流域水文干旱等级预测问题的研究

Study on drought class prediction problem in basin

采用滦河潘家口水库控制流域9个水文站1958-2009年的逐月流量数据,通过计算各站12个月时间尺度的标准化径流指数(SRI),得到了实测水文干旱等级序列。根据干旱等级转换的观测频数,提出了基于三维列联表的对数线性干旱预测模型,并计算了干旱等级转换间的优势及其置信区间,以估算干旱等级转换的概率,实现了滦河潘家口水库控制流域预见期为1个月和2个月的水文干旱等级预测。选取2000-2009年的水文干旱演变过程对预测模型进行了验证。结果表明:干旱等级转换的三维列联表呈现一种强对角线趋势,各站三维对数线性模型的检验显著性水平p均大于检验水准α(0.05),观测频数与期望频数吻合程度高,且预见期为1个月的水文干旱等级预测精度较高,可用于滦河潘家口水库控制流域短期水文干旱预警。但预见期为2个月的水文干旱等级的预测误差较大,需根据流域的气候特征及下垫面的物理状态对模型进行进一步的改进。因此,如何综合应用气象和水文干旱监测信息,并采用对数线性预测法建立新的模型对水文干旱进行提前预警是需要解决的重要问题。

Based on the monthly streamflow data from 9 hydrologic stations during the period of 1958-2009 in the Panjiakou Reservoir catchment of the Luanhe River Basin,drought class time series were derived from standardized runoff index(SRI) time series computed in a 12-month time scale. Loglinear drought prediction models based on three-dimensional contingency tables were propased according to the observed frequencies of drought class transitions. Odds and respective confidence intervals for drought class transitions were also calculated to estimate the drought class transition probabilities. Thus the drought classes with 1 and 2 months prevision were predicted and the short term hydrological drought class prediction was achieved in the studied area. The validation of the predictions was performed for the evalutionary process of hydrological drought in the period of 2000-2009. The results show that the three-dimensional contingency tables of drought class transitions present a strong diagonal tendency and results for all sites show a similar agreement between observed and expected frequencies with all the models presenting a test p-value exceeding the chosen significance level of a = 0. 05. Three-dimensional loglinear modeling method is applicable to predict the hydrological drought classes with 1 month prevision. And results of modelingpresent better results when comparing predicted and observed drought classes for those 9 stations. It could be concluded that loglinear prediction of drought class transitions is useful for short term hydrological drought warning in the Panjiakou Reservoir catchment. However,the prediction error is bigger when predicting the hydrological drought classes with 2 month prevision. The model should be further improved according to the climate characteristics and physical state of underlying surface in the catchment. Therefore,how to establish a new model for early warning of hydrologic drought through loglinear modeling method with comprehensive application of meteorological and hydrological drought monitoring information is one of the important problems need to be solved.