SWAT-CS: Revision and testing of SWAT for Canadian Shield catchments

- SWAT-CS significantly improves daily streamflow model performance.
- Land cover and soil properties are transferable between sub-catchments.
- With SWAT-CS, efficient modeling of SWE and streamflow are positively correlated.
- Reservoir-based target releasing method is efficient in modeling natural lake.

SummaryCanadian Shield catchments are under increasing pressure from various types of development (e.g., mining and increased cottagers) and changing climate. Within the southern part of the Canadian Shield, catchments are generally characterized by shallow forested soils with high infiltration rates and low bedrock infiltration, generating little overland flow, and macropore and subsurface flow are important streamflow generation processes. Large numbers of wetlands and lakes are also key physiographic features, and snow-processes are critical to catchment modeling in this climate. We have revised the existing, publicly available SWAT (version 2009.10.1 Beta 3) to create SWAT-CS, a version representing hydrological processes dominating Canadian Shield catchments, where forest extends over Precambrian Shield bedrock. Prior to this study, very few studies applying SWAT to Canadian Shield catchments exist (we have found three). We tested SWAT-CS using the Harp Lake catchment dataset, an Ontario Ministry of Environment research station located in south-central Ontario. Simulations were evaluated against 30 years of observational data, including streamflow from six headwater sub-catchments (0.1-1.9 km2), outflow from Harp Lake (5.4 km2) and five years of weekly snow water equivalent (SWE). The best Nash-Sutcliffe efficiency (NSE) results for daily streamflow calibration, daily streamflow validation, and SWE were 0.60, 0.65, and 0.87, respectively, for sub-catchment HP4 (with detailed land use and soil data). For this range of catchment scales, land cover and soil properties were found to be transferable across sub-catchments with similar physiographic features, namely streamflow from the remaining five sub-catchments could be modeled well using sub-catchment HP4 parameterization. The Harp Lake outflow was well modeled using the existing reservoir-based target release method, generating NSEs of 0.72 and 0.67 for calibration and verification periods respectively. With significant changes to the infiltration module (introducing macropore flow and reduced bedrock percolation), more than 90% of interflow was generated close to the soil-bedrock interface and the contribution of groundwater flow to total runoff was reduced to small amounts, consistent with hydrological process understanding in this terrain. These two changes also allowed for a positive linear relationship between NSE of SWE and Q, whereas prior to these changes there was a negative relationship. With these key revisions to the infiltration and bedrock percolations modules, it is concluded that SWAT-CS can reasonably capture key hydrological processes within Canadian Shield catchments. Further testing will examine water quality modeling and larger-scale applications.