基于时空统计方法的多源定量遥感产品融合方法研究

The high level land,water and atmospheric remote sensing products that derived from a single sensor observation has to face problems of the spatio-temporal incompleteness, low accuracy, and less continuity in space and time.However, products from different sensors may be complementary in uncertainties and spatio-temporal completeness. It is potential to merging these complementary products to generate an improved suite of products with improved accuracy and spatio-temporal completeness.However, the method on merging multi-source high level remote sensing products to generate high accuracy and spatio-temporally complete products with various spatial resoution is less investigated so far. The overall objective of this project is develop the methods for merging the multi-source high level remote sensing products based on the spatio-temporal statistics.Specifically, we will develop the remote sensing products merging methods based on two spatio-temporal statistical models. The first one is the BME framework that is based on the modern spatio-temporal geostatistics, and the second one is the hiararchical bayesian modeling.we aim to build the efficient workflows for merging the multi-source high level remote sensing products in both BME and the HBM frameworks.The key techniques involved in this study are:(1) How to assess the uncertainties in the to be merged products and how to incoporting this uncertainties in the merging procedures; (2)how to acquire,express and integrate various knowledges in relevance to the variables uncer study in the merging procedure? and (3) how to modeling the hidden spatio-temporal process in spatio-temporal statistical merging procedures? we plan to develop prototype system for the multi-source remote sensing products merging. dataset of merged products on LAI,SST and AOD in our study area will be produced and validated.

由单一传感器反演得到的地表特征参量定量遥感产品存在时空完整性低、精度不够高、反映地表时空变化不连续、空间分辨率单一、不同传感器反演得到的数据间缺少物理意义上的一致性等问题。如何利用多传感器遥感产品在时空完整性、精度等方面的互补性，融合多传感器的定量遥感产品生成时空完整、高精度、一致、时空连续的遥感产品，是当前定量遥感产品生成中需要解决的重要问题。本项目针对这一问题，研究不依赖于动力学过程模型，基于时空统计方法融合多源定量遥感产品的方法。具体地，项目将以LAI、SST和AOD参量的多源遥感产品融合为例，分别研究基于时空地统计学贝叶斯最大熵方法（BME）和层次贝叶斯方法（BHM）两种时空统计方法的多源遥感数据融合方法，重点研究时空统计多源遥感产品融合中不确定性的表达、先验知识的获取与表达、时空过程模型的构建和计算方法，生成研究区样例数据集，并开发软件原型。

由单一传感器反演得到的地表特征参量定量遥感产品存在时空完整性低、精度不够高、反映地表时空变化不连续、空间分辨率单一、缺少一致性等问题。融合多传感器的定量遥感产品生成时空完整、高精度、一致、时空连续的遥感产品，是当前定量遥感产品生成中需要解决的重要问题。本项目针对这一问题，研究不依赖于动力学过程模型，基于时空统计方法融合多源定量遥感产品的方法。项目以SST、LST、AOD、LAI等水、地表、大气、植被等具有不同时空特征的地表参量多源遥感产品融合为例开展研究，发展了基于贝叶斯最大熵方法的多源、多时空尺度遥感产品时空融合理论框架，针对SST、LST、AOD和LAI分别构建了融合数据处理和技术流程，开发了相关程序模块，生成了样例数据集；发展了基于时空数据层次贝叶斯统计的多源、多时空尺度遥感产品融合方法，成功用于多尺度SST卫星产品的时空融合；针对点位观测和 遥感数据的融合问题，分别发展了发展了基于BME和基于地统计学回归的点-面数据融合方法，在LAI点位观测数据与遥感数据的融合中得到了较高的融合精度。本项目发展的多源遥感产品融合算法和点位观测与遥感数据的融合算法，对于有效协同多卫星观测生成时空完整、多尺度、高精度定量遥感产品，提高定量遥感产品尺度转换和精度验证，推动遥感产品应用的广度和深度具有重要的理论意义和应用价值。