Land surface temperature assimilation into a soil moisture-temperature model for retrieving farm-scale root zone soil moisture

Abstract

• Data are assimilated into soil moisture and temperature modules of HYDRUS-1D. • For SSM retrieval, assimilation of SSM data is superior to LST assimilation. • For retrieval of RZSM, LST assimilation is better than SSM assimilation. • Assimilation of LANDSAT8-LST data leads to a more accurate RZSM retrieval. • Assimilation of MODIS-LST results in more efficient estimation of soil parameters. Thermal infrared remote sensing have been extensively applied to estimate global- or regional- extent surface soil moisture. Meanwhile, potentials of the remotely sensed data for farm-scale retrieval of root zone soil moisture (RZSM) as well as estimation of soil hydraulic parameters, have been rarely investigated. Using Ensemble Kalman Filter, we propose a new methodology to assimilate land surface temperature (LST) of both MODIS and LANDSAT-8, into the soil temperature module of HYDRUS-1D model. The main objectives are to estimate soil hydraulic parameters and to retrieve RZSM with high spatiotemporal resolution independent of any in-situ measurements of soil temperature or moisture. However, we consider some modeling scenarios by which we assimilate in-situ measurements of soil moisture into the soil moisture module of the HYDRUS-1D model to provide a reference to compare with results of the LST assimilation scenarios. We apply the proposed methodology to a farm located in Moghan irrigation district, Ardabil province of Iran, which has in-situ soil moisture measurements. Even in the least accurate scenario of ours by which MODIS-LST was assimilated, RMSE varies in the range of 0.012–0.013 cm 3 ·cm −3 demonstrated to be superior compared to preceding recent works in the literature of satellite soil moisture retrieval. Moreover, the scenario of assimilating LANDSAT-LST data leads to higher parameter uncertainty compared to the assimilation of solely in-situ soil moisture or MODIS-LST which is related to higher temporal resolution of both in-situ and MODIS data compared to LANDSAT data and the error stems from the algorithm of deriving LANDSAT-LST. Accordingly, our study recommend that assimilation of the satellite-based land surface temperature of both LANDSAT-8 and MODIS are appropriate alternatives for expensive in-situ measurement.