Abstract ID: 179
Analog Post-processing of Week 2-3 Probabilistic Precipitation Forecasts over Taiwan
Lead Author: Hui-Ling Chang
Central Weather Bureau, Taiwan (R.O.C.)
Keywords: probabilistic precipitation forecasts, Analog Post-processing, reliability, discrimination, potential economic value
Abstract: The predictability of precipitation is limited due to the important role finer-scale processes play. However, demand for extended-range (10-to-30 days) precipitation forecasts by users in agriculture and water resource management has grown significantly. Therefore, the goal of this study is to predict the conditional climatology of precipitation given the forecast of the large-scale circulation conditions, which still retain predictability in the extended range.
In this study, we focus on week 2-3 precipitation forecasts over Taiwan, and use Analog Post-processing (AP) to produce posterior ensembles with reasonable spread to effectively mitigate the problem of under-dispersion, which is very common for ensemble prediction systems. The AP forecast ensembles are derived from observed high-resolution precipitation patterns corresponding to historical forecast analogs that most resemble the current precipitation forecast. Frequency counting is then applied to the AP ensembles to produce well-calibrated and downscaled week 2-3 probabilistic precipitation forecasts.
Forecast evaluation confirms that the raw ensemble is under-dispersive with an obvious wet bias. In contrast, the AP ensemble distribution is well calibrated with most of the bias removed. Compared to the raw forecasts, the AP-based probabilistic forecasts have better reliability and higher skill in discrimination in the winter and Mei-yu seasons. Evaluation of potential economic value demonstrates that users with a much wider spectrum of cost/loss ratio can benefit from the calibrated forecasts in decision making as compared to the raw forecast, with a significantly higher gain.
Hui-Ling Chang (Central Weather Bureau, Taiwan)
Zoltan Toth(Global Systems Laboratory, National Oceanic and Atmospheric Administration, USA)
Shih-Chun Chou(Central Weather Bureau, Taiwan)
Chih-Yung Feng(Manysplended Infotech Ltd, Taiwan)
Yi-Shan Liao(Central Weather Bureau, Taiwan)
Ting-Hsuan Lee(Central Weather Bureau, Taiwan)
Meng-Shih Chen(Central Weather Bureau, Taiwan)
Tony Liao(Global Systems Laboratory, National Oceanic and Atmospheric Administration, USA)