Assessment of Red Sea temperatures in CMIP5 models for present and future climate

the performance of marine organisms and the regulation of ecosystems processes, which are usually acclimated to the local thermal range. Hence, changes in the temperature field (e.g. warming) affect marine organisms, typically by shifting their distribution poleward and advancing their phenology [5,6]. Besides the long-term changes, extreme events like marine heatwaves (MHWs), defined as prolonged periods of anomalously high temperatures, can have severe impacts on marine ecosystems. For instance, coral bleaching in tropical ecosystems [2], reduced surface chlorophyll levels due to increased surface layer stratification, mass mor- tality of marine invertebrates due to heat stress, and fishery collapse [7,8]. Semi-enclosed seas are particularly vulnerable to ocean warming, which is often amplified in these seas [9], and where the scope for organisms to adapt by migrating poleward is restricted by the presence of land masses [10]. This is the case of the Red Sea, which is considered to be almost closed in its northern boundary in terms of water exchanges, although the Suez Canal has offered a limited escape route into the Mediterranean, where many Red Sea species are becoming established [11]. Consequently, there is an urgency to develop reliable thermal projections throughout the 21st century for marine ecosystems, particularly so for semi-enclosed seas. Although being one of the global hot spots for climate-vulnerable coral reefs, the assessment of these projections is limited to few works. One of them use CMIP3 GCMs without doing any model assessment or

selection before providing future temperatures [12]. The other one projects the future temper- ature in the Red Sea using an unique GCM after a rather simple selection procedure [13]. Long-term temperature projections are usually based on global climate models (GCMs) run under different scenarios of greenhouse gas (GHG) emissions. The most used ensemble of GCMs is the one produced by the fifth phase of the Climate Model Intercomparison Project (CMIP5, [14]), which has been used for a large number of studies of climate change of both the atmosphere and ocean. However, these global models have weaknesses when applied to regional seas [15]. Their coarse spatial resolution (typically from 0.2˚ to 2˚ in the ocean) pre- cludes them from solving key processes that may have a leading role in the local response to global warming [16]. But for some processes characterized by simpler dynamics, GCMs may be enough to reproduce the local response to global warming. In the case of the Red Sea, a recent study by [17] has shown that interannual variations of the heat content in the 0–100 m layer are mostly driven by changes in the atmospheric temperature. Thus, GCMs (at least some of them), may be able to represent the heat uptake variability in the region. Therefore, whether GCMs are well suited for a particular region or process must be analysed case by case [18]. In any case it is clear that regional high-resolution simulations partially overcome the shortcomings of global models and are, therefore, expected to be more reliable for regional

seas (e.g MedCORDEX in the Mediterranean Sea; [19]). Unfortunately, regional projections for the Red Sea are not available at present, so the only source of projections for the evolution of sea temperature throughout the 21st century is the global simulations. Here we focus on conducting a skill assessment of the CMIP5 models in the characteriza- tion of the Red Sea temperature. Namely, in a first step, different comparisons with an observa- tional dataset are performed aiming at assessing the quality of the models in simulating different aspects of the ocean heat uptake. Additionally, we compare the CMIP5 simulations with two state-of-the-art simulations, a medium resolution global reanalysis [20] and a high resolution regional model [21]. That comparison provides a benchmark for the expected qual- ity of model results in the Red Sea. These analyses allow to discard models showing low skills in reproducing the ocean heat uptake in the region. The models that are kept are then used to analyse the projected 3D temperature evolution under different emissions scenarios. It is clear that GCMs are not the best option for that purpose, but until an ensemble of high resolution ocean models is available, the projections from the selected GCMs can provide a first approxi- mation to it. Lastly, we focus on marine heatwaves (MHWs). The reason is they have a strong PLOS ONE Assessment of Red Sea temperatures in CMIP5 models for present and future climate PLOS ONE | https://doi.org/10.1371/journal.pone.0255505 July 30, 2021 2 / 26 (KAUST) through funds provided to S.A. and C.M.D

(BAS/1/1072-01-01). M.A. has been partly funded by the European Union’s Horizon 2020 research and innovation programme under grant agreement No 776661 (SOCLIMPACT project), and "Proyecto Intramurales Especial REDTEM - CSIC (202030E234)" Each named author has substantially contributed to conducting the underlying research and drafting this manuscript. Additionally, to the best of our knowledge, the named authors have no conflict of interest, financial or otherwise. Competing interests: The authors have declared that no competing interests exist.