中国西部典型区无资料水文数据估计方法研究

The study on hydrological data estimation method of ungauged basins is a very important question in PUB (Predictions in Ungauged Basins) plan. This project mainly study the typical areas of the source region of the Yangtze River Basin and the inland river basin(Heihe River, Shule River) in Western China, where hydrological data is less recorded. Taking the hydrological prosesses and characteristics in typical regions as the research object, combined with the characteristics of vegetation distribution difference of land surface conditions and runoff, with the integrated application of modern statistical new methods, by using PSO(Particle Swarm Optimization algorithm), genetic algorithm, artificial neural network, support vector machine et al, the hydrological process of long sequences hydrologic station will be simulated and hydrological characteristics will be analysed. Then hiddened the long sequence of hydrological data and make them into a short sequence data, by repeated testing, the mathematical model with the time scale stability and reliability will be established. Using newly established model, the annual and monthly runoff in data-deficiency site will be estimated. The runoff estimation and reduction target can be realized in ungauged basins or data-deficiency regions. The hydrologic estimation ability for ungauged basins or data-deficiency regions can be improved. The reach results of the project will innovate hydrological mathematical methods and modeling, provide a theoretical basis and reference for hydrological research of ungauged basins, and have higher academic value and very good application prospect.

无资料水文数据估计方法研究是PUB(Predictions in Ungauged Basins)计划中非常重要的问题。本项目选择典型区为中国西部水文数据缺乏的长江源区流域和内陆河典型流域（黑河、疏勒河），以典型区水文过程和水文特征为研究对象，结合典型区域植被分布差异显著的陆面条件和产汇流特性，综合应用现代概率统计的新方法，借助PSO、遗传算法、人工神经网络、支撑向量机等新技术，模拟长序列水文站水文过程，分析其水文特征，再把长序列水文资料人为隐去，使其变成短序列资料，反复检验，建立具有时间尺度稳定性和可靠性的数学模型，用该模型来估计资料短缺站点年月径流量，从而实现无资料或资料缺乏流域径流估计和还原的目标，提高无资料地区或资料缺乏地区水文估算能力。该项目研究将实现水文数学方法和建模的创新，可为无资料流域水文研究提供理论依据和可借鉴的成果，具有较高的学术价值和很好的应用前景。

我们选择典型区为中国西部水文数据缺乏的长江源区流域和内陆河典型流域（黑河、疏勒河），以典型区水文过程和水文特征为研究对象，结合典型区域植被分布差异显著的陆面条件和产汇流特性，对长江源区流域的直门达站和沱沱河站的站点资料、黑河水系的莺落峡水文站和札马什克水文站资料及疏勒河水系的昌马堡水文站资料，用现代统计推断的新方法，借助PSO、集合经验模式分解（EEMD），径向基函数神经网络（RBFN），和支持向量机（SVM）、X12季节调整，季节性自回归移动平均（SARIMA），双指数平滑（DES）、遗传算法等新的算法组合优化模型的参数，模拟长序列水文站水文过程，分析其水文特征，再把长序列水文资料人为隐去，使其变成短序列资料，反复检验，建立具有时间尺度稳定性和可靠性的数学模型，用该模型来估计资料短缺站点年月径流量，从而实现无资料或资料缺乏流域径流估计和还原的目标，提高无资料地区或资料缺乏地区水文估算能力。我们建立了几个新的混合预测模型，分别对一些典型站点的降水、气温、径流量等数据进行分析和预测，得到了比较好的预测结果。我们的结果将为研究寒区无资料流域水文研究提供重要的理论依据，同时实现了水文数学方法和模型研究和创新，具有较高的学术价值和很好的应用前景。