GIS and Modeling in Ecological Studies: Analysis of Beaver Pond Impacts on Runoff and its Quality

Abstract

The ARC/INFO GRID module was used to derive watershed variables for input to AGNPS, a cell-based runoff model that estimates water volume, peak flow, eroded and delivered sediment, chemical oxygen demand, and nutrient export from watersheds (Young et al. 1987). The boundary of a 534 ha watershed in Voyageurs National Park was hand-digitized from 1:24,000 topographic maps, and used to clip elevation data from a 7~ minute U.S.G.S. Digital Elevation Model (DEM) with 30 m mesh-point spacing. ARC/INFO GRID was used to generate slope, slope shape, and field slope length for each of the 90x90 m cells used to subdivide the watershed. A surface runoff network was then generated using the FLOWDIRECTION, FLOWACCUMULATION, and STREAMLINE hydrologic modeling tools in ARC/INFO GRID. Each of the 90x90 m cells was uniquely numbered, and receiving cell numbers were derived for each source cell based on FLOWDIRECTION results. A 1:24,000 land cover map (Allen et al. 1993) was digitized and gridded to derive Manning's roughness coefficient, surface condition constant, and chemical oxygen demand factor for each cell. Detailed soil maps have never been made for the wilderness study site used, so land cover classes were coupled with information about soil series from nearby mapped sites to estimate soil texture, soil erodibility factor, and hydrologic group (to derive SCS curve numbers). The drainage area for each beaver impoundment in the watershed was derived from a digital database of subwatersheds. All variables were exported from ARC/INFO into the Microsoft Excel spreadsheet program, which was used to generate a data file in the appropriate format for AGNPS. The methodology was applied to the 534 ha, third order stream watershed to determine the influence of beaver ponds on water quality and quantity. Beavers influence runoff by: 1) constructing dams that retard the flow of water, 2) creating ponds that promote sediment deposition and increase phosphorus retention, and 3) changing forest land cover to water and wetland vegetation. We ran the model for a range of storms with average 24-hour rainfalls equivalent to a 1 yr, 2 yr, 5 yr, 10 yr, 25 yr, 50 yr, and 100 yr storm, based on National Weather Service records for the region. Model runs for the beaver-impounded landscape ("with ponds") were compared with those for the same watershed without the influence of beaver ("no ponds"), based on historical vegetation and adjacent forest types. The ''with ponds" scenario resulted in slightly increased water flow at the mouth of the watershed. This is because, assuming that the pond is full, 100% of the rain that falls onto it will flow off of it. This caused a 10% increase in runoff at the lowest rainfall intensity, but only a 1% difference during the 100 yr storm. The runoff contribution of individual cells changed relatively little between the two scenarios, but there were easily discernable differences in accumulated runoff per cell ii: with distance downstream. Sediment deposition in the beaver ponds also had an effect that accumulated downstream, so that the ''with ponds" scenario yielded 7 to 12% less sediment than the ''no ponds" scenario, an effect that increased with storm intensity. The model predicted a 4% decrease in watershed nitrogen output with ponds for a 1 yr storm, but there was no effect for storms with a 10 to 50 yr frequency, and a net increase for a 100 yr storm. This implies that while beaver ponds may retain N during low-intensity storms, there may be a flushing of that retained N during high-intensity storms. This pattern is visible on GIS maps for "with ponds" scenario with low intensity storm: higher nitrogen concentrations were observed at the locations, where no ponds were situated, and nitrogen content in runoff had remarkable alterations in cells adjacent to ponds. Chemical oxygen demand (COD) showed the largest effect of any of the parameters predicted: the presence of beaver ponds was associated with a 10 to 17% reduction of COD, depending on the storm intensity. This is because a forest has a lot more primary productivity than a pond, and therefore contributes more organic matter (and therefore COD) to the system. This model-based approach provided insight into the landscape scale influence of beaver ponds that could not have been derived using conventional field techniques. The modeling was done at a spatial level of detail that would have been impractical using manual data entry to AGNPS. Automating the derivation and interchange of variables with a GIS made this research possible.