|dc.description.abstract||In order to accurately model runoff with GIS software, areas that are capable of
contributing runoff to surface waters must be correctly identified. Identifying areas
capable of contributing runoff to streams is problematic in regions containing many
topographically-closed, internally-drained depressions and low relief over large
distances. Both of these complicating factors are present in the Upper Midwest of the
Traditional delineation methods fill sinks in the terrain under the assumption that most
are results of data errors. This assumption can cause inaccuracies when modeling
runoff in extensively internally-drained watersheds. Three delineation methods were
compared for ten Wisconsin and Minnesota watersheds: filled digital elevation models,
removing filled sinks from modeled area, and delineated potential contributing source
areas from an unfilled digital elevation model outward from an initial contributing area.
The delineations were used to model runoff in ArcMap using the Curve Number
equation. Runoff producing rain events were modeled from summers with total
precipitation within one standard deviation of the mean. Modeled runoff was compared
to USGS discharge data. Results from watersheds of varied sizes were compared using
normalized error, defined as the model error divided by the filled watershed area.
Normalized error was compared to drainage density, percent internally drained area,
and land cover types.
Results from ten watersheds indicate that models perform similarly for all delineations.
Within watersheds models perform differently for large storms and small storms, with
large storms defined as over 0.2 feet precipitation. Modeled storms under 0.2 feet tend
to underestimate runoff with increasing drainage density and with increasing area.
Large storms show no discernible differences when comparing normalized error to other