【作者】 王萍；

【导师】 彭勇；

【作者基本信息】 大连理工大学 ， 水文学及水资源， 2013， 硕士

【摘要】 耦合降水预报信息的洪水预报可以延长预见期,为防洪调度决策赢得宝贵的时间。但是降水预报信息具有不确定性,即基于“单一的”降水预报结果进行的洪水预报具有不确定性,所以基于“单一的”降水预报结果做出的决策可能不是最优的,甚至可能会出现较大的偏差。针对以上问题,本文以桓仁水库以上流域为试验流域,首先检验评价了UKMO(英国气象局),ECMWF (欧洲中尺度天气预报中心),NCEP (美国国家环境预测中心)和CMC(加拿大气象中心)这四个中心的集合预报数据,然后研究了这四个中心的集合预报数据在洪水预报中的可利用性,最后将ECMWF的集合降水预报数据驱动新安江水文模型进行了洪水预报研究。主要研究内容如下：(1)通过TS评分,BS评分和Talagrand分布对四个中心的集合预报系统进行检验评价,结果表明四个集合预报都具有一定的预报技巧,并且效果相当。由Talagrand分布检验表明,四个中心集合预报的发散度还不够,存在着对小量级降水预报偏大,而对大量级降水(如强降水事件)的预报偏小的情况。其中对于本次试验,UKMO和CMC的Talagrand分布相对较好。(2)通过统计四个中心集合平均降水预报的误差特征值(准确率,空报率和漏报率)来进行其应用于洪水预报的可行性分析。结果显示无雨预报和小雨预报的精度较高。另外,通过分析对比雨量站点预报降雨的三种估算方法(双线性插值法,反距离加权法和相关系数法),提出了集成双线性插值法和反距离加权法的耦合估算方法。(3)针对桓仁水库以上流域的流域特征,选择三水源新安江水文模型作为桓仁水库的洪水预报模型,并应用遗传算法对水文模型进行了参数优选。(4)将集合降水预报驱动水文模型模拟洪水过程,得到了模拟径流的区间范围,为决策者提供了更多有用的风险信息。同时,根据集合平均降水预报的特点,对集合平均降水数据进行修正。并将修正后的集合平均数据模拟洪水过程,其效果有了很大的改善,具有较好的效果。更多还原

【Abstract】 Flood forecasting coupled with precipitation forecast information can prolong forecast period, thus winning precious time for making decisions of flood control operation. However, flood forecasting based on the results of "single" precipitation forecast has the uncertainty, which results from the uncertainty of precipitation forecast information. Decisions based on the above information with uncertainty may not be the best or even larger deviations may occur. For above problems, this paper took the basin above Huanren reservoir as a test basin, testing and evaluating ensemble forecast data of the four centers:UKMO, ECMWF, NCEP and CMC. Availability was researched whether the four centers’data can be used in flood forecasting and ensemble precipitation forecast of ECMWF was adopted to drive the Xinanjiang model for flood forecasting research. Main research contents and conclusions are as follows:(1)The ensemble forecast systems of the four centers were tested and evaluated by using TS score, BS score and Talagrand distribution. Results showed that each of the four ensemble forecasts had a certain prediction technique and the effects were equivalent. Talagrand distribution tests showed that the divergence of ensemble forecast of each center was not enough. Small-level precipitation predictions were bigger and large-level (such as heavy rain events) were smaller. In this test, UKMO and CMC Talagrand distributions were relatively better.(2)Based on statistical analysis of error characteristic values (accuracy rate, vacancy rate and missing rate) of ensemble average precipitation forecast of the four centers, feasibility analysis was taken for their applications in flood forecasting. The results showed that the accuracy of no rain forecast and light rain forecast were higher. Moreover, three estimation methods (bi-linear interpolation, inverse distance weighting method and the correlation coefficient method) were compared and the results showed that there was no absolutely optimal way. So integration of bi-linear interpolation method and inverse distance weighting was put forward for hydrometeorological coupling research.(3)According to basin characteristics above Huanren reservoir, Xinanjiang Model (3Components) was chosen as the flood forecast model and genetic algorithm was applied for parameter optimization.(4)Ensemble precipitation forecast were respectively applied to simulating flood process. And ensemble precipitation forecast had the advantage since it can get the range of runoff simulation, which can provide more useful risk information for decision makers. According to the characteristics of ensemble average precipitation forecast, the average precipitation data were revised and then adopted to simulating the flood process, resulting in a significant improvement of the effect.更多还原