Sensitivity of a regional evapotranspiration model to the spatial averaging of parameters

Date of Completion

January 1996


Agriculture, Forestry and Wildlife|Physics, Atmospheric Science|Remote Sensing




The purpose of this research was to quantify the statistical relationship between spatial resolution error in a regional surface flux model and surface parameter heterogeneity. Three studies were carried out. In the first, resolution error in albedo, canopy height, canopy resistance, Normalized Difference Vegetation Index (NDVl) and calculated evapotranspiration was quantified for a site in Ellington, Connecticut. These values corresponded well with errors predicted by an equation based on spatial autocorrelations. In the second study, the effect of surface cover patch size and patch contrast on resolution error was investigated using six artificial landscapes. The error prediction equation appeared sensitive to these scene-dependent components of spatial heterogeneity. In the third study, the evapotranspiration model's relative sensitivity to the spatial averaging error in its input parameters was quantified using four methods: (1) simple algebraic test, (2) varying parameter resolutions, (3) predicting sensitivity using partial derivatives and (4) analysis of probable error vectors. Error vector results were presented using a color compositing data visualization technique. All methods indicated that error in canopy resistance was mainly responsible for resolution error in calculated ET. The scene-dependence of the model's sensitivity appears to be driven by thematic misclassification probabilities and potential parameter error vectors. ^