U.S. Environmental Protection Agency


Date of this Version



Published in Journal of Hydrology 309 (2005) 258–276.


To test a conceptual model of non-linear response of hydrologic regimes to watershed characteristics, we selected 48 secondand third-order study sites on the North and South Shores of western Lake Superior, MN(USA) using a random-stratified design based on hydrogeomorphic region, fraction mature forest, and fraction watershed storage (lakeCwetland area/watershed area). We calculated several commonly used hydrologic indices from discharge and velocity estimates, including daily flow indices, overall flood indices, low flow variables, and ratios or ranges of flow percentiles reflecting the nature of cumulative frequency distributions. Four principal components (PCs) explained 85.9 and 88.6% of the variation of flow metrics among second- and third-order stream sites, respectively. Axes of variation corresponded to a runoff vs. baseflow axis, flow variability, mean flow, and contrasts between flood duration and frequency. Analysis of velocity metrics for third-order streams yielded four PCs corresponding to mean or maximum velocity, Froude number, and inferred shear velocity, as well as spate frequencies vs. intervals associated with different velocity ranges.

Using discriminant function analysis, we could discriminate among watershed classes based on region, mature forest, or watershed storage as a function of flow metrics. For second-order streams, median flow (Qs50) increased as watershed storage increased. North Shore streams showed a more skewed distribution and greater spread of discharge values than did South Shore streams for both stream orders, while third-order North Shore streams exhibited a higher frequency of spates. Independent of regional differences, loss of mature forest increased the range of variation between baseflow and peak flows, and depressed baseflow. Consistent with our initial model for watershed classification, Classification and Regression Tree (CART) analysis confirmed significant thresholds of change in flow metrics averaging between 0.506 and 0.636 for fraction mature forest and between 0.180 and 0.258 for fraction watershed storage.