Modeling the Natural Drainage Network of the Grand River in Southern Ontario: Agriculture May Increase Total Channel Length of Low-Order Streams

The Grand River watershed is an important agricultural area in southern Ontario, with several large and growing municipalities. Based on digital elevation models (DEMs), the natural drainage network was modelled to predict flow paths. Channel lengths and locations of the predicted network were compared with a ground-truthed channel network to determine efficacy of the models. Approximately 5% of predicted channels lay >40 m from actual channel locations. This amounted to 388 km of channel that had no corresponding channels in reality. The model was unable to predict, based on topography, 2535 km of actual channel present in the watershed. Channels not anticipated by topography were mostly first-order, with low sinuosity, were most common in areas with high agricultural land use, and are likely excavated extensions to headwater streams to facilitate drainage. In addition, this study showed that Soil and Water Assessment Tool (SWAT) models produced using different DEM resolutions did not predict significantly different stream flows, even when resolution was as low as 200 m. However, these low resolution DEMs did result in under-prediction of sediment export entering Lake Erie, most likely because the low resolution maps failed to account for small localized areas of high slope that would have relatively higher rates of erosion