e.g., Anthoff et al., 2016), and local losses can amount to a few percent of gross domestic product (GDP), however under a complete shutdown (Link and Tol, 2011). Declines in Barents Sea fish species could lead to economic losses (Link and Tol, 2009), but more comprehensive economic studies are lacking.

The numerous potential impacts of AMOC weakening (see Section 6.7.2) require adaptation responses. A specific adaptation action is a monitoring and early warning system using observation and prediction systems, which can help to respond in time to effects of an AMOC decline. Although it is difficult to warn very early for large changes in AMOC to come, notably due to large natural decadal variability of the AMOC (Boulton et al., 2014), the observation arrays that are in place may allow the development of such an early warning system. Nevertheless, the prospects for its operational use for early warnings have not yet been fully developed. In this respect, developing early warning systems that do not depend on statistical timeseries analysis of long observational record might be seen as an important research goal in the future.

Decadal prediction systems can help fill this gap. Skilful prediction of AMOC variation has been demonstrated on the multi-annual scale (Matei et al., 2012) and retrospective prediction experiments have demonstrated that the large changes in the SPG seen in the mid-1990s could have been foreseen several years in advance (Wouters et al., 2013; Msadek et al., 2014). The World Climate Research Programme’s grand challenge of launching decadal predictions every year (Kushnir et al., 2019) is an important step towards anticipating rapid changes in the near term and can drive decadal-scale climate services. For example, a few studies have already shown that small variations anticipated by decadal predictions (e.g., Sheen et al., 2017) can be useful for the development of climate services, notably for agriculture in south and east Africa (Nyamwanza et al., 2017). Decadal predictions also match the decision making time horizons of many users of the ocean (Tommasi et al., 2017b) and are expected to play an increasingly important role in this sector in the future (Payne et al., 2017).

Compound events refer to events that are characterised by multiple failures that can amplify overall risk and/or cause cascading impacts (Helbing, 2013; Gallina et al. 2016; Figure 6.1). These impacts may be triggered by multiple hazards that occur coincidently or sequentially and can lead to substantial disruption of natural or human systems (Leonard et al., 2014; Oppenheimer et al., 2014; Gallina et al., 2016; Zscheischler et al., 2018). These concepts are illustrated in a series of recent case studies that show how compound events interact with multiple elements of the ecosystem and society to create compound risk and cascading impacts (Box 6.1).

Compound events and cascading impacts are examples of deep uncertainty because data deficiency often prevents the assessment of probabilities and consequences of the risks from compound events. Furthermore, climate drivers that contribute to compound events could cross tipping points in the future (e.g., Cai et al., 2016; Cross-Chapter Box 4 in Chapter 1). Concepts and methods for addressing compound events and cascading impacts have a solid foundation in disaster risk reduction frameworks (Scolobig, 2017) where they may be assessed with scenarios, risk mapping, and participatory governance (Marzocchi et al., 2012; Komendantova et al., 2014). However, these approaches have tended to not consider the effects of climate change, rather considering hazards and vulnerability as stationary entities (Gallina et al., 2016). Trends in geophysical and meteorological extreme events and their interaction with more complex social, economic and environmental vulnerabilities overwhelm existing governance and institutional capacities (Shimizu and Clark, 2015) because of the aggregated cascading impacts. 

Understanding regions where changes in the climate system could increase the likelihood or severity of multiple hazards is relevant to understanding compound events (Figure 6.1). Several recent studies have highlighted coastal regions that are becoming more susceptible to multiple hazards from changes in regional climate. Warming and poleward expansion of the warm western boundary current regions (WBCs; Yang et al., 2016a) together with intensified cyclogenesis in these WBC regions; the Gulf Stream (Booth et al., 2012), the Kuroshio (Hirata et al., 2016) and the East Australian Current (EAC; Pepler et al., 2016a) can increase the likelihood of multiple hazards. These include increased rates of SLR (Brunnabend et al., 2017; Zhang et al., 2017b) together with increases in severe rainfall, storm surges and associated flooding (Thompson et al., 2013; Oey and Chou, 2016; Pepler et al., 2016a). WBCs have undergone an intensification and poleward expansion in all but the Gulf Stream where the weakening of the AMOC cancelled this effect (Seager and Simpson, 2016; Yang et al., 2016a). 

Acknowledging the dual role of regional SLR and TCs frequency and intensity changes for future flood risk, Little et al. (2015) developed a flood index that takes account of local projected SLR along with TC frequency and intensity changes in a CMIP5 multi-model ensemble. They find that relative to 1986–2005, the Flood Index is 4–75 times higher by 2080–2099 for RCP2.6 (10–90th percentile range) and 35–350 times higher for RCP8.5. In the vicinity of the East Australian Current, Pepler et al. (2016b) found warmer SSTs boost the intensification of weak to moderate ETC’s. Neglecting the compounding effects of flood and extreme sea level drivers can cause significant underestimation of flood risk and projected failure probability (Wahl et al., 2016; Moftakhari et al., 2017). 

Over the last decade, several efforts have been made to address long-term shoreline change driven by the cascading impact of SLR, waves and MSL. Ranasinghe et al. (2012) presented the Probabilistic Coastline Recession model, which provides probabilistic estimates of coastline recession in response to both storms and SLR in the 21st century. Dune recession is estimated for each storm considering the recovery between storms, which is obtained empirically. More recently, Toimil et al. (2017) developed a methodology to address shoreline change over this century due to the action of waves, storm surges, astronomical tides in combination with SLR. The methodology considers the generation of thousands of multi-variate hourly time series of waves and storm surges to reconstruct future shoreline evolution probabilistically, which enables estimates of extreme recessions and long-term coastline change to be obtained. The model proposed by Vitousek et al. (2017) integrates longshore and cross-shore transport induced by GCM-projected waves and SLR, which allows it to be applied to both long and pocket sandy beaches. The analysis provides only one instance of what coastline change over the 21st century may be.

To summarise, new studies highlight regions such as coasts including those adjacent to WBCs, that are experiencing larger changes to multiple phenomena simultaneously such as SLR and cyclone intensity linked to higher SST increases (medium confidence), which increases the likelihood of extremes from multiple hazards occurring (medium confidence). Failing to account for the multiple factors responsible for extreme events will lead to an underestimation of the probabilities of occurrence (high confidence).

Damage and loss of ecosystems (mangrove, coral reefs, polar deserts, wetlands and salt marshes); or regime shifts in ecosystem communities lead to reduced resilience of all the ecosystems and possible flow-on effects to human systems. For example, recent studies showed that living corals and reef structures have experienced significant losses from human-related drivers such as coastal development; sand and coral mining; overfishing, acidification, and climate-related storms and bleaching events (Smith, 2011; Nielsen et al., 2012; Hilmi et al., 2013; Graham et al., 2015; Lenoir and Svenning, 2015; Hughes et al., 2017b). As a consequence, reef flattening is taking place globally due the loss of corals and from the bio-erosion and dissolution of the underlying reef carbonate structures (Alvarez-Filip et al., 2009). Reef mortality and flattening due to non-climate and climate-related drivers trigger cascading impacts and risks due to the loss of the protection services provided to coastal areas. High emission scenarios are expected to lead to almost the complete loss of coral cover by 2100, although policies aiming to lower the combined aerosol-radiation interaction and aerosol-cloud interaction (e.g., IPCC RCP 6.0) may partially limit the impacts on coral reefs and the associated habitat loss, thereby preserving an estimated 14 to 20 billion USD in consumer surplus 2100 (2014 USD, 3% discount; Speers et al., 2016). Moreover, projected SLRwill increase flooding risks, and these risks will be even greater if reefs that now help protect coasts from waves are lost due to bleaching-induced mortality. 

Impacts of compound events also have significant multi-effects in the societal system. Cascading impacts are particularly driven by the loss or (temporary) disruption of critical infrastructure (Pescaroli and Alexander, 2018), such as communications, transport, and power supply, on housing, dams and flood protection; as well as health provision. Repeated extreme and compound events are leading to critical transitions in social systems (Kopp et al., 2016) which may cause the disruption of (local) communities, creating cascading impacts consisting of short-term impacts as well as long-lasting economic effects, and in some cases migration. When the responses of the economic sector to short term weather variations are applied to long term-climate projections, risks associated with climate change on different sectors are projected to result in an average 1.2% of decrease of US GDP per degree Celsius of warming. Furthermore, broad geographical discrepancies generate a large transfer of value northwards and westwards with the expected consequence of increased economic inequality (Hsiang et al., 2017). The severity and intensity of the cascading impacts also depend on the affected societies’ vulnerability and resilience. For example, the intensity and influence of compound events are dependent on the size and scale of the affected society and the percentage of economy or GDP impacted (Handmer et al., 2012 in IPCC SREX). Smaller countries and especially small islands face the challenge of being unable to ‘hedge’ the risk through geographical redistribution (see Cross-Chapter Box 9). 

Impacts from the natural system can descend into a cascade of disasters. For example, in 2005, Hurricane Katrina led to heavy flooding in the coastal area, dike breaches, emergency response failures, chaos in evacuation (traffic jams) and social disruption. Flooding in Thailand in 2011 led to the closure of many factories which not only impacted on the country’s economy but impaired the global automobile and electronic industry (Kreibich et al., 2014). Female-owned establishments are more challenged with failures than businesses owned by men due to less experience, shorter duration and smaller size of businesses (Haynes et al., 2011; Marshall et al., 2015). The impact of compound events on ecosystems can also, in the long run, have devastating impacts on societal systems, for example, impacts from tropical storms can lead to coral degradation, which leads to increased wave impact and subsequent accelerated coastal erosion and impacts on fishing resources. This subsequently can have an impact on local economies, potentially leading to social disruption and migration (Saha, 2017). Impacts on marine ecosystems and habitats will also affect subsistence and commercial fisheries and, as a result, food security (Barrow et al., 2018). Climate-induced community relocations in Alaska stem from repeated extreme weather events coupled with climate change-induced coastal erosion and these impact the habitability of the whole community (Bronen, 2011; Durrer and Adams, 2011; Marino, 2011; Marino, 2012; Bronen and Chapin, 2013; see also Cross-Chapter Boxes 2 and 5 in Chapter 1).

The management of compound events and cascading impacts in the context of governance poses challenges, partly because it is place dependent and heavily influenced by local parameters such as hazard experience and cultural values. Moreover, in some cases, people perceive that their community or country is less affected than others, leading to a ‘spatial optimism bias’ that delays or reduces the scope of actions (Nunn et al., 2016). In other cases it is unclear who will take responsibility when compound events and cascading impacts occur (Scolobig, 2017), although for some compound risks (e.g.,. na-tech disasters – when natural hazards trigger technological disasters), the private sector cooperates with governments to manage and respond to risks (Krausmann et al., 2017). Considerable variations exist among and inside countries. The level of engagement depends on the process of cascading impacts and the role of governance arrangement at the country level (Lawrence et al., 2018) countries’ capacity to develop integrated risk and disaster frameworks and regulations, viable multi-stakeholder and public-private partnership in the case of multiple technological and natural hazards (Gerkensmeier and Ratter, 2018), the initiatives of local governments to exercise compound risk operations, and experience in interagency cooperation (Scolobig, 2017). The importance of local knowledge and traditional practices in disaster risk prevention and reduction is widely recognised (Hiwasaki et al., 2014; Hilhorst et al., 2015; Audefroy and Sánchez, 2017) (high confidence). The need to strengthen DRM is evident and can be improved and communicated effectively by integrating local knowledge such as Inuit’s indigenous knowledge and local knowledge in Alaska (Pearce et al., 2015; Cross-Chapter Box 3 in Chapter 1) since it is easier for communities to accept than pure science-based DRM (Ikeda et al., 2016). 

Despite difficulties of governance and decision making, many researchers and policy makers have recognised the need to study combined climatic and other hazards and their impacts. Several methods are now being employed to assess climatic hazards and compound events simultaneously, and also in combination (Klerk et al., 2015; van den Hurk et al., 2015; Wahl et al., 2015; Zscheischler and Seneviratne, 2017; Wu et al., 2018; Zscheischler et al., 2018). Policy makers can also begin to plan for disaster risk reduction and adaptation, based on these analyses of compound events and risks. Addressing limitations in understanding the compound hazards, as well as adequate mechanisms of the cascading impacts is needed. Finally, there are limits to resources to study these complex interactions in sufficient detail, as well as limits to data and information on past events that would allow the simulation of these effects, including economic impacts.

A small number of studies (Lontzek et al., 2015; Cai et al., 2016; Lemoine and Traeger, 2016) use different versions of the Dynamic Integrated Climate-Economy assessment model (Nordhaus, 1992; Nordhaus, 2017) to assess the impact of diverse sets of tipping points and causal interactions between them on the socially optimal reduction of gas emissions and the present social cost of carbon, representing the economic cost caused by an additional ton of CO2 emissions or its equivalent.

Cai et al. (2016) consider five interacting, stochastic, potential climate tipping points: reorganisation of the AMOC; disintegration of the GIS; collapse of the WAIS; dieback of the Amazon Rain Forest; and shift to a more persistent El Niño regime. The deep uncertainties associated with the likelihood of each of these tipping points and the dependence of them on the state of the others is addressed through expert elicitation. There is limited evidence, but high agreement that present costs of carbon are clearly underestimated. Double (Lemoine and Traeger, 2016), triple (Ceronsky et al., 2011), to eightfold (Cai et al., 2016) increase of the carbon price are suggested, depending on the working hypothesis. Cai et al. (2016) indicate that with the prospect of multiple interacting tipping points, the present social cost of carbon increases from 15–116 USD per tonne of CO2, and conclude that stringent efforts are needed to reduce CO2 emission if these impacts are to be avoided.

As outlined earlier in this report, several approaches exist for adaptive responses towards climate change impacts. Other sections that deal with adaptation responses to extremes include Section 1.5.2, Section 4.4 (SLR and coastal flooding), Cross Chapter Box 4 in Chapter 1 and Section 5.5.2.5 in Chapter 5 (adaptation limits for coastal infrastructure and ecosystems). Here, we address adaptation responses especially to abrupt and extreme changes (for responses to special abrupt changes (e.g., AMOC; see also Section 6.7).

Since AR5, growing discussions have advocated for transformative adaptation, implying that they support fundamental societal shift towards sustainability and climate-resilient development pathways (Moloney et al., 2017; IPCC, 2018; Morchain, 2018). Successful adaptation to abrupt change and extreme events incorporates climate change concerns and the impact of climate extremes on vulnerable populations taking into account community participation and local knowledge (Tozier de la Poterie and Baudoin, 2015). These interventions reduce risk and enhance resilience, and contribute to the SDGs and social justice (Mal et al., 2018). Temporal scales denote before and after abrupt changes and extreme events (prevention and post-event response), long- and short-term adaptation measures, and the lag time between forecast, warning and event (Field et al., 2012; IPCC, 2012). Spatial dimensions include local risk management and adaptation as well as regional and international coordination to prepare for unexpected extremes tackling the impacts at multiple geographic scales (Devine-Wright, 2013; Barnett et al., 2014; Lyth et al., 2016; Barange et al., 2018). 

Decision making about abrupt change or extreme events is not autonomous; it is constrained by formal and informal institutional processes such as regulatory structures, property rights, as well as culture, traditions and social norms (Field et al., 2012; IPCC, 2012). Efforts in various countries and large cities to improve resilience and adaptation are growing, and these efforts are linked to a global network of research, information and best practices (e.g., Aerts et al., 2014). In both northern and southern high latitudes, extreme climatic conditions and remoteness from densely populated regions constrain human choices. The question is whether responses to extremes and abrupt changes require approaches that are different from the anticipatory management of adaptation to changes in climate and weather extremes. While there are several impact studies on extreme events and abrupt change, very few focus on the necessity of dedicated individual, governmental or business adaptive responses (Tol et al., 2006; Anthoff et al., 2010; Anthoff et al., 2016).

Making appropriate decisions to manage abrupt change and extreme events given deep uncertainty is challenging (Weaver et al., 2013; see Cross-Chapter Boxes 4 and 5 in Chapter 1). This requires the construction of new models integrating different uncertainties under extreme or abrupt scenarios and evaluation of value for money (Weaver et al., 2013). Examples include the inclusion of rapid SLR for assessing coastal impacts and adaptation options (Ranger et al., 2013; Haasnoot et al., 2018; see Sections 6.4 and 6.7). Decision analysis frameworks such as ‘Robust Decision Making’, ‘Decision Scaling’, ‘Assess Risk of Policy’, ‘Info-gap’, ‘Dynamic Adaptation Policy Pathways’, ‘Dynamic Adaptive Pathways Planning’, ‘Multi-Criteria Decision Analysis’, ‘Real Options Analysis’ and ‘Context-First’ accommodate a wide range of uncertainties with subsequent socio-ecological impact (Weaver et al., 2013). The central question remains, however, how one can overcome path dependencies which may cause technical lock-ins in the current system. Monitoring systems of climatic and derived variables, in order to predict necessary shifts in adaptation policies are in development (Haasnoot et al., 2015). However, these frameworks have so far been mostly applied to more gradual shifts of climate change, rather than extreme events and abrupt changes.

Request for the use of ‘actionable’ information and communication based on climate science and modelling will increase (McNie, 2007; Moser and Boykoff, 2013). Such information can only be effective when it is perceived as ‘credible, salient, and legitimate’ (Paton, 2007; Paton, 2008; Dilling et al., 2015). Since SREX (IPCC, 2012), there is medium confidence that trust in the information and the institution (Hardin, 2002; Townley and Garfield, 2013) that governs extreme events and abrupt change (Malka et al., 2009; Birkmann et al., 2011; Schoenefeld and McCauley, 2016) is important. Trust in expert and scientific knowledge helps people make sense of climate change impact and engage with adaptation measures (Moser and Boykoff, 2013; Yeh, 2016). Without such knowledge, people have little recourse to believe and evaluate relevant information (Bråten et al., 2011). Individuals who trust their government can be complacent and do not prepare for the consequences of extremes (Simpson, 2012; Edmondson and Levy, 2019), and shift the responsibility to the government (Edmondson and Levy, 2019). Familiarity with and information about hazards, community characteristics, as well as the relationship between people and government agencies influence the level of trust (Paton, 2007).

Recent literature shows that there are crucial differences between the ethical challenges of mitigation and those of adaptation (Wallimann-Helmer, 2015; Wallimann-Helmer, 2016) in their dealings with Loss and Damage (L&D); and the ongoing analysis disputes how to distribute responsibilities between mitigation and adaptation based on climate justice criteria (Wallimann-Helmer et al., 2019). The Warsaw International Mechanism on L&D under the United Nations Framework Convention on Climate Change (UNFCCC) addresses irreversible changes and limits to adaptation at the global scale (see also Cross-Chapter Box 1 in Chapter 1). This is in contrast to national and local policies, addressing impacts and adaptation. Within the SROCC report, several of the documented and projected irreversible or unavoidable and thus residual impacts beyond adaptation would potentially fall under this category (e.g., Warner and van der Geest, 2013; Huggel et al., 2019; Mechler et al., 2019), including impacts from SLR, land erosion and reduced freshwater resources on small islands, changes in high mountains and cryosphere changes, as well as changes in ocean species and resources. Apart from climate hazards, risks for L&D are also determined by increasing exposure and vulnerability (Birkmann and Welle, 2015). Such impacts can be assessed using conventional frameworks, but the debate on the precise scope of such impacts remains, including those from anthropogenic climate change impacts as well as natural climate variability and extremes (e.g., James et al., 2014). More work is required to explore the range of activities available for responding to L&D resulting from slow onset processes in the scope of the SROCC report such as ocean acidification (Harrould-Kolieb and Hoegh-Guldberg, 2019) and mountain cryosphere changes (Huggel et al., 2019).

Under the same L&D mechanism, risk transfer mechanisms and insurance have been suggested as a specific adaptation policy option. Several forms of ‘climate change’ insurance have been proposed recently, but their potential for adaptation has met with criticism, importantly because of the costs of formal insurance and other risk transfer options, as well as issues with sustainability given the lack of loss prevention and adaptation (Surminski et al., 2016; Linnerooth-Bayer et al., 2019). A compensation mechanism for low-lying small islands inclusive of L&D proposal is in progress (Adelman, 2016). Insurance (see also Section 4.4.4) can help absorb extreme shocks for both individuals, using traditional insurance and parametric insurance. Sovereign insurance mechanisms can help governments absorb large losses (Linnerooth-Bayer et al., 2019), but eventually they need to be coupled with other incentives for adaptation and risk reduction measures to be cost-effective (Botzen, 2013) (medium confidence).

There is a consensus that investing in disaster risk reduction has economic benefits, although there is medium evidence about the range of the estimated benefits which varies from a global estimate of two to four dollars saved for each dollar invested (Kull et al., 2013; Mechler, 2016) to about 400 EUR per invested 1 EUR in the case of flood early warning systems in Europe (Pappenberger et al., 2015). The US Federal Emergency Management Agency indicated that a 1% increase in annual investment in flood management decreases flood damage by 2.1% (Davlasheridze et al., 2017). Conserving ecosystems that provide services for risk reduction also has monetary benefits. Wetlands have been observed to reduce damages during storms. Wetlands and floodplains in Otter Creek (Vermont, USA) reduced damages caused by storms by 54–78% and 84–95%, respectively, for Tropical Storm Irene (Watson et al., 2016). For the whole of the USA, wetlands provide 23.2 billion USD yr-1 in storm protection services and the loss of 1 hectare of wetland is estimated to correspond to an average 33,000 USD increase in storm damage from specific storms (Costanza et al., 2008). Engineered structures are also expected to reduce risks. In Europe, to maintain the coastal flood loss constant relative to the size of the economy, flood defence structures need to be able to protect coastal areas for a projected increase of sea level between 0.5–2.5 m. Without these risk reduction actions, the expected damages from coastal floods could increase by two or three degrees of magnitude compared to the present (Vousdoukas et al., 2018). Although risk reduction actions are generally considered an effective way to reduce the damages by shifting the loss-exceedance curve, cost-benefit analysis of disaster risk reduction actions faces several challenges, including its limited role in informing decisions, spatial and temporal uncertainty scales, and discounting and choice of discount rate that affect cost-benefit analysis results heavily (Mechler, 2016).

Governance for effective adaptation defined as changes in practice, process and structure (Smit et al., 2001) considers equity, legitimacy and co-benefits (Patterson et al., 2018) appropriate to the issue (Young, 2002). Countries, sectors and localities place different values and perspectives on these categories, and they can change over time (Plummer et al., 2017; see Cross-Chapter Boxes 1 and 2 in Chapter 1). Transformative governance embraces a wider application of climate change-induced mitigation and adaptation strategies to generate fundamental change. It is society-wide and goes beyond the goals of climate change policies and measures (IPCC, 2013; Patterson et al., 2018). It is distinguished from conventional strategies and solutions, as it includes both natural and human systems and intertwines with the SDGs (Fleurbaey et al., 2014; Tàbara et al., 2019). Transformational adaptation is also needed when incremental adaptation to extreme events and abrupt changes are insufficient (Kates et al., 2012). Planned retreat from SLR and climate refugees illustrate the need for transformative governance as the current coastal and risk management regimes do not have the capacity to handle these issues adequately. Inclusion of bilateral and regional agreements related to climate-induced migration (McAdam, 2011), land use planning frameworks to respond to policy, institutional and cultural implications of migration (Matthews and Potts, 2018), and identification of beneficiaries of managed retreat (Hino et al., 2017) along with positive opportunities for migrants to diversify income and avoid being in harm’s way (Gemenne, 2015) are steps towards transformative governance. Retreat and migration entail local responses that include indigenous and local knowledges and perspectives that can be applied to solve these issues (Farbotko and Lazrus, 2012; Hilhorst et al., 2015; Tharakan, 2015; Iloka, 2016; Nunn et al., 2016; see also Cross-Chapter Boxes 2 and 5 in Chapter 1). Another example is the Polar region which has started to pursue transformative governance given the potential for increased tourism and cooperation that require changed governance structure (see Sections 3.5.2; 3.5.5 and Table 3.7 in Chapter 3). Accountability for transformations and transitions has been identified as a crucial factor to support responsible action and strengthen climate governance (Edmondson and Levy, 2019).

Though discourse abounds related to transformative governance, it falls short of its ideal in climate change action plans as it is unclear whether communities have the capacity to engage in substantive change to build low-carbon and resilient communities (Burch et al., 2014). The results of a study on the USA by Tang and Dessai (2012) indicate that climate adaptation and mitigation plans’ treatment of extreme climate conditions and disaster preparedness is limited. Moreover, risk communication with the public is part of an integrated disaster warning system, but behavioural response to disaster warnings are often governed by personal beliefs about the nature of the hazard; and ultimately swaying individual decisions to comply with or ignore the warning message (Mayhorn and McLaughlin, 2014). New approaches such as the ‘first mile’ of early warning systems, built on the specific needs from beneficiary communities instead of on technological progress, are being implemented (Zommers et al., 2017); but they have not yet been assessed. 

Coupling disaster risk reduction and management with climate change adaptation effort—following the set targets of UNFCCC and the Sendai Framework—has shown progress since SREX and AR5 (e.g., Lawrence and Saunders, 2017). Substantial literature exists on the topic, but there is little assessment of practices on the ground in the implementation of integrated disaster management and climate change adaptation (Nalau et al., 2016) including health (Banwell et al., 2018). Mainstreaming disaster risk reduction and climate change adaptation within and across sectors is considered essential to ensure administrative coordination and coherence across sectoral plans and policies (Shimizu and Clark, 2015) (medium confidence). Financial and technological support and capacity building especially related to public works, savings or loans enable households to build assets and improve livelihoods (Ulrichs et al., 2019). No assessment is available so far of the efficiency and effectiveness of mainstreaming especially related to the integration of climate change adaptation and disaster risk reduction, let alone for abrupt and extreme impacts.

Case studies of integration note major problems, for example, weak coordination among government agencies (Seidler et al., 2018); lack of data and user-friendly information to guide decision making at the local level (Jones et al., 2017) and the need for the central governmental support for data availability (Putra et al., 2018); fragmentation due to competing local objectives (Forino et al., 2017); dependence on regional and international frameworks in the absence of a national framework (Rivera and Wamsler, 2014); limited availability of formal training in integration (Hemstock et al., 2017); and turf wars between responsible government agencies (Nemakonde and Van Niekerk, 2017). The case of Pacific islands such as Vanuatu is indicative of these problems. Though they have coupled disaster risk reduction with climate change adaptation, problems manifest in relationships, responsibilities, capacity and expectations between government agencies and other actors (e.g., international donors and non-governmental organisations), as analysed by Vanuatu’s response to the Category 5 TC Pam (Nalau et al., 2017). Some solutions are proposed such as getting all the actors on the same page and focusing on reducing vulnerability to longer-term environmental hazards (Schipper et al., 2016); focussing on specific goals, objectives and strategies (Organization of American States, 2014); assigning a single department to handle integration (APEC, 2016); and citing real-life decision examples in national guidelines (Bell et al., 2017). Place-based responses also entail the inclusion and the acknowledgement of indigenous and local knowledge for an enhanced resilience pathway (Hilhorst et al., 2015; Tharakan, 2015; Iloka, 2016; Nunn et al., 2016). 

Given the significance of disaster risk reduction to enhance climate change adaptation regardless of the integration of the two, the Sendai Framework for Disaster Risk Reduction 2015–2030 focuses on seven targets and four priorities that foster participation beyond information sharing and include partnerships and collaborations within society (UNISDR, 2015). Inclusion of, and coordination between, different stakeholders is a key component for managing risks of extreme events, including in a changing climate (medium confidence). In the Wadden Sea coastal area, for example, crucial parts of coordinating disaster risk reduction, include (i) responsibility-sharing among authorities, sectors and stakeholders, (ii) all-of-society engagement and partnership with empowerment and inclusive participation, and (iii) development of international, regional, subregional and transboundary cooperation schemes (González-Riancho et al., 2017). In India, a change in the coordination structure was pivotal in reducing fatalities from over 10,000 to 45 between cyclones Orissa (Odisha) in 1999 and Phailin in 2013. In this case, the Disaster Management Act of 2005 established a comprehensive policy and command and control system during disaster response that empowered the most qualified government officials regardless of their rank. This system provides authority to and holds accountability for those in charge of ground operations. Though this rigid system may sometimes be questioned, a unified and top-down command structure works better when there is a lack of mature disaster management system (Pal et al., 2017).

In sum, limiting the risk from the impact of extreme events and abrupt changes leads to successful adaptation to climate change if climate-affected sectors and disaster management relevant agencies coordinate well (high confidence). Transformative governance, including successful integration of disaster risk management and climate change adaptation, empowerment of vulnerable groups, accountability of governmental decisions, and longer-term planning promotes climate-resilient development pathways (high confidence). An enhanced understanding of the institutional capacity as well as the legal framework addressing abrupt changes and extreme events is especially important (medium confidence). 

Knowledge gaps limit the identification of the most relevant actions to achieve and pursue climate-resilient development pathways. Since SREX and AR5, there is little research on indirect impacts of climatic extremes on ecosystems and consequences on poverty and livelihoods critical to the SDGs. For example, adaptation solutions and limitations, including governance challenges, for the ocean do not include extreme events (Sections 5.5.2 and 5.5.3 in Chapter 5). Further, there is only scant literature on L&D, including non-economic impacts, resulting from well-documented processes such as MHWs (Section 6.4), SLR impacts on low-lying coasts (Section 4.3), and cryosphere changes (Section 2.3; Chapter 3) (high confidence). Limited information is available concerning the cost-benefit and effectiveness of risk-reduction measures. Coupling risk transfer and insurance mechanisms with risk reduction measures, for example, can enhance the cost-effectiveness of adapting to climate change (medium confidence).