Regression Kriging-Based Upscaling of Soil Moisture Measurements From a Wireless Sensor Network and Multiresource Remote Sensing Information Over Heterogeneous Cropland

The ground truth estimated by in situ measurements is important for accurately evaluating retrieved remote sensing products, particularly over heterogeneous land surfaces. This letter analyzes the role of multisource remote sensing observations on the upscaling of soil moisture observed by a wireless sensor network at the pixel scale via the regression kriging (RK) method. Three types of auxiliary remote sensing information are employed, including Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Temperature Vegetation Dryness Index (TVDI; 90 m), Polarimetric L-band Multiband Radiometer brightness temperature (700 m), and Moderate Resolution Image Spectroradiometer TVDI (1000 m). Moreover, a comparison with the ordinary kriging method is analyzed. The spatial inferences show that the RK method is more accurate and that its spatial pattern is more consistent with the auxiliary data when the trend is successfully removed, particularly when spatial continuity is destroyed by irrigation. The ASTER TVDI has a higher resolution and stronger correlation with soil moisture and yields more accurate interpolation results than the other types of remote sensing information. Although medium-resolution data do not substantially contribute to capture the spatial patterns of soil moisture, such data may still improve the prediction accuracy.