Editing IPCC:AR6/WGI/Chapter-10 (section)

== 10.4 Interplay Between Anthropogenic Change and Internal Variability at Regional Scales ==

<div id="h1-5-siblings" class="h1-siblings"></div>

This section focuses on the assessment of the methodologies used to identify the physical causes of past and future regional climate change in the context of the ongoing anthropogenic influence on the global climate. The main foci are the attribution of past regional-scale changes (Sections 10.4.1–2) and the robustness and future emergence of the regional-scale response to anthropogenic forcing ( [[#10.4.3|Section 10.4.3]] ).

In this chapter, regional-scale attribution is defined as the process of evaluating the relative contributions of multiple causal factors (or drivers) to regional climate change (Cross-Working Group Box: Attribution in Chapter 1; [[#Rosenzweig--2013|Rosenzweig and Neofotis, 2013]] ; [[#Shepherd--2019|Shepherd, 2019]] ). Attribution at regional scale builds upon the usual definition of attribution used in the AR5 (Cross-Working Group Box: Attribution in Chapter 1; [[#Hegerl--2010|Hegerl et al., 2010]] ). However, in contrast with global-scale attribution methods where internal variability might be considered as a noise problem ( [[IPCC:Wg1:Chapter:Chapter-3#3.2|Section 3.2]] ), the preliminary detection step is not always required to perform regional-scale attribution since causal factors of regional climate change may also include internal modes of variability in addition to external natural and anthropogenic forcing. Importantly, regional-scale (or process-based) attribution also seeks to determine the physical processes and uncertainties involved in the causal factor’s influence (Cross-Working Group Box: Attribution in Chapter 1).

( [[#10.4.1|Section 10.4.1]] describes regional-scale attribution methodologies and assesses their application to regional changes of temperature and precipitation. [[#10.4.2|Section 10.4.2]] presents three illustrative attribution examples that illustrate a number of specific regional-scale challenges and methodological aspects. [[#10.4.3|Section 10.4.3]] focuses on methodologies used to assess the robustness and emergence of the regional climate response to anthropogenic forcing. A basic description of future regional climate change for all regions considered in the report (as defined in [[IPCC:Wg1:Chapter:Chapter-1#1.4.5|Section 1.4.5]] ) appears in the Atlas.

<div id="10.4.1" class="h2-container"></div>

<span id="methodologies-for-regional-climate-change-attribution"></span>
=== 10.4.1 Methodologies for Regional Climate Change Attribution ===

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

Attribution at sub-continental and regional scales is usually more complicated than at the global scale due to various factors: a larger contribution from internal variability, an increased similarity among the responses to different external forcings leading to a more difficult discrimination of their effects, the importance at regional scale of some omitted forcings in global model simulations, and model biases related to the representation of small-scale phenomena ( [[#Zhai--2018|Zhai et al., 2018]] ). Since AR5 and in addition to standard optimal fingerprint regression-based approaches ( [[IPCC:Wg1:Chapter:Chapter-3#3.2.1|Section 3.2.1]] and Zhai et al. 2018), several emerging methodologies have been increasingly used for regional-scale climate change attribution. These include several statistical approaches that differ in their use or omission of spatiotemporal co-variance information. Dynamical adjustment and pattern recognition techniques fall into the category of spatiotemporal methods while univariate detection and attribution methods rely on single grid-point analysis. Finally, the development, evaluation and use of all these methodologies rely upon the availability of multiple and high-quality observational datasets ( [[#10.2|Section 10.2]] ) as well as multi-model simulations of the historical period constrained by different external forcing combinations, including single-forcing experiments and single-model initial-condition large ensembles (SMILEs).

<div id="10.4.1.1" class="h3-container"></div>

<span id="optimal-fingerprinting-methods"></span>
==== 10.4.1.1 Optimal Fingerprinting Methods ====

<div id="h3-38-siblings" class="h3-siblings"></div>

Optimal fingerprint regression-based methods have been applied to detection and attribution of mean temperature anthropogenic signal in several regions of the world such as Canada, India, central Asia, northern and western China, Australia, and North Africa ( [[#Xu--2015|Xu et al., 2015]] ; [[#Li--2017|]] [[#Li--2017|C. Li et al., 2017]] ; [[#Dileepkumar--2018|Dileepkumar et al., 2018]] ; Y. [[#Wang--2018|]] [[#Wang--2018|Wang et al., 2018]] ; [[#Peng--2019|Peng et al., 2019]] ; [[#Wan--2019|Wan et al., 2019]] ). The influence of anthropogenic forcing, and in particular that of greenhouse gases (GHGs), is robustly detected in annual and seasonal mean temperatures for all considered regions. Most of the observed regional temperature changes since the mid-twentieth century can only be explained by external forcings, with anthropogenic influence being the dominant factor. GHG increase is found to be the primary factor of the anthropogenic-induced warming while the aerosol forcing leads to a cooling offsetting a fraction of the GHG change ( [[#Li--2016|]] [[#Li--2016|]] [[#Li--2016|]] [[#Li--2016|C. Li et al., 2016]] , 2017). While the influence of external natural forcing can often be detected as well, its contribution to observed changes is usually much smaller ( [[#Li--2017|]] [[#Li--2017|C. Li et al., 2017]] ; [[#Wan--2019|Wan et al., 2019]] ). Temperature detection results are found to be robust to the use of different observational datasets and detection methodologies ( [[#Dileepkumar--2018|Dileepkumar et al., 2018]] ).

Detection of mean precipitation changes caused by human influence is much more difficult, due to a larger role of internal variability at regional to local scales, as well as substantial modelling and observational uncertainty ( [[#Wan--2015|Wan et al., 2015]] ; [[#Sarojini--2016|Sarojini et al., 2016]] ; [[#Li--2017|]] [[#Li--2017|C. Li et al., 2017]] ). However, multi-decadal precipitation changes due to anthropogenic forcing have been detected for several regions. [[#Ma--2017b|Ma et al. (2017b)]] show that anthropogenic forcing has strongly contributed to the observed shift of China daily precipitation towards heavy precipitation. The observed weakening of the East Asia summer monsoon, also known as the southern flooding and northern drought pattern has been partially linked to anthropogenic forcing ( [[IPCC:Wg1:Chapter:Chapter-8#8.3.2.4.2|Section 8.3.2.4.2]] ; [[#Song--2014|Song et al., 2014]] ; [[#Zhou--2017|Zhou et al., 2017]] ; [[#Tian--2018|Tian et al., 2018]] ). Changes in GHGs lead to increasing precipitation over southern China, while changes in anthropogenic aerosols over East Asia are the dominant factors determining drought conditions over northern China ( [[#Song--2014|Song et al., 2014]] ; [[#Tian--2018|Tian et al., 2018]] ). Based on all-forcing and single-forcing simulation ensembles with a high-resolution model, [[#Delworth--2014|Delworth and Zeng (2014)]] found that the observed long-term regional austral autumn and winter rainfall decline over southern and particularly south-west Australia is partially reproduced in response to anthropogenic changes in GHGs and ozone in the atmosphere, whereas anthropogenic aerosols do not contribute to the simulated precipitation decline. In contrast, the observed increase of north-west Australian summer rainfall since 1950 has been partially attributed to anthropogenic aerosol based on CMIP5 detection and attribution single-forcing simulations ( [[IPCC:Wg1:Chapter:Chapter-8#8.3.2.4.6|Section 8.3.2.4.6]] ; [[#Dey--2019a|Dey et al., 2019a]] , [[#Dey--2019b|b]] ).

It is noteworthy that these methods require a very significant reduction of spatial and temporal dimensions in order to reliably estimate the co-variance matrix of internal variability (an entire region is thus often considered as being only one or a few spatial points that represent the spatial average of the whole region or a few sub-regions; time samples are often 5- or 10-year averages). Finally, model bias is rarely considered in statistical models used in detection and attribution regional studies, while it has been shown to have a strong impact on the stability of detection results and their associated confidence intervals when increasing the spatial dimension ( [[#Ribes--2013|Ribes and Terray, 2013]] ). New statistical methods are emerging to provide some alternative to standard optimal fingerprinting but they have not yet been evaluated and applied at regional scales ( [[IPCC:Wg1:Chapter:Chapter-3#3.2.2|Section 3.2.2]] ).

<div id="10.4.1.2" class="h3-container"></div>

<span id="other-spatiotemporal-statistical-methods-for-isolating-regional-climate-responses-to-external-forcing"></span>
==== 10.4.1.2 Other Spatiotemporal Statistical Methods for Isolating Regional Climate Responses to External Forcing ====

<div id="h3-39-siblings" class="h3-siblings"></div>

The primary objective of any attribution method is to optimally separate the influences of external forcing and internal variability on a global or regional climate record. In a multi-model ensemble context, the estimation of the externally-forced climate response has been typically performed by ensemble averaging of linear trends or regional domain spatial average, thus not taking into account the available and complete space and time co-variance information. Since AR5, methods using spatiotemporal information have been further developed and used to improve the separation between external and internal drivers in multiple or single historical climate realizations performed by a given global model.

The typical ensemble size of CMIP historical climate simulations for a given model traditionally range between one and ten members, with three often being the default choice. At the regional scale, a simple ensemble average with such sample sizes does not provide robust estimates of the response patterns to external forcing ( [[#Maher--2019|Maher et al., 2019]] ; [[#Deser--2020|Deser et al., 2020]] ). Since AR5, pattern filtering methods such as signal-to-noise maximizing empirical orthogonal functions ( [[#Ting--2009|Ting et al., 2009]] ) have been shown to improve the identification of forced response patterns when few model members are available ( [[#Wills--2020|Wills et al., 2020]] ). Using SMILEs as a test bed, it has been shown that pattern filtering strongly reduces the number of ensemble members needed to estimate the forced response pattern compared to simple ensemble averaging. Pattern filtering allows the identification of low signal-to-noise signals such as the El Niño-like response to volcanic eruptions ( [[#Khodri--2017|Khodri et al., 2017]] ; [[#Wills--2020|Wills et al., 2020]] ).

Methods to extract the response to external forcing in an observed or simulated single realization include dynamical adjustment ( [[#Smoliak--2015|Smoliak et al., 2015]] ; [[#Deser--2016|Deser et al., 2016]] ; [[#Sippel--2019|Sippel et al., 2019]] ) and time scale separation methods ( [[#DelSole--2011|DelSole et al., 2011]] ; [[#Wills--2018|Wills et al., 2018]] , 2020). Dynamical adjustment seeks to isolate changes in surface air temperature or precipitation that are due purely to atmospheric circulation changes. The residual can then be analysed and attributed to internal changes in both land or ocean surface conditions and the thermodynamical response to external forcing. [[#Smoliak--2015|Smoliak et al. (2015)]] performed their dynamical adjustment using partial least squares regression of temperature to remove variations arising from sea level pressure changes. [[#Deser--2016|Deser et al. (2016)]] used constructed atmospheric circulation analogues and resampling to estimate the dynamical contribution to changes in temperature. [[#Sippel--2019|Sippel et al. (2019)]] used machine learning techniques known as regularized linear regression to provide estimates of circulation-induced components of precipitation and temperature variability from global to local scales. It is noteworthy that the dynamical adjustment method by itself cannot account for the component of the forced response associated with circulation changes that project onto atmospheric internal variability. However, this component can be estimated within a model framework by averaging the dynamical contribution across multiple members of a SMILE ( [[#Deser--2016|Deser et al., 2016]] ).

Dynamical adjustment methods have been used by, for instance, [[#Deser--2016|Deser et al. (2016)]] , [[#Saffioti--2016|Saffioti et al. (2016)]] , [[#O’Reilly--2017|O’Reilly et al. (2017)]] , [[#Gong--2019|Gong et al. (2019)]] , and R. [[#Guo--2019|]] [[#Guo--2019|Guo et al. (2019)]] . [[#Deser--2016|Deser et al. (2016)]] focused on the causes of observed and simulated multi-decadal trends in North American temperature. They demonstrated that the main advantage of this technique is to narrow the spread of temperature trends found by the model ensemble and to bring the dynamically-adjusted observational trend much closer to the forced response estimated by the model ensemble mean. Similar results were obtained by [[#Saffioti--2016|Saffioti et al. (2016)]] regarding recent observed winter temperature and precipitation trends over Europe. Similarly, [[#O’Reilly--2017|O’Reilly et al. (2017)]] applied dynamical adjustment techniques to more carefully determine the influence of the Atlantic Multi-decadal Variability (AMV; Annex IV.2.7) on continental climates. Over Europe, summer temperature anomalies induced thermodynamically by the warm phase of the AMV are further reinforced by circulation anomalies; meanwhile, precipitation signals are largely controlled by dynamical responses to the AMV. Based on a partial least-squares approach, [[#Gong--2019|Gong et al. (2019)]] showed that recent winter temperature 30-year trends over northern East Asia are strongly influenced by internal variability linked to decadal changes of the Arctic Oscillation. Using dynamical adjustment purely on precipitation observations, R. [[#Guo--2019|]] [[#Guo--2019|Guo et al. (2019)]] showed that human influence has led to increased winter precipitation across north-eastern North America, as well as a small region of north-western North America, and to an increase in precipitation across much of north-western and north central Eurasia. The latter results confirm previous findings obtained by standard optimal fingerprinting methods ( [[#Wan--2015|Wan et al., 2015]] ).

Time scale separation methods such as the low-frequency component analysis and ensemble empirical mode decomposition methods take advantage of the longer time scale associated with anthropogenic external forcing compared to that of most internal modes of variability. The low-frequency component analysis method tries to find low-frequency variability patterns by searching for linear combinations of a moderate number of empirical orthogonal functions that maximize the ratio of low-frequency to total variance. It has first been used to separate internal modes of interannual and decadal variability from slowly varying and externally-forced variability in the Pacific and Atlantic oceans ( [[#Wills--2018|Wills et al., 2018]] , 2019). The methodology has also been applied to patterns of observed surface air temperature to isolate the slow components of observed changes that are consistent with the expected response to anthropogenic greenhouse gas and aerosol forcing ( [[#Wills--2020|Wills et al., 2020]] ).

The ensemble empirical mode decomposition method ( [[#Wu--2009|Wu and Huang, 2009]] ; [[#Wilcox--2013|Wilcox et al., 2013]] ; [[#Ji--2014|Ji et al., 2014]] ; [[#Qian--2014|Qian and Zhou, 2014]] ) decomposes data, such as time series of historical temperature and precipitation, into independent oscillatory modes of decreasing frequency. The last step of the method leaves behind a smooth and low-frequency residual time series. Typically, the non-linear anthropogenic trend (e.g., of 20th-century temperature) can be reconstructed by summing the long-term mean, the residual, and eventually the lowest-frequency mode to account for a multi-decadal forced signal, for instance associated with anthropogenic aerosol forcing. The ensemble empirical mode decomposition method is an example of a data-driven, non-parametric approach that can be used to directly provide an estimate of the forced response without the need for model data ( [[#Qian--2016|Qian, 2016]] ).

<div id="10.4.1.3" class="h3-container"></div>

<span id="other-regional-scale-attribution-approaches"></span>
==== 10.4.1.3 Other Regional-scale Attribution Approaches ====

<div id="h3-40-siblings" class="h3-siblings"></div>

The univariate detection method does not use spatial pattern information, but compares observed trends in gridded datasets with distributions of trends from ensembles of simulations during the historical period ( [[#Knutson--2013|Knutson et al., 2013]] ; [[#Knutson--2018|Knutson and Zeng, 2018]] ). The trends arising from simulations constrained by natural forcing-only and all-forcing are compared with distributions of trends purely due to internal variability and derived from long simulations with constant pre-industrial external forcing. Consistency between observed and simulated historical trends is also assessed with statistical tests that can be applied independently over a large number of grid points. The fraction of area over a given region where the change is classified as detectable, attributable, or consistent/inconsistent, is then finally estimated. The method can be viewed as a simple consistency test for both amplitude and pattern of observed versus simulated trends. Its application to CMIP3 and CMIP5 models suggests that 80% of the Earth’s surface has a detectable anthropogenic warming signal ( [[#Knutson--2013|Knutson et al., 2013]] ). Regarding regional land precipitation changes over the 1901–2010 and 1951–2010 periods, application of the univariate detection method based on CMIP5 models suggests attributable anthropogenic changes at several locations such as increases over regions of the north-central USA, southern Canada, Europe, and southern South America and decreases over parts of the Mediterranean region, northern tropical Africa and south-western Australia ( [[#Delworth--2014|Delworth and Zeng, 2014]] ; [[#Knutson--2018|Knutson and Zeng, 2018]] ).

Another regional attribution technique is based on the similarity of past changes between observations and one or several simulations of a large ensemble that share the same time evolution for a suggested driver of these changes. [[#Huang--2020b|Huang et al. (2020b)]] used a perturbed physics ensemble to attribute the drying trend of the Indian monsoon over the latter half of the 20th century to decadal forcing from the Pacific Decadal Variability (PDV; Annex IV.2.6). The ensemble members predicted different trends in PDV behaviour across the 20th century and the negative precipitation trend was only replicated in those members with a strong negative-to-positive PDV transition across the 1970s, consistent with the observed PDV behaviour (see also the detailed case study in [[#10.6.3|Section 10.6.3]] ). In a similar manner, [[#Cvijanovic--2017|Cvijanovic et al. (2017)]] addressed the possible influence of Arctic sea ice loss on the North Pacific pressure ridge and, consequently, on south-western USA precipitation. They sampled the uncertainties in selected sea ice physics parameters to achieve a ‘low Arctic sea ice’ state in their perturbed simulations. They then compared the latter with control simulations representative of sea ice conditions at the end of the 20th century to assess changes purely due to sea ice loss.

New methods aiming to remove underlying model biases before performing detection and attribution, for instance related to precipitation changes, are emerging based on image transformation techniques such as warping ( [[#Levy--2014a|Levy et al., 2014a]] ). By correcting location and seasonal precipitation biases in CMIP5 models, [[#Levy--2014b|Levy et al. (2014b)]] showed that the agreement between observed and fingerprint patterns can be improved, further enhancing the ability to attribute observed precipitation changes to external forcings. The improvement mainly relies on the assumption that precipitation changes are tied to the underlying climatology, which has been shown to be a reasonable assumption in regions of the world where intensification of the hydrological cycle is expected ( [[#Held--2006|Held and Soden, 2006]] ).

Importantly, evidence that the models employed in regional-scale attribution are fit for purpose is essential in order to estimate the degree of confidence in the attribution results ( [[#10.3.3|Section 10.3.3]] ). For example, models need to be evaluated and assessed in their ability to simulate internal variability modes that are known to be important drivers of regional climate change (Sections 3.7 and 10.3.3.3 and Annexes IV.2 and IV.3). Models are likely to have different performance in different regions and therefore their evaluation needs to be performed in terms of key physical processes and mechanisms relevant to the climate of the region under consideration ( [[#10.3.3|Section 10.3.3]] ).

To conclude, there is ''very'' ''high confidence'' ( ''robust evidence'' and ''high agreement'' ) that the use of diverse and independent attribution methods, multiple model ensemble types and observed datasets strengthens the robustness of results of regional-scale attribution studies. Since AR5, multiple SMILEs have provided an adequate testbed for new attribution methodologies aimed at separating forced signals from internal variability in observational records as well as small-size single-model ensembles.

<div id="10.4.2" class="h2-container"></div>

<span id="regional-climate-change-attribution-examples"></span>
=== 10.4.2 Regional Climate Change Attribution Examples ===

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

This section focuses on three illustrative examples that span different regions, time scales, and attribution methods, without aiming at being comprehensive. These examples illustrate attribution statements that are based upon multiple lines of evidence, combining multiple observational datasets, different generations and types of models, process understanding and assessment of various sources of uncertainty. Detection and attribution assessments for all AR6 regions and specific variables can be found in the Atlas.

<div id="10.4.2.1" class="h3-container"></div>

<span id="the-sahel-and-west-african-monsoon-drought-and-recovery"></span>
==== 10.4.2.1 The Sahel and West African Monsoon Drought and Recovery ====

<div id="h3-41-siblings" class="h3-siblings"></div>

The Sahel, fed by the West African monsoon, has experienced severe decadal rainfall variations (Figure 10.11a). Abundant rainfall in the 1950s–1960s was followed by a large negative trend (Figure 10.11b) until at least the 1980s, over which annual rainfall fell by 20–30% ( [[#Hulme--2001|Hulme, 2001]] ). The subsequent partial recovery ( [[#Wang--2021|]] [[#Wang--2021|B. Wang et al., 2021]] ) is more uncertain: rain-gauge studies suggest a return to long-term positive anomalies in the western Sahel in the early 2000s ( [[#Panthou--2018|Panthou et al., 2018]] ), while CHIRPS merged satellite/gauge data show a wetter western Sahel since 1981 ( [[#Bichet--2018a|Bichet and Diedhiou, 2018a]] , b). The recovery has been more significant over the central rather than the western Sahel ( [[#Lebel--2009|Lebel and Ali, 2009]] ; [[#Maidment--2015|Maidment et al., 2015]] ; Sanogo et al., 2015) and a multiple-gauge record supports a greater recovery to the eastern side ( [[#Nicholson--2018|Nicholson et al., 2018]] ). In this attribution example, drivers of the long-term drought and subsequent partial recovery are discussed, including anthropogenic GHG and aerosol emissions, and sea surface temperature (SST) variations that, in part, relate to internal variability. The reader is also referred to assessment in [[IPCC:Wg1:Chapter:Chapter-8#8.3.2.4|Section 8.3.2.4]] . We define the Sahel within 10°N–20°N across to 30°E, consistent with the eastern boundary used in Chapter 8, and the rainy season as spanning June to September.

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

[[File:1db32c7ae496c02335d5444aa0a60a8c IPCC_AR6_WGI_Figure_10_11.png]]

'''Figure 10.11''' '''|''' '''Attribution of historic precipitation change in the Sahelian West African monsoon during June to September. (a)''' Time series of CRU TS precipitation anomalies (mm day <sup>–1</sup> , baseline 1955–1984) in the Sahel box (10°N–20°N, 20°W–30°E) indicated in panel '''(b)''' applying the same low-pass filter as that used in Figure 10.10. The two periods used for difference diagnostics are shown in grey columns. (b) Precipitation change (mm day <sup>–1</sup> ) in CRU TS data for 1980–1990 minus 1950–1960 periods. '''(c)''' Precipitation difference (mm day <sup>–1</sup> ) between 1.5× and 0.2× historical aerosol emissions scaling factors averaged over 1955–1984 and five ensemble members of HadGEM3 experiments after [[#Shonk--2020|Shonk et al. (2020)]] . '''(d)''' Sahel precipitation anomaly time series (mm day <sup>–1</sup> , baseline 1955–1984) in Coupled Model Intercomparison Project Phase 6 (CMIP6) for 49 historical simulations with all forcings (red), and thirteen for each of greenhouse gas-only forcing (light blue) and aerosol-only forcing (grey), with a thirteen-point weighted running mean applied (a variant on the binomial filter with weights [1-6-19-42-71-96-106-96-71-42-19-6-1]). The CMIP6 subsample of all forcings matching the individual forcing simulations is also shown (pink). '''(e)''' Precipitation linear trend (% per decade) for (left) decline (1955–1984) and (right) recovery periods (1985–2014) for ensemble means and individual CMIP6 historical experiments (including single-forcing) as in panel (d) plus 34 CMIP5 models (dark blue). Box-and-whisker plots show the trend distribution of the three coupled and the d4PDF atmosphere-only single-model initial-condition large ensembles (SMILEs) used throughout (Chapter 10 and follow the methodology used in Figure 10.6. The two black crosses represent observational estimates from GPCC and CRU TS. Trends are estimated using ordinary least-squares regression. Further details on data sources and processing are available in the chapter data table (Table 10.SM.11).

The role of SST forcing in the rainfall decline is assessed first. Competing mechanisms from equatorial Atlantic SSTs and inter-hemispheric SST gradients regulate decadal variability in the Sahel ( [[#Nicholson--2013|Nicholson, 2013]] ), alternatively explained by tropical warming leading to Sahel drought, while North Atlantic warming promotes increased rainfall ( [[#Rodríguez-Fonseca--2015|Rodríguez-Fonseca et al., 2015]] ). The SST influence has been formalized in an AMV framework ( [[#Giannini--2013|Giannini et al., 2013]] ; [[#Martin--2014|Martin and Thorncroft, 2014]] ; [[#Martin--2014|Martin et al., 2014]] ; [[#Park--2015|Park et al., 2015]] ), suggesting that relative North Atlantic SST warming increases the Northern Hemisphere differential warming, enhancing Sahel rainfall. The AMV influence is supported by CMIP5 initialized decadal hindcasts ( [[#Gaetani--2013|Gaetani and Mohino, 2013]] ; [[#Mohino--2016|Mohino et al., 2016]] ; [[#Sheen--2017|Sheen et al., 2017]] ), which outperform empirical predictions based on persistence. Some caution is needed since the full magnitude of internal variability is not captured in most CMIP5 models, as poor resolution prevents reproduction of AMV teleconnection responses ( [[#Vellinga--2016|Vellinga et al., 2016]] ), and the magnitude of AMV-related SST variation may be underestimated in CMIP5 ( [[IPCC:Wg1:Chapter:Chapter-3#3.7.7|Section 3.7.7]] , which also assesses that the AMV may be partially forced). The influence of PDV has been studied to a lesser extent, with the PDV positive phase having a negative impact on Sahel rainfall in combined observational/CMIP5 analysis ( [[#Villamayor--2015|Villamayor and Mohino, 2015]] ). The closer match between the observed rainfall declining trend and those in an atmosphere-only SMILE, in which SSTs are matched to observations, compared to three coupled SMILEs in which they are not, suggests that the underlying ocean surface might be essential in driving the decline (Figure 10.11e).

In terms of anthropogenic emissions, regional aerosol emissions from Europe, and to a lesser extent from Asia, have been shown in a global model to weaken Sahel precipitation either through a weakened Saharan heat low or via the Walker circulation ( [[#Dong--2014|Dong et al., 2014]] ). Greenhouse gases (GHGs) and anthropogenic aerosol can be considered together to control ITCZ position based on temperature asymmetry at the hemispheric scale. GHGs increase Sahel precipitation, while aerosol reduces it (in coupled slab-ocean model experiments by [[#Ackerley--2011|Ackerley et al. (2011)]] following [[#Biasutti--2006|Biasutti and Giannini (2006)]] ). This effect is stronger when models account for aerosol–cloud interactions ( [[#Allen--2015|Allen et al., 2015]] ). Perturbed physics GCM ensembles suggests that aerosol emissions were the main driver of observed drying over 1950–1980 ( [[#Ackerley--2011|Ackerley et al., 2011]] ), supported by CMIP5 single-forcing experiments ( [[#Polson--2014|Polson et al., 2014]] ). A coherent drying signal in CMIP5 over the extended 1901–2010 period has also been found, although smaller than the observed trend ( [[#Knutson--2018|Knutson and Zeng, 2018]] ). By applying aerosol scaling factors to the historical period in order to sample the uncertainty in CMIP5 aerosol radiative forcing, [[#Shonk--2020|Shonk et al. (2020)]] found differences of 0.5 mm day <sup>–1</sup> for Gulf of Guinea rainfall between strong and weak aerosol experiments as illustrated in Figure 10.11c, although the drying appears further south than observed due to model bias.

For the partial recovery in West African monsoon and Sahel rainfall since the late 1980s, a detection study using three reanalyses ( [[#Cook--2015|Cook and Vizy, 2015]] ) shows a connection to increasing Saharan temperatures at a rate two to four times greater than the tropical mean, also confirmed by multiple observational and satellite-based data ( [[#Zhou--2016|Zhou and Wang, 2016]] ; [[#Vizy--2017|Vizy and Cook, 2017]] ) and the review of [[#Cook--2019|Cook and Vizy (2019)]] . Reanalyses are also noted to significantly underestimate the Saharan warming ( [[#Zhou--2016|Zhou and Wang, 2016]] ). Saharan warming causes a stronger thermal low and more intense monsoon flow, providing more moisture to the central and eastern Sahel, supported by CMIP5 models ( [[#Lavaysse--2016|Lavaysse et al., 2016]] ), although not all models capture the observed rainfall–heat–low relationship. Sahel rainfall is also incorrectly located in prototype versions of a few CMIP6 models, related to tropospheric temperature biases ( [[#Martin--2017|Martin et al., 2017]] ). Amplified Saharan warming has increased the wind shear, leading to a tripling of extreme storms since 1982, which may partially explain the recovery ( [[#Taylor--2017|Taylor et al., 2017]] ). Instead, observations, multiple models and SST-sensitivity experiments with AGCMs have suggested that stronger Mediterranean Sea evaporation enhances low-level moisture convergence to the Sahel, increasing rainfall ( [[#Park--2016|Park et al., 2016]] ). Meanwhile, an AGCM study suggested that GHGs alone (in the absence of SST warming) could cause Sahel rainfall recovery, with an additional role for anthropogenic aerosol ( [[#Dong--2015|Dong and Sutton, 2015]] ); recent changes in North Atlantic SSTs, although substantial, did not exert a significant impact on the recovery. Large spread in the recovery in a five-member AGCM ensemble suggests that atmospheric internal variability cannot be discounted ( [[#Roehrig--2013|Roehrig et al., 2013]] ).

Consistent timing of the southward ITCZ shift during the decline period in CMIP3 and CMIP5 historical simulations supports the role of external forcing, chiefly anthropogenic aerosol ( [[#Hwang--2013|Hwang et al., 2013]] ). The evolution of the observed decline and recovery is largely followed by the CMIP5 multi-model mean, further supporting the role of external drivers ( [[#Giannini--2019|Giannini and Kaplan, 2019]] ). Updated results from CMIP6 for historical simulations with all and single forcings are represented in Figure 10.11d,e showing smaller trends than those observed. [[#Giannini--2019|Giannini and Kaplan (2019)]] attempted to unify the driving mechanisms for decline and recovery based on singular-value decomposition of observed and modelled SSTs. Since the 1950s, tropical warming arising from GHGs and North Atlantic cooling from aerosol led to regional stabilization, suppressing Sahel rainfall. The subsequent reduction in aerosol emissions then led to North Atlantic warming and recovery of Sahel rainfall. Such mechanisms continue into the near-term future in idealized and modified RCP experiments, with scenarios featuring more aggressive reductions in aerosol emissions, or including aerosol–cloud interactions, favouring a greater northward shift of rainfall ( [[#Allen--2015|Allen, 2015]] ; [[#Westervelt--2017|Westervelt et al., 2017]] , 2018; [[#Scannell--2019|Scannell et al., 2019]] ). There is paleoclimate evidence of changes to Sahel rainfall in the past, in particular with enhancement of the West African monsoon during the mid-Holocene. However, the mechanisms governing such a change have been shown to be largely dynamical in nature ( [[#D’Agostino--2019|D’Agostino et al., 2019]] ), suggesting that the mid-Holocene cannot be used to inform the credibility of changes due to greenhouse warming.

There is ''very high confidence'' ( ''robust evidence'' and ''high agreement'' ) that patterns of 20th-century ocean and land surface temperature variability have caused the Sahel drought and subsequent recovery by adjusting meridional gradients. There is ''high confidence'' ( ''robust evidence'' and ''medium agreement'' ) that the changing temperature gradients that perturb the West African monsoon and Sahel rainfall are themselves driven by anthropogenic emissions: warming by GHG emissions was initially restricted to the tropics but suppressed in the North Atlantic due to nearby emissions of sulphate aerosols, leading to a reduction in rainfall. The North Atlantic subsequently warmed following the reduction of aerosol emissions, leading to rainfall recovery.

<div id="10.4.2.2" class="h3-container"></div>

<span id="the-south-eastern-south-america-summer-wetting"></span>
==== 10.4.2.2 The South-Eastern South America Summer Wetting ====

<div id="h3-42-siblings" class="h3-siblings"></div>

A positive trend in summer (December to February) precipitation has been detected in multiple observational sources in south-eastern South America since the beginning of the 20th century ( [[#Gonzalez--2013|Gonzalez et al., 2013]] ; [[#Vera--2015|Vera and Díaz, 2015]] ; [[#Wu--2016|Wu et al., 2016]] ; H. [[#Zhang--2016|]] [[#Zhang--2016|]] [[#Zhang--2016|Zhang et al., 2016]] ; [[#Díaz--2017|Díaz and Vera, 2017]] ; [[#Saurral--2017|Saurral et al., 2017]] ). Sedimentary records from the Mar Chiquita lake indicate that the last quarter of the 20th century was wetter than any period during the last 200 years ( [[#Piovano--2004|Piovano et al., 2004]] ). In this attribution example the drivers contributing to the positive trend for the period 1951–2014 are discussed (Figure 10.12a). Precipitation anomalies of Climatic Research Unit Time Series (CRU TS) as well as for the two members of a SMILE with the most negative and positive trends for 1951–2014 are displayed in Figure 10.12b. The trend for 1951–2014 using CRU TS and GPCC is illustrated in Figure 10.12c, and for the region defined by the black quadrilateral, it amounts to 2.8 (CRU TS) – 3.5 (GPCC) mm per month and decade (see black crosses in Figure 10.12d) while the mean summer monthly precipitation for the same period is 104 (CRU TS) –109 (GPCC) mm. The trend is also detectable in daily and monthly extremes ( [[#Re--2009|Re and Barros, 2009]] ; [[#Marengo--2010|Marengo et al., 2010]] ; [[#Penalba--2010|Penalba and Robledo, 2010]] ; [[#Doyle--2012|Doyle et al., 2012]] ; Donat et al., 2013; [[#Lorenz--2016|Lorenz et al., 2016]] ).

<div id="_idContainer040" class="Basic-Text-Frame"></div>

[[File:9df6b6493db188e1f655e2700e811c40 IPCC_AR6_WGI_Figure_10_12.png]]

'''Figure 10.1''' '''2 |''' '''South-Eastern South America positive mean precipitation trend and its drivers during 1951–2014. (a)''' Mechanisms that have been suggested to contribute to South-Eastern South America summer wetting. '''(b)''' Time series of austral summer (December to February) precipitation anomalies (%, baseline 1995–2014) over the South-Eastern South American region (26.25°S–38.75°S, 56.25°W–66.25°W), black quadrilateral in the first map of panel '''(c)''' . Black, brown and green lines show low-pass filtered time series for CRU TS), and the members with driest and wettest trends of the MPI-ESM single-model initial-condition large ensemble (SMILE; between 1951–2014), respectively. The filter is the same as the one used in Figure 10.10. (c) Mean austral summer precipitation spatial linear 1951–2014 trends (mm per month and decade) from CRU TS and GPCC. Trends are estimated using ordinary least squares regression. '''(d)''' Distribution of precipitation 1951–2014 trends over South-Eastern South America from GPCC and CRU TS (black crosses), CMIP6 all-forcing historical (red circles) and MIROC6, CSIRO-Mk3-6-0, MPI-ESM and d4PDF SMILEs (grey box-and-whisker plots). Grey squares refer to ensemble mean trends of their respective SMILE and the red circle refers to the CMIP6 multi-model mean. Box-and-whisker plots follow the methodology used in Figure 10.6. Further details on data sources and processing are available in the chapter data table (Table 10.SM.11).

The influence of SST anomalies on south-eastern South America precipitation have been studied extensively on interannual to multi-decadal time scales ( [[#Paegle--2002|Paegle and Mo, 2002]] ). The positive phase of El Niño–Southern Oscillation (ENSO; Annex IV.2.3) is related to stronger mean and extreme rainfall over south-eastern South America ( [[#Ropelewski--1987|Ropelewski and Halpert, 1987]] ; [[#Grimm--2009|Grimm and Tedeschi, 2009]] ; [[#Robledo--2016|Robledo et al., 2016]] ). The ENSO influence may be modulated by the PDV ( [[#Kayano--2007|Kayano and Andreoli, 2007]] ; [[#Fernandes--2018|Fernandes and Rodrigues, 2018]] ) and the AMV ( [[#Kayano--2014|Kayano and Capistrano, 2014]] ). PDV and AMV also influence the south-eastern South American climate independently of ENSO ( [[#Barreiro--2014|Barreiro et al., 2014]] ; [[#Grimm--2015|Grimm and Saboia, 2015]] ; [[#Robledo--2020|Robledo et al., 2020]] ). While Pacific SSTs dominate the overall influence of oceanic variability in the region, the Atlantic variability seems to dominate on multi-decadal time scales and has been proposed as a driver for the long-term positive trend ( [[#Seager--2010|Seager et al., 2010]] ; [[#Barreiro--2014|Barreiro et al., 2014]] ). Based on experiments designed to test how south-eastern South America precipitation is modulated by tropical Atlantic SSTs, [[#Seager--2010|Seager et al. (2010)]] showed that cold anomalies in the tropical Atlantic favour wetter conditions by inducing an upper-tropospheric flow towards the equator, which, via advection of vorticity, leads to ascending motion over south-eastern South America (Figure 10.12a). [[#Monerie--2019|Monerie et al. (2019)]] supported this argument showing a negative relationship between south-eastern South America precipitation and the AMV index ( [[#Huang--2015|Huang et al., 2015]] ) using an AGCM coupled to an ocean mixed-layer model with nudged SSTs.

The positive trend of precipitation has also been attributed to anthropogenic GHGemissions and stratospheric ozone depletion. CMIP5 models only show a positive trend when including anthropogenic forcings ( [[#Vera--2015|Vera and Díaz, 2015]] ). These results were supported by [[#Knutson--2018|Knutson and Zeng (2018)]] based on univariate detection/attribution analysis of annual mean trends for the 1901–2010 and 1951–2010 periods. However, the main features of summer mean precipitation and variability of South America are still not well-represented in all CMIP5 and CMIP6 models ( [[#Gulizia--2015|Gulizia and Camilloni, 2015]] ; [[#Díaz--2017|Díaz and Vera, 2017]] ; [[#Díaz--2021|Díaz et al., 2021]] ). This motivates the construction of ensembles that exclude the worst performing models ( [[#10.3.3.4|Section 10.3.3.4]] ). The construction of ensembles of CMIP5 historical simulations with realistic representation of precipitation anomalies with opposite sign over south-eastern South America and eastern Brazil showed that the trend since the 1950s could be related to changes in precipitation characteristics only when simulations included anthropogenic forcings ( [[#Díaz--2017|Díaz and Vera, 2017]] ). GHG emissions have been related to increased precipitation in south-eastern South America through three different mechanisms (Figure 10.12a). First, GHG warming induces a non-zonally uniform pattern of SST warming that includes a warming pattern over the Indian and Pacific oceans that excites wave responses over South America ( [[#Junquas--2013|Junquas et al., 2013]] ). Zonally uniform SST patterns of warming alone lead to precipitation signals opposite to those observed in an AGCM ( [[#Junquas--2013|Junquas et al., 2013]] ). Second, GHG radiative forcing drives an expansion of the Hadley cell so that its descending branch moves poleward from the region, generating anomalous ascending motion and precipitation (H. [[#Zhang--2016|]] [[#Zhang--2016|]] [[#Zhang--2016|Zhang et al., 2016]] ; [[#Saurral--2019|Saurral et al., 2019]] ). The third mechanism by which increased GHG can contribute to increased precipitation in the region is through a delay of the stratospheric polar vortex breakdown. As depicted in Figure 10.12a, both stratospheric ozone depletion and increased GHGs have contributed to the later breakdown of the polar vortex in recent decades ( [[#McLandress--2010|McLandress et al., 2010]] ; [[#Wu--2017|Wu and Polvani, 2017]] ; [[#Ceppi--2019|Ceppi and]] [[#Shepherd--2019|Shepherd, 2019]] ). [[#Mindlin--2020|Mindlin et al. (2020)]] developed future atmospheric circulation storylines ( [[#10.3.4.2|Section 10.3.4.2]] , Box 10.2) for Southern Hemisphere mid-latitudes with the CMIP5 models and found that for south-eastern South America summer precipitation, increases are related to the late-spring breakdown of the stratospheric polar vortex. The connecting mechanism is through a lagged southward shift of the jet stream ( [[#Saggioro--2019|Saggioro and]] [[#Shepherd--2019|Shepherd, 2019]] ), which enhances cyclonic activity over the region ( [[#Wu--2017|Wu and Polvani, 2017]] ).

A common feature among the above discussed studies is that even if global models simulate positive trends when forced with GHG and/or stratospheric ozone, these trends are in general smaller than those observed (e.g., CMIP6 trends in red open circles in Figure 10.12d). [[#Díaz--2021|Díaz et al. (2021)]] showed that to capture the observed trend a multi-model ensemble of SMILEs is needed. Out of the 12 large ensembles examined (with ensemble size varying in the 16–100 range), only seven simulated the observed trend within their range. This could partly be explained by model biases in mean precipitation and its interannual variability. In the sub-ensemble of six models that reproduce reasonably well the observed spatial patterns of mean precipitation and interannual variability, the ensemble mean spread is lower, and the forced response, taken as the multi-model ensemble mean, is slightly more positive than that of the six poorly performing models. The signal-to-noise ratio, estimated as the ratio of the forced response to the spread due to internal variability, is also slightly higher for the best-performing models, suggesting that selecting the best-performing models may have an influence on both attribution of the observed trend and emergence of the forced response in future ( [[#10.4.3|Section 10.4.3]] ).

There is ''high confidence'' that South-Eastern South America summer precipitation has increased since the beginning of the 20th century. Since AR5, science has advanced in the identification of the drivers of the precipitation increase in South-Eastern South America since 1950, including GHG through various mechanisms, stratospheric ozone depletion and Pacific and Atlantic variability. There is ''high confidence'' that anthropogenic forcing has contributed to the South-Eastern South America summer precipitation increase since 1950, but ''very low confidence'' on the relative contribution of each driver to the precipitation increase.

<div id="10.4.2.3" class="h3-container"></div>

<span id="the-south-western-north-america-drought"></span>
==== 10.4.2.3 The South-western North America Drought ====

<div id="h3-43-siblings" class="h3-siblings"></div>

Persistent hydroclimatic drought in south-western North America remains a much-studied event. Drought is a regular feature of the south-western North America’s climate regime, as can be seen in both the modern record, and through paleoclimate reconstructions ( [[#Cook--2010|Cook et al., 2010]] ; [[#Woodhouse--2010|Woodhouse et al., 2010]] ; [[#Williams--2020|Williams et al., 2020]] ), as well as in future climate model projections ( [[#Cook--2015a|Cook et al., 2015a]] ). Since the early 1980s, which were relatively wet in terms of precipitation and streamflow, the region has experienced major multi-year droughts such as the turn-of-the-century drought that lasted from 1999 to 2005, and the most recent and extreme 2012–2014 drought that in certain locations is perhaps unprecedented in the last millennium ( [[IPCC:Wg1:Chapter:Chapter-8#8.3.1.6|Section 8.3.1.6]] ; [[#Griffin--2014|Griffin and Anchukaitis, 2014]] ; [[#Robeson--2015|Robeson, 2015]] ). Shorter dry spells also happened between these multi-year droughts making 1980 to present a period with an exceptionally steep trend from wet to dry (Figure 10.13a), leading to strong declines in Rio Grande and Colorado river flows ( [[#Lehner--2017b|Lehner et al., 2017b]] ; [[#Udall--2017|Udall and Overpeck, 2017]] ). While robust attribution of this trend is complicated by the large natural variability in this region, the 20th century warming has been suggested to increase the chances for hydrological drought periods by lowering runoff efficiency ( [[#Woodhouse--2016|Woodhouse et al., 2016]] ; [[#Lehner--2017b|Lehner et al., 2017b]] ; [[#Woodhouse--2018|Woodhouse and Pederson, 2018]] ) and affecting evapotranspiration ( [[#Williams--2020|Williams et al., 2020]] ). There is some evidence suggesting that the Last Glacial Maximum, a period of low atmospheric CO <sub>2</sub> , about 21 ka ago, has a thermodynamically-driven zonal mean precipitation response similar to that of the current state with relatively high CO <sub>2</sub> levels when compared with the pre-industrial period. Pluvial conditions at that time and a reduction in precipitation from the Last Glacial Maximum to the pre-industrial period are consistent with drying trends for the region in models with GHG concentrations exceeding pre-industrial levels. However, the dominant large-scale drivers responsible for the precipitation changes observed during these two transitions are markedly different: mainly ice-sheet retreat and increasing insolation on one hand, increasing GHGs on the other hand. This suggests that the Last Glacial Maximum correspondence is fortuitous which strongly limits its use to capture future hydrological cycle changes ( [[IPCC:Wg1:Chapter:Chapter-8#8.3.2.4.4|Section 8.3.2.4.4]] ; [[#Morrill--2018|Morrill et al., 2018]] ; [[#Lowry--2019|Lowry and Morrill, 2019]] ). Furthermore, the conclusion of the Last Glacial Maximum drying versus wetting seems to strongly depend on the physical property of interest, hydrologic or vegetation indicators ( [[#Scheff--2017|Scheff et al., 2017]] ). Droughts are characterized by deficits in total soil moisture content that can be caused by a combination of decreasing precipitation and warming temperature, which promotes greater evapotranspiration. Regional-scale attribution of the prevalence of south-western North America drought since 1980 then mostly focuses on the attribution of change in these two variables.

<div id="_idContainer042" class="Basic-Text-Frame"></div>

[[File:724be120d5dc061c3a1e9a1bf1604e0e IPCC_AR6_WGI_Figure_10_13.png]]

'''Figure 10.13''' '''|''' '''Attribution of the south-western North America precipitation decline during the 1983–2014 period. (a)''' Water year (October to September) precipitation spatial linear trend (in percent per decade) over North America from 1983 to 2014. Trends are estimated using ordinary least squares. Top row: observed trends from CRU TS, REGEN, GPCC, and the Global Precipitation Climatology Project (GPCP). Middle row: driest, mean and wettest trends (relative to the region enclosed in the black quadrilateral, bottom row) from the 100 members of the MPI-ESM coupled SMILE. Bottom row: driest, mean and wettest trends relative to the above region from the 100 members of the d4PDF atmosphere-only SMILE. '''(b)''' Time series of water year precipitation anomalies (%, baseline 1971–2000) over the above south-western North America region for CRU TS (grey bar charts). Black, brown and green lines show low-pass filtered time series for CRU TS, driest and wettest members of the d4PDF SMILE, respectively. The filter is the same as the one used in Figure 10.10. '''(c)''' Distribution of south-western region-averaged water-year precipitation 1983–2014 trends (in percent per decade) for observations (CRU TS, REGEN, GPCC and GPCP, black crosses), CMIP6 all-forcing historical simulations (red circles), the MIROC6, CSIRO-Mk3-6-0, MPI-ESM and d4PDF SMILEs (grey box-and-whisker plots). Grey squares refer to ensemble mean trends of their respective SMILE and the red circle refers to the CMIP6 multi-model mean. Box-and-whisker plots follow the methodology used in Figure 10.6. Further details on data sources and processing are available in the chapter data table (Table 10.SM.11).

The observed south-western North America drying fits the narrative of what might happen in response to increasing GHG concentrations due to a poleward expansion of the subtropics, that is conducive to drying trends over subtropical to mid-latitude regions ( [[#Hu--2013b|Hu et al., 2013b]] ; [[#Birner--2014|Birner et al., 2014]] ; [[#Lucas--2014|Lucas et al., 2014]] ). However, several studies based on modern reanalyses and CMIP5 models have recently shown that the current contribution of GHGs to Northern Hemisphere tropical expansion is much smaller than in the Southern Hemisphere and will remain difficult to detect due to large internal variability, even by the end of the 21st century ( [[IPCC:Wg1:Chapter:Chapter-3#3.3.3.1|Section 3.3.3.1]] ; [[#Garfinkel--2015|Garfinkel et al., 2015]] ; [[#Allen--2017|Allen and Kovilakam, 2017]] ; [[#Grise--2018|Grise et al., 2018]] , 2019). In addition, the widening of the Northern Hemisphere tropical belt exhibits strong seasonality and zonal asymmetry, particularly in autumn and the North Atlantic ( [[#Amaya--2018|Amaya et al., 2018]] ; [[#Grise--2018|Grise et al., 2018]] ). Therefore, it seems that the recent Northern Hemisphere tropical expansion results from the interplay of internal and forced modes of tropical width variations and that the forced response has not robustly emerged from internal variability (Sections 3.3.3.1 and 10.4.3).

A second possible causal factor is the role for ocean-forced or internal atmospheric circulation change. Analysis of observed and CMIP5-simulated precipitation indicates that the drought prevalence since 1980 is linked to natural, internal variability in the climate system ( [[#Knutson--2018|Knutson and Zeng, 2018]] ). Based on observations and ensembles of SST-driven atmospheric simulations, [[#Seager--2014|Seager and Hoerling (2014)]] suggested that robust tropical Pacific and tropical North Atlantic forcing drove an important fraction of annual mean precipitation and soil moisture changes and that early 21st century multi-year droughts could be attributed to natural decadal swings in tropical Pacific and North Atlantic SSTs. A cold state of the tropical Pacific would lead by well-established atmospheric teleconnections to anomalous high pressure across the North Pacific and southern North America, favouring a weaker jet stream and a diversion of the Pacific storm track away from the south-west ( [[#Delworth--2015|Delworth et al., 2015]] ; [[#Seager--2017|Seager and Ting, 2017]] ). The multi-year drought of 2012–2016 has been linked to the multi-year persistence of anomalously high atmospheric pressure over the north-eastern Pacific Ocean, which deflected the Pacific storm track northward and suppressed regional precipitation during California’s rainy season ( [[#Swain--2017|Swain et al., 2017]] ). Going into more detail, [[#Prein--2016a|Prein et al. (2016a)]] used an assessment of changing occurrence of weather regimes to judge that changes in the frequency of certain regimes during 1979–2014 have led to a decline in precipitation by about 25%, chiefly related to the prevalence of anticyclonic circulation patterns in the north-east Pacific. Finally, the moderate model performance in representing Pacific SST decadal variability and its remote influence ( [[IPCC:Wg1:Chapter:Chapter-3#3.7.6|Section 3.7.6]] ) as well as its change under warming may affect attribution results of observed and future precipitation changes ( [[#Seager--2019|Seager et al., 2019]] ).

It has also been suggested that the ocean-controlled influence is limited and internal atmospheric variability has to be invoked to fully explain the observed history of drought on decadal time scales ( [[#Seager--2014|Seager and Hoerling, 2014]] ; [[#Seager--2017|Seager and Ting, 2017]] ). From roughly 1980 to the present, the regional climate signals show an interesting mix between forced and internal variability. [[#Lehner--2018|Lehner et al. (2018)]] used a dynamical adjustment method and large ensembles of coupled and SST-forced atmospheric experiments to suggest that the observed south-western North America rainfall decline mainly results from the effects of atmospheric internal variability, which is in part driven by a PDV-related phase shift in Pacific SST around 2000 (Figure 10.13b,c). Based upon four SMILEs (three using a GCM and another one an AGCM constrained by observed SSTs) and a CMIP6 multi-model suite constrained by observed external forcings, Figure 10.13 shows, in agreement with [[#Lehner--2018|Lehner et al. (2018)]] , that observed SSTs with their associated atmospheric response are the main drivers of the south-western North America precipitation decrease during the 1983–2014 period. Once aspects of the internal variability are removed by dynamical adjustment, the observed precipitation change signal and simulated anthropogenically-forced components look more similar ( [[#Lehner--2018|Lehner et al., 2018]] ).

Importantly, as the AR6 assessment views the PDV as being mostly driven by internal variability ( [[IPCC:Wg1:Chapter:Chapter-3#3.7.6|Section 3.7.6]] ), the lines of evidence cited above suggest that the contribution of natural and anthropogenic forcings to the precipitation decline has a small amplitude. Unlike the precipitation deficit, the accompanying south-western North America warming is driven primarily by anthropogenic forcing from GHGs rather than atmospheric circulation variability and may help to enhance the drought through increased evapotranspiration ( [[#Knutson--2013|Knutson et al., 2013]] ; [[#Diffenbaugh--2015|Diffenbaugh et al., 2015]] ; [[#Williams--2015|Williams et al., 2015]] , [[#Williams--2020|Williams et al., 2020]] ; [[#Lehner--2018|Lehner et al., 2018]] , [[#Lehner--2020|2020]] ).

To conclude, there is ''high confidence'' ( ''robust evidence'' and ''medium agreement'' ) that most (&gt;50%) of the anomalous atmospheric circulation that caused the south-western North America negative precipitation trend can be attributed to teleconnections arising from tropical Pacific SST variations related to PDV. There is ''high confidence'' ( ''robust evidence'' and ''medium agreement'' ) that anthropogenic forcing has made a substantial contribution (about 50%) to the south-western North America warming since 1980.

<div id="10.4.2.4" class="h3-container"></div>

<span id="assessment-summary"></span>
==== 10.4.2.4 Assessment Summary ====

<div id="h3-44-siblings" class="h3-siblings"></div>

The robustness of regional-scale attribution differs strongly between temperature and precipitation changes. While the influence of anthropogenic forcing on regional temperature long-term change has been detected and attributed in almost all land regions, a robust detection and attribution of human influence on regional precipitation change has not yet fully occurred for many land regions ( [[#10.4.3|Section 10.4.3]] ). Although the contribution of anthropogenic forcing to long-term regional precipitation change has been detected in some regions, a robust quantification of the contributions of different drivers remains elusive. The delayed emergence of the anthropogenic precipitation fingerprint with respect to temperature is likely due to the opposing sign of the fast and slow land precipitation forced responses and time-dependent SST change patterns (Sections 8.2.1 and [[#10.4.3|Section 10.4.3]] ), stronger internal variability ( [[#10.3.4.3|Section 10.3.4.3]] ) as well as larger observational uncertainty ( [[#10.2|Section 10.2]] ) and impact of model biases. The contribution of internal variability to the observed changes can also be very sensitive to the period length and level of spatial aggregation for the region under scrutiny ( [[IPCC:Wg1:Chapter:Chapter-4#4.4.1|Section 4.4.1]] and Cross-Chapter Box 3.1; [[#Kumar--2016|Kumar et al., 2016]] ). Finally, even in the case of temperature changes at multi-decadal time scale, internal variability can still be a substantial driver of regional changes due to cancellation between different external forcings ( [[#Nath--2018|Nath et al., 2018]] ).

To conclude, it is ''virtually certain'' ( ''robust evidence'' and ''high agreement'' ) that anthropogenic forcing has been a major driver of temperature change since 1950 in many sub-continental regions of the world. There is ''high confidence'' ( ''robust evidence'' and ''medium agreement'' ) that anthropogenic forcing has contributed to multi-decadal mean precipitation changes in several regions, for example western Africa, south-east South America, south-western Australia, northern central Eurasia, and South and East Asia. However, at regional scale, the role of internal variability is stronger while uncertainties in observations, models and external forcing are all larger than at the global scale, precluding a robust assessment of the magnitude of the relative contributions of greenhouse gases, including stratospheric ozone, and different aerosol species.

<div id="10.4.3" class="h2-container"></div>

<span id="future-regional-changes-robustness-and-emergence-of-the-anthropogenic-signal"></span>
=== 10.4.3 Future Regional Changes: Robustness and Emergence of the Anthropogenic Signal ===

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

Regional climate projections are one key element of the multiple lines of evidence that are used for climate risk assessments as well as for adaptation and policy decisions at regional scales (Sections 10.3.3.9 and 10.5). Regional climate projections can be separated into two components: the regional-scale forced response or regional-scale climate sensitivity when normalized by the global mean temperature change ( [[#Seneviratne--2020|Seneviratne and Hauser, 2020]] ) and the climate internal variability characterizing the future period or global warming level under scrutiny. This section assesses a few methodological aspects related to robustness and emergence properties of the regional-scale forced response as well as the possible influence of internal variability on the emergence of the anthropogenic signal.

<div id="10.4.3.1" class="h3-container"></div>

<span id="robustness-of-the-anthropogenic-signal-at-regional-scale"></span>
==== 10.4.3.1 Robustness of the Anthropogenic Signal at Regional Scale ====

<div id="h3-45-siblings" class="h3-siblings"></div>

Standard methodologies to derive the regional forced response include pattern-scaling and the time-shift or epoch approach ( [[IPCC:Wg1:Chapter:Chapter-4#4.2.4|Section 4.2.4]] ; [[#Tebaldi--2014|Tebaldi and Arblaster, 2014]] ; [[#Vautard--2014|Vautard et al., 2014]] ; [[#Herger--2015|Herger et al., 2015]] ; [[#Tebaldi--2018|Tebaldi and Knutti, 2018]] ; [[#Christensen--2019|Christensen et al., 2019]] ). Pattern-scaling assumes that the spatial patterns of regional change, often based on a time-averaged 20- or 30-year period at the end of the 21st century, are roughly constant in time, and simply scale linearly with global mean warming. The time-shift approach defines a target in terms of global warming level (GWL) and locates the time segment, usually 20 or 30 years, in historical or scenario simulations in which global mean warming matches the required GWL ( [[#10.1.2|Section 10.1.2]] and Cross-Chapter Box 11.1). Physical consistency between multiple variables and space-time co-variance are fully preserved in the time-shift approach, which is not the case for pattern-scaling ( [[#Herger--2015|Herger et al., 2015]] ). Importantly, pattern scaling cannot account for the non-linearity arising from either interacting quasi-linear processes ( [[#Chadwick--2013|Chadwick and Good, 2013]] ) and purely non-linear mechanisms, which have been shown to be present in CMIP5 models for high GWL (4°C) and affect precipitation more than temperature at the regional-scale ( [[IPCC:Wg1:Chapter:Chapter-8#8.5.3.1|Section 8.5.3.1]] ; [[#Good--2015|Good et al., 2015]] , 2016). The time-shift approach can also be used to test whether regional climate change patterns depend on the rate of global mean warming and external forcing pathways, in addition to global warming magnitude. A global evaluation of both approaches in projecting the forced temperature and precipitation response for a highly mitigated scenario based on a moderately mitigated one has been performed using a perfect-model framework ( [[#Tebaldi--2018|Tebaldi and Knutti, 2018]] ). The amplitude of errors for both approaches appears to be substantially smaller than model uncertainty approximated by the CMIP5 multi-model spread.

Based on large and coordinated modelling exercises such as CMIP5 and CORDEX, the time-shift approach has been largely used to assess differences in regional climate impacts for different GWLs, with a strong focus on 1.5°C versus 2°C ( [[#Karmalkar--2017|Karmalkar and Bradley, 2017]] ; [[#Dosio--2018|Dosio and Fischer, 2018]] ; [[#Karnauskas--2018|Karnauskas et al., 2018]] ; W. [[#Liu--2018|]] [[#Liu--2018|Liu et al., 2018]] ; [[#Taylor--2018|Taylor et al., 2018]] ; [[#Weber--2018|Weber et al., 2018]] ; Chapter 3, SR1.5, [[#Hoegh-Guldberg--2018|Hoegh-Guldberg et al., 2018]] ). Comparisons between pattern-scaling and time-shift approaches allow assessment of the scalability of the regional climate change signal and the extent to which pattern-scaling assumptions still hold at regional scale for a wide range of GWL. This was the approach followed by [[#Matte--2019|Matte et al. (2019)]] in their assessment of the scalability of European regional climate projections. Based on EURO-CORDEX projections, they performed a detailed comparison between the pattern scaling and the GWL spatial patterns (GWL range: 1°C, 2°C and 3°C) for different seasons, regional model resolutions, and both temperature and precipitation. High pattern correlation values (greater than 0.9) are found between the scaled pattern and all GWL patterns for temperature. In the case of precipitation, the correspondence is slightly lower, especially in summer, for high GWLs (2°C and 3°C) and much lower for 1°C.

Figure 10.14 illustrates a similar comparison based on the CMIP6 multi-model ensemble forced with the scenario SSP5-8.5 and applied to two large-scale continental areas. The forced response to anthropogenic forcing is simply taken as the CMIP6 multi-model mean of future regional climate change relative to the 1850–1900 reference period. Robustness of the forced response is based on both significance of the change and model agreement about the sign (direction) of change (Cross-Chapter Box Atlas.1; Figure 10.14). Caution has to be exercised against a too literal interpretation of lack of robust change given that significance and sign agreement can be sensitive to spatial and temporal aggregation (Cross-Chapter Box Atlas.1, Figure 2) and lack of a robust change does not necessarily translate to lack of regional-scale climate change impacts ( [[#McSweeney--2013|McSweeney and Jones, 2013]] ; [[#Hibino--2016|Hibino and Takayabu, 2016]] ).

<div id="_idContainer044" class="Basic-Text-Frame"></div>

[[File:1c60cae6202f2ab8868a790176e35a5e IPCC_AR6_WGI_Figure_10_14.png]]

'''Figure 10.14''' '''|''' '''Robustness and scalability of anthropogenic signals at regional scale. (a)''' Spatial patterns of European and African summer (June to August) surface air temperature change (in °C °C <sup>–1</sup> ) from the Coupled Model Intercomparison Project Phase 6 (CMIP6) multi-model mean (45 models, one member per model, historical simulations and scenario SSP5-8.5) at different global warming levels (GWLs) and the end-21st century scaling pattern estimated from the multi-model mean difference between 2081–2100 and the pre-industrial period (1850–1900) divided by the corresponding global mean warming. The scale of all GWL patterns has been adjusted to a global mean warming of 1°C (for example, the resulting 3°C spatial pattern has been divided by three). The scales of the GWL patterns have to be multiplied by their threshold values to obtain the actual simulated warming. The metrics shown in the bottom left corner of the GWL pattern plots indicate the spatial pattern correlation and the root-mean-square difference between the GWL patterns and the scaling pattern. The number in bold just above the metrics gives the number of used CMIP6 models (out of 45) that have reached the GWL threshold. Areas with robust change (at least 66% of the models have a signal-to-noise ratio greater than one and 80% or more of the models agree on the sign of the change) are coloured with no pattern overlaid (Cross-Chapter Box Atlas.1). Areas with a significant change (at least 66% of the models have a signal-to-noise ratio greater than one) and lack of model agreement (meaning that less than 80% of the models agree on the sign of the change) are marked by cross-hatching. Areas with no change or no robust change (less than 66% of the models have a signal-to-noise ratio greater than one) are marked by negatively sloped hatching. '''(b)''' Same as (a) but for North, Central and South America annual mean precipitation relative change (percent °C <sup>–1</sup> ). The baseline for precipitation climatology is 1850–1900. Further details on data sources and processing are available in the chapter data table (Table 10.SM.11).

If projected regional mean temperature (Figure 10.14a) and precipitation (Figure 10.14b) changes were to scale linearly with global mean warming, the adjusted spatial patterns would be congruent with each other at different GWLs. While pattern scaling seems to be a reasonable first-order approximation for both temperature and precipitation changes in tropical and high latitude regions (high pattern correlation values), there are a number of regions exhibiting substantial amplitude differences at different GWLs (northern Africa and Middle East, southern and eastern Europe for temperature; south-western North America, Chile and north-eastern Brazil for precipitation). These differences hint at the possible influence of non-linear mechanisms ( [[#Good--2015|Good et al., 2015]] ), including soil-moisture feedbacks ( [[#Seneviratne--2010|Seneviratne et al., 2010]] ; [[#Vogel--2017|Vogel et al., 2017]] ), a time-dependent balance between the different contributions of fast and slow response to greenhouse gas forcing as well as changing SST response patterns ( [[#Long--2014|Long et al., 2014]] ; [[#Good--2016|Good et al., 2016]] ; [[#Ceppi--2018|Ceppi et al., 2018]] ; [[#Zappa--2020|Zappa et al., 2020]] ). Decreasing spatial pattern amplitude with increasing GWL suggests that the initial transient regional response overshoots the long-term change in regions such as northern Africa for summer temperature and south-western South America for precipitation ( [[#Zappa--2020|Zappa et al., 2020]] ). In the latter region, long simulations with stabilized GHG concentrations even suggest a change of sign when near-equilibrium is reached ( [[#Sniderman--2019|Sniderman et al., 2019]] ). The reverse behaviour, increasing pattern amplitude with increasing GWL, is seen for summer temperature in southern and eastern Europe and for precipitation in south-western North America ( [[#Sniderman--2019|Sniderman et al., 2019]] ; [[#Zappa--2020|Zappa et al., 2020]] ), suggesting that, in these regions, the initial transient response is lagging global mean warming and final regional climate change will be reached once GHG concentrations are stabilized.

There is ''high confidence'' that the time-evolving contribution of different mechanisms operating at different time scales can modify the amplitude of the regional-scale response of temperature, and both the amplitude and sign of the regional-scale response of precipitation, to anthropogenic forcing. These mechanisms include non-linear temperature, precipitation and soil-moisture feedbacks, and slow and fast response of SST patterns and atmospheric circulation changes to increasing GHGs.

<div id="10.4.3.2" class="h3-container"></div>

<span id="emergence-of-the-anthropogenic-signal-at-regional-scale"></span>
==== 10.4.3.2 Emergence of the Anthropogenic Signal at Regional Scale ====

<div id="h3-46-siblings" class="h3-siblings"></div>

This section provides an assessment of the different approaches used in emergence studies as well as sensitivities to methodological choices. The section then focuses on the possible influence of internal variability on future emergence of the simulated mean precipitation anthropogenic signal at regional scales with some illustrative examples.

In climate science, emergence or distinguishability of a signal refers to the appearance of a persistent change in the probability distribution and/or temporal properties of a climate variable compared with that of a reference period ( [[IPCC:Wg1:Chapter:Chapter-1#1.4.2|Section 1.4.2]] ; [[#Giorgi--2009|Giorgi and Bi, 2009]] ; [[#Mahlstein--2011|Mahlstein et al., 2011]] , [[#Mahlstein--2012|2012]] ; [[#Hawkins--2012|Hawkins and Sutton, 2012]] ). Similar to anthropogenic climate change detection (Cross-Working Group Box: Attribution in Chapter 1), signal emergence can be detected, at least initially, without identifying the physical causes of the emergence ( [[IPCC:Wg1:Chapter:Chapter-1#1.4.2|Section 1.4.2]] ). In the context of human influence on climate, the objective of emergence studies is the search for the appearance of a signal characterizing an anthropogenically-forced change relatively to the climate variability of a reference period, defined as the noise.

Precise definitions of signal and noise as well as a metric to measure the relative importance of the signal are key ingredients of the emergence framework and depend on the framing question. In particular, emergence study results can depend on the specific definitions of signal and noise such as the level of spatial and temporal aggregation ( [[#McSweeney--2013|McSweeney and Jones, 2013]] ). For instance, grid-point scale emergence will likely be delayed compared with region-average emergence ( [[IPCC:Wg1:Chapter:Chapter-11#11.2.4|Section 11.2.4]] and [[IPCC:Wg1:Chapter:Atlas-1-figure-2|Cross-Chapter Box Atlas.1, Figure 2]] ; [[#Fischer--2013|Fischer et al., 2013]] ; [[#Maraun--2013b|Maraun, 2013b]] ; [[#Lehner--2017a|Lehner et al., 2017a]] ). The signal is often estimated by a running mean multi-decadal average or probability distribution function of the physical variable under scrutiny in order to avoid false emergence due to manifestation of multi-decadal internal variability ( [[#King--2015|King et al., 2015]] ). In the case of extremes such as climate records, a notion of multi-year persistence or recurrence can also be used to fully characterize the anthropogenic signal and its emergence ( [[#Christiansen--2013|Christiansen, 2013]] ; [[#Bador--2016|Bador et al., 2016]] ).

Emergence is also sensitive to the noise characteristics: assuming a common signal definition, larger signal-to-noise values and earlier emergence will arise if the noise is based on decadal mean variability rather than interannual variability ( [[#Kusunoki--2020|Kusunoki et al., 2020]] ). Depending on the framing question, the noise can include or omit external natural forcing such as volcanic and solar forcing ( [[#Zhang--2018|Zhang and Delworth, 2018]] ; [[#Silvy--2020|Silvy et al., 2020]] ). Furthermore, emergence results are very sensitive to the choice and length of the reference period ( [[IPCC:Wg1:Chapter:Chapter-1#1.4.1|Section 1.4.1]] ). The reference period can be the pre-industrial, the very recent past or even a time-evolving baseline, depending on both the framing and assumption that adaptation to the current climate has already occurred ( [[#King--2015|King et al., 2015]] ; [[#Zhang--2018|Zhang and Delworth, 2018]] ; [[#Brouillet--2020|Brouillet and Joussaume, 2020]] ). These choices will then determine the type of simulations and periods that will be used to construct the noise distribution. Finally, the permanence of future emergence cannot be taken for granted when emergence occurs in the late-21st century based on simulations ending in 2100 ( [[#Hawkins--2014|Hawkins et al., 2014]] ; [[#King--2015|King et al., 2015]] ; [[#Lehner--2017a|Lehner et al., 2017a]] ).

Robust assessments and comparisons of past emergence between observations and models are strengthened by the use of consistent definitions of signal and noise ( [[#Lehner--2017a|Lehner et al., 2017a]] ; [[#Hawkins--2020|Hawkins et al., 2020]] ). In the case of future emergence under increasing greenhouse gas emissions, two main approaches have been followed to assess emergence. The first is based on estimating the signal and noise (and sometimes the signal-to-noise ratio as well) in individual models before using the resulting distribution median or mean to construct the final emergence metric ( [[#Hawkins--2012|Hawkins and Sutton, 2012]] ; [[#Maraun--2013b|Maraun, 2013b]] ; [[#Sui--2014|Sui et al., 2014]] ; [[#Barrow--2019|Barrow and Sauchyn, 2019]] ). The second method first estimates the signal as a multi-model mean change and the noise variance as a combination of internal variability and model structural differences ( [[#Giorgi--2009|Giorgi and Bi, 2009]] ; [[#Mariotti--2015|Mariotti et al., 2015]] ; [[#Nguyen--2018|Nguyen et al., 2018]] ). The first approach allows the definition of emergence of the signal relative to internal variability only and treats model error as source of uncertainty ( [[#Maraun--2013b|Maraun, 2013b]] ; [[#Lehner--2017a|Lehner et al., 2017a]] ). The second assumes that the multi-model mean is the optimal estimate of the signal and confounds internal variability and model structural differences in the noise estimate. It is noteworthy that most emergence studies implicitly assume model independence ( [[#Annan--2017|Annan and Hargreaves, 2017]] ; [[#Boé--2018|Boé, 2018]] ; Box 4.1) and therefore sensitivity of emergence results to model selection or weighting is rarely performed ( [[#Akhter--2018|Akhter et al., 2018]] ).

Metrics can vary from a simple signal-to-noise ratio to statistical distributional tests ( [[#King--2015|King et al., 2015]] ; [[#Gaetani--2020|Gaetani et al., 2020]] ) and give median estimates and uncertainty bounds for the date (or time of emergence) corresponding to the exceedance of specific thresholds by the emergence metric. Reconciling future emergence results among different studies is challenging due to their many methodological differences including the choice of the reference period, the selected climate models and scenario, the precise definition of signal and noise and the choice of different signal-to-noise thresholds to characterize robust emergence. Contrasting with binary yes/no statements, emergence can also be viewed as a continuous process characterized by an amplitude or level, for example the value of the signal-to-noise ratio, that is a function of time or global warming level.

Since AR5, the development and production of SMILEs (Sections 4.2.5 and 10.3.4.3) has allowed the assessment of the influence of internal variability on anthropogenic signal emergence. The influence of internal variability, and specifically of the unforced atmospheric circulation, on temperature signal emergence can delay or advance the time of emergence by a decade or two in mid- to high-latitude regions ( [[#Lehner--2017a|Lehner et al., 2017a]] ; [[#Koenigk--2020|Koenigk et al., 2020]] ). Internal variability can also result in small or decreasing decadal to multi-decadal heatwave frequency trends under the historical anthropogenic forcing over most regions, thereby delaying emergence of unprecedented heatwave frequency trends relative to the pre-industrial trend distribution (Sections 11.2–11.3; [[#Perkins-Kirkpatrick--2017|Perkins-Kirkpatrick et al., 2017]] ).

Regional precipitation future changes are much more impacted by internal variability than their temperature counterpart ( [[#Monerie--2017b|Monerie et al., 2017b]] ; [[#Dai--2019|Dai and Bloecker, 2019]] ; [[#Singh--2019|Singh and AchutaRao, 2019]] ; [[#von%20Trentini--2019|von Trentini et al., 2019]] ; [[#Koenigk--2020|Koenigk et al., 2020]] ). Relative to mean temperature changes, this larger influence of internal variability on mean precipitation changes contributes, among other factors ( [[#Sarojini--2016|Sarojini et al., 2016]] ), to a much delayed emergence of the forced precipitation response in observations ( [[#Hawkins--2020|Hawkins et al., 2020]] ). Based on the CMIP6 multi-model ensemble forced with the scenario SSP5-8.5, we assess the future emergence of mean precipitation forced change as a function of GWLs for all AR6 land regions (Figure 10.15a). The methodology is a straightforward adaptation of the standard approach ( [[#Hawkins--2012|Hawkins and Sutton, 2012]] ). While the standard method is only based on the signal-to-noise ratio exceedance of a specified threshold (taken as one), the approach used here assumes that grid-point emergence occurs when the forced change is considered robust following the AR6 WGI definition of robustness for projected changes (Cross-Chapter Box Atlas.1). At a GWL of 1°C, emergence only occurs in high-latitude regions ( [[#Wan--2015|Wan et al., 2015]] ; R. [[#Guo--2019|]] [[#Guo--2019|Guo et al., 2019]] ), albeit with only small (less than 30%) area fraction with robust change. Robust changes in tropical and subtropical regions only appear from GWLs of 1.5°C, for example in south-western South America ( [[#Boisier--2016|Boisier et al., 2016]] ), western Africa ( [[#Hawkins--2020|Hawkins et al., 2020]] ; [[#10.4.2.1|Section 10.4.2.1]] ) and southern Australia ( [[#Delworth--2014|Delworth and Zeng, 2014]] ). Substantial (taken here simply as area fraction greater than 50%) emergence only occurs in some tropical, subtropical and mid-latitude regions when high GWLs (3°C–4°C) are reached. Importantly, even at these high GWL values, there are still a large number of these regions with robust changes covering less than 50% of their area. In contrast, most high-latitude regions have an area fraction with robust changes greater than 80% at GWLs of 3°C and above.

<div id="_idContainer046" class="Basic-Text-Frame"></div>

[[File:1c824a91cc0b7392bb857583d40348dc IPCC_AR6_WGI_Figure_10_15.png]]

'''Figure 10.15''' '''|''' '''Future emergence of anthropogenic signal at regional scale. (a)''' Percentage area of land regions with robust annual mean precipitation change as a function of increasing global warming levels (GWLs). Robustness of the precipitation change is first estimated at each grid-point followed by the estimation of the AR6 region area with robust changes. For each Coupled Model Intercomparison Project Phase 6 (CMIP6) model considered (45 models, one member per model, historical simulations and scenario SSP5-8.5), the annual mean precipitation change is based on the difference between a 20-year average centred on the GWL crossing year and the mean precipitation during the pre-industrial period (1850–1900) taken as a reference. The change is considered to be robust when at least 66% of the models (30 out of 45) have a signal-to-noise ratio greater than one and at least 80% of them (36 out of 45) agree on the sign of change. The signal-to-noise ratio is estimated for each model from the ratio between the change and the standard deviation of non-overlapping 20-year means of the corresponding pre-industrial simulation (scaled by square root of 2 times 1.645). '''(b)''' Time evolution of the percentage area of land region with robust annual mean precipitation change for five AR6 land regions. Thick solid lines represent precipitation changes based on the same CMIP6 ensemble as in (a). Thin solid, dotted and dashed lines represent changes based on the three coupled single-model initial-condition large ensembles (SMILEs) used in Chapter 10, illustrating the influence of internal variability on the emergence of robust change. The change is estimated from the difference between all consecutive 20-year periods from 1900–1919 up to 2081–2100 and the pre-industrial period. The line colour indicates the sign of the robust change given by the multi-model mean (CMIP6) or ensemble mean (SMILE) change: brown (decreasing precipitation) and dark green (increasing precipitation). Further details on data sources and processing are available in the chapter data table (Table 10.SM.11).

We now illustrate the potential influence of internal variability on late or lack of emergence for a few AR6 land regions (Figure 10.15b). For each of these AR6 regions, the time evolution of the percentage area with robust annual mean precipitation change is estimated for both the CMIP6 multi-model ensemble and the three coupled SMILEs used throughout Chapter 10. Similarity in percentage area time evolution between CMIP6 and the three coupled SMILEs suggests that internal variability can substantially influence the timing of emergence. For example, internal variability could explain the mid-21st century emergence (percentage area greater than 50%) of the drying and wetting signal over the Mediterranean and South Asia (see also [[#10.6.3|Section 10.6.3]] ) regions, respectively. Internal variability can also contribute to the late and moderate emergence over South-Eastern South America (see also [[#10.4.2|Section 10.4.2]] ) and West South Africa (see also [[#10.6.2|Section 10.6.2]] ). In contrast, it cannot explain the lack of robust changes (percentage area less than 30%) over Western Africa at the end of the 21st century, suggesting that model differences are also contributing to the lack of emergence ( [[#Monerie--2017a|Monerie et al., 2017a]] , b). In addition to different forced signals, the differences of time evolution between the three SMILEs, in particular for African regions, point to the issue of global model performance in accurately representing internal variability and its future changes. While overestimation and underestimation of internal variability in current models have been reported ( [[#Eade--2014|Eade et al., 2014]] ; [[#Laepple--2014|Laepple and Huybers, 2014]] ), methodological challenges to assess the magnitude and spatial pattern of model biases in simulating internal variability, still remain [[#10.3.4.3|Section 10.3.4.3]] ). Therefore, the existence of model biases and the limited knowledge of their characteristics lead to limitations about a precise quantification of internal variability influence on delayed regional-scale emergence.

There is ''high confidence'' that consistency in definitions of signal and noise, choice of the reference period and signal-to-noise threshold, is important to robustly assess the future emergence of anthropogenic signals across different types or generations of models, as well as comparing past emergence results between observations and models. There is ''high confidence'' that internal variability can delay the emergence of the regional-scale mean precipitation anthropogenic signal in many regions, mainly located in the tropics, subtropics and mid-latitudes. An accurate estimation of the delay in regional-scale emergence caused by internal variability remains challenging due to global model biases in their representation of internal variability as well as methodological difficulties to precisely estimate these biases ( ''high confidence'' ).

<div id="10.5" class="h1-container"></div>

<span id="combining-approaches-to-constructing-regional-climate-information"></span>