Construction of precipitation index based on ensemble forecast and heavy precipitation forecast in the Hanjiang River Basin, China

Abstract

Uncertainty about the occurrence of extreme precipitation events has increased significantly under global climate change. The Hanjiang River Basin (HRB) is located at the junction of the north and the south of China, and extreme precipitation events occur frequently due to the impact of climate change. Extreme precipitation is the leading factor causing flood disasters in the HRB, and effective forecasting of extreme precipitation is of great significance for preventing flood disasters. In this study, we take full advantage of the uncertainty information in THORPEX Interactive Grand Global Ensemble (TIGGE) ensemble forecasts datasets and perform statistical postprocessing through the Bayesian model averaging (BMA) method. Then, the extreme precipitation index (EPI) is constructed using the cumulative distribution functions (CDFs) of historical model precipitation and the forecast precipitation. The research results show that the constructed EPI index can well reflect the occurrence intensity of precipitation in the HRB. According to the Spearman correlation coefficient, the CMA model has the best effect, followed by the control forecast (CF) model, and the JMA model has the worst effect. As indicated by the area under curve (AUC) indicator at the 25 mm precipitation threshold, the forecast effect of CMA model is still the best, followed by the CF model, and the third is the control forecast and perturbed forecast (CFPF) model. The comprehensive effect of the JMA model is significantly worse than other models. Under the precipitation threshold of 50 mm, the prediction effect of the NCEP model is the best, followed by the ECMWF model, and the third is the CF model. The least effective is also the JMA model, followed by the UKMO model. Under the precipitation threshold of 25 mm, the prediction effect of all models is better than that of the precipitation threshold of 50 mm. The prediction stability of NCEP and ECMWF models is the strongest, and the effects of CF and CFPF models integrating multiple models are in the middle. This study provides an important reference for the theory and application of extreme precipitation forecasting in the HRB. Show more