PROBABILISTIC FORECAST OF RAIN EVENT IN BANDUNG USING ENSEMBLE PREDICTION SYSTEM

• Main Author: FAUZAN PRATAMA - Nim: 12813040, ANGGA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/21092
• Institution: Institut Teknologi Bandung

This research conducts an operational experiment of probabilistic forecast of rain event in Bandung using ensemble prediction system. The ensemble method used is time-lagged ensemble. The ensemble prediction system is built by using Weather and Climate Laboratory ITB operational deterministic forecast outputs. Weather and Climate Laboratory ITB runs weather forecast twice a day, with each prediction is for 48 hours, so the ensemble prediction consists of 5 ensemble members. Rain probability is calculated from the percentage of ensemble member that predicts rainfall above a threshold. For evaluation purpose, the probabilistic forecast is compared to the rain observation data in ITB, Antapani, and Majalaya in April 2017 <br /> <br /> <br /> <br /> In this research, four threshold value for rain event is experimented, which are 0.5 mm/3 hour; 1.0 mm/3 hour; 1.5 mm/3 hour; and 3.0 mm/3 hour. Experiment shows that the threshold that is able to detect rain event well is 3.0 mm/3 hour. Meanwhile, for heavy rain, three threshold value is experimented, which are 6.0 mm/3 hour, 9.0 mm/3 hour, and 12.0 mm/3 hour. The result shows that 6.0 mm/3-hour threshold is a good threshold value to predict heavy rain in Bandung. Both threshold value is used for the future probabilistic forecast operational. Operational experiment shows that probabilistic forecast system has a good discrimination ability, although the prediction result tends to be overforecast. Probabilistic forecast also has lower number of mistakes than deterministic prediction, because of the lower number of miss and false alarm.