The representation of summer monsoon rainfall over northeast India: assessing the performance of CORDEX-CORE model experiments

Abstract

In this study, the performance of the latest high-resolution CORDEX-CORE model simulations is assessed with respect to the corresponding gridded Indian Meteorological Department (IMD) and ERA5 observations in representing the monsoon rainfall over northeast India during the historical period (1979–2005). Three different RCM model simulations (COSMO, RegCM4.7, and REMO) downscaled from the global data over the South Asian CORDEX domain are used in this study. Their corresponding RCM ensembles using boundary conditions from ERA-Interim re-analysis and CMIP5 GCMs and a combined ensemble of all the RCMs were also evaluated to assess their performance. The analysis shows that the COSMO model experiments resemble closer to both the observations while comparing with the other two RCMs. With respect to IMD and ERA5 observations, the COSMO model experiments show moderate wet bias across the study area. In contrast, the RegCM model experiments show very wet bias and the REMO model experiments show more dry bias across a significant part of the study region. Quantitatively, in comparison to IMD, the COSMO suite of models shows a slight overestimation ranging between 7 and 13%, and with ERA5, an underestimation of about 14 to 18% is observed. In comparison to both observations, the REMO model experiments underestimate (15 to 50%), whereas the RegCM model experiments overestimate (15 to 80%) the monsoon rainfall. Overall, these model experiments replicate the monsoon rainfall over the study region but with biases that differ spatially.