异质性地表反照率产品真实性检验中的尺度效应研究

Validation of satellite-derived products is an essential precursor to scientific research and subsequent environment monitoring. However, it is challenged by the large spatial scale difference between in situ measurements and satellite product pixel, as well as the widely distributed surface heterogeneity. Based on the theories of spatio-temporal analysis, geostatistics as well as validation, this project aims to investigate and quantify the uncertainties during the process of validation based on in situ measurements, and develop a spatio-temporal upscaling method specified for single site measurement by integrating the spatial information provided by auxiliary high-resolution images and the temporal information provided by long time series in situ observations. As a result, the pixel scale albedo “truth” can be derived continuously to support validation over heterogeneous surfaces. The achievements of this project are expected to improve the reliability and rigor of the validation results over heterogeneous surfaces.

真实性检验由于能提高遥感产品的定量化水平和应用精度而越来越受到重视。目前，遥感产品的真实性检验受到地面观测与遥感产品像元尺度不匹配以及地表异质性的影响，使真实性检验结果的准确性和可信度降低。本项目拟结合时空分析、地统计、真实性检验等领域的理论和方法，重点研究基于地面站点观测进行真实性检验的不确定性来源以及基于单个站点数据的时空升尺度方法，实现长时间序列、连续地面观测数据的升尺度转换，从而为在异质性地表检验中低分辨率遥感反照率产品提供时间连续的像元尺度“真值”。本课题的研究有望提高异质性地表反照率产品真实性检验的可信度和准确度。

本项目主要研究了遥感产品真实性检验中由于地面观测跟卫星像元尺度不匹配和地表广泛分布的时空异质性引起的尺度效应。考虑地表动态变化和地面单个站点观测数据的有限空间代表性，建立了基于先验知识趋势面的单个站点尺度转换模型，从根本上解决了地面观测信息空间代表性不够、与待验证像元尺度不匹配的问题。以怀来遥感试验站的地面站点为例，升尺度后的RMSE从0.017降低到了0.012，Bias从-0.012降低到了0。升尺度方法能够提高地面单个站点的代表度，尤其适用于异质性站点。能够大大提高稀疏观测网中地面站点在真实性检验中的利用率。在项目的资助下，在遥感、地学类顶级期刊JGR-Atmosphere（1篇）、IEEE TGRS（2篇）、ESR（1篇），ESSD（1篇）发表文章5篇。