Determining the sensitive parameters of the Weather Research and Forecasting (WRF) model for the simulation of tropical cyclones in the Bay of Bengal using global sensitivity analysis and machine learning

Abstract. The present study focuses on identifying the parameters from Weather Research and Forecasting (WRF) model that strongly influence simulation of tropical cyclones over Bay Bengal (BoB) region. Three global sensitivity analysis (SA) methods, namely, Morris One-at-A-Time (MOAT), multivariate adaptive regression splines (MARS), surrogate-based Sobol', are employed to identify most sensitive out 24 tunable corresponding seven parameterization schemes WRF model. Ten across different categories, such as cyclonic storms, severe very storms BoB between 2011 2018, selected in this study. scores evaluated for eight meteorological variables. parameter results consistent three SA methods all variables, 8 contribute 80 %–90 % overall scores. It is found Sobol' method with Gaussian progress a surrogate can produce reliable when available samples exceed 200. which simulations have least RMSE values compared observations considered optimal parameters. Comparing default shows set yield 16.74 improvement 10 m wind speed, 3.13 surface air temperature, 0.73 pressure, 9.18 precipitation