降水观测系列集合插补及可靠性评价方法研究

Precipitation is an important part of the hydrology cycle, is the key input of the land surface hydrological fluxes simulation. The qualified rainfall dada, both in accuracy and spatial - temporal resolution, takes great significant in hydrology cycle research and rational utilization of water resources. Affected by natural conditions and human activities, rainfall stations often exist such problems as lack measurement, short measurement and isolated measurement, should be extended, interpolated and filled. However, traditional reference station single linear regression interpolation extension methods have some disabilities such as simplified factors, lower recognition, poor reliability and data disposal, etc. This study intends to introduce large data set analysis thought, using Bayesian level statistics, genetic programming data mining and multivariate data correction method to develop an ensemble predication algorithm and model for precipitation observational data interpolation extension. Addressing the influences of precipitation and rainfall-runoff physical mechanism, the study is to explore the site–site, site-satellite, site–satellite-runoff imputation interpolation method. Conception of ensemble forecast for reference, put forward the site observation precipitation both original dataset and extended interpolation dataset into a reliable probability evaluation space to establish a site observation precipitation data holographic expression method. Taking Qilian Mountains as the study area, the study will comprehensively evaluate and extended interpolate the precipitation dataset which can enhance the potential of the existed hydrology and meteorology observation data. The study will enrich the methodology of hydro-metrology data collating and management.

降雨是水文循环的重要环节，是陆面水文通量模拟的关键输入。具有一定精度和时空分辨率的降雨资料对水文循环科学研究和水资源合理利用具有重要意义。受自然条件及人为影响，降水测站常存在缺测、短测和孤测等问题，需插补延长。然而，传统的参照站线性回归插补延长方法存在因素单一、辨识率低、可靠性差和数据丢弃多等不足。本研究拟引入大数据集合分析思想，采用贝叶斯层次统计、遗传规划数据挖掘、多元数据交叉校正等方法，发展降水观测数据的集合预报插补延长算法和模型；从降雨本身影响因素和降水径流物理机制出发，探索考虑多重因素的站点-站点、站点-卫星、站点-卫星-径流的降雨融合插补方法；借鉴集合预报概念，提出降水原始观测数据集合和插补延长数据集合的可靠度概率空间评价方法，提出降水测站数据可用度全息表达方法。以祁连山区南北两麓为研究区域，全面评价和插补延长降水数据集合，挖掘原始水文气象观测数据潜力，丰富水文气象数据整理方法。

降水是水文循环的重要环节，是陆面水文通量模拟的关键要素，而降水时空观测序列及其可靠性是水文循环研究和水资源合理利用的基础，具有十分重要的意义。受自然条件和其他各种因素影响，地表降水测站常存在缺测、短测和孤测等问题，需插补延长。针对传统降水序列插补延长方法存在因素单一、辨识率低、可靠性差和数据丢弃多等问题，引入大数据集合分析思想，采用贝叶斯层次统计、遗传规划数据挖掘、多元数据交叉校正等方法，发展降水观测数据的集合预报插补延长算法和模型；从降雨本身影响因素和降水径流物理机制出发，探索考虑多重因素的站点-站点、站点-卫星、站点-卫星-径流的降雨融合插补方法；借鉴集合预报概念，提出降水原始观测数据集合和插补延长数据集合的可靠度概率空间评价方法，提出降水测站数据可用度全息表达方法。以祁连山区南北两麓为研究区域，全面评价和插补延长降水数据集合，挖掘原始水文气象观测数据潜力，丰富水文气象数据整理方法。