基于贝叶斯最大熵与先验知识的被动微波土壤水分产品真实性检验研究

项目名称： 基于贝叶斯最大熵与先验知识的被动微波土壤水分产品真实性检验研究

项目编号： No.41501400

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 天文学、地球科学

项目作者： 王春梅

作者单位： 中国科学院遥感与数字地球研究所

项目金额： 20万元

中文摘要： 随着全球气候变化和陆面数据同化研究对土壤水分反演的精度要求不断提高，被动微波土壤水分产品的真实性检验变得极为重要。如何获取能够代表卫星观测尺度、并揭示空间异质性的土壤水分数字地图，是这一研究的关键问题。. 本项目使用地面样点数据与多源先验知识，基于BME方法框架，以采样数据作为硬数据，运用普通克里格法得到土壤水分的先验概率分布；根据土壤水分与先验知识的定量关系得到土壤水分的模糊分布，作为软数据；然后在软硬数据和先验概率分布的基础上，运用贝叶斯条件概率公式得到后验概率分布，制作土壤水分数字地图。在此基础上，建立一种基于多源数据类型的被动微波遥感土壤水分产品真实性检验方法，并开展国产FY-3产品及国外SMAP产品的检验和评价。该研究不但能够丰富遥感真实性检验学科理论和方法，而且能够提高低分辨率土壤水分产品的精度，提升在相关行业领域的应用价值。

中文关键词： 被动微波遥感；土壤水分；真实性检验；贝叶斯最大熵；先验知识

英文摘要： The validation of soil moisture product from passive microwave remote sensing takes a key role in the study of climate change and data assimilation. The most important questions are how to get the high-precision digital map, which can represent the pixel mean and reveal the spatial heterogeneity. . A new approach for getting digital map of soil moisture is proposed based on the theory of Bayesian maximum entropy (BME) with the ground sampling data and multi-sources of prior knowledge. In the study, with sampling data as hard data, the prior probability distribution of soil moisture is obtained by using ordinary Kriging. And the fuzzy distribution of soil moisture estimated with the quantitative relationship between soil moisture and prior knowledge, as soft data; then based on hard data and prior probability distribution, probability distribution by using the Bayesian conditional probability formula is obtained, making the soil moisture of digital map. On this basis, the validation method of passive microwave remote sensing of soil moisture products is built. And we carry out inspection and evaluation of domestic FY-3 products and foreign SMAP products. The research can not only enrich validation discipline theory and method, but also can improve the precision of low resolution soil moisture products and promote the application value in related industry.

英文关键词： passive microwave remote sensing;soil moisture;validation;bayesian maximum entropy;priori knowledge