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

Criterion 3: Models with a reasonable representation of the spatial structure of the surface variability at seasonal and interannual scales and showing a present climate trend compatible with observations are kept. I.e. in this criterion more weight is given to the spatial structure of the heat distribution. Criterion 4: Models fulfilling all the criteria except the one on trends are kept. This will be a version of Criterion 1 without the constrain in the trends. The models included following each criterion are listed in Table 2. In this work, we are mainly interested in the evolution of the heat content of the basin (i.e. the averaged 3D temperature) and the vertical structure of the temperature field. So, we have followed Criterion 2 to do the GCM selection, as a compromise between model reliability and keeping enough models to do the projections. In other words, following Criterion 2 we sacri- fice the quality in the description of the horizontal structure of the surface temperature field in order to keep enough models for robust ensemble averaging. There are 10 models fulfilling cri- terion 2, and 8 of them are available for the two RCP scenarios (Table 2, S3 Fig). As mentioned above, there is an unavoidable degree of subjectivity when selecting the criterion, so the impli- cations of this choice on the final results are addressed in the Discussion section. Temperature projections Sea surface temperature The ensemble of selected models projects Red Sea annual mean sea surface warming ranging

between 2˚C and 4˚C by the end of the century (2080–2100) under the RCP8.5 scenario (Fig 7, Tables 3 and S3). The highest warming is projected by CMCC simulations (CMCC-CM, CMCC-CMS and CMCC-CESM) and IPSL-CM5A-MR, with projected values above 3.5˚C. In contrast, GISS-E2-RCC projects the smallest warming (around +2˚C) followed by NOR- ESM1-M (+2.5–3˚C). In most cases warming is predicted to be quite homogeneous across the Red Sea basin, except for the MPI simulations (MPI-ESM-LR and specially MPI-ESM-MR), which show a clear north to south gradient, with warming in the southern Red Sea being approximately 0.6˚C greater than in the northern part. In order to obtain the absolute value of projected temperatures we added the projected anomalies to the observed mean annual temperatures averaged over the period 1985–2005. The ensemble average of SST under scenario RCP8.5 (Fig 8) shows a temperature gradient Table 2. List of GCMs that meet at least one of the criterion. The last two columns indicate whether the model has been run under scenario RCP8.5 and RCP4.5. In bold, the models that have been selected for the analysis of temperature projections (Criterion 2 and results for the two scenarios). Model Cr 1 Cr 2 Cr 3 Cr 4 RCP 8.5 RCP 4.5 GISS-E2-R-CC ✔ ✔ ✔ ✔ ✔ ✔ IPSL-CM5A-MR ✔ ✔ ✔ ✔ ✔ ✔ MIROC4H ✔ ✔ ✔ ✔ ✘ ✔ MPI-ESM-LR ✘ ✔ ✘ ✘ ✔ ✔ MPI-ESM-MR ✘ ✔ ✘ ✘ ✔ ✔ CMCC-CESM ✘ ✔ ✘ ✘ ✔ ✔ CMCC-CM ✘ ✔ ✘ ✘ ✔ ✔ CMCC-CMS ✘ ✔ ✘ ✘ ✔ ✔ MPI-ESM-P ✘ ✔ ✘ ✘ ✘ ✔ NORESM1-M ✘ ✔ ✘ ✘ ✔ ✔ CNRM-CM5-2 ✘ ✘ ✘ ✔ ✘ ✔ IPSL-CM5A-LR ✘ ✘ ✘ ✔ ✔ ✔

https://doi.org/10.1371/journal.pone.0255505.t002 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 12 / 26