Modeling large-scale fluvial erosion in geographic information systems

Variants of the stream power model have become standard for large-scale erosion modeling in geographic information systems (GIS) because they can be applied over broad areas without the need for detailed knowledge of stream characteristics. GIS-based implementations of the shear stress, stream power per unit length and stream power per unit area models are closely related to one another and related also to empirical sediment yield models derived from continental-scale factor analyses. Based on a detailed examination of the implementation of stream power analyses at the scale of continental mountain ranges, we demonstrate that: (1) the careful selection of a digital elevation model (DEM) projection can minimize length and area distortion when analyzing large portions of the earth (such as the Himalaya or Andes) in the two-dimensional plane of a DEM. (2) The area-discharge proxy frequently employed in GIS-based stream power studies may not be appropriate for rivers that flow through significant rain shadows or climatic zones. (3) Decreasing the resolution of a DEM from 30- to the 900-m typical for studies of large extent decreased the mean slopes of 15 rivers in the Olympic mountains by 65%, increased the mean drainage basin size by 14%, and caused a 17% reduction in median main-stem channel length. (4) The coefficients k, m and n common to different versions of the Stream Power Law are themselves sensitive to grid resolution when determined from an analysis of area–slope plots. (5) Stream power per unit area decreased in the Olympics mountains as grid resolution decreased.