Climate Variability and Change—Hydrological Impacts (Proceedings of the Fifth FRIEND World Conference held at Havana, Cuba, November 2006), IAHS Publ. 308, 2006, 275–280.

---

Application of a statistical method for medium-term rainfall prediction

ZHONGMIN LIANG¹, BO LU² & XIAOFAN ZENG³

1 State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Hohai University, Nanjing 210098, P.R. China

zmliang@hhu.edu.cn

2 Department of Civil & Architectural Engineering, Drexel University, Pennsylvania 19104, USA

3 Nanjing Institute of Geography and Limnology, Nanjing 210098, P.R. China

Abstract This paper presents a statistical approach for prediction of medium-term rainfall class based on the correlation between rainfall and meteorological indicators, including geopotential height, temperature, dew-point deficit, wind direction and wind velocity. Total rainfall of a 10-day period during the rain season (July and August) is classified into two types (dry or wet) using the K-mean method. Meteorological indicators that influence rainfall are selected by the F-test method for rainfall type and rainfall class, and then used with a bi- or multi-discriminant method to establish prediction models. This procedure is applied to predict rainfall class with a three-day lead time for the Yuecheng basin in China. Results show that it is effective in medium-term rainfall prediction with relatively little data requirement.

Key words medium-term rainfall prediction; antecedent influencing factor (AIF); K-mean method; F-test method; multi-discriminant method

---