Performance assessment of Coupled Model Intercomparison Project Phase 5 models in tropical South America using TOPSIS-based method

The use if the general circulation model (GCM) as a boundary condition for the dynamic regionalization process, without robust technical criteria, is one of the problems responsible for increasing the uncertainties of simulations and projections of climate scenarios, mainly affecting the observed trends and consequently the representation of future trends. Based on this premise, the present study evaluated the historical ability of 30 models that make up phase 5 of the Coupled Model Intercomparison Project using the Technique for Order Performance by Similarity to Ideal Solution (TOPSIS) method. The meteorological variables used in this assessment were precipitation and near-surface air temperature from 1975 to 2005 in the southern sectors of the Amazon Basin (SAMZ), Eastern Northeast Brazil (ENEB), and the intersection area between these regions, called MATOPIBA, was used to select the best GCM. The annual cycle, Taylor diagram and TOPSIS method were used as metrics of similarity between the models and the reference dataset. In general, the models simulated the temperature more accurately than the precipitation, with lower dispersions. According to the TOPSIS, over SAMZ, the HadGEM2-ES (MIROC5) model was ranked with greater (less) ability to represent precipitation. In turn, for temperature the ensmean_cmip5 (IPSL-CM5A-MR) was the best (worst). Over ENEB, the model that showed the greatest (lowest) ability to simulate precipitation was CSIRO-ACESS1.0 (HadGEM2-AO). For temperature, the NorESM-ME model (INMCM4) was ranked first (last). Over MATOPIBA, the CSIRO-ACESS1.0 (NorESM-ME) model was selected as having the best performance and the MPI-ESM-LR (MPI-ESM-LR) as the worst performance when representing precipitation (temperature). The similarities and discrepancies in the capacity of the GCM presented in the different metrics covered in this study can assist in the selection of more appropriate climate models in other regions for future studies of climate change.

Automated summary

This study evaluated the historical ability of 30 GCM models to simulate precipitation and temperature in different regions. The results showed that the models performed better in simulating temperature than precipitation, with lower dispersion. There were differences in the simulation abilities of different models in different regions.