ISSN 1000-0526
CN 11-2282/P
CN 11-2282/P
Improvement of the Optimal Percentile Fusion Method Based on Ensemble Forecast of Typhoon Rainfall in Guangdong Province
- Article
- Figures
- Metrics
- Preview PDF
- Reference
- Related
- Cited by
- Materials
Abstract:
Based on ECMWF ensemble forecast, the optimal percentile fusion product of typhoon rainfall in Guangdong Province is developed by using customized fusion parameters. Verification indicates that threat score (TS) of severe precipitation is significantly improved than the ensemble average product. The longer the forecast lead time, the greater the increment, but Bias and false alarm ratio (FAR) also become greater accordingly. The over-estimation of severe precipitation forecast is related to the divergence of typhoon path forecast. Since the fusion product uses high percentile field for mapping in large rainfall grades, the fusion result of severe precipitation is close to the union of severe rainfall locations of each member. When the typhoon paths are more dispersed, the spatial location of the severe precipitation area of each member is generally more divergent, resulting in the large range error of the severe rainfall area predicted by the fusion product. In order to solve this problem, a forecast index describing the probability of forecasting a certain precipitation threshold is introduced, which can effectively identify the heavy rainfall forecast samples with large false alarms but few hit alarms. Using the index, we are likely able to improve the optimal percentile fusion product. Under the condition of maintaining TS, the improved fusion products have Bias decreasing from 1.27 to 1.03 and FAR from 0.51 to 0.43 during the test period. Meanwhile, the longer the forecast time, the greater the improvement effect of Bias and TS of the fusion products. Therefore, the revised products can provide grid quantitative precipitation forecast with more appropriate severe precipitation scope and more accurate rainfall location in forecasting operation.
Keywords:
Project Supported:
Clc Number:
Mobile website
