Maximizing spatial congruence of observed and DEM-delineated overland flow networks

Model simulated overland flow is traditionally routed by the terrain-based algorithms that read 90 to 10-m pixel data, yet observed flow is often influenced by sub-meter micro-terrain features. While sub-meter terrain maps are not available for most watersheds, routing algorithms that bifurcate flow through multiple pixel boundaries may function to match observed runoff patterns. This paper assesses the spatial congruence between observed overland flow from two agricultural hillslopes in New Jersey with 30-m pixel flowpaths predicted by five common routing algorithms known as D8. Multiple Flow, 2D-Lea, (a building block in DEMON), 2D-Jensen, and D-Infinity. The lowest congruence ratings were assigned to D8, which constrained flow to a single neighbour, and MF, which bifurcated flow into all neighbours lower in elevation. 2D-Lea and D-Infinity, algorithms that bifurcated flow to a maximum of two neighbours, achieved the highest accuracy rankings. Simple algorithm modifications that increased D8 bifurcation, and constrained MF bifurcation, resulted in congruence rankings that rivaled the more sophisticated 2D-Lea. 2D-Jensen, and D-Infinity results. These experiments support the use of flow bifurcation schemes that pass flow into a maximum of between two and three pixels when routing polluted runoff.