基于启发式分割算法的水文变异分析研究

Change Point Analysis of Hydrologic Time Series Based on Heuristic Segmentation Method

通过构建理想样本探讨了启发式分割算法在水文变异诊断中的可行性,比较了其与传统变异检验方法在检验性能上的差异,同时结合珠江三角洲河网三水站1900-2004年年最大洪峰流量序列检验其实际应用效果。研究表明该方法能较好排除虚假变异点干扰,较准确识别变异点数目及其发生时间,检验性能优于传统方法,适用于水文变异分析,特别是对复杂非线性非平稳水文序列,实际应用效果较好。通过变异分析发现三水站年最大洪峰流量序列在1993年发生跳跃变异,且为均值向上跳跃,表明区域防洪形势将更加趋于严峻。

The feasibility of hydrologic variation analysis with heuristic segmentation was discussed with constructed ideal samples in this paper. And its performance was compared with traditional change point analysis methods. The annual maximum flood peak discharge data from 1900 to 2004 of Shanshui station in river network of pearl river delta was used to test the practical effect in change point analysis. The results show that the heuristic segmentation method can eliminate interference of illusive change point, identify the change point and its occurring time. It is suitable to deal with change point in hydrologic time series, especially for nonlinear nonstationary time series. The practical application effect is satisfactory. The change point in Shanshui station is 1993. The mean of the series jumped up after 1993, and it goes against the flood defense situation.