Evaluation of NASA’s NEX-GDDP-simulated summer monsoon rainfall over homogeneous monsoon regions of India

Abstract

The current research aimed to evaluate the predictive skill of statistically downscaled National Aeronautics Space Administration (NASA) Earth Exchange Global Daily Downscaled Projection (NEX-GDDP) data in simulating the Indian summer monsoon rainfall (ISMR) for the period of 1961–2005 over the individual homogeneous monsoon regions of India (HMRI). For the purpose, five models are selected, as these models (in GCM) have shown better performance in the simulation of ISMR by the researcher. The spatial characteristics and statistical scores (annual cycle, percentage bias, Taylor score, probability distribution function) are used to evaluate the performance of each model in simulating rainfall over land points of individual HMRI. In the spatial analysis, it seems that models of NEX-GDDP can simulate the ISMR, pretty well in comparison to APHRODITE (observation), and show a moderate to significantly high correlation (grid point) over each of the HMRI particularly to core monsoon region, except over few parts of PI. The Taylor statistics suggest that the model CanESM performs very well over the regions of PI, NWI, and WCI. The models MPI-ESM-LR and NorESM perform well in simulating the ISMR over CNI, followed by ACCESS, CanESM, and CCSM4. The models have varying bias in predicting the rainfall; however, ACCESS does perform well and shows the minimum bias (ranges from ~ 1 to ~ 14% only) among others. The models CanESM and NorESM (except over CNI) performed relatively better. The NEX-GDDP models overcome the global climate models (GCMs) in the retrospective simulation of ISMR over the land points of India. It is concluded that the models have good predictability of JJAS rainfall but unable to catch daily rainfall variability.