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

==== [[#Atlas.5.5.3|Atlas.5.5.3]] Assessment of Model Performance ====

<div id="h3-36-siblings" class="h3-siblings"></div>

There is ''limited evidence'' about the performance of GCMs and RCMs in representing the current climate of South West Asia due to very few studies evaluating models over this region, but literature is now emerging particularly on CMIP5/CMIP6 and CORDEX simulations.

Over ARP, surface temperature biases for 18 of 30 CMIP5 models are within one standard deviation of the observed variability ( [[#Almazroui--2017|Almazroui et al., 2017]] ). A warm bias in summer and a cold bias for other months along with an underestimation of wet-season precipitation and an overestimation in the dry season have been reported in 26 CMIP5 models ( [[#Lelieveld--2016|Lelieveld et al., 2016]] ). Thirty CMIP6 GCMs have limited skill in simulating annual precipitation patterns, annual cycle statistics and long-term precipitation trends over Central Asia partially due to considerable wet biases of up to 100% in the southern Xinjiang and Hexi Corridor regions ( [[#Guo--2021|Guo et al., 2021]] ). Also, CMIP6 models display a wide range of performance in reproducing ENSO teleconnections that influence the region ( [[#Barlow--2021|Barlow et al., 2021]] ).

RCM simulations using the CORDEX-MENA domain reproduce the main features of the mean surface climatology over ARP with moderate biases ( ''high confidence'' ). RegCM4 driven by five GCMs (HadGEM2, GFDL, CNRM, CanESM2 and ECHAM6) showed an ensemble-mean cold bias of about –0.7°C and a dry bias of –13% over ARP ( [[#Almazroui--2016|Almazroui, 2016]] ) with a cold (warm) bias over western (south-eastern) areas ( [[#Syed--2019|Syed et al., 2019]] ). Temperature biases in 30-year historical simulations with WRF using three different radiation parametrizations were within ±2°C and mostly caused by surface long-wave radiation errors which affected nighttime minimum temperatures over 70% of the domain ( [[#Zittis--2017|Zittis and Hadjinicolaou, 2017]] ). Mean absolute errors in COSMO-CLM driven by ERA-Interim were about 1.2°C for temperature, 15 mm per month for precipitation and 9% for total cloud cover, and with new parametrizations of albedo and aerosols optimized for the region the RCM simulated the main climate features of this very complex area ( [[#Bucchignani--2016|Bucchignani et al., 2016]] ). RegCM4.4 also simulated the main features of the observed climatology (especially for dry regions) with temperature biases within ±3.0°C. Annual precipitation was overestimated with winter and spring underestimated ( [[#Ozturk--2018|Ozturk et al., 2018]] ).

Four RCMs (REMO, RegCM4.3.5, ALARO-0, and COSMO-CLM5.0) driven by ERA-Interim, NCEP2 reanalyses and two different GCMs reproduced reasonably well the spatio-temporal patterns for temperature and precipitation though underestimated diurnal temperature range and had cold biases over mountainous and high plateau regions in all seasons. There is ''low confidence'' in this result because of low station density and a lack of high-elevation stations, and with biases dependent on the choice of the observational dataset. However, the performance of both GCMs and RCMs is better than reanalyses when compared to available observations ( [[#Mannig--2013|Mannig et al., 2013]] ; [[#Ozturk--2017|Ozturk et al., 2017]] ; [[#Russo--2019|Russo et al., 2019]] ; [[#Top--2021|Top et al., 2021]] ).

<div id="Atlas.5.5.4" class="h3-container"></div>

<span id="atlas.5.5.4-assessment-and-synthesis-of-projections"></span>