Effect and uncertainty of digital elevation model spatial resolutions on predicting the topographical factor for soil loss estimation

ABSTRACT:

Soil erosion is very sensitive to the topographical factor LS (as a product of slope length L and steepness S) in the Revised Universal Soil Loss Equation (RUSLE). Improving prediction of LS by assessing uncertainty is thus very important. In this study, digital elevation models (DEMs) at different spatial resolutions obtained by interpolation were used to derive the slope and the up-slope contributing area required in a physically based LS equation and to obtain LS maps. The effect of spatial resolution in predicting LS was investigated by comparing the maps for overall differences, spatial distribution, and spatial variability of each estimated variable. Spatial error budgets were generated for LS by modeling uncertainty propagation from slope, up-slope contributing area, and model parameters with a variance partitioning method. The results showed that the uncertainty in predicting LS came mainly from slope in gentle areas and from up-slope contributing area in steep areas. The effect of spatial resolution for LS was primarily explained by uncertainty propagation from up-slope contributing area. The coarse resolutions led to extremely large predicted values and variances of up-slope contributing area, hence large uncertainty in LS. The interpolation of a DEM into finer resolution provides more spatial information without degrading elevation accuracy, resulting in a rapid decrease of variance for predicting up-slope contributing area and LS. For the case presented in this study, a DEM with a lower resolution than 5 m (16.4 ft) was considered useless for predicting LS due to large variances from up-slope contributing areas.