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

anomalies are constant throughout the year (~-2.5˚C). The shape of the profile also showed a seasonal evolution in the upper 100 m, being more homogeneous in winter and strongly strati- fied in summer and autumn. The reference simulations GLORYS and KAUST showed small RMSE (0.30˚C and 0.34˚C, respectively) and a good representation of the seasonal changes of the profile shape (averaged correlation of 0.99 for both hindcasts). Regarding the GCMs, two models deviated greatly from the rest: “ACCES 1–0” and “ACCES 1–3” (S1 Table). For the rest of models, the range of RMSE was 0.5˚C -3.5˚C, and the correlations fluctuated between 0.65 and 0.99 (Fig 3, S1 Table). In the Outer Region the surface anomalies were larger in spring (up to +8˚C) and the small- est in winter (~+6.2˚C), when the mixed layer reached up to 80 m compared to the stronger stratification found in summer (S1 Fig). GLORYS showed good skills in reproducing the shape and the seasonal evolution of the anomaly profiles, with an averaged RMSE of 0.32˚C and a correlation of 0.99. Most GCMs models had a temperature anomaly between 1 and 5˚C lower than observations, from surface up to 100 meters of depth. Almost all the models exhib- ited between 1 and 5˚C higher temperature values than observed between 100 and 300 meters depth. Hence, most GCMs were unable to properly reproduce the heat uptake in the upper lay- ers in winter and spring. In autumn and summer, the GCMs performed better but most mod- els showed a reduced vertical gradient of temperature (S1 Fig).

Fig 2. Scatter that represents area (x-axis) against volume (y-axis) for each GCM (grey dots) and the reference simulations (GLORYS in red and KAUST in green). The observed values are indicated by the black dots and the acceptble threshold (±25% with respect to the observed value) is represented by the dashed red line. The results are shown for the Red Sea basin (left panel) and the Gulf of Aden (right panel). Note that KAUST simulation does not cover completely the Gulf of Aden and is therefore not included in the right panel. https://doi.org/10.1371/journal.pone.0255505.g002 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 7 / 26