Abstract:

The method that analyzes the period characteristics of annual runoff with continuous wavelet transformation and wavelet variance has been widely used.However it only gives qualitative prediction results, but not quantitative results. Based on the primary qualitative method, we propose an annual runoff prediction model using weighted sum of wavelet coefficients of major periods to predict the periodic components.The method effectively takes advantage of the multiresolution characteristic of wavelet analysis, and can reduce prediction errors caused by trend components.The annual runoff data of Qianhe basin are used to build and check the model.The proposed method can provide better results useful in predicting annual runoff time series.

Key words: wavelet coefficients, weighted summation, annual runoff prediction, time series