Evaluation of Precipitation Forecasts from Single models and Ensemble Prediction Systems

Behiyan Motlagh, Soudabeh; Honarbakhsh, Afshin

doi:10.22059/ije.2026.404547.1892

Evaluation of Precipitation Forecasts from Single models and Ensemble Prediction Systems

Articles in Press

Document Type : Research Article

Authors

¹ Shahrekord University

² Natural Engineering-Faculty of Natural Resources and Earth Sciences

10.22059/ije.2026.404547.1892

Abstract

Using numerical weather prediction for precipitation forecasting can yield significant results. This study aims to compare the accuracy of ensemble prediction systems with that of single models. Precipitation forecasts from the TIGGE database were utilized, specifically from the ECMWF, NCEP, and UKMO numerical prediction models. Three precipitation events between March 1, 2018, and May 2019 were analyzed.

The results show that the effect of the number of members in ensemble models with a number of members of approximately 20, such as NCEP, has little effect, but the effect of the number of members was significant in the ECMWF with 50 members. Time series evaluation results showed that the ensemble predictions from both the ECMWF and NCEP models closely matched the observed events. Additionally, increasing spatial resolution did not improve the forecast quality. The ensemble forecasts achieved 100% precipitation detection. According to the RMSE and correlation coefficient (CC) statistical indices, the ensemble models from NCEP and ECMWF outperformed the single models.

Finally, by combining single and ensemble models, a series of precipitation data was obtained that demonstrated acceptable performance based on statistical indicators and time series analysis. The precipitation data generated in this study can be applied in flood warning systems.

Keywords