基于多模型 M CP 方法的洪水概率预报

洪水概率预报通过提供具有一定置信度的预报区间,评估预报结果的可靠度,为防洪调度提供重要依据。以淮河关键防洪断面王家坝为研究对象,分别采用API和新安江(XAJ)确定性模型进行初始的确定性预报,在此基础上,再采用模型条件处理器(MCP)推求不同量级洪水预报流量的条件概率分布函数,实现洪水概率预报。分别从中位数的确定性精度评价和概率预报的可靠度评价两方面对预报结果进行分析,结果表明:MCP洪水概率预报结果不仅具有较高的可靠度,而且其中位数预报与确定性模型结果相比,预报精度整体有所提高,说明MCP具备一定的校正预报能力。

Probabilistic flood forecasting based on multi-model MCP

The pr obabilistic flo od fo recasting can pro v ide a predictio n inter val w ith a certain reliabilit y, and can be used to e2 v aluat e the reliability of fo recasting results. It can pro v ide an im portant basis fo r floo d co ntro l scheduling . We too k Wa ng jia2 ba cr oss2sectio n, a key f lo od contr ol sectio n of H uaihe River, as the resear ch object. Based on the predict ion results of API and XAJ mo dels, using t he M odel condit ional pr ocesso r ( MCP) to deduce the co nditio na l pro ba bility distributio n function of the for ecasting runo ff of floo ds o f different m agnitudes, w e r ealized pro babilistic flo od fo recasting . T he pr edictio n results w ere analyzed in terms of the dete rministic precision evaluatio n o f m edian number and the r eliability ev aluat ion of pro babilis2 tic for ecasting . The r esults show ed that the M CP pr obabilistic floo d fo recasting has a high r eliability, a nd its median number predictio n has a hig her pr edictio n accur acy t ha n the determinist ic mo del, indicating that M CP has a certain ability of co rr ec2 tion and prediction.