IMPACTS OF ADDITIONAL WARM-START GENERATED ENSEMBLE MEMBERS ON THE SKILL OF WRF PROBABILISTIC FORECAST OF HEAVY

• Main Author: Arum Ayuningtyas, Sekar
• Format: Final Project
• Language: Indonesia
• Subjects: Geologi, hidrologi & meteorologi
• Online Access: https://digilib.itb.ac.id/gdl/view/76092
• Institution: Institut Teknologi Bandung

This study attempts to improve an existing Weather Research and Forecasting (WRF) Ensemble Prediction System (EPS), which utilizes a Time-Lagged Ensemble (TLE) scheme to generate eight ensemble members from four daily forecast cycles. Previous studies show that at least ten ensemble members are needed for better probabilistic forecasts of 24-hour accumulated precipitation. The WRF weather modeling package facilitates modification of initial condition with Data Assimilation (WRFDA) module, which can be used to increase ensemble size of the EPS. In this case, a warm start scheme has been successfully setup and additional ensemble members are obtained by assimilating synthetic radar reflectivity from previous cycles into a WRF forecast run. Moreover, four experiments are carried out to produce 10, 20, 30 and 40 ensemble members. Impact of each ensemble members in each experiment is evaluated using Continuous Ranks Probability Score (CRPS) with samples aggregated over two areas i.e., Semarang City and Serayu Water Shed, against satellite rainfall data. It is found that TLE with 10 or more members show better probability distribution as indicated by smaller CRPS compared to the original EPS with only 8 members. However, EPS with 10 members generated from a cycle with the nearest lead time, shows the best improvement. Improvement on the skill of probabilistic forecast is also evaluated using Brier Score (BS) and Brier Skill Score (BSS), which shows better results for the forecasts of light and moderate rainfall events. However, heavy rainfall events cannot be predicted in the experiments due the lack of samples. Considering that spatially aggregated rainfall data have been used in this case, criteria of heavy rainfall events may need to be modified in future studies.