Editing IPCC:AR6/WGI/Chapter-11 (section)

== 11.2 Data and Methods ==

<div id="h1-3-siblings" class="h1-siblings"></div>

This section provides an assessment of observational data and methods used in the analysis and attribution of climate change specific to weather and climate extremes. It also introduces some concepts used in presenting future projections of extremes. Later sections (Sections 11.3–11.8) also provide additional assessments on relevant observational datasets and model validation specific to the type of extremes to be assessed. General background on climate modelling is provided in Chapters 4 and 10.

<div id="11.2.1" class="h2-container"></div>

<span id="definition-of-extremes"></span>
=== 11.2.1 Definition of Extremes ===

<div id="h2-18-siblings" class="h2-siblings"></div>

In the literature, an event is generally considered extreme if the value of a variable exceeds (or lies below) a threshold. The thresholds have been defined in different ways, leading to differences in the meaning of extremes that may share the same name. For example, two sets of metrics for the frequency of hot/warm days have been used in the literature. One set counts the number of days when maximum daily temperature is above a relative threshold defined as the 90th or higher percentile of maximum daily temperature for the calendar day over a base period. An event based on such a definition can occur at any time of the year, and the impact of such an event would differ depending on the season. The other set counts the number of days in which maximum daily temperature is above an absolute threshold such as 35°C, because exceeding this temperature can sometimes cause health impacts (however, these impacts may depend on location and whether ecosystems and the population are adapted to such temperatures). While both types of hot extreme indices have been used to analyse changes in the frequency of hot/warm events, they represent different events that occur at different times of the year, possibly affected by different types of processes and mechanisms, and possibly also associated with different impacts.

Changes in extremes have also been examined from two perspectives: changes in the frequency for a given magnitude of extremes; or changes in the magnitude for a particular return period (frequency). Changes in the probability of extremes (e.g., temperature extremes) depend on the rarity of the extreme event that is assessed, with a larger change in probability associated with a rarer event (e.g., [[#Kharin--2018|Kharin et al., 2018]] ). However, changes in the magnitude represented by the return levels of the extreme events may not be as sensitive to the rarity of the event. While the answers to the two different questions are related, their relevance may differ for distinct audiences. Conclusions regarding the respective contribution of greenhouse gas forcing to changes in magnitude versus frequency of extremes may also differ ( [[#Otto--2012|Otto et al., 2012]] ). Correspondingly, the sensitivity of changes in extremes to increasing global warming is also dependent on the definition of the considered extremes. In the case of temperature extremes, changes in magnitude have been shown to often depend linearly on global surface temperature ( [[#Seneviratne--2016|Seneviratne et al., 2016]] ; [[#Wartenburger--2017|Wartenburger et al., 2017]] ), while changes in frequency tend to be nonlinear and can, for example, be exponential for increasing global warming levels ( [[#Fischer--2015|Fischer and Knutti, 2015]] ; [[#Kharin--2018|Kharin et al., 2018]] ). When similar damage occurs once a fixed threshold is exceeded, it is more important to ask a question regarding changes in the frequency. But when the exceedance of this fixed threshold becomes a normal occurrence in the future, this can lead to a saturation in the change of probability ( [[#Harrington--2018a|Harrington and Otto, 2018a]] ). Also, if the impact of an event increases with the intensity of the event, it would be more relevant to examine changes in the magnitude. Finally, adaptation to climate change might change the relevant thresholds over time, although such aspects are still rarely integrated in the assessment of projected changes in extremes. Framing is considered when forming the assessments of this Chapter, including how extremes are defined and how the questions are asked in the literature ''.''

<div id="11.2.2" class="h2-container"></div>

<span id="data"></span>
=== 11.2.2 Data ===

<div id="h2-19-siblings" class="h2-siblings"></div>

Studies of past and future changes in weather and climate extremes, and in the mean state of the climate, use the same original sources of weather and climate observations, including in situ observations, remotely sensed data, and derived data products such as reanalyses. Sections 2.3 and 10.2 assess various aspects of these data sources and data products from the perspective of their general use, and in the analysis of changes in the mean state of the climate in particular. Building on these previous chapters, this subsection highlights particular aspects that are related to extremes and are most relevant to the assessment of this Chapter. The SREX (Chapter 3, [[#Seneviratne--2012|Seneviratne et al., 2012]] ) and AR5 (Chapter 2, [[#Hartmann--2013|Hartmann et al., 2013]] ) addressed critical issues regarding the quality and availability of observed data and their relevance for the assessment of changes in extremes.

Extreme weather and climate events occur on time scales of hours (e.g., convective storms that produce heavy precipitation) to days (e.g., tropical cyclones, heatwaves), to seasons and years (e.g., droughts). A robust determination of long-term changes in these events can have different requirements for the spatial and temporal scales and sample size of the data. In general, it is more difficult to determine long-term changes for events of fairly large temporal duration, such as ‘megadroughts’ that last several years or longer (e.g., [[#Ault--2014|Ault et al., 2014]] ), because of the limitations of the observational sample size. Literature that studies changes in extreme precipitation and temperature often uses indices representing specifics of extremes that are derived from daily precipitation and temperature values. Station-based indices would have the same issues as those for the mean climate regarding the quality, availability, and homogeneity of the data. For the purpose of constructing regional information and/or for comparison with model outputs, such as model evaluation, and detection and attribution, these station-based indices are often interpolated onto regular grids. Two different approaches, involving two different orders of operation, have been used in producing such gridded datasets.

In some cases, such as for the HadEX3 dataset ( [[#Dunn--2020|Dunn et al., 2020]] ), indices of extremes are computed using time series directly derived from stations first, and are then gridded over the space. As the indices are computed at the station level, the gridded data products represent point estimates of the indices averaged over the spatial scale of the grid box. In other instances, daily values of station observations are first gridded (e.g., [[#Contractor--2020a|Contractor et al., 2020a]] ), and the interpolated values can then be used to compute various indices. Depending on the station density, values for extremes computed from data gridded this way represent extremes of spatial scales anywhere from the size of the grid box to a point. In regions with high station density (e.g., North America, Europe), the gridded values are closer to extremes of area means and are thus more appropriate for comparisons with extremes estimated from climate model output, which is often considered to represent areal means ( [[#Chen--2008|Chen and Knutson, 2008]] ; [[#Gervais--2014|Gervais et al., 2014]] ; [[#Avila--2015|Avila et al., 2015]] ; [[#Di%20Luca--2020b|Di Luca et al., 2020b]] ). In regions with very limited station density (e.g., Africa), the gridded values are closer to point estimates of extremes. The difference in spatial scales among observational data products and model simulations needs to be carefully accounted for when interpreting the comparison among different data products. For example, the average annual maximum daily maximum temperature (TXx) over land computed from the original ERA-Interim reanalysis (at 0.75° resolution) is about 0.4°C warmer than that computed when the ERA-Interim dataset is upscaled to the resolution of 2.5° × 3.75° ( [[#Di%20Luca--2020|Di Luca et al., 2020]] ).

Extreme indices computed from various reanalysis data products have been used in some studies, but reanalysis extreme statistics have not been rigorously compared to observations ( [[#Donat--2016a|Donat et al., 2016a]] ).

In general, changes in temperature extremes from various reanalyses were most consistent with gridded observations after about 1980, but larger differences were found during the pre-satellite era ( [[#Donat--2014b|Donat et al., 2014b]] ). Overall, lower agreement across reanalysis datasets was found for extreme precipitation changes, although temporal and spatial correlations against observations were found to be still significant. In regions with sparse observations (e.g., Africa and parts of South America), there is generally less agreement for extreme precipitation between different reanalysis products, indicating a consequence of the lack of an observational constraint in these regions ( [[#Donat--2014b|Donat et al., 2014b]] , 2016a). More recent reanalyses, such as ERA5 ( [[#Hersbach--2020|Hersbach et al., 2020]] ), seem to have improved over previous products, at least over some regions (e.g., [[#Mahto--2019|Mahto and Mishra, 2019]] ; [[#Gleixner--2020|Gleixner et al., 2020]] ; [[#Sheridan--2020|Sheridan et al., 2020]] ). Caution is needed when reanalysis data products are used to provide additional information about past changes in these extremes in regions where observations are generally lacking.

Satellite remote sensing data have been used to provide information about precipitation extremes because several products provide data at sub-daily resolution for precipitation, for example, Tropical Rainfall Measuring Mission (TRMM; [[#Maggioni--2016|Maggioni et al., 2016]] ) and clouds, for example, Himawari (Bessho et al., 2016; [[#Chen--2019|Chen et al., 2019]] ). However, satellites do not observe the primary atmospheric state variables directly and polar orbiting satellites do not observe any given place at all times. Hence, their utility as a substitute for high-frequency (i.e., daily) ground-based observations is limited. For instance, [[#Timmermans--2019|Timmermans et al. (2019)]] found little relationship between the timing of extreme daily and five-day precipitation in satellite and gridded station data products over the USA.

<div id="box-11.3" class="h2-container box-container"></div>

Box 11.3 | Extremes in Paleoclimate Archives Compared to Instrumental Records

<div id="h2-20-siblings" class="h2-siblings"></div>

Examining extremes in pre-instrumental information can help to put events occurring in the instrumental record (referred to as ‘observed’) in a longer-term context. This box focuses on extremes in the Common Era (CE, the last 2000 years), because there is generally higher confidence in pre-instrumental information gathered from the more recent archives from the Common Era than from earlier evidence. It addresses evidence of extreme events in paleoreconstructions, documentary evidence (such as grape harvest data, religious documents, newspapers, and logbooks) and model-based analyses, and whether observed extremes have or have not been exceeded in the Common Era. This box provides overviews of: (i) AR5 assessments; (ii) types of evidence assessed here; evidence of: (iii) droughts; (iv) temperature extremes; (v) paleofloods; and (vi) paleotempests; and (vii) a summary.

( [[IPCC:Wg1:Chapter:Chapter-5|Chapter 5]] of AR5 ( [[#Masson-Delmotte--2013|Masson-Delmotte et al., 2013]] ) concluded with ''high confidence'' that droughts of greater magnitude and of longer duration than those observed in the instrumental period occurred in many regions during the preceding millennium. There was ''high confidence'' in evidence that floods during the past five centuries in northern and Central Europe, the western Mediterranean region, and eastern Asia were of a greater magnitude than those observed instrumentally, and ''medium confidence'' in evidence that floods in the Near East, India and Central North America were comparable to modern observed floods. While AR5 assessed 20th century summer temperatures compared to those reconstructed in the Common Era, it did not assess shorter duration temperature extremes.

Many factors affect confidence in information on pre-instrumental extremes. First, the geographical coverage of paleoclimate reconstructions of extremes is not spatially uniform ( [[#Smerdon--2016|Smerdon and Pollack, 2016]] ) and depends on both the availability of archives and records, which are environmentally dependent, and also the differing attention and focus from the scientific community. In Australia, for example, the paleoclimate network is sparser than for other regions, such as Asia, Europe and North America, and synthesized products rely on remote proxies and assumptions about the spatial coherence of precipitation between remote climates ( [[#Cook--2016c|Cook et al., 2016c]] ; [[#Freund--2017|Freund et al., 2017]] ). Second, pre-instrumental evidence of extremes may be focused on understanding archetypal extreme events, such as the climatic consequences of the 1815 eruption of Mount Tambora, Indonesia (Veale and Endfield, 2016). These studies provide narrow evidence of extremes in response to specific forcings (M. [[#Li--2017|]] [[#Li--2017|]] [[#Li--2017|]] [[#Li--2017|Li et al., 2017]] ) for specific epochs. Third, natural archives may provide information about extremes in one season only and may not represent all extremes of the same types.

Evidence of shorter duration extreme event types, such as floods and tropical storms, is further restricted by the comparatively low chronological controls and temporal resolution (e.g., monthly, seasonal, yearly, multiple years) of most archives compared to the events (e.g., minutes to days). Natural archives may be sensitive only to intense environmental disturbances, and so only sporadically record short-duration or small spatial-scale extremes. Interpreting sedimentary records as evidence of past short-duration extremes is also complex and requires a clear understanding of natural processes (Wilhelm et al. , 2019) . For example, paleoflood reconstructions of flood recurrence and intensity produced from geological evidence (e.g., river and lake sediments), speleothems ( [[#Denniston--2017|Denniston and Luetscher, 2017]] ), botanical evidence (e.g., flood damage to trees, or tree ring reconstructions), and floral and faunal evidence (e.g., diatom fossil assemblages) require understanding of sediment sources and flood mechanisms. Pre-instrumental records of tropical storm intensity and frequency (also called paleotempest records) derived from overwash deposits of coastal lake and marsh sediments are difficult to interpret. Many factors have an impact on whether disturbances are deposited in archives ( [[#Muller--2017|Muller et al., 2017]] ) and deposits may provide sporadic and incomplete preservation histories (e.g., [[#Tamura--2018|Tamura et al., 2018]] ).

Overall, the most complete pre-instrumental evidence of extremes occurs for long-duration, large spatial-scale extremes, such as for multi-year meteorological droughts or seasonal- and regional-scale temperature extremes. Additionally, more precise insights into recent extremes emerge where multiple studies have been undertaken, compared to the confidence in extremes reported at single sites or in single studies, which may not necessarily be representative of large-scale changes, or for reconstructions that synthesize multiple proxies over large areas (e.g., drought atlases). Multiproxy synthesis products combine paleoclimate temperature reconstructions and cover sub-continental- to hemispheric-scale regions to provide continuous records of the Common Era (e.g., Ahmed et al. , 2013; Neukom et al. , 2014 fo r temperature).

There is ''high confidence'' in the occurrence of long-duration and severe drought events during the Common Era for many locations, although their severity compared to recent drought events differs between locations and the lengths of reconstruction provided. Recent observed drought extremes in some regions – such as the eastern Mediterranean Levant ( [[#Cook--2016a|Cook et al., 2016a]] ), California in the USA( [[#Cook--2014b|Cook et al., 2014b]] ; [[#Griffin--2014|Griffin and Anchukaitis, 2014]] ), and in the Andes (Garreaud et al. , 2017; Domínguez-Castro et al. , 2018) – do not have precedents within the multi-century periods reconstructed in these studies, in terms of duration and/or severity. In some regions (in south-western North America ( [[#Asmerom--2013|Asmerom et al., 2013]] ; [[#Cook--2015|Cook et al., 2015]] ), the Great Plains region ( [[#Cook--2004|Cook et al., 2004]] ), the Middle East ( [[#Kaniewski--2012|Kaniewski et al., 2012]] ), and China ( [[#Gou--2015|Gou et al., 2015]] )), recent drought extremes may have been exceeded in the Common Era. In further locations, there is conflicting evidence for the severity of pre-instrumental droughts compared to observed extremes, depending on the length of the reconstruction and the seasonal perspective provided (see Cook et al. , 2016c; Freund et al. , 2017 for Australia). There can also be differing conclusions for the severity, or even the occurrence, of specific individual pre-instrumental droughts when different evidence is compared (e.g., [[#Wetter--2014|Wetter et al., 2014]] ; [[#Büntgen--2015|Büntgen et al., 2015]] ).

There is ''medium confidence'' that the magnitudes of large-scale, seasonal-scale extreme high temperatures in observed records exceed those reconstructed over the Common Era in some locations, such as Central Europe. In one example, multiple studies have examined the unusualness of present-day European summer temperature records in a long-term context, particularly in comparison to the exceptionally warm year of 1540 CE in Central Europe. Several studies indicate that recent extreme summers (2003 and 2010) in Europe have been unusually warm in the context of the last 500 years ( [[#Barriopedro--2011|Barriopedro et al., 2011]] ; [[#Wetter--2013|Wetter and Pfister, 2013]] ; [[#Wetter--2014|Wetter et al., 2014]] ; [[#Orth--2016b|Orth et al., 2016b]] ), or longer ( [[#Luterbacher--2016|Luterbacher et al., 2016]] ). Others studies show that summer temperatures in Central Europe in 1540 were warmer than the present-day (1966–2015) mean, but note that it is difficult to assess whether or not the 1540 summer was warmer than observed record extreme temperatures ( [[#Orth--2016b|Orth et al., 2016b]] ).

There is ''high confidence'' that the magnitude of floods over the Common Era exceeded observed records in some locations, including Central Europe and eastern Asia. Recent literature supports the AR5 assessments of floods ( [[#Masson-Delmotte--2013|Masson-Delmotte et al., 2013]] ). For example, high temporally resolved records provide evidence of Common Era floods exceeding the probable maximum flood levels in the Upper Colorado River, USA ( [[#Greenbaum--2014|Greenbaum et al., 2014]] ) and peak discharges that are double gauge levels along the middle Yellow River, China ( [[#Liu--2014|Liu et al., 2014]] ). Further studies demonstrate pre-instrumental or early instrumental differences in flood frequency compared to the instrumental period, including reconstructions of high and low flood frequency in the European Alps (e.g., [[#Swierczynski--2013|Swierczynski et al., 2013]] ; [[#Amann--2015|Amann et al., 2015]] ) and Himalayas ( [[#Ballesteros%20Cánovas--2017|Ballesteros Cánovas et al., 2017]] ). The combination of extreme historical flood episodes determined from documentary evidence also increases confidence in the determination of flood frequency and magnitude, compared to using geomorphological archives alone ( [[#Kjeldsen--2014|Kjeldsen et al., 2014]] ). In regions, such as Europe and China, that have rich historical flood documents, there is strong evidence of high-magnitude flood events over pre-instrumental periods (Kjeldsen et al., 2014; [[#Benito--2015|Benito et al., 2015]] ; [[#Macdonald--2017|Macdonald and Sangster, 2017]] ). A key feature of paleoflood records is variability in flood recurrence at centennial timescales ( [[#Wilhelm--2019|Wilhelm et al., 2019]] ), although constraining climate-flood relationships remains challenging. Pre-instrumental floods often occurred in considerably different contexts in terms of land use, irrigation, and infrastructure, and may not provide direct insight into modern river systems, which further prevents long-term assessments of flood changes being made based on these sources.

There is ''medium confidence'' that periods of both more and less tropical cyclone activity (frequency or intensity) than observed occurred over the Common Era in many regions. Paleotempest studies cover a limited number of locations that are predominantly coastal, and hence provide information on specific locations that cannot be extrapolated basin-wide (see [[#Muller--2017|Muller et al., 2017]] ). In some locations, such as the Gulf of Mexico and the New England, USA, coast, similarly intense storms to those observed recently have occurred multiple times over centennial timescales ( [[#Donnelly--2001|Donnelly et al., 2001]] ; [[#Bregy--2018|Bregy et al., 2018]] ). Further research focused on the frequency of tropical storm activity. Extreme storms occurred considerably more frequently in particular periods of the Common Era, compared to the instrumental period in north-east Queensland, Australia ( [[#Nott--2009|Nott et al., 2009]] ; [[#Haig--2014|Haig et al., 2014]] ), and the Gulf Coast (e.g., [[#Brandon--2013|Brandon et al., 2013]] ; [[#Lin--2014|Lin et al., 2014]] ).

The probability of finding an unprecedented extreme event increases with a longer length of past record-keeping, in the absence of longer-term trends. Thus, as a record is extended to the past based on paleoreconstruction, there is a higher chance of very rare extreme events having occurred at some time prior to instrumental records. Such an occurrence is not, in itself, evidence of a change, or lack of a change, in the magnitude or the likelihood of extremes in the past or in the instrumental period at regional and local scales. Yet, the systematic collection of paleoclimate records over wide areas may provide evidence of changes in extremes. In one study, extended evidence of the last millennium from observational data and paleoclimate reconstructions using tree rings indicates that human activities affected the worldwide occurrence of droughts as early as the beginning of the 20th century ( [[#Marvel--2019|Marvel et al., 2019]] ).

In summary, there is ''low confidence'' in overall changes in extremes derived from paleo-archives. There is ''high confidence'' that long-duration and severe drought events occurred at many locations during the last 2000 years. There is also ''high confidence'' that high-magnitude flood events occurred at some locations during the last 2000 years, but overall changes in infrastructure and human water management make the comparison with present-day records difficult. But these isolated paleo-drought and paleo-flood events are not evidence of a change, or lack of a change, in the magnitude or the likelihood of relevant extremes.

<div id="11.2.3" class="h2-container"></div>

<span id="attribution-of-extremes"></span>
=== 11.2.3 Attribution of Extremes ===

<div id="h2-21-siblings" class="h2-siblings"></div>

Attribution science concerns the identification of causes for changes in characteristics of the climate system (e.g., trends, single extreme events). A general overview and summary of methods of attribution science is provided in the Cross-Working Group Box 1.1. Trend detection using optimal fingerprinting methods is a well-established field, and has been assessed in AR5 (Chapter 10, [[#Bindoff--2013|Bindoff et al., 2013]] ), and [[IPCC:Wg1:Chapter:Chapter-3#3.2.1|Section 3.2.1]] of this Report. There are specific challenges when applying optimal fingerprinting to the detection and attribution of trends in extremes and on regional scales where the lower signal-to-noise ratio is a challenge. In particular, the method generally requires the data to follow a Normal (Gaussian) distribution, which is often not the case for extremes. However, recent studies showed that extremes can be transformed to a Gaussian distribution, for example, by averaging over space, so that optimal fingerprinting techniques can still be used ( [[#Wen--2013|Wen et al., 2013]] ; [[#Zhang--2013|Zhang et al., 2013]] ; [[#Wan--2019|Wan et al., 2019]] ). Non-stationary extreme value distributions, which allow for the detailed detection and attribution of regional trends in temperature extremes, have also been used (Z. [[#Wang--2017|]] [[#Wang--2017|Wang et al., 2017]] a).

Apart from the detection and attribution of trends in extremes, new approaches have been developed to answer the question of whether, and to what extent, external drivers have altered the probability and intensity of an individual extreme event ( [[#NASEM--2016|NASEM, 2016]] ). In AR5, there was an emerging consensus that the role of external drivers of climate change in specific extreme weather events could be estimated and quantified in principle, but related assessments were still confined to particular case studies, often using a single model, and typically focusing on high-impact events with a clear attributable signal.

However, since AR5, the attribution of extreme weather events has emerged as a growing field of climate research with an increasing body of literature (see series of supplements to the annual State of the Climate report ( [[#Peterson--2012|Peterson et al., 2012]] , 2013a; [[#Herring--2014|Herring et al., 2014]] , 2015, 2016, 2018), including the number of approaches to examining extreme events(described in [[#Easterling--2016|Easterling et al., 2016]] ; [[#Otto--2017|Otto, 2017]] ; [[#Stott--2016|Stott et al., 2016]] )). A commonly used approach – often called the risk-based approach in the literature, and referred to here as the ‘probability-based approach’ – produces statements such as ‘anthropogenic climate change made this event type twice as likely ’ or ‘anthropogenic climate change made this event 15% more intense’. This is done by estimating probability distributions of the index characterizing the event in today’s climate, as well as in a counterfactual climate, and either comparing intensities for a given occurrence probability (e.g., 1-in-100-year event) or probabilities for a given magnitude (see FAQ 11.3). There are a number of different analytical methods encompassed in the probability-based approach, building on observations and statistical analyses (e.g., van Oldenborgh et al., 2012), optimal fingerprint methods ( [[#Sun--2014|Sun et al., 2014]] ), regional climate and weather forecast models (e.g., [[#Schaller--2016|Schaller et al., 2016]] ), global climate models (GCMs) (e.g., [[#Lewis--2013|Lewis and Karoly, 2013]] ), and large ensembles of atmosphere-only GCMs (e.g., [[#Lott--2013|Lott et al., 2013]] ). A key component in any event attribution analysis is the level of conditioning on the state of the climate system. In the least conditional approach, the combined effect of the overall warming and changes in the large-scale atmospheric circulation are considered and often utilize fully coupled climate models ( [[#Sun--2014|Sun et al., 2014]] ). Other more conditional approaches involve prescribing certain aspects of the climate system. These range from prescribing the pattern of the surface ocean change at the time of the event (e.g., [[#Hoerling--2013|Hoerling et al., 2013]] , 2014), often using Atmospheric Model Intercomparison Project (AMIP) style global models, where the choice of sea surface temperature and ice patterns influences the attribution results ( [[#Sparrow--2018|Sparrow et al., 2018]] ), to prescribing the large-scale circulation of the atmosphere and using weather forecasting models or methods (e.g., [[#Pall--2017|Pall et al., 2017]] ; [[#Patricola--2018|Patricola and Wehner, 2018]] ; [[#Wehner--2018a|Wehner et al., 2018a]] ). These highly conditional approaches have also been called ‘storylines’ (Cross-Working Group Box 1.1; [[#Shepherd--2016|Shepherd, 2016]] ) and can be useful when applied to extreme events that are too rare to otherwise analyse, or where the specific atmospheric conditions were central to the impact. These methods are also used to enable the use of very high-resolution simulations in cases were lower-resolution models do not simulate the regional atmospheric dynamics well ( [[#Shepherd--2016|Shepherd, 2016]] ; [[#Shepherd--2018|Shepherd et al., 2018]] ). However, the imposed conditions limit an overall assessment of the anthropogenic influence on an event, as the fixed aspects of the analysis may also have been affected by climate change. For instance, the specified initial conditions in the highly conditional hindcast attribution approach often applied to tropical cyclones (e.g., [[#Takayabu--2015|Takayabu et al., 2015]] ; [[#Patricola--2018|Patricola and Wehner, 2018]] ) permit only a conditional statement about the magnitude of the storm if similar large-scale meteorological patterns could have occurred in a world without climate change, thus precluding any attribution statement about the change in frequency if used in isolation. Combining conditional assessments of changes in the intensity with a multi-model approach does allow for the latter as well ( [[#Shepherd--2016|Shepherd, 2016]] ).

The outcome of event attribution is dependent on the definition of the event ( [[#Leach--2020|Leach et al., 2020]] ), as well as the framing ( [[#Otto--2016|Otto et al., 2016]] ; [[#Christidis--2018|Christidis et al., 2018]] ; [[#Jézéquel--2018|Jézéquel et al., 2018]] ) and uncertainties in observations and modelling. Observational uncertainties arise in estimating the magnitude of an event as well as its rarity ( [[#Angélil--2017|Angélil et al., 2017]] ). Results of attribution studies can also be very sensitive to the choice of climate variables ( [[#Sippel--2014|Sippel and Otto, 2014]] ; [[#Wehner--2016|Wehner et al., 2016]] ). Attribution statements are also dependent on the spatial (Uhe et al., 2016; [[#Cattiaux--2018|Cattiaux and Ribes, 2018]] ; Kirchmeier‐Young et al., 2019) and temporal ( [[#Harrington--2017|Harrington, 2017]] ; [[#Leach--2020|Leach et al., 2020]] ) extent of event definitions, as events of different scales involve different processes (W. [[#Zhang--2020|]] [[#Zhang--2020|]] [[#Zhang--2020|]] [[#Zhang--2020|Zhang et al., 2020]] ) and large-scale averages generally yield higher attributable changes in magnitude or probability due to the smoothing out of noise. In general, confidence in attribution statements for large-scale heat and lengthy extreme precipitation events have higher confidence than shorter and more localized events, such as extreme storms, an aspect also relevant for determining the emergence of signals in extremes or the confidence in projections (see also Cross-Chapter Box Atlas.1).

The reliability of the representation of the event in question in the climate models used in a study is essential ( [[#Angélil--2016|Angélil et al., 2016]] ; [[#Herger--2018|Herger et al., 2018]] ). Extreme events characterized by atmospheric dynamics that stretch the capabilities of current-generation models ( [[IPCC:Wg1:Chapter:Chapter-10#10.3.3.4|Section 10.3.3.4]] ; [[#Shepherd--2014|Shepherd, 2014]] ; [[#Woollings--2018|Woollings et al., 2018]] ) limit the applicability of the probability-based approach of event attribution. The lack of model evaluation, in particular in early event attribution studies, has led to criticism of the emerging field of attribution science as a whole ( [[#Trenberth--2015|Trenberth et al., 2015]] ) and of individual studies ( [[#Angélil--2017|Angélil et al., 2017]] ). In this regard, the storyline approach ( [[#Shepherd--2016|Shepherd, 2016]] ) provides an alternative option that does not depend on the model’s ability to represent the circulation reliably. In addition, several ways of quantifying statistical uncertainty ( [[#Paciorek--2018|Paciorek et al., 2018]] ) and model evaluation ( [[#Lott--2016|Lott and Stott, 2016]] ; [[#Philip--2018b|Philip et al., 2018b]] , 2020) have been employed to evaluate the robustness of event attribution results. For the unconditional probability-based approach, multi-model and multi-approach (e.g., combining observational analyses and model experiments) methods have been used to improve the robustness of event attribution ( [[#Hauser--2017|Hauser et al., 2017]] ; [[#Otto--2018a|Otto et al., 2018a]] ; [[#Philip--2018b|Philip et al., 2018b]] , 2019, 2020; [[#van%20Oldenborgh--2018|van Oldenborgh et al., 2018]] ; [[#Kew--2019|Kew et al., 2019]] ).

In the regional tables provided in [[#11.9|Section 11.9]] , the different lines of evidence from event attribution studies and trend attributions are assessed alongside one another to provide an assessment of the human contribution to observed changes in extremes in all AR6 regions.

<div id="11.2.4" class="h2-container"></div>

<span id="projecting-changes-in-extremes-as-a-function-of-global-warming-levels"></span>
=== 11.2.4 Projecting Changes in Extremes as a Function of Global Warming Levels ===

<div id="h2-22-siblings" class="h2-siblings"></div>

The most important quantity used to characterize past and future climate change is global warming relative to its pre-industrial level. Changes in global warming are linked quasi-linearly to global cumulative carbon dioxide (CO <sub>2</sub> ) emissions (IPCC, 2013), and for their part, changes in regional climate, including many types of extremes, scale quasi-linearly with changes in global warming, often independently of the underlying emissions scenarios (SR1.5 Chapter 3; [[#Seneviratne--2016|Seneviratne et al., 2016]] ; [[#Matthews--2017|Matthews et al., 2017]] ; [[#Wartenburger--2017|Wartenburger et al., 2017]] ; [[#Kharin--2018|Kharin et al., 2018]] ; Y. [[#Sun--2018a|]] [[#Sun--2018|Sun et al., 2018]] a ; [[#Tebaldi--2018|Tebaldi and Knutti, 2018]] ; [[#Beusch--2020|Beusch et al., 2020]] ; [[#Li--2021|Li et al., 2021]] ). In addition, the use of global warming levels in the context of global policy documents – in particular the 2015 Paris Agreement ( [[#UNFCCC--2016|UNFCCC, 2016]] ) implies that information on changes in the climate system, and specifically extremes, as a function of global warming are of particular policy relevance. Cross-Chapter Box 11.1 provides an overview on the translation between information at global warming levels (GWLs) and scenarios.

The assessment of projections of future changes in extremes as function of GWL has an advantage in separating uncertainty associated with the global warming response (see Chapter 4) from the uncertainty resulting from the regional climate response as a function of GWLs ( [[#Seneviratne--2020|Seneviratne and Hauser, 2020]] ). If the interest is in the projection of regional changes at certain GWLs, such as those defined by the Paris Agreement, projections based on time periods and emissions scenarios have unnecessarily larger uncertainty due to differences in model global transient climate responses. To take advantage of this feature and to provide easy comparison with SR1.5, assessments of projected changes in this chapter are largely provided in relation to future GWLs, with a focus on changes at +1.5°C, +2°C, and +4°C of global warming above pre-industrial levels (e.g., Tables 11.1 and 11.2 and regional tables in [[#11.9|Section 11.9]] ). These encompass a scenario compatible with the lowest limit of the Paris Agreement (+1.5°C), a scenario slightly overshooting the aims of the Paris Agreement (+2°C), and a ‘worst-case’ scenario with no mitigation (+4°C). Cross-Chapter Box 11.1 provides a background on the GWL sampling approach used in AR6, for the computation of GWL projections from climate models contributing to Phase 6 of the Coupled Model Intercomparison Project (CMIP6) as well as for the mapping of existing scenario-based literature for CMIP6 and the CMIP Phase 5 (CMIP5) to assessments as function of GWLs (see also [[#11.9|Section 11.9]] . and Table 11.3 for an example).

While regional changes in many types of extremes do scale robustly with global surface temperature, generally irrespective of emissions scenarios ( [[#11.1.4|Section 11.1.4]] , Figures 11.3, 11.6 and 11.7 and Cross-Chapter Box 11.1), effects of local forcing can distort this relation. For example, emissions scenarios with the same radiative forcing can have different regional extreme precipitation responses resulting from different aerosol forcing (Z. [[#Wang--2017|]] [[#Wang--2017|Wang et al., 2017]] b). Another example is related to forcing from land-use and land cover changes ( [[#11.1.6|Section 11.1.6]] ). Climate models often either overestimate or underestimate observed changes in annual maximum daily maximum temperature, depending on the region and considered models ( [[#Donat--2017|Donat et al., 2017]] ; [[#Vautard--2020|Vautard et al., 2020]] ). Part of the discrepancies may be due to the lack of representation of some land forcings, in particular crop intensification and irrigation (N.D. [[#Mueller--2016|]] [[#Mueller--2016|Mueller et al., 2016]] ; [[#Findell--2017|Findell et al., 2017]] ; [[#Thiery--2017|Thiery et al., 2017]] , 2020). Since these local forcings are not represented, and their future changes are difficult to project, these can be important caveats when using GWL scaling to project future changes for these regions. However, these caveats also apply to the use of scenario-based projections.

The SR1.5 (Chapter 3) assessed different climate responses at +1.5°C of global warming, including transient climate responses, short-term stabilization responses, and long-term equilibrium stabilization responses, and their implications for future projections of different extremes. Indeed, the temporal dimension – that is, when the given GWL occurs – also matters for projections, in particular beyond the 21st century, and for some climate variables related to components of the climate system associated with large inertia (e.g., sea level rise and associated extremes). Nonetheless, for assessments focused on conditions within the next decades, and for the main extremes considered in this chapter, derived projections are relatively insensitive to details of climate scenarios and can be well-estimated based on transient simulations (Cross-Chapter Box 11.1; see also SR1.5).

An important question is the identification of the GWL at which a given change in a climate extreme can begin to emerge from climate noise. Figure 11.8 displays analyses of the GWLs at which emergence in hot extremes – annual maximum daily temperature represented by TXx and heavy precipitation represented by Rx1day is identified in AR6 regions for the whole CMIP5 and CMIP6 ensembles. Overall, signals for extremes emerge very early for TXx, already below 0.2°C in many regions (Figure 11.8a,b), and at around 0.5°C in most regions. This is consistent with conclusions from the SR1.5 ( [[IPCC:Wg1:Chapter:Chapter-3|Chapter 3]] for less-rare temperature extremes (TXx on the yearly time scale), which shows that a difference as small as 0.5°C of global warming – for example, between +1.5°C and +2°C of global warming – leads to detectable differences in temperature extremes in TXx in most WGI AR6 regions in CMIP5 projections (e.g., [[#Wartenburger--2017|Wartenburger et al., 2017]] ; [[#Seneviratne--2018b|Seneviratne et al., 2018b]] ). The GWL emergence for Rx1day is also largely consistent with analyses for less-extreme heavy precipitation events (Rx5day on the yearly time scale) in SR1.5 (see Chapter 3).

<div id="_idContainer035" class="Basic-Text-Frame"></div>

[[File:cbeef96f6cb682637c8719e08f470d31 IPCC_AR6_WGI_Figure_11_8.png]]

'''Figure 11.8 |''' '''Global and regional-scale emergence of changes in temperature (a) and precipitation (b) extremes for the globe (glob.), global oceans (oc.), global lands (land), and the AR6 regions.''' Colours indicate the multi-model mean global warming level at which the difference in 20-year means of the annual maximum daily maximum temperature (TXx) and the annual maximum daily precipitation (Rx1day) become significantly different from their respective mean values during the 1850–1900 base period. Results are based on simulations from the Coupled Model Intercomparison Project Phases 5 and 6 (CMIP5 and CMIP6) multi-model ensembles. See Atlas.1.3.2 for the definition of regions. Adapted from [[#Seneviratne--2020|Seneviratne and Hauser (2020)]] under the terms of the Creative Commons Attribution licence.

To some extent, analyses as functions of GWLs replace the time axis with a global surface temperature axis. Nonetheless, information on the timing of given changes in extremes is obviously also relevant. (For information on the time frame at which given GWLs are reached, see Cross-Chapter Box 11.1 and [[IPCC:Wg1:Chapter:Chapter-4#4.6|Section 4.6]] ). Figure 11.5 provides a synthesis of attributed and projected changes in extremes as function of GWLs (see also Figures. 11.3, 11.6 and 11.7 for regional analyses).

<div id="cross-chapter-box-11.1" class="h2-container box-container"></div>

'''Cross-Chapter Box 11.1 | Translating Between Regional Information at Global Warming Levels''' '''Versus Scenario''' '''s for End Users'''

<div id="h2-23-siblings" class="h2-siblings"></div>

'''Contributors:''' Erich Fischer (Switzerland), Mathias Hauser (Switzerland), Sonia I. Seneviratne (Switzerland), Richard Betts (United Kingdom), José M. Gutiérrez (Spain), Richard G. Jones (United Kingdom), June-Yi Lee (Republic of Korea), Malte Meinshausen (Australia/Germany), Friederike Otto (United Kingdom/Germany), Izidine Pinto (Mozambique), Roshanka Ranasinghe (The Netherlands/Sri Lanka/Australia), Joeri Rogelj (Germany/Belgium), Bjørn Samset (Norway), Claudia Tebaldi (United States of America), Laurent Terray (France)

'''Background'''

Traditionally, projections of climate variables are summarized and communicated as function of time and emissions scenarios. Recently, quantifying global and regional climate at specific global warming levels (GWLs) has become widespread, motivated by the inclusion of explicit GWLs in the long-term temperature goal of the Paris Agreement ( [[IPCC:Wg1:Chapter:Chapter-1#1.6.2|Section 1.6.2]] ). GWLs, expressed as changes in global surface temperature relative to the 1850–1900 period (see Cross-Chapter Box 2.3), are used in SR1.5 and in the assessment of Reasons for Concerns in the WGII reports (see also Cross-Chapter Box 12.1). Cross-Chapter Box 11.1, Figure 1 illustrates how the assessment of the climate response at GWLs relates to the uncertainty in scenarios regarding the timing of the respective GWLs, as well as to the uncertainty in the associated regional climate responses, including extremes and other climatic impact-drivers (CIDs). For many (but not all) climate variables and CIDs, the response pattern for a given GWL is consistent across different scenarios (Chapters 1, 4, 9, 11 and Atlas). GWLs are defined as long-term means (e.g., 20-year averages) compared to the pre-industrial period, are commonly used in the literature, and were also underlying main assessments of SR1.5 (Chapter 3).

<div id="_idContainer038" class="Basic-Text-Frame"></div>

[[File:b7a56ba76272d0ff7a301f5255e3b28c IPCC_AR6_WGI_CCBox_11_1_Figure_1.png]]

'''Cross-Chapter Box 11.1, Figure 1 |'''

'''Schematic representation of relationship between emissions scenarios, global warming levels (GWLs), regional climate responses, and impacts.''' The illustration shows the implied uncertainty problem associated with differentiating between 1.5°C, 2°C, and other GWLs. Focusing on GWLs raises questions associated with emissions pathways to get to these temperatures (scenarios), as well as regional climate responses and the associated impacts at the corresponding GWL (the impacts question). Adapted from [[#James--2017|James et al. (2017)]] and [[#Rogelj--2013|Rogelj (2013)]] under the terms of the Creative Commons Attribution licence.

Numerous studies have compared the regional response to anthropogenic forcing at GWLs in annual and seasonal mean values and extremes of different climate and impact variables across different multi-model ensembles and/or different scenarios (e.g., Frieler et al. , 2012; Schewe et al. , 2014; Herger et al. , 2015; Schleussner et al. , 2016; Seneviratne et al. , 2016; Wartenburger et al. , 2017; Betts et al. , 2018; [[#Dosio--2018|Dosio and Fischer, 2018]] ; Samset et al. , 2019; Tebaldi et al. , 2020 ; see Sections 4.6.1, 8.5.3, 9.3.1, 9.5, 9.6.3, 10.4.3 and 11.2.4 for further details). The regional response patterns at given GWLs have been found to be consistent across different scenarios for many climate variables (Cross-Chapter Box 11.1 Figure 2; Pendergrass et al. , 2015; Seneviratne et al. , 2016; Wartenburger et al. , 2017; [[#Seneviratne--2020|Seneviratne and Hauser, 2020]] ) . The consistency tends to be higher for temperature-related variables than for variables in the hydrological cycle or variables characterizing atmospheric dynamics, and for intermediate to high-emissions scenarios than for low-emissions scenarios (e.g., for mean precipitation in the Representative Concentration Pathway (RCP) 2.6 scenario: [[#Pendergrass--2015|Pendergrass et al., 2015]] ; [[#Wartenburger--2017|Wartenburger et al., 2017]] ). Nonetheless, Cross-Chapter Box 11.1 Figure 2 illustrates that, even for mean precipitation, which is known to be forcing dependent (Sections 4.6.1 and 8.5.3), scenario differences in the response pattern at a given GWL are smaller than model uncertainty and internal variability in many regions ( [[#Herger--2015|Herger et al., 2015]] ). The response pattern is further found to be broadly consistent between models that reach a GWL relatively early, and those that reach it later under a given Shared Socio-economic Pathway (SSP; see Cross-Chapter Box 11.1, Figure 2g,h).

<div id="_idContainer041" class="Basic-Text-Frame"></div>

[[File:ae46a5e11d09f48a6a3483bfefac89b4 IPCC_AR6_WGI_CCBox_11_1_Figure_2.png]]

'''Cross-Chapter Box 11.1, Figure 2 |'''

'''(a–c) Coupled Model Intercomparison Project Phase 6 (CMIP6) multi-model mean precipitation change at 2°C global warming level (GWL) (20-year mean) in three different Shared Socio-economic Pathway (SSP) scenarios relative to 1850–1900.''' All models reaching the corresponding GWL in the corresponding scenario are averaged. The number of models averaged across is shown at the top right of the panel. The maps for the other two SSP scenarios SSP1-1.9 (five models only) and SSP3-7.0 (not shown) are consistent. '''(d–f)''' Same as (a–c) but for annual mean temperature. '''(g)''' Annual mean temperature change at 2°C in CMIP6 models with high warming rate reaching the GWL in the corresponding scenario before the earliest year of the assessed very likely range ( [[IPCC:Wg1:Chapter:Chapter-4#4.3.4|Section 4.3.4]] ). '''(h)''' Climate response at 2°C GWL across all SSP1-1.9, SSP2-2.6, SSP2-4.5. SSP3-7.0 and SSP5-8.5 in all other models not shown in (g). The close agreement of (g) and (h) demonstrates that the mean temperature response at 2°C is not sensitive to the rate of warming, and thereby the global mean surface air temperature (GSAT) warming of the respective models in 2081–2100. Uncertainty is represented using the advanced approach: No overlay indicates regions with robust signal, where ≥66% of models show change greater than the variability threshold and ≥80% of all models agree on the sign of change; diagonal lines indicate regions with no change or no robust signal, where &lt;66% of models show a change greater than the variability threshold; crossed lines indicate regions with conflicting signal, where ≥66% of models show change greater than the variability threshold and &lt;80% of all models agree on the sign of change. For more information on the advanced approach, please refer to the Cross-Chapter Box Atlas.1.

In contrast to linear pattern scaling ( [[#Mitchell--2003|Mitchell, 2003]] ; [[#Collins--2013|Collins et al., 2013]] ), the use of GWLs as a dimension of integration does not require linearity in the response of a climate variable. It is therefore useful even for metrics that do not show a linear response, such as the frequency of heat extremes over land and oceans ( [[#Fischer--2015|Fischer and Knutti, 2015]] ; [[#Frölicher--2018|Frölicher et al., 2018]] ; [[#Kharin--2018|Kharin et al., 2018]] ; [[#Perkins-Kirkpatrick--2017|Perkins-Kirkpatrick and Gibson, 2017]] ) if the relationship of the variable of interest to the GWL is scenario independent. The latter means that the response is independent of the pathway and relative contribution of various radiative forcings. For some more complex indices like warm-spell duration, or for regions with strong aerosol changes, discrepancies can be larger (Z. [[#Wang--2017|]] [[#Wang--2017|Wang et al., 2017]] b; [[#King--2018|King et al., 2018]] ; [[#Tebaldi--2020|Tebaldi et al., 2020]] ). (See also the subsection below on GWLs vs scenarios for further caveats.)

The limited scenario dependence of the GWL-based response for many variables implies that the regional response to emissions scenarios can be split in almost independent contributions of: (i) the transient global warming response to scenarios (see Chapter 4); and (ii) the regional response as function of a given GWL, which has also been referred to as ‘regional climate sensitivity’ ( [[#Seneviratne--2020|Seneviratne and Hauser, 2020]] ). This property has also been used to develop regionally resolved emulators for global climate models, using global surface temperature as input ( [[#Beusch--2020|Beusch et al., 2020]] ; [[#Tebaldi--2020|Tebaldi et al., 2020]] ). Analyses of the CMIP6 and CMIP5 multi-model ensembles shows that the GWL-based responses are very similar for temperature and precipitation extremes across the ensembles ( [[#Seneviratne--2020|Seneviratne and Hauser, 2020]] ; [[#Wehner--2020|Wehner, 2020]] ; Li et al. , 2021 ). This is despite their difference in global warming response (Chapter 4), confirming a substantial decoupling between the two responses (global warming vs GWL-based regional response) for these variables. Thus, the GWL approach isolates the uncertainty in the regional climate response from the global warming uncertainty induced by scenario, global mean model response and internal variability (Cross-Chapter Box, Figure 1).

'''Mapping between GWL- and scenario-based responses in model analyses'''
To map scenario-based climate projections into changes at specific GWLs, first, all individual Earth system model (ESM) simulations that reach a certain GWL are identified. Second, the climate response patterns at the respective GWL are calculated using an approach termed here ‘GWL-sampling’ – sometimes also referred to as epoch analysis, time shift, or time sampling approach – taking into account all models and scenarios (Cross-Chapter Box, Figure 3). Note that the range of years when a given GWL is reached in the CMIP6 ensemble is different from the AR6 assessed range of projected global surface temperature ( [[IPCC:Wg1:Chapter:Chapter-4#4.3.4|Section 4.3.4]] ; Table 4.5). The latter further takes into account different lines of evidence, including the assessed observed warming between pre-industrial and present day, information from observational constraints on CMIP6, and emulators using the assessed transient climate response (TCR) and equilibrium climate sensitivity (ECS) ranges ( [[IPCC:Wg1:Chapter:Chapter-4#4.3.4|Section 4.3.4]] ). Hence the [[IPCC:Wg1:Chapter:Chapter-4|Chapter 4]] assessed range (Table 4.5) is the reference to determine when a given GWL is ''likely'' reached under given scenarios, while the mapping between scenarios/time frames and GWLs is used to assess the respective regional responses happening at these time frames (which also allows accounting for the global surface temperature assessment, rather than using scenarios analyses directly from CMIP6 output).

In the model-based asssessment of Chapters 4, 8, 10, 11, 12 and the Atlas, the estimation of changes at GWLs are generally defined as the 20-year time period in which the mean global surface air temperature (GSAT; Cross-Chapter Box 2.3) first exceeds a certain anomaly relative to 1850–1900 – for simulations that start after 1850, relative to all years up to 1900 (Cross-Chapter Box Figure 3). The years when each individual model reaches a given GWL for CMIP6 and CMIP5 can be found in [[#Hauser--2021|Hauser et al. (2021)]] . The changes at given GWLs are identified for each ensemble member (for all scenarios) individually. Thereby, a given GWL is potentially reached a few years earlier or later in different realizations of the same model due to internal variability, but the temperature averaged across the 20-year period analysed in any simulation is consistent with the GWL. Instead of blending the information from the different scenarios, the Interactive [[IPCC:Wg1:Chapter:Atlas|Atlas]] can be used to compare the GWL spatial patterns and timings across the different scenarios (see Section ( [[IPCC:Wg1:Chapter:Atlas|Atlas]] 1.3.1).

<div id="_idContainer043" class="Basic-Text-Frame"></div>

[[File:bb5854aa2ebde4943d2aca103b5e727e IPCC_AR6_WGI_CCBox_11_1_Figure_3.png]]

'''Cross-Chapter Box11.1, Figure 3'''

|

'''Illustration of the AR6 global warming level (GWL) sampling approach to derive the timing and the response at a given GWL for the case of Coupled Model Intercomparison Project Phase 6 (CMIP6) data.''' For the mapping of scenarios/time slices into GWLs for CMIP6, please refer to Table 4.2. Respective numbers for the CMIP5 multi-model experiment are provided in [https://www.ipcc.ch/report/ar6/wg1/chapter/chapter-11 Chapter 11] Supplementary Material (11.SM.1). Note that the time frames used to derive the GWL time slices can also include a different number of years (e.g., 30 years for some analyses).

'''Mapping between GWL- and scenario-based responses for literature'''
A large fraction of the literature considers scenario-based analyses for given time slices. When GWL-based information is required instead, an approximated mapping of the multi-model mean can be derived based on the known GWL in the given experiments for a particular time period. As a rough approximation, CMIP6 multi-model mean projections for the near-term (2021–2040) correspond to changes at about 1.5°C, and projections for the high-end scenario (SSP5-8.5) for the long-term (2081–2100) correspond to about 4°C–5°C of global warming (see Table 4.2 for changes in the CMIP6 ensemble and the [https://www.ipcc.ch/report/ar6/wg1/chapter/chapter-11 Chapter 11] Supplementary Material (11.SM.1) and [[#Hauser--2021|Hauser (2021)]] for details on other time periods and CMIP5). These approximated changes are used for some of the GWL-based assessments provided in the [https://www.ipcc.ch/report/ar6/wg1/chapter/chapter-11 Chapter 11] regional tables ( [[#11.9|Section 11.9]] and Table 11.3) when literature based on scenario projections is used to assess estimated changes at given GWLs.

'''GWLs versus scenarios'''
The use of scenarios remains a key element to inform mitigation decisions (Cross-Chapter Box 1.4), to assess which emissions pathways are consistent with a certain GWL (Cross-Chapter Box 1.4, Figure 1), to estimate when certain GWLs are reached ( [[IPCC:Wg1:Chapter:Chapter-4#4.3.4|Section 4.3.4]] ), and to assess for which variables it is meaningful to use GWLs as a dimension of integration. The use of scenarios is also essential for variables whose climate response strongly depends on the contribution of radiative forcing (e.g., aerosols) or land-use and land management changes, are time and warming rate dependent (e.g., sea level rise), or differ between transient and quasi-equilibrium states. Furthermore, the use of concentration or emission-driven scenario simulations is required if regional climate assessments need to account for the uncertainty in GSAT changes or climate-carbon feedbacks.

Forcing dependence of the GWL response is found for global mean precipitation ( [[IPCC:Wg1:Chapter:Chapter-8#8.4.3|Section 8.4.3]] ), but less for regional patterns of mean precipitation changes (Cross-Chapter Box 11.1, Figure 2). Limited dependence is found for extremes, as highlighted above. In the cryosphere, elements that are quick to respond to warming like sea ice area, permafrost and snow, show little scenario dependence (Sections 9.3.1.1, 9.5.2.3 and 9.5.3.3), whereas slow-responding variables such as ice volumes of glaciers and ice sheets respond with a substantial delay and, due to their inertia, the response depends on when a certain GWL is reached. This also applies to some extent for sea level rise where, for example, the contributions of melting glaciers and ice sheets depend on the pathway followed to reach a given GWL ( [[IPCC:Wg1:Chapter:Chapter-9#9.6.3.4|Section 9.6.3.4]] ).

In addition to the lagged effect, the climate response at a given GWL may differ before and after a period of overshoot, for example in the Atlantic Meridional Overturning Circulation (e.g., Palter et al. 2018). Finally, as assessed in IPCC SR1.5, there is a difference in the response even for temperature-related variables if a GWL is reached in a rapidly warming transient state or in an equilibrium state when the land–sea warming contrast is less pronounced (e.g., King et al. 2020). However, in this Report, GWLs are used in the context of projections for the 21st century when the climate response is mostly not in equilibrium and where projections for many variables are less dependent on the pathway than for projections beyond 2100 ( [[IPCC:Wg1:Chapter:Chapter-9#9.6.3.4|Section 9.6.3.4]] ).

'''Key conclusions on assessments based on GWLs'''

GWL-based projections can inform society and policymakers on how climate would change under GWLs consistent with the aims of the Paris Agreement (stabilization at 1.5°C/well below 2°C), as well as on the consequences of missing these aims and reaching GWLs of 3°C or 4°C by the end of the century. The AR6 assessment shows that every bit of global warming matters and that changes in global warming of 0.5°C lead to statistically significant changes in mean climate and climate extremes on global scale and for large regions (Sections 4.6.2, 11.2.4, 11.3, 11.4, 11.6 and 11.9, Figures 11.8 and 11.9, [[IPCC:Wg1:Chapter:Atlas|Atlas]] and Interactive Atlas), as also assessed in IPCC SR1.5.

<div id="11.3" class="h1-container"></div>

<span id="temperature-extremes-1"></span>