Characteristic length scale of input data in distributed models: implications for modeling grid size

The appropriate spatial scale for a distributed energy balance model was in vestigated by: (a) determining the scale of variability associated with the remotely sensed and GIS-generated model input data; and (b) examining the effects of input data spatial aggregation on model response. The semi-vario gram and the characteristic length calculated from the spatial autocorrelat ion were used to determine the scale of variability of the remotely sensed and GIS-generated model input data. The data were collected from two hillsi des at Upper Sheep Creek, a sub-basin of the Reynolds Creek Experimental Wa tershed, in southwest Idaho. The data were analyzed in terms of the semivar iance: and the integral of the autocorrelation. The minimum characteristic length associated with the variability of the data used in the analysis was 15 m. Simulated and observed radiometric surface temperature fields at dif ferent spatial resolutions were compared. The correlation between agreement simulated and observed fields sharply declined after a 10 x 10 m(2) modeli ng grid size. A modeling grid size of about 10 x 10 m(2) was deemed to be t he best compromise to achieve: (a) reduction of computation time and the si ze of the support data; and (b) a reproduction of the observed radiometric surface temperature. (C) 2000 Elsevier Science B.V. All rights reserved.