Improving the Estimation of Soil Moisture in Semi-Arid Regions Using Data from Different Remote Sensing Techniques

Satellite-derived soil moisture fields received attention due to their large spatial coverage and spatial resolution that suits many applications. The sensors used vary from passive (e.g., LANDSAT-8) to active (e.g., SENTINEL-1) with varying accuracy problems. Passive sensing can only determine relative indices between pixels within a vegetation class and not the real value of moisture. Active sensing suffers from the sensitivity of its detecting behaviour to the level of moisture (anomalous backscatter). The above problems impose limitations on the application without frequent ground-based calibration. The paper investigates possible models to improve the estimation of soil moisture using the powers of the two sensors. In addition, a Hydrologic Surface Moisture indicator (HSM) is included as a third source of information. The paper tests modeling combinations of the three soil moisture predictors (Landsat-8, Sentinel-1, and HSM). The models are validated using in-situ measurements. The results showed that Landsat-8 data can be rescaled using HSM to provide the actual soil moisture in the soil. On the other side, it is possible to remove the anomaly from the Sentinel-1 backscatter using either Landsat-8 or HSM. The elimination of the above problems explained a significant portion of the differences between the two sensors.