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

SST seasonality The seasonal anomaly patterns are shown in Fig 4. In winter, the largest negative anomalies (up to -4˚C) in the Red Sea basin, relative to the basin-average annual mean, were found in the northern Red Sea, while the smallest anomalies (around -1˚C) were located around 18˚N (Fig 4). In summer the seasonal pattern is rather similar, with the lowest anomalies in the northern Red Sea (approximately 0˚C) and the highest in the southern part (around +4˚C). The refer- ence runs show a very similar behaviour, with an averaged RMSE of 0.23˚C and 0.29˚C for GLORYS and KAUST, respectively and an averaged spatial correlation of 0.98 and 0.99 respec- tively. The GCMs show RMSE values ranging from 0.5˚C to 2.8˚C (see S1 Table), and spatial correlations ranging from -0.60 to 0.95. No significant differences in the model skills have been found for different seasons. For illustrative purposes we focus on the results of two models (Fig 4). The MPI-ESM-LR model showed good skills in the Red Sea (RMSE 0.70˚C and correlation 0.88): the seasonal evolution is very well captured, the local winter maximum is in the right place around 18˚N, and the range of values also match well the observations in both seasons. On the other hand, GFDL-ESM2G showed one of the poorest skills across all models tested for this diagnostic in the Red Sea (RMSE 1.10˚C and correlation 0.69). This is mainly because of the spatial structure of the anomalies which show the local maximum was placed northwards. Also, the winter values were about 1˚C cooler than observed, while the summer values were about

1˚C warmer. However, the seasonal transition was well captured even in this case (see S1 Table). Fig 3. Seasonally averaged vertical profile of temperature anomalies in the Red Sea for the observations (black line), CMIP5 models (grey lines), GLORYS (blue line), KAUST (red line), and the average of the selected GCMs (see “Model Selection” section, green line). https://doi.org/10.1371/journal.pone.0255505.g003 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 8 / 26