Validating remotely sensed land surface fluxes in heterogeneous terrain with large aperture scintillometry

Abstract

The Large Aperture Scintillometer (LAS) has emerged as one of the best tools for quantifying areal averaged fluxes over heterogeneous land surfaces. This is particularly useful as a validation of surface energy fluxes derived from satellite sources. We examine how changes in surface source area contributing to the scintillometer and eddy covariance measurements relate to satellite derived estimates of sensible heat flux. Field data were collected on the Konza Prairie in Northeastern Kansas, included data from two eddy covariance towers: one located on an upland, relatively flat homogeneous area, and the second located in a lowland area with generally higher biomass and moisture conditions. The large aperture scintillometer spanned both the upland and lowland areas and operated with a path length of approximately 1 km specifically to compare to Moderate Resolution Imaging Spectroradiometer (MODIS) derived estimates of surface fluxes. The upland station compares well with the LAS (correlation of 0.96), with the lowland station being slightly worse (correlation of 0.84). Data from the MODIS sensor was used to compute surface fluxes using the ‘triangle’ method which combines the remotely sensed data with a soil-vegetation-atmosphere-transfer scheme and a fully developed atmospheric boundary layer model. The relative contribution to the surface observations is estimated using a simple footprint model. As wind direction varies, the relative contribution of upland and lowland sources contributing to the LAS measurements varies while the MODIS pixel contribution remains relatively constant. With the footprint model, we were able to evaluate the relationship between the LAS observations and the remotely sensed estimates of the surface energy balance. The MODIS derived sensible heat flux values correspond better to the LAS measurements (percentage error: 0.04) when there was a larger footprint compared to a time with a smaller footprint (percentage error: −0.13). Results indicate that the larger the footprint, the better the agreement between satellite and surface observations.