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

=== 10.3.1 Model Types ===

<div id="h2-15-siblings" class="h2-siblings"></div>

Regional climate change information may be derived from a hierarchy of different model types covering a wide range of spatial scales and processes (Figure 10.5). The application of any model relies on assumptions, depending on the specific model as well as the application. Table 10.1 gives an overview of the generic assumptions of the different model types discussed here for generating regional climate information. The violation of these assumptions will affect the model performance, which is discussed in [[#10.3.3|Section 10.3.3]] .

<div id="_idContainer024" class="Basic-Text-Frame"></div>

'''Table''' '''10.1 |''' '''Assumptions underlying different model types in simulating regional climate and climate change. Violating these assumptions will affect model performance (see links to different subsections for details).''' All assumptions regarding future climate are in addition to those regarding present climate and predicated on the driving global model simulating a plausible global climate sensitivity ( [[IPCC:Wg1:Chapter:Chapter-1#1.3.5|Section 1.3.5]] , Chapters 4 and 7). The assumptions listed for future climate applications of perfect prognosis statistical downscaling and bias adjustment are often called the ‘stationarity assumption’. Numbers in curly brackets refer to chapters and sections assessing these assumptions.

{| class="wikitable"
|-
| Model Type

| Scale at Which the Assumption Applies

| Assumptions to Realistically Simulate Present Regional Climate

| Additional Assumptions to Be Fit for Simulating Future Regional Climate

|-
| rowspan="2"| Global model i.e., atmosphere-only general circulation model, global climate model, Earth system model (AGCM, GCM or ESM; not bias adjusted) ( [[#10.3.1.1|Section 10.3.1.1]] )

| Large (&gt;1000 km)

| Global model includes all relevant large-scale forcings and realistically simulates relevant large-scale circulation (Sections 3.3.3, 8.5.1 and 10.3.3.3).

| Global model realistically simulates processes controlling large-scale changes. Parametrizations are valid in future climate (Chapter 3, and Sections 4.2, 4.5, 8.5.1 and 10.3.3.9).

|-
| Regional (&lt;1000 km)

| Global model includes all relevant regional forcings and realistically simulates all relevant regional-scale processes and feedbacks and their dependence on large-scale climate (Sections 8.5.1, 10.3.3.4–10.3.3.6 and 10.3.3.8).

| Global model realistically simulates processes controlling regional changes. Parametrizations are valid in future climate (Sections 8.5.1 and 10.3.3.9).

|-
| rowspan="2"| Dynamical downscaling of global model with regional climate model (RCM; not bias adjusted) ( [[#10.3.1.2|Section 10.3.1.2]] )

| Large

| Driving global model includes all relevant large-scale forcings and realistically simulates relevant large-scale circulation, RCM does not deteriorate global simulations. Feedbacks from regional into large-scale processes are negligible (Sections 3.3.3, 8.5.1 and 10.3.3.3).

| Driving global model realistically simulates processes controlling large-scale changes, RCM does not deteriorate global model changes. Parametrizations are valid in future climate ( [[IPCC:Wg1:Chapter:Chapter-3|Chapter 3]] and Sections 4.2, 4.5. 8.5.1 and 10.3.3.9).

|-
| Regional

| RCM includes all relevant regional forcings and realistically simulates all relevant regional-scale processes and feedbacks and their dependence on large-scale climate (Sections 10.3.3.4–10.3.3.6 and 10.3.3.8).

| RCM realistically simulates processes controlling regional changes. Parametrizations are valid in future climate ( [[#10.3.3.9|Section 10.3.3.9]] ).

|-
| rowspan="2"| Perfect prognosis statistical downscaling of GCM ( [[#10.3.1.3|Section 10.3.1.3]] )

| Large

| Global model realistically simulates all relevant large-scale predictors. The predictors are bias free and represent the regional variability at all desired time scales (Sections 3.3.3, 8.5.1 and 10.3.3.3).

| Global model realistically simulates processes controlling changes in the predictors. The predictors represent the response to external forcing ( [[IPCC:Wg1:Chapter:Chapter-3|Chapter 3]] and Sections 4.2, 4.5. 8.5.1 and 10.3.3.9).

|-
| Regional

| The statistical model structure is adequate to represent the predictor influence on regional-scale variability. There is no relevant feedback involving the predictands ( [[#10.3.3.7|Section 10.3.3.7]] ).

| The statistical model structure is adequate under the required extrapolation ( [[#10.3.3.9|Section 10.3.3.9]] ).

|-
| rowspan="2"| Bias adjustment of dynamical model (GCM or RCM) ( [[#10.3.1.3|Section 10.3.1.3]] )

| Large

| As per driving model.

| As per driving model.

|-
| Regional

| As per driving model, apart from adjustable biases. The gap between driving model resolution and target resolution is minor (Sections 10.3.3.4–10.3.3.6 and 10.3.3.8, and Cross-Chapter Box 10.2).

| As per driving model, apart from adjustable biases. The chosen bias adjustment is applicable in a future climate ( [[#10.3.3.9|Section 10.3.3.9]] and Cross-Chapter Box 10.2).

|-
| rowspan="2"| Delta change approach applied to dynamical model ( [[#10.3.1.3|Section 10.3.1.3]] )

| Large

| Not applicable

| As per driving model. There are no changes altering the non-changed statistics (e.g., no circulation changes that alter temporal structure) ( [[IPCC:Wg1:Chapter:Chapter-3|Chapter 3]] and Sections 4.2, 4.5, 8.5.1 and 10.3.3.9).

|-
| Regional

| Not applicable

| As per driving model. There are no changes altering the non-changed statistics. The gap between driving model resolution and target resolution is minor ( [[#10.3.3.9|Section 10.3.3.9]] ).

|-
| rowspan="2"| Change factor weather generator applied to dynamical model ( [[#10.3.1.3|Section 10.3.1.3]] )

| Large

| Not applicable

| As per driving model.

|-
| Regional

| The weather generator structure is adequate ( [[#10.3.3.7|Section 10.3.3.7]] ).

| As per driving model. The weather generator structure is adequate in a future climate. Change factors are adequately incorporated for all changing weather aspects. The gap between driving model resolution and target resolution is minor ( [[#10.3.3.9|Section 10.3.3.9]] ).

|}

<div id="_idContainer023" class="Basic-Text-Frame"></div>

[[File:cf76c3bd9073f355518d95d0c24d2e5b IPCC_AR6_WGI_Figure_10_5.png]]

'''Figure 10.5''' '''|''' '''Typical model types and chains used in modelling regional climate.''' The dashed lines indicate model chains that might prove useful but have not or only rarely been used. Hybrid approaches combining the model types shown have been developed.

<div id="10.3.1.1" class="h3-container"></div>

<span id="global-models-including-high-resolution-and-variable-resolution-models"></span>
==== 10.3.1.1 Global Models, Including High-resolution and Variable Resolution Models ====

<div id="h3-16-siblings" class="h3-siblings"></div>

Model-based regional climate projections are all based upon some type of global model, including state-of-the-art Earth system models (ESMs), coupled atmosphere–ocean general circulation models (GCMs) or atmosphere-only general circulation models (AGCMs) (see [[IPCC:Wg1:Chapter:Chapter-1#1.5.3.1|Section 1.5.3.1]] ). They are collectively referred to as global models.

State-of-the-art global models are generally used to derive climate information at continental to global scales both for past and future climates (e.g., Chapters 3 and 4). The nominal horizontal resolution in CMIP5 global models is typically 100–200 km. The effective resolution, for which the shape of the kinetic energy spectrum is simulated correctly, is about three to five times larger ( [[#Klaver--2020|Klaver et al., 2020]] ), and a similar relationship also applies to RCMs ( [[#Skamarock--2004|Skamarock, 2004]] ). This strongly limits their ability to resolve local details. Since AR5 the progress in reducing biases and providing more credible regional projections by global models has been moderate in spite of the more realistic representation of a number of processes and the increase in resolution of some models. For AR6, several of the new CMIP6 ( [[#Eyring--2016a|Eyring et al., 2016a]] ) model intercomparison projects (MIPs) address some of these limitations. The list of MIPs is provided in [[IPCC:Wg1:Chapter:Chapter-1|Chapter 1]] (Table 1.3). High-Resolution MIP (HighResMIP; [[#Haarsma--2016|Haarsma et al., 2016]] ) and Global Monsoons MIP (GMMIP; [[#Zhou--2016|Zhou et al., 2016]] ) specifically address the regional climate challenge using global models. HighResMIP focuses on producing global climate projections at a horizontal resolution of around 50 km grid spacing or finer while GMMIP aims at better understanding and predicting the monsoons.

An alternative to increasing resolution everywhere is offered by variable resolution global models, that is, with regionally finer resolution. They have been developed since the 1970s ( [[#Li--1999|Li, 1999]] ), resulting in a first coordinated effort (SGMIP) by Fox-Rabinovitz et al. (2006, 2008). They are expected to offer the finest resolution possible in the region of interest, while still resolving the climate processes at the global scale (although at lower resolution). An overview of recent developments is in [[#McGregor--2015|McGregor (2015)]] . This is a rapidly developing field ( [[#Krinner--2014|Krinner et al., 2014]] ; [[#Ferguson--2016|Ferguson et al., 2016]] ; [[#Huang--2016|Huang et al., 2016]] ) that will possibly contribute to improved future regional projections.

<div id="10.3.1.2" class="h3-container"></div>

<span id="regional-climate-models"></span>
==== 10.3.1.2 Regional Climate Models ====

<div id="h3-17-siblings" class="h3-siblings"></div>

Regional climate models (RCMs) are dynamical models similar to global models that are applied over a limited area, but with a horizontal resolution higher than that of standard global models. They are the basis for dynamical downscaling to produce sub-continental climate information (e.g., Chapters 11, 12 and Atlas) but are also often used for process understanding. At lateral and, if applicable, lower boundaries, RCMs take their values from a driving dataset, which could be a global model or a reanalysis. RCMs are typically one-way nested: they do not feed back into the driving model, although two-way nested global model-RCM simulations have been performed that examine regional influence on large-scale climate, potentially improving it ( [[#Lorenz--2005|Lorenz and Jacob, 2005]] ; [[#Harris--2013|Harris and Lin, 2013]] ; [[#Junquas--2016|Junquas et al., 2016]] ). Spectral nudging ( [[#Kida--1991|Kida et al., 1991]] ; [[#Waldron--1996|Waldron et al., 1996]] ; [[#von%20Storch--2000|von Storch et al., 2000]] ; [[#Kanamaru--2007|Kanamaru and Kanamitsu, 2007]] ) can increase consistency with the driving model, whereby selected variables, such as the wind field, are forced to closely follow a prescribed large-scale field over a specified range of spatial scales. RCMs can inherit biases from the driving global model in addition to producing biases themselves ( [[#Hall--2014|Hall, 2014]] ; [[#Hong--2014|Hong and Kanamitsu, 2014]] ; [[#Dosio--2015|Dosio et al., 2015]] ; [[#Takayabu--2016|Takayabu et al., 2016]] ). The consistency between the circulation features simulated by the RCM and those inherited through the boundary conditions depends on (i) the relative importance of the large-scale forcing compared to local-scale phenomena, and (ii) the size of the RCM domain (e.g., [[#Diaconescu--2013|Diaconescu and Laprise, 2013]] ). Large domains also allow the RCM to generate much of its own internally generated unforced variability ( [[#Nikiema--2017|Nikiema et al., 2017]] , and references therein; [[#Sanchez-Gomez--2018|Sanchez-Gomez and Somot, 2018]] ).

The Coordinated Regional Climate Downscaling Experiment (CORDEX) initiative ( [[#Giorgi--2009|Giorgi et al., 2009]] ; [[#Giorgi--2015|Giorgi and Gutowski, 2015]] ; [[#Gutowski%20Jr.--2016|Gutowski Jr. et al., 2016]] ) provides ensembles of high-resolution historical (starting as early as 1950) and future climate projections for various regions. RCMs in CORDEX typically have a horizontal resolution between 10 and 50 km. But much finer spatial resolution is required to fully resolve deep convection, an important cause of precipitation in much of the world. Therefore, an emerging strand in dynamical downscaling employs simulations at convection permitting scales, at horizontal resolutions of a few kilometres, where deep-convection parametrizations can be switched off, approximately simulating deep convection ( [[#Prein--2015|Prein et al., 2015]] ; [[#Stratton--2018|Stratton et al., 2018]] ; [[#Coppola--2020|Coppola et al., 2020]] ). A recent study indicates that switching off the deep-convection parametrization may be beneficial also in simulations performed at coarser resolutions ( [[#Vergara-Temprado--2020|Vergara-Temprado et al., 2020]] ). Alternatively, some RCMs make use of scale-aware parametrizations that are able to adapt to increasing resolution without switching off the convection scheme ( [[#Hamdi--2012|Hamdi et al., 2012]] ; [[#De%20Troch--2013|De Troch et al., 2013]] ; Plant and Yano, 2015; [[#Giot--2016|Giot et al., 2016]] ; [[#Termonia--2018|Termonia et al., 2018]] ; [[#Yano--2018|Yano et al., 2018]] ).

RCMs have often consisted of atmospheric and land components that do not include all possible Earth system processes and therefore neglect important processes such as air-sea coupling (in standard RCMs sea surface temperatures, SSTs, are prescribed from global model simulations or reanalyses) or the chemistry of aerosol–cloud interaction (aerosols prescribed with a climatology), which may influence regional climate projections. Therefore, some RCMs have been extended by coupling to additional components like interactive oceans, sometimes with sea ice ( [[#Kjellström--2005|Kjellström et al., 2005]] ; [[#Somot--2008|Somot et al., 2008]] ; [[#Van%20Pham--2014|Van Pham et al., 2014]] ; [[#Sein--2015|Sein et al., 2015]] ; [[#Ruti--2016|Ruti et al., 2016]] ; [[#Zou--2016a|Zou and Zhou, 2016a]] ; [[#Zou--2017|Zou et al., 2017]] ; [[#Samanta--2018|Samanta et al., 2018]] ), rivers ( [[#Sevault--2014|Sevault et al., 2014]] ; [[#Lee--2015|Lee et al., 2015]] ; [[#Di%20Sante--2019|Di Sante et al., 2019]] ), glaciers ( [[#Kotlarski--2010|Kotlarski et al., 2010]] ), and aerosols ( [[#Zakey--2006|Zakey et al., 2006]] ; [[#Zubler--2011|Zubler et al., 2011]] ; [[#Nabat--2015|Nabat et al., 2015]] ). The coupling of these components allows for the investigation of additional climate processes such as regional sea level change ( [[#Adloff--2018|Adloff et al., 2018]] ), ocean–land interactions ( [[#Lima--2019|Lima et al., 2019]] ; [[#Soares--2019a|Soares et al., 2019a]] ), or the impact of high-frequency ocean–atmosphere coupling on the climatology of Mediterranean cyclones ( [[#Flaounas--2018|Flaounas et al., 2018]] ).

<div id="10.3.1.3" class="h3-container"></div>

<span id="statistical-approaches-to-generate-regional-climate-projections"></span>
==== 10.3.1.3 Statistical Approaches to Generate Regional Climate Projections ====

<div id="h3-18-siblings" class="h3-siblings"></div>

An alternative or addition to dynamical downscaling is the use of statistical approaches to generate regional projections. In AR5 these methods were collectively referred to as statistical downscaling, but their performance assessment has received little attention. A major conclusion was that a wide range of different methods exist and a general assessment of their performance is difficult ( [[#Flato--2014|Flato et al., 2014]] ). Since AR5, several initiatives have been launched to improve the understanding of statistical approaches such as VALUE (Validating and Integrating Downscaling Methods for Climate Change Research, now merged into the EURO-CORDEX activities; [[#Maraun--2015|Maraun et al., 2015]] ), STaRMIP (Statistical Regionalization Models Intercomparisons and Hydrological Impacts Project; [[#Vaittinada%20Ayar--2016|Vaittinada Ayar et al., 2016]] ) and BADJAM (Bias ADJustment of climate scenarios for Agricultural Model applications; [[#Galmarini--2019|Galmarini et al., 2019]] ). The performance of different implementations of these approaches will be assessed in [[#10.3.3.7|Section 10.3.3.7]] .

<div id="10.3.1.3.1" class="h4-container"></div>

<span id="perfect-prognosis"></span>
===== 10.3.1.3.1 Perfect prognosis =====

<div id="h4-1-siblings" class="h4-siblings"></div>

Perfect-prognosis models are statistical models calibrated between observation-based large-scale predictors (e.g., from reanalysis) and observed local-scale predictands ( [[#Maraun--2018b|Maraun and Widmann, 2018b]] ). Regional climate projections are then generated by replacing the quasi-observed predictors by those from climate model (typically global model) projections. Predictor patterns that are common to observations and climate model data can be defined by common empirical orthogonal functions ( [[#Benestad--2011|Benestad, 2011]] ). The perfect prognosis approach can either be used to generate daily (or even sub-daily) time series, or local weather statistics (e.g., [[#Benestad--2018|Benestad et al., 2018]] ).

Regression-like models ( [[#Maraun--2018b|Maraun and Widmann, 2018b]] ) rely on a transfer function linking an observed local statistic (such as the temperature at a given day) to some set of large-scale predictors. Recent developments include stochastic regression models to explicitly simulate local variability ( [[#San-Martín--2017|San-Martín et al., 2017]] ; those explicitly modelling temporal dependence are assessed in [[#10.3.1.3.4|Section 10.3.1.3.4]] ). The use of machine learning techniques has been reinvigorated, including genetic programming to construct a data-driven model structure ( [[#Zerenner--2016|Zerenner et al., 2016]] ) and deep and convolutional neural networks ( [[#Reichstein--2019|Reichstein et al., 2019]] ; [[#Baño-Medina--2020|Baño-Medina et al., 2020]] ).

Analogue methods ( [[#Martin--1996|Martin et al., 1996]] ; [[#Maraun--2018b|Maraun and Widmann, 2018b]] ) compare a simulated large-scale atmospheric field with an archive of observations and select, using some distance metric, the closest observed field in the archive. The downscaled atmospheric field is then chosen as the local atmospheric field observed on the instant the analogue occurred. New analogue methods have been developed to simulate unobserved values including a rescaling of the analogue ( [[#Pierce--2014|Pierce et al., 2014]] ) or by combining analogues and regression models ( [[#Chardon--2018|Chardon et al., 2018]] ).

<div id="10.3.1.3.2" class="h4-container"></div>

<span id="bias-adjustment"></span>
===== 10.3.1.3.2 Bias adjustment =====

<div id="h4-2-siblings" class="h4-siblings"></div>

Bias adjustment is a statistical post-processing technique used to pragmatically reduce the mismatch between the statistics of climate model output and observations. The approach estimates the bias or relative error between a chosen simulated statistical property (such as the long-term mean or specific quantiles of the climatological distribution) and that observed over a calibration period; the simulated statistic is then adjusted taking into account the simulated deviation. Bias adjustment methods are regularly applied on a spatial scale similar to that of the simulation being adjusted, but they are often used as a simple statistical downscaling method by calibrating them between coarse resolution (e.g., global) model output and finer observations ( [[#Maraun--2018b|Maraun and Widmann, 2018b]] ).

Typical implementations of bias adjustment are (i) additive adjustments, where the model data is adjusted by adding a constant, (ii) rescaling, where the model data is adjusted by a factor, and (iii) more flexible quantile mapping approaches that adjust different ranges of a distribution individually. Hempel et al. (2013), [[#Pierce--2015|Pierce et al. (2015)]] , [[#Switanek--2017|Switanek et al. (2017)]] , and [[#Lange--2019|Lange (2019)]] developed variants of quantile mapping that preserve trends in the mean or even further distributional statistics. Multivariate bias adjustment extends univariate methods, which adjust statistics of individual variables separately, to joint adjustment of multiple variables simultaneously. Implementations remove biases in (i) specific measures of multivariate dependence, like correlation structure, via linear transformations ( [[#Bárdossy--2012|Bárdossy and Pegram, 2012]] ; [[#Cannon--2016|Cannon, 2016]] ), or, more flexibly, (ii) the full multivariate distribution via non-linear transformations ( [[#Vrac--2015|Vrac and Friederichs, 2015]] ; [[#Dekens--2017|Dekens et al., 2017]] ; [[#Cannon--2018|Cannon, 2018]] ; [[#Vrac--2018|Vrac, 2018]] ; [[#Robin--2019|Robin et al., 2019]] ). Other research strands focus on the explicit separation of bias adjustment and downscaling ( [[#10.3.1.3.5|Section 10.3.1.3.5]] ), or the integration of process understanding ( [[#Maraun--2017|Maraun et al., 2017]] ), such as by conditioning the adjustment on the occurrence of relevant phenomena ( [[#Addor--2016|Addor et al., 2016]] ; [[#Verfaillie--2017|Verfaillie et al., 2017]] ; [[#Manzanas--2019|Manzanas and Gutiérrez, 2019]] ). Some authors suggest to mitigate the influence of large-scale temperature or circulation biases by performing a bias adjustment of the driving fields prior to dynamical downscaling ( [[#Colette--2012|Colette et al., 2012]] ; [[#Hernández-Díaz--2013|Hernández-Díaz et al., 2013]] , 2019). Issues that may arise when using bias adjustment are discussed in Cross-Chapter Box 10.2.

<div id="10.3.1.3.3" class="h4-container"></div>

<span id="delta-change-approaches"></span>
===== 10.3.1.3.3 Delta-change approaches =====

<div id="h4-3-siblings" class="h4-siblings"></div>

In the delta change approach, selected observations are modified according to corresponding changes derived from dynamical model simulations. Traditionally, only long-term means have been adjusted, but recently approaches to modify temporal dependence ( [[#Webber--2018|Webber et al., 2018]] ) have been developed, as well as quantile mapping approaches that individually adjust quantiles of the observed distribution ( [[#Willems--2011|Willems and Vrac, 2011]] ). By construction, the approach cannot modify the spatial and temporal dependence structure of the input observations ( [[#Maraun--2016|Maraun, 2016]] ).

<div id="10.3.1.3.4" class="h4-container"></div>

<span id="weather-generators"></span>
===== 10.3.1.3.4 Weather generators =====

<div id="h4-4-siblings" class="h4-siblings"></div>

Weather generators are statistical models that simulate weather time series of arbitrary length. They are calibrated to represent observed weather statistics, in particular daily or even sub-daily variability. One variant of these models are advanced stochastic perfect-prognosis methods, conditioned on large-scale atmospheric predictors on a daily basis, for instance multisite generalized linear models ( [[#Chandler--2020|Chandler, 2020]] ). Another widely used variant is change-factor weather generators: the weather generator parameters are calibrated against present and future climate model simulations, and the climate change signals are then applied to the parameters calibrated to observations. Recent research has mainly focussed on multi-site Richardson type (Markov-chain) weather generators ( [[#Keller--2015|Keller et al., 2015]] ; [[#Dubrovsky--2019|Dubrovsky et al., 2019]] ), some explicitly modelling extremes and their spatial dependence ( [[#Evin--2018|Evin et al., 2018]] ).

<div id="10.3.1.3.5" class="h4-container"></div>

<span id="hybrid-approaches-and-emulators"></span>
===== 10.3.1.3.5 Hybrid approaches and emulators =====

<div id="h4-5-siblings" class="h4-siblings"></div>

A wide variety of approaches has been proposed to combine the advantages of different statistical approaches. For instance, to overcome the scale mismatch between climate model output and observations, bias adjustment has been combined with stochastic downscaling ( [[#Volosciuk--2017|Volosciuk et al., 2017]] ; [[#Lange--2019|Lange, 2019]] ) or rescaled analogues ( [[#Pierce--2014|Pierce et al., 2014]] ). Other approaches known as emulators have been developed to emulate an RCM using a statistical model and also applied to a range of driving global models ( [[#Déqué--2012|Déqué et al., 2012]] ; [[#Haas--2012|Haas and Pinto, 2012]] ; [[#Walton--2015|Walton et al., 2015]] , 2017; [[#Beusch--2020|Beusch et al., 2020]] ; [[#Erlandsen--2020|Erlandsen et al., 2020]] ).

<div id="10.3.2" class="h2-container"></div>

<span id="types-of-model-experiments"></span>