The influence of spatial resolution on the estimation of land surface heat fluxes from remote sensing is poorly understood. In this study, the effects of aggregation from fine (< 100 m) to medium (approx. 1 km) scales are investigated using high resolution Landsat 5 overpasses. A temporal sequence of satellite imagery and needed meteorological data were collected over an agricultural region, capturing distinct variations in crop stage and phenology. Here, we investigate both the impact of aggregating the input forcing and of aggregating the derived latent heat flux. In the input aggregation scenario, the resolution of the Landsat based radiance data was increased incrementally from 120 m to 960 m, with the land surface temperature calculated at each specific resolution. Reflectance based land surface parameters such as vegetation height and leaf area index were first calculated at the native 30 m Landsat resolution and then aggregated to multiple spatial scales. Using these data and associated meteorological forcing, surface heat fluxes were calculated at each distinct resolution using the Surface Energy Balance System (SEBS) model. Results indicate that aggregation of input forcing using a simple averaging method has limited effect on the land surface temperature and available energy, but can reduce evapotranspiration estimates at the image scale by up to 15%, and at the pixel scale by up to 50%. It was determined that the predominant reason for the latent heat flux reduction in SEBS was a decrease in the aerodynamic resistance at coarser resolutions, which originates from a change in the roughness length parameters of the land surface due to the aggregation. In addition, the magnitude of errors in surface heat flux estimation due to input aggregation was observed to be a function of the heterogeneity of the land surface and evaporative elements. In examining the response of flux aggregation, fine resolution (120 m) heat fluxes were aggregated to coarser resolutions using a range of common spatial interpolation algorithms. Results illustrate that a simple averaging scheme provides the best choice for flux aggregation compared to other approaches such as nearest neighbour, bilinear interpolation or bicubic interpolation, as it not only preserves the spatial distribution of evapotranspiration, but most importantly also conserves the mass balance of evaporated water across pixel and image scales.
LandsatMODISSurface Energy Balance System (SEBS)Flux aggregationInput aggregationUpscalingLand surface temperatureUncertaintyRoughnessAerodynamic resistance