Abstract
Calibration of the cumulus parameterization scheme for localized areas is one method that can improve numerical weather prediction rainfall forecast accuracy. Calibration for model development, however, is a time-consuming procedure that requires numerous simulations. The utilization of an efficient method for calibrating complex dynamical models can help mitigate the heavy computational costs involved. In this study, five parameters in the Weather Research and Forecasting (WRF) Kain–Fritsch cumulus scheme were calibrated based on rainfall verification in the northwest and the southeast regions of the Philippines. Two optimization methods—Multiple Very Fast Simulated Annealing (MVFSA) and Adaptive Surrogate Modeling-Based Optimization (ASMO)—were used to find the best parameter set values. Both methods generated a higher coefficient of downdraft (Pd), lower entrainment (Pe), and longer convective available potential energy (CAPE) consumption time (Pc), which were found to result in better skill scores than the default WRF. Precipitation amount in both calibrated models decreased leading to an overall less wet bias. Precipitation skill score in the northwestern Philippines significantly improved by 35%, while that of the southeastern Philippines only increased by 3%. In addition, model calibration had no significant effect on the simulated temperature and wind speed. The results show that calibration of the cumulus parameterization scheme yields better results for convective rainfall rather than rain from stratiform clouds, which is expected since the cumulus parameterization scheme represents the effects of sub-grid-scale convective processes.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have