An Approach to Estimating Spatially Correlated Error Covariance of Remote Sensing Retrieved Soil Moisture

We have designed an approach TC_Cov based on triple collocation covariance, which can be used to estimate spatially correlated error covariance of three soil moisture data sets retrieved from satellite data. In theory, we formalize each grid into one random variable and soil moisture time series of two grids into samples of two random variables, then the covariance of two random variables are estimated. An element of multivariate covariance matrix takes bivariate covariance. The experimental results are analyzed in a qualitative way. In the experimental region of Murrumbidgee catchment located in Australia, the error covariances of soil moisture data sets retrieved from ASCAT scatterometer, TRMM radiometer and ERA-Interim reanalysis data are estimated, and experimental results obey the Tobler's first law well and explain real soil moisture error covariance estimation effectively.