Overland flow networks are fundamental to both hydrologic model and water quality model predictions. Hydrologic models mostly use flow pathways to predict river hydrographs while water quality models depend on pathways to predict pollutant fate and transport. While hydrographs are considered insensitive to the various formulations of overland flow path algorithms, water quality models are sensitive (e.g. small changes in flow path formulation result in large changes in predicted pollutant loading) and it is therefore important to reconcile modeled and observed flow paths. Observed flow, which is often directed into paths that are transverse to the maximum slope by 0.1 to 1-m scale vegetative, soil, and terrain heterogeneities, is distinct from the 1 to 90-m pixel scale modeled runoff which simply routes flow according to terrain. Although various routing algorithm exist that allow for flow to pass into different combinations of downslope directions other than the path of steepest descent, there have been few studies that assess the accuracy of such flow networks. In this study we compared the predicted paths of the D8, DEMON, D-Infinity, and Multiple Flow algorithms, along with seven hybrid algorithms, with observed paths from two agricultural hillslopes in Princeton, New Jersey. Error matrices that identified "errors of omission" "errors of commission" as well as a K(hat)-based "overall accuracy" were used to rank the various routing algorithms.

The D8 algorithm, which constrains flow to a single path, had the highest omission error and three modifications to D8 that increased flow bifurcation significantly decreased these errors of omission while simultaneously increasing errors of commission. MF, which allows flow to divide into all eight lower neighbors, had the lowest omission error but also had the highest commission error. Modifications to MF that limited flow to fewer neighbors increased omission error while significantly reducing errors of commission. DEMON, which divides flow into two adjacent cardinal neighbors, had the second lowest omission error and the lowest commission error. D-Infinity, which is less dispersive than DEMON and divides flow into adjacent cardinal-diagonal pairs, had larger errors of omission and smaller errors of commission than DEMON. MF and D8, which respectively use the maximum and minimum available flow directions, were determined to have the lowest overall accuracy due to MF’s large commission error and D8’s large omission error. DEMON, D-Infinity the other hybrid flow algorithms utilizing a maximum of two-flow directions were ranked highest for overall accuracy and individual algorithm rankings varied with the terrain type. Simulation of overland flow paths in directions other than steepest descent appears a reasonable surrogate for simulating the observed flow redirection and bifurcation in overland runoff networks and is recommended for improving water quality model predictive accuracy.