基于门限交互式回归模型的感潮河段潮位预报

Application of threshold interactive regression model in tidal reach water level forecasting

感潮河段水流情势复杂,河道冲淤多变,地形资料获取困难。依据潮位变化成因,结合统计分析方法,提出利用门限交互式回归模型TIR对特征潮位进行预报。TIR模型将交互式回归置于门限类模型框架中,避免洪潮要素综合过程中线性叠加假定的限制和缺陷,同时考虑洪潮驱动要素的时空差异及潮位非线性变化特征。依据门限分段理论,针对特定潮位站的不同水位等级,分别建立了不同的潮位预报相关关系。将TIR模型分别应用于南京、镇江及江阴3个潮位站,结果表明,各站预报效果良好。门限交互式回归模型框架合理、适应性强、资料依赖程度低、可获得较高的预报精度。

The flow regime and sediment deposition- scouring in a tidal reach are complicated,so it is difficult to obtain the topographical data of a tidal channel. According to the causes of tidal level variation,by using the statistical analysis technique,the threshold interactive regression model is put forward to forecast the tidal level. The threshold interactive regression model places interactive regression into threshold model framework,which can avoid the limitation and default of the linear superposition assumption in a flood- tidal integration process and can consider the spatial and temporal differences of tidal driving factors and the nonlinear variation characteristics of tidal level. According to threshold sectional theory,the staged tidal level forecasting relationships are established for different tidal level stages of some selected tidal stations. The model has been used in three hydrological stations（ Nanjing,Zhenjiang and Jiangyin） respectively. The results showed that the model works well in each station. With reasonable framework and strong adaptation,the TIR model can achieve good performance in forecasting and have low dependence to data. This model has practical significance and promotion value.