USGS Home
Contact USGS
Search USGS

We stratified the PPR into four regions based on wetland density. A high-density region approximately coincides with the Missouri Coteau. This is a large belt of glacial moraine generally trending northwest to southeast across the Dakotas and characterized by collapsed hummocky topography or dead-ice moraine (Bluemle, 1977). A medium-density region approximately coincides with the drift plain, an area of glacial drift with less relief than, and lying east of, the Missouri Coteau. We recognized two low-density regions. The first is the Red River Valley of eastern North Dakota and western Minnesota. The Red River Valley is not a valley, but mostly the bed of glacial Lake Agassiz. The second is southern Minnesota, an area physiographically similar to the drift plain region except that most basin wetlands are drained.

We selected wetlands under the assumption that those in cropland-dominated watersheds would be in a more degraded state than those in grassland-dominated watersheds. We then tested indicators by their discriminate between wetlands at the extremes of that proxy. Therefore, we used the ratio of area of cropland to total upland on plots as a proxy for wetland quality. We selected wetlands from an existing large sample (n = 422) of 10.4-km² (4 mi²) plots used for a mallard simulation model (Cowardin et al. 1988). We mapped wetlands on each plot according to the classification Cowardin et al. (1979), and the uplands according to a simplified classification that included grassland and cropland (Cowardin et al., 1988).

We calculated the ratio of area of cropland to total area of upland on each plot. Plots with the largest ratios (maximum amounts of cropland) and smallest ratios (maximum amounts of grassland) represented poor-and good-quality watersheds, respectively. From plots within each of the four regions (Figure 1), we selected four, the two with the largest and the two with the smallest ratios of cropland.

Figure 1. Prairie pothole region of Minnesota and the Dakotas showing strata based on wetland density and location of 10.4-km2(4mi2) plots containing study wetlands.

We further constrained plot selection by the rule that each pair of plots must contain at least two temporary, two seasonal, and two semipermanent wetland basins (Stewart & Kantrud,1971). These correspond to palustrine emergent temporarily flooded, palustrine emergent seasonally flooded, and palustrine emergent semipermanently flooded wetlands of Cowardin et al. (1979). These basin wetlands are identified on National Inventory (NWI) maps and are the most common wetlands in the PPR. We maintained separation between landscapes of high and low cropland:upland ratios in all regions except the Red River Valley where there were few available plots and few semipermanent wetlands basins. In that region it was impossible to maintain a meaningful separation, so we dropped the requirement of having semipermanent wetland basins.

We modified the sample because of refusal of access by several landowners. When refusal resulted in inability to obtain two wetland basins of each class in each extreme pair of plots, we selected the next most extreme plot, redrew wetland basins and again contacted landowners. We repeated this procedure using our original constraints until the sample was complete. We further altered the process at the end of the 1992 field season because drought and access denials resulted in almost no data from the low-wetland-density strata. We dropped those strata from the sample and selected two new samples from the high and the medium wetland density strata for study the following year.

We classified wetland basins (Cowardin, 1982) using a modification of Stewart & Kantrud's (1971) classification. We first grouped all wetland polygons (mapping units used by NWI) classified in the data set according to Cowardin et al. (1979) into basins. We then used the polygon with the most permanent water regime to determine the Stewart & Kantrud (1971) class for the group of polygons included in the basin. We assigned a random number to each wetland basin in each wetland class in each region. We prepared a sorted list of these random numbers for the pairs of plots representing good and poor quality. We selected the top two wetlands in each wetland basin class if they met the following constraints: (1) were not mapped as linear or point features (mostly drainage ditches and dugouts constructed for watering livestock), (2) did not contain dugouts or lacustrine wetland (permanently flooded lakes), or (3) were not temporary or seasonal wetlands that overlapped the plot boundary. This procedure biases the selection process against large temporary and seasonal basins but was unavoidable because we could not classify multi-polygon basins outside the plot boundary. We included semipermanent basins that were partly within the plot because we knew the basin class by default. Exclusion of these wetlands would bias the sample against large basins. We also rejected all basins in good-quality plots surrounded by cropland and basins in poor-quality plots surrounded by upland other than cropland.

A few landowners rescinded permission after fieldwork began. In these cases, we gathered remaining data from the nearest available wetland basin of the same class. For purposes of analysis, we redefined watershed quality for some wetlands based on the cropland:upland ratio for the individual basins. This was necessary because of classification errors in the original data used in the selection process and major landscape changes. All plots and basins were in Minnesota, North Dakota, or South Dakota. Further details on the plot and basin selection process are in "Report on pilot test of indicators of wetland condition" (1994; unpublished draft) available from U.S. Environmental Protection Agency, Environmental Research Laboratory, 200 S.W. 35th St., Corvallis, OR 97333.