EVALUATING PREDICTIONS OF THE SOIL MOISTURE MODEL WITH DATA ASSIMILATION BY THE TRIPLE COLLOCATION METHOD

Abstract

For a long time, ground measurements are considered the most accurate method of regular monitoring of soil moisture. However, ground stations are expensive, require local calibration and thus often not practical to use. Other, more affordable means of soil moisture monitoring can be developed with the recent advancement of Earth remote sensing technologies. In this paper, we describe a nonlinear problem of soil moisture transfer problem with addition of satellite soil moisture measurements. The mathematical model is based on the Richards equation for soil moisture transport, and solved with the finite difference method on implicit iterative scheme. Satellite moisture retrievals are acquired by combining active and passive sensors data with the decomposition algorithm. The satellite data are incorporated into the model with the data assimilation algorithm called Newtonian nudging. This method adds a special ‘nudging’ term into the model governing equation. This assures that the model is corrected by satellite measurements without affecting the process physics. Moreover, we look into the nudging factor problem, and propose a simple empirical relation based on the soil properties for more universally stable work of the method. For validation purpose, we conduct a massive numerical experiments over all registered ground stations in the USA. Evaluation is done by the use of triple collocation method, which allows assessing the errors of three independent data sources. The data sources used for evaluation are the model results, ground station measurements and ERA5 satellite observations. The results demonstrate that the presented model is capable of producing results with close accuracy to the ground station measurements.