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

selection of the diagnostics presented above (Table 1). In all cases, those models showing a bad representation of the basin topography in terms of surface and/or volume are rejected and not further discussed here. Then, for each criterion (i.e. subset of diagnostics), we discard those models that show a RMSE three times larger than the RMSE of the reference simulation in any of the selected diagnostics. Additionally, for some of the criteria those models showing trends outside the range defined by the observed trend plus ± minus the uncertainty were discarded. Namely the different criteria are the following (Table 1): Criterion 1: Models that fulfill all diagnostics. Criterion 2: Models with a reasonable representation of the vertical structure and variability of the temperature field at seasonal and interannual time scales and showing a present climate trend compatible with observations are kept. I.e. in this criterion more weight is given to the vertical transfer of heat. Fig 6. Annual Mean SST anomalies (˚C) in the Red Sea (Top) and in the Gulf of Aden (bottom) for the observations (black line), KAUST (red line), GLORYS (blue line), GCMs (grey lines) and the average of the selected GCMs (see “Model Selection” section; green line). Anomalies are computed with respect to the 1985–2005 average. https://doi.org/10.1371/journal.pone.0255505.g006 Table 1. Definition of the diagnostics considered in each selection criterion: Area/Volume; Seasonal Average Vertical Profile Temperature (SAVPT); Seasonal

Cycle of the Sea Surface Temperature (SCSST); Vertical profile of interannual std (STDVPT); Interannual std of the Sea Surface Temperature (STDSST) and long term Trends in the present climate. Area/Vol SAVPT SCSST STDVPT STDSST Trends Cr 1 ✔ ✔ ✔ ✔ ✔ ✔ Cr 2 ✔ ✔ ✘ ✔ ✘ ✔ Cr 3 ✔ ✘ ✔ ✘ ✔ ✔ Cr 4 ✔ ✔ ✔ ✔ ✔ ✘ https://doi.org/10.1371/journal.pone.0255505.t001 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 11 / 26