High resolution dynamical downscaling of global products using spectral nudging for improved simulation of Indian monsoon rainfall

Abstract

Societal wellbeing and risk management demand the evolution of rainfall during the Indian summer monsoon (ISM) at various spatio-temporal scales. The ISM rainfall (ISMR) is simulated for 1982–2018 using the Weather Research and Forecasting (WRF) model at 15 km grid-spacing by forcing with ERA5 reanalysis data. Two experiments are conducted without spectral-nudging (NSN) and with spectral-nudging (SN) in which the wind and temperature fields are updated. The seasonal mean rainfall over India and its homogeneous regions has realistically captured in SN experiment with less root mean square error (4.5 mm day−1) and high correlation (0.62) than the NSN run (6.0 mm day−1 and 0.35), with significant improvements in the first 60-days' simulation. Very-light, light, moderate, and heavy rain categories are well detected in the SN in 5 homogeneous regions out of 6. About 57–89% of years match the observed contribution of these rainfall categories in the SN experiment's seasonal rainfall, while 21–54% in the NSN. Better simulation of moderate rain from SN run in 24 years leads to improved contribution to the seasonal rainfall. NSN runs mostly underestimate the rainrate contribution in recent years. High threat scores (0.2–0.6) for these rainfall categories indicate the consistent performance of SN run in all monsoon months than the NSN (0.1–0.45). The SN could capture the interannual variability of observed ISMR, which aids in detecting the normal, excess, and deficit years at ∼75%, ∼86%, and ∼ 70% rate with high threat scores (0.60, 0.55, and 0.58, respectively), unlike NSN run. The SN experiment is superior in capturing frequent rainfall in extreme years, improving performance at monthly and seasonal scales. While NSN mainly underestimates rainfall in all extreme years. Thus, the study highlights the importance of spectral-nudging to high-resolution downscaling for improved long-range simulations.