Northern Prairie Wildlife Research Center
The effects of watershed drainage area, stream length, and average annual discharge on species richness of streams were examined by simple regression analyses. Data were collected from 26 streams. The Red River main stem was omitted because it was a severe outlier due to high discharges. Species richness included all native species reported from mouth to headwater reaches. Watershed drainage areas (mi2) to the stream mouth were obtained from Simons and King (1922), Maclay et al. (1972), and U.S. Army Corps of Engineers (1980). Total length of each stream was obtained from USGS 1:250,000 scale quadrangle maps digitized using a Summagraphics Microgrid II digitizing pad and TOSCA (1993) software. The digitized streams (line vectors) were imported into an IDRISI (1993) geographic information system (GIS) where they were converted to raster format. To determine stream lengths, the number of pixels making up each stream was counted and compared to that of a rasterized line of known length. Total length (mi) of each stream included that of the main stem and all non-intermittent lower-order tributaries. Average annual discharge (cfs) of streams was obtained from Water Resources Data for Minnesota (USGS 1972a-1994a) and for North Dakota (USGS 1972b-1994b) and from Surface Water Supply Papers (USGS 1953, 1956, 1958, 1960, 1971, and 1976c) for sites near the mouth of each stream.
Tests of the hypotheses that the four variables were random samples from normal distributions were conducted using the PROC UNIVARIATE procedure (SAS Institute 1990a). Only the variable species richness initially passed the test, Shapiro-Wilk statistic W=0.93 (P<W 0.09); and the distribution of the variables watershed area, stream length, and annual discharge was approximately normal after log-transformation (W>0.94, P<W 0.15 or greater). Regression analyses were conducted using the PROC REG procedure (SAS Institute 1990b), with species richness entered as the dependent variable and the log-transformed variables each separately entered as independent. Significance of linear models produced was determined by ANOVA F-values and associated probabilities.