Comparison of spatial and temporal performance of two Regional Climate Models in the Amazon and La Plata river basins

Abstract

This study aims to present a comprehensive spatio-temporal performance analysis of two water balance variables, precipitation and evapotranspiration, simulated by two Regional Climate Models (RCM) multi-model ensembles. We evaluate simulations from two RCMs from the Coordinated Regional Climate Downscaling Experiment in the two largest hydrological basins in South America, the Amazon River basin and La Plata River basin. Information was analyzed at 0.5°x0.5° resolution, compared with their driving Global Climate Models (GCM) from the Coupled Models Intercomparison Project (CMIP5) and reference datasets from observations and reanalysis for the period 1986–2005. Spatial and temporal model performance metrics were calculated for monthly datasets, at the annual scale. Annual cycles and the spatial distribution of mean fields are well represented by the RCMs, whereas the empirical frequency distributions reveal discrepancies in lower frequencies. In terms of spatial performance of the RCMs, the quantitative metrics showed that mean fields are better simulated by RCA4 and only RegCM4 offers added value with respect to GCM. Statistical metrics for temporal variability assessment presented a better performance over the La Plata basin from both RCMs. Over the Amazon basin, temporal variability of precipitation is better represented by RegCM4, and the opposite occurs for evapotranspiration with lower differences. According to the overall assessment of spatio-temporal results and performance metrics, the RegCM4 ensemble shows a better performance even though the number of members is smaller.