Intraseasonal oscillation of heavy rainfall over Haihe River Basin and its extended-range forecast

Summer heavy rainfall over Haihe River Basin frequently triggered flood disasters in recent years. To improve the accuracy of heavy rainfall forecast, this study applied the Fourier harmonic analysis, multivariate empirical orthogonal function, and convolutional neural network regression to analyze the characteristics of summer heavy rainfall and its critical circulations on the intraseasonal timescale over Haihe River Basin. A prediction model was constructed by integrating statistical methods and deep learning algorithm. Results show that, (1) the 10–30 d intraseasonal oscillation contributes more to the interannual variation of pentad-mean precipitation than those of the other timescales. When heavy rainfall occurs, the North China cyclone in the lower troposphere promotes the convergence of warm and humid water vapor, and the “west low, east high” pattern of geopotential height anomaly in the mid-latitudinal Eurasian significantly enhances the dynamic ascent. (2) The North China vorticity index at 850 hPa and the Northeast Asia height difference index at 500 hPa both have significant leading correlation with heavy rainfall, with a longer forecast period of heavy rain than that of moderate rain. (3) Based on the optimal forecast day, a heavy rainfall prediction model on extended-range timescale is constructed, which achieves 72.0% and 66.0% accuracy for heavy rain at 8- and 18- lead days, respectively, and 64.3% for moderate rain 8-day in advance. This study reveals the impact of atmospheric intraseasonal circulations on heavy rainfall in the Haihe River Basin, provides a scientific basis for improving the prediction accuracy, and enhances the early warning ability of extreme weather as well as climate events in flood season over northern China.