Real-time Flood Forecasting: A Model Based on Volterra Series and its Application in Semiarid Area

摘要

Mechanism identification and model selection for hydrological modeling in semiarid river basin are of great importance and urgency, especially in ungauged basins. On the basis of the present models with systematic approach, a total runoff non-linear response model based on Volterra series (Volterra-TNLR model) is proposed, and its expressions in matrix are deduced, together with its solving process. A 3-layer artificial neural network (ANN) model is introduced to simulate the self-relative relationships of error series between observed and calculated runoff. And then the predicted runoff can be calibrated dynamically by ANN model in the process of realtime flood forecasting. Both the Volterra-TNLR and ANN model have clear structure and high capability of non-linear mapping, and {hey are easy to be programmed and integrated into a real-time flood forecasting system. A semiarid watershed (Qian River) in northwestern China was considered as an example to verify the Volterra-TNLR model. The results showed that the model has a high precision, and it can be used in the similar semiarid river basins.