Soil Moisture Index Model for Retrieving Soil Moisture in Semiarid Regions of China

There have been some limitations in acquiring an accurate representation of remotely sensed data-derived soil moisture. Here, we propose a simplified thermal inertia model (STIM). This model requires only the albedo and surface maximum temperature easily obtained from satellite imagery, such as that of the moderate resolution imaging spectroradiometer (MODIS). In this study, we defined a soil moisture index (SMI) using the MODIS imagery to obtain a simplified thermal inertia-ratio vegetation index spectral feature space. SMI results from STIM were validated at several locations in the study area of western Inner Mongolia and compared with those from the apparent thermal inertia (ATI) model. Our results showed that our SMI model could explain 71% of the variance in the surface soil moisture, approximately 5% higher than that of the ATI model. In a comparison of field-measured soil moisture data with data simulated using two methods, the SMI and ATI, the SMI showed better retrieval accuracy by lessening the effective error due to the vegetation by 4.2%-10.8%, whereas soil moisture data simulated with ATI showed an effective error of 4.5%-17.0%. The SMI model was also used to map soil moisture; a relative root-mean-square error of 7.67% was recorded for the region, implying the ability of the model to map soil moisture over large areas. Here, the proposed SMI model was proven to be more suitable for estimating soil moisture in locations in which the vegetation index values ranged from 0 to 3. Thus, the proposed SMI model provides a new approach using remote sensing thermal inertia methods to quantify soil moisture at the regional scale.