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

monthly 3D temperature fields for the period 1958–2017. More details on the TEMPERSEA product can be found in https://doi.pangaea.de/10.1594/PANGAEA.909472. Satellite data. The monthly time resolution of TEMPERSEA makes it unsuitable for the characterization of extreme events such as heatwaves. For that purpose, we used daily satellite SST observations which have already been used in the Red Sea [28]. The remote dataset was obtained from the National Ocean and Atmosphere Agency (NOAA), based on AVHRR (Advanced Very High-Resolution Radiometer) imagery over the period 1981–2017. These data have a spatial resolution of 0.25˚x0.25˚ and can be obtained at daily temporal resolution from the National Center for Environmental Information (NCDC-NOAA, https://www.ncdc. noaa.gov/isd/data-access). These data are used to characterize the high-frequency temperature variability in the Red Sea (section 4.3). It must be noted that the agreement between the first layer of the TEMPERSEA product and the satellite SST is very high at monthly time scales [17]. Thus, for clarity, the comparison of GCMs with monthly satellite SST fields has not been included as the results are very similar to those obtained with TEMPERSEA. Atmospheric reanalysis. JRA55 atmospheric reanalysis is used to assess the role of the atmosphere in the sea temperature variations. It has a global coverage with a spatial resolution of 1.25˚ and covers the period 1958–2014, coinciding with the establishment of the global radiosonde observing system [29]. The 3D air temperature fields have been freely downloaded

from https://rda.ucar.edu/datasets/ds628.1/?hash=access#!access. Numerical simulations The CMIP5 ensemble of global climate models. The ensemble of GCMs from the fifth phase of the Coupled Model Intercomparison Project (CMIP5) was used in this work. The goal of CMIP5 is to produce a set of coordinated climate model experiments that focus on major gaps in understanding past and future changes in climate [14]. The GCMs in the CMIP5 ensemble have a different degree of complexity but all of them include at least an ocean and an atmosphere component interactively coupled. The spatial resolution of CMIP5 models ranges from 0.2˚ to 2˚ in the ocean. From the 49 models initially considered, 6 of them were automatically discarded as their land-sea mask did not include the Red Sea (see S1 Table). Three different sets of simulations were analyzed. First, the historical simulations that cover the industrial period (mid-nineteenth century to 2005) only forced by observed GHG and aerosol concentrations. As no data assimilation is involved, one cannot expect that the chro- nology of events matches the observed ones, so we analysed those simulations in terms of sta- tistical moments (i.e., mean and standard deviation). The other two sets of simulations correspond to future projections in which the models are forced with specified atmospheric greenhouse gas concentrations, known as RCPs (Representative Concentration Pathways [14] throughout the 21st century (2005–2100). In particular, we analysed a high emissions scenario

(RCP8.5), also known as business-as-usual scenario, and a midrange emissions scenario (RCP4.5), which represents a scenario with moderate mitigation policies [30]. The data were obtained from (https://pcmdi.llnl.gov/mips/cmip5/). In particular, we analyzed ensembles of 43 historical simulations, 26 RCP8.5 simulations and 39 RCP4.5 simulations (see S1 Table). Reference state-of-the-art simulations. In order to have a benchmark of the state-of-the- art quality in Red Sea modelling, two high quality simulations of present climate are used. One is a regional high resolution hindcast (KAUST) and the other is a global reanalysis with medium resolution (GLORYS). The King Abdullah University of Science and Technology (KAUST) Red Sea hindcast is based on the MIT ocean general circulation model [31] tuned to the regional processes of the Red Sea as in [21]. The model domain covers the entire Red Sea 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 4 / 26