Topographically based spatially averaging of SAR data improves performance of soil moisture models

Spatial averaging schemes have often been used to improve empirical models that relate radar backscatter coefficient to soil moisture. However, reducing the noise in backscatter response not related to soil moisture often results in signal losses that are related to soil moisture. In this study we tested whether a spatial averaging scheme based on topographic features improved regressions relating backscatter coefficient and soil moisture on the low relief landscape of the Prairie Pothole Region of Canada. Soil moisture data were collected along hillslope transects within pothole drainage basins at intervals coincident with RADARSAT-1 satellite overpass. Spatial averaging schemes were designed at four scales: pixel, topographic feature (uplands, sideslopes, and lowlands), pothole drainage basin, and landscape (0.8 km × 1.6 km). The relationship between soil moisture and backscatter coefficient improved with increasing area of spatial averaging from a pixel (R2 = 0.18, P < 0.005), to the pothole drainage basin (R2 = 0.36, P < 0.005), to the landscape (R2 = 0.66, P < 0.005). However, the strongest relationship (R2 = 0.72, P < 0.005) was obtained by spatially averaging radar images based on topographic features. These findings indicate that topographically based spatial averaging of RADARSAT-1 imagery improves empirical models that are created to map the complex patterns of soil moisture in prairie pothole landscapes.

► Challenge was remotely sensing soil moisture in the Prairie Pothole Region.
► LiDAR models defined topographic features representing hydrologic processes.
► RADARSAT-1 backscatter spatially averaged by topographic feature.
► Observed strong relationships between ground- and radar-based data.
► Maps show spatial and temporal variation in soil moisture.