Study on Variation and Prediction of Flood and Drought in Rainy Season in the Pi River Valley
淠河流域汛期旱涝变化的周期性与旱涝等级状态预测

Droughts and floods occurr frequently in the rainy season in the Pi River Valley, and they seriously affect the life and properties of local inhabitants as well as the sustainable development of economy in this region. Therefore, analysis on the changing patterns and prediction of floods and droughts are of great importance for taking precautions against and fight natural adversities in the Pi River Valley. Wavelet transforms in Mexican Hat Function and overlay Markov chain analysis are used in the paper to explore the changing patterns of floods and droughts, so as to predict the conditions during the rainy seasons in future years in Pi River Valley. The research reveals the periodic changes of floods and droughts in Pi River Valley in recent 50 years on different time scales. Since 1980s, the frequency and intensity of floods and waterlogs have become greater. The quasi-fluctuations of 2a is most noticeable from mid-1960s to the end of 1970s and the quasi-fluctuations of 3-4a show predominance after 1993. The two major periods of floods and droughts variations in the rainy seasons of Pi River Valley in the past five decades are 2a and 9a respectively. The risks of floods and waterlogs will be high in the 10 to 12 years after 2003. Overlay Markov chain can reliably forecast the grade of floods and droughts in future years. The results show that precipitation in the rainy seasons in 2009 and 2010 should be normal. The probability of a year with normal precipitation during the rainy season is the largest and the multiyear return period is only 2.02 years. The probability of each flood grade is greater than it of the corresponding drought grade, and the multiyear return periods of each flood grade is shorter than it of the corresponding drought grade. Wavelet transforms offer trend and background for Markov chain analysis, and Overlay Markov chain can forecast grades of floods and droughts in medium short-term correctly. The combination of wavelet transforms and Markov chain analysis can improve the correctness and reliability of prediction of floods and droughts since they are complementary to each other.