This study evaluates the performances of all forty different global climate models (GCMs) that participate in the Coupled Model Intercomparison Project Phase 5 (CMIP5) for simulating climatological temperature and precipitation for Southeast Asia. Historical simulations of climatological temperature and precipitation of the 40 GCMs for the 40-year period of 1960–1999 for both land and sea and those for the century of 1901–1999 for land are evaluated using observation and reanalysis datasets. Nineteen different performance metrics are employed. The results show that the performances of different GCMs vary greatly. CNRM-CM5-2 performs best among the 40 GCMs, where its total error is 3.25 times less than that of GCM performing worst. The performance of CNRM-CM5-2 is compared with those of the ensemble average of all 40 GCMs (40-GCM-Ensemble) and the ensemble average of the 6 best GCMs (6-GCM-Ensemble) for four categories, i.e., temperature only, precipitation only, land only, and sea only. While 40-GCM-Ensemble performs best for temperature, 6-GCM-Ensemble performs best for precipitation. 6-GCM-Ensemble performs best for temperature and precipitation simulations over sea, whereas CNRM-CM5-2 performs best over land. Overall results show that 6-GCM-Ensemble performs best and is followed by CNRM-CM5-2 and 40-GCM-Ensemble, respectively. The total errors of 6-GCM-Ensemble, CNRM-CM5-2, and 40-GCM-Ensemble are 11.84, 13.69, and 14.09, respectively. 6-GCM-Ensemble and CNRM-CM5-2 agree well with observations and can provide useful climate simulations for Southeast Asia. This suggests the use of 6-GCM-Ensemble and CNRM-CM5-2 for climate studies and projections for Southeast Asia.

No abstract

No abstract

This study evaluates the performance of 13 global climate models (GCMs) from the Coupled Model Intercomparison Project Phase 6 (CMIP6) for simulating the temperature over Thailand during 2000-2014, for land-only, sea-only, and both land and sea. Both observation and reanalysis datasets are employed to compare with the GCMs, evaluated by five performance metrics including mean annual temperature, mean bias errors, mean seasonal cycle amplitude, correlation coefficient, and root mean square error. GCMs are ranked by relative error of all performance metrics. Results show that the temperatures from most GCM simulations are below the mean reference data (i.e., average of ground-based and reanalysis datasets), with north to south gradient in the range from 19 C to 33 C. In addition, all the GCM biases range from -0.07 C to 2.78 C and show severity of the temperature changes in spatial pattern ranging from -5 C to 15 C. The correlations of most GCMs range from 0.70 to 0.95, while the magnitudes of error are less than 2 C. Study cases point out that the 13-MODEL ENSEMBLE, CESM2, and CNRM-CM6-1 perform better than the other models in simulating the temperature over Thailand for land-only and sea-only, and both land and sea cases, respectively, while MIROC6 performs the worst for all study cases in this study area. From the designed methodology, CNRM-CM6-1 has the best performance and is the most appropriate choice to simulate the temperature for the overall Thailand area.

No abstract