Author(s): Hui Wan; Jun Xia; Liping Zhang; Sidong Zeng; Hong Du; Xiaoyan Zhai
Linked Author(s): Jun Xia
Keywords: MTVGM; Real-time; Flood forecasting; Hydrological system approach
Abstract: Time Variant Gain Model (TVGM) describing the nonlinear relationship between rainfall and runoff is widely used in flood forecasting, which plays a significant role in non-structural measurement of flood control decision and management. However, this model with single surface flow source has a drawback in simulating flow peak and decreasing stage. Thus in this paper, a multi-source TVGM (MTVGM) is developed, taking groundwater source into consideration based on TVGM, and also inheriting the main feature that is the gain factor in the runoff generation process varies with the Antecedent Precipitation Index (API) with an explicit equation. Because of the time-variance, non-linearity, uncertainty and complexity of hydrological system, real-time flood forecasting is becoming an increasing effective approach to improve the accuracy in operational flood forecasting, which makes flood mitigation more efficient. A separated calibration approach for renewing parameters used in the runoff generation process and the flow routing process is proposed. The 3-hour continuous data from 2002 to 2008 for 13 subbasins above Bengbu in Huaihe River Basin in east China is used to calibrate and validate the model. The water balance efficiency, Nash-Sutcliffe efficiency (R2) and the relative peak error are applied in the evaluation of the performance of model. Calibration and validation show MTVGM with simple relationship between API and gain factor satisfactorily simulates the flood in Huaihe River Basin area. And the results of real-time flood forecasting by using the self-adaptive Kalman Filtering algorithm indicate that the on-line identification is computationally flexible and adaptable and obtain higher predicting accuracy with higher R2 than that of off-line approach. The efficiency of real-time updating method is showed to be the same with that of the Recursive Least-Square method.
Year: 2013