Estimating tallgrass prairie soil moisture using active satellite microwave imagery and optical sensor inputs

Recent advances in active microwave remote sensing techniques provide the potential for monitoring soil moisture conditions at the spatial and temporal scales required for detailed local modeling efforts. The goal of this research was to produce accurate and spatially distributed estimates of soil moisture using a time series of ERS-2 images for the Konza Prairie, a tallgrass environment in northeast Kansas. The methods used in this research involve field data collection of soil moisture, digital image interpretation of optical (NOAA AVHRR and LANDSAT TM) and radar (ERS-2) imagery, and environmental modeling in a raster GIS environment.; To accomplish the research goals, the effect of variable terrain on radar image backscatter values was quantified and reduced. Next, the scattering behavior of the overlying vegetation canopy was simulated using a water cloud model that estimated the contribution of vegetation backscatter (sigma oveg) to the total backscatter coefficient (sigma ototal). Critical to this process were estimates of aboveground primary production made using the normalized difference vegetation index from a combination of AVHRR and LANDSAT TM images. With sigmao veg removed from the amount of backscatter contributed by the soil surface (sigmaosoil) was calculated and the linear relationship between sigmaosoil and volumetric soil moisture was determined. This regression model was then inverted and solved for volumetric soil moisture to quantify near surface soil moisture conditions across the study area.; Local incidence angle had the strongest relationship on SAR image backscatter values (r = -0.35) and when used in an empirical correction function reduced image variance by at least 8%. Backscatter modeling to separate the vegetation and soil components of the radar signal performed worse than expected, resulting in a weak correlation between composite sigmao soil and volumetric soil water content (r = 0.21) and different values for burned and unburned watersheds (r = 0.09 and r = 0.32, respectively). Soil backscatter values were estimated without accounting for canopy and litter layer moisture conditions, causing a reduction in the effectiveness of the cloud model output. The model performed very well, however, on a day basis with single date correlations for burned and unburned watersheds being among the highest yet reported when using radar satellite data. While many studies have questioned the sensitivity of C-band radars, operating at moderate incidence angles, to near surface soil moisture conditions, results here show that the ERS-2 data are capable of monitoring soil moisture conditions over even dense natural vegetation characteristic of tallgrass prairie.