Application Of The Functional-Coefficient Auto-Regressive Model In Hydrologic Forecast

Hydrologic forecast is one of the most important parts of hydrology research. It is the key issue in flood control, water resource usage and protection, hydraulic structure design reservoir operation, and industry agriculture production. Owing to be influenced by climatic factor, monthly discharge sequence takes the year as the cyclical seasonal fluctuation. Its statistical characteristic varies along with season. It is one kind of seasonal non-stable time series, taking on a stronger randomness and non-stability. Toward to the formulation of production plan, the flood prevention, drought-defying, the electricity generation, the water resources program management and the comprehensive utilization, the monthly discharge forecast has the extremely vital significance.The time series analysis method plays an important role in the hydrologic forecast. The seasonal AR model and the functional-coefficient autoregressive model that belong to non-linear time series model are established. The functional-coefficient autoregressive model is estimated by using polynomial spline method, including the selection of knot number and position, the selection of threshold variable and significant variables. The two established models are utilized to fitting and predict for the data of Yangtze River da-tong hydrologic station monthly discharge. The results show that the forecast for da-tong hydrologic station monthly discharge of functional-coefficient autoregressive model is quite to be good. Especially when flood season existence, the forecast precision of functional-coefficient autoregressive model is higher than the seasonal AR model. It reflects the fluctuation tendency of monthly discharge very good.