Assimilation of soil moisture using surface temperature data and Kalman Filter

Abstract

Initialization of soil moisture is important and difficult problem in numerical weather prediction model. In this study, we use thermal data to estimate soil moisture in land surface model (SiBUC). Temperature and soil moisture are closely related each other through evapotranspiration term which appears in both prognostic equations. So we use Kalman filter to connect these variables in this highly non-linear system.
First step is to formulate filtering system to be used in SiBUC model. State variables are three temperatures (Tc, Tg, Td) and three soil moistures (W1, W2, W3). System equations are prognostic equations of these variables. Observation vector is surface temperature (Tc, Tg) which are expected to be derived from satellite Infra-red image (e.g.GMS5-IR1, IR2).
Second step is to test the system by using the result of control run as observation vector. Changing the initial value of soil moisture, several simulations were executed. In all cases, the system worked to reduce the initial soil moisture error. The system is also tested in JSM-SiBUC coupled model.