Downscaling 1-km Topographic Index Distributions to a Finer Resolution for the TOPMODEL-Based GCM Hydrological Modeling

Abstract

TOPMODEL predictions of surface runoff and subsurface flow are fundamentally developed on the basis of the topographic index distribution (TID). The scale dependency of the TID [i.e., dependency on the resolution of the digital elevation model (DEM) data used to compute the topographic indexes] determines that downscaling of the TID computed from a coarser resolution DEM to a finer resolution is needed before the TOPMODEL concepts can be applied to simulate hydrological processes at some larger scales than the scale of hillslopes. It was found that adjusting only the mean values cannot achieve an accurate downscaling of the TID because the difference between 2-m TIDs and the downscaled TIDs from a coarser resolution to 2 m through adjusting only mean values resulted in overestimation of the fraction of the saturation area and surface runoff under wet conditions and underestimation under dry conditions. It was found that downscaling by correcting for scale-dependencies in the first three moments of TIDs produced better predictions. A series of empirical relationships among mean, standard deviation, and coefficient of skewness of TIDs of nine catchments in eastern Tennessee at resolutions of 2, 10, and 100 m and 205 watersheds across the contiguous United States at resolutions of 10 m and 1 km were developed for downscaling TIDs from 1 km to 2 m through approximating TIDs by a 3-parameter gamma distribution function. The errors in the downscaled TIDs from 1 km to 10 m over 205 watersheds across the contiguous United States decreased with increasing watershed size and approached a minimum (approximately 6%) as the watershed drainage area was larger than approximately