Nierji reservoir flood forecasting based on a Data-Based Mechanistic methodology

Abstract

The Nierji Basin, in the north-east of China, is one of the most important basins in the joint operation of the entire Songhua River, containing a major reservoir used for flood control. It is necessary to forecast the flow of the basin during periods of flood accurately and with the maximum lead time possible. This paper presents a flood forecasting system, using the Data Based Mechanistic (DBM) modeling approach and Kalman Filter data assimilation for flood forecasting in the data limited Nierji Reservoir Basin (NIRB). Examples are given of the application of the DBM methodology using both single input (rainfall or upstream flow) and multiple input (rainfalls and upstream flow) to forecast the downstream discharge for different sub-basins. Model identification uses the simplified recursive instrumental variable (SRIV) algorithm, which is robust to noise in the observation data. The application is novel in its use of stochastic optimisation to define rain gauge weights and identify the power law nonlinearity. It is also the first application of the DBM methodology to flood forecasting in China. Using the methodology allows the forecasting with lead times of 1-day, 2-day, 3-day, 4-day, 5-day with 98%, 97%, 96%, 96% and 93% forecast coefficient of determination respectively, which is sufficient for the regulation of the reservoirs in the basin.