Using Landscape Information Approaches to Increase Duck Recruitment in the Prairie Pothole Region

Study Area and Methods

Wetland Data

Digital data for wetlands classified as part of the National Wetland Inventory (NWI) (Cowardin et al. 1979) were obtained from the U.S. Fish and Wildlife Service, NWI Office, Saint Petersburg, Florida, for Stutsman and Wells counties in North Dakota. Duck pair data, used for constructing our pair/wetland regression models, were derived from surveys where duck pairs were counted on a sample of wetland basins of various classes with class equal to the deepest water regime of all wetland polygons in the basin, as described by Cowardin et al. (1995). Therefore, prior to applying the model to the wetlands in our study area, we converted NWI digital wetland data into a basin class coverage using the same classification scheme from which our model data were derived. The classes were temporary, seasonal, semi-permanent, lake and riverine. Basically, this process resulted in a classification similar to that described by Stewart and Kantrud (1971) with an additional class—riverine.

Duck Pair Estimates

Basin wetlands. Breeding duck pairs associated with each wetland basin in our study area were estimated using regression models for five upland-nesting duck species: blue-winged teal (Anas discors), gadwall (A. strepera), mallard (A. platyrhynchos), northern pintail (A. acuta) and northern shoveler (A. clypeata). Baseline regression equations were constructed from survey data collected at the Arrowwood Wetland Management District, North Dakota, during 1982 through 1984 (Cowardin et al. 1995), and included our two-county study area. The form of the equations was:

These equations were adjusted to account for variation among areas and years. This adjustment (Y [gamma]) was derived from waterfowl surveys in which 2,200 to 2,800 sample wetland basins were visited each year in May, 1987 to 1993. During these surveys, the area covered by water in each basin was estimated, and the number of duck pairs and indicated pairs was determined using techniques described by Hammond (1969) and Dzubin (1969). For a more thorough discussion of the field procedure and sample basin selection see Cowardin et al. (1995). Y is calculated as the ratio of pairs counted during field surveys to pairs predicted from the base regression (Cowardin et al. 1995). The adjustment (Y) was calculated for each of 14 geographical districts (Wetland Management Districts) in the PPJV area by averaging values from data collected in each district during the period 1987 through 1993. Due to a phase-in period of the survey and persistent drought in some areas, data were not consistently available for all years in all districts. We assumed that data available for each district reflected long-term average values for that district. Y was found to vary with geographical location (universal transverse mercator [UTM] coordinates), as measured from the geographical center of each district, for gadwall (F_{2, 11} = 19.06, P = 0.001), mallard (F_{1, 12} = 18.27, P = 0.001) and northern pintail (F_{1, 12} = 18.22, P = 0.001). The Y values used in our pair models for these species were modeled using the UTM coordinates for the center of each wetland basin in our study area. For blue-winged teal and northern shoveler, Y did not vary with location and was calculated as the average of the 14 district values. Because of temporal and seasonal variation in precipitation and evapotranspiration rates, some wetland basins do not pond water during May in all years. This is particularly true for shallow basins that have a water regime other than "lake" in our basin class scheme. Therefore, to prevent overestimating duck pair numbers for these shallow basins, we adjusted the pair estimates based on the proportion of basins for each type that would be expected to pond water in an average year, as described by Cowardin et al. (1988).

Riverine wetlands. Duck pair data for the riverine wetland class were derived from surveys conducted in the PPJV area of North Dakota during May, 1983 to 1986. Survey procedures followed those described by Hammond (1969). Pair counts were conducted on 338 stream sample miles (544 km) (excluding Missouri River) during the period 1983 through 1986 and results were translated into pairs per acre of riverine wetland class using NWI digital wetland data. The model for riverine wetlands took the form:

Procedure

All data processing, analyses and cartographic output were completed using Arc/lnfo GIS software (Environmental Systems Research Institute, Redlands, California) on UNIX workstations.

The first step in the process involved using the previously discussed models for the five species to calculate a breeding duck pair value for each wetland basin in the digital database for our study area. This step estimated the average number of duck pairs that would be expected to settle in each wetland basin in May. Because these species nest primarily in uplands in our study area (Klett et al. 1988), we needed to determine the number of hens available to different landscape units based on the number that settled in the surrounding wetlands. This required knowledge of the distance hens would potentially travel from their primary core wetland (Poston 1974) to nesting cover. The distances used in our models and the source of information are presented in Table 2.

Next, we divided the study area into arbitrarily sized units 1,320 by 1,320 feet (402 x 402 m). For each unit, we summed the number of duck pairs that was estimated to occupy the wetlands within a specific distance (buffer) from the center of each unit. The buffer distance for each species equaled the travel distance presented in Table 2. Pair values for wetlands that were only partially within the buffer area were prorated based on the proportion of the wetland inside the buffer.

The process provided an estimate of duck pairs per square mile (pairs/km² x 2.59) within the buffer area of each unit in the study area. These values were translated into management priority levels and were displayed as a map for our study area. Finally, we obtained a Public Land Survey System digital file for our study area from the U.S. Geological Survey and overlaid the township and section (not displayed in Figure 1) boundaries to facilitate locating the position of units within the study area.

Results

Figure 1 shows the results of our model procedure for Stutsman and Wells counties, North Dakota. Each shade on the map represents a range of breeding duck pair density values scaled as a percentile of the land area. We reiterate that the value assigned to each unit does not represent the duck pairs modeled for the wetland basins in that unit, but rather the modeled pairs for the wetlands in that unit and those within a specific buffer distance of that unit. Additionally, because buffer zones of one unit overlap those of other nearby units, each unit cannot be considered independent of neighboring units.

Figure 1. Map of Wells and Stutsman counties in North Dakota that shows areas of different management priority based on breeding duck density for upland nesting ducks. Example "A" and "B" are circled.

Management Application

The primary purpose of the duck pair density map is to allow managers to identify "high priority" sites where the most potential benefits can be achieved from treatments designed to reduce predation on nesting hens and their nests. The type of treatment applied to a site requires additional information that must be obtained from other sources. A summary of treatments, and their application guidelines, identified as useful in the PPJV area are outlined in the U.S. Prairie Pothole Joint Venture Implementation Plan Update (Anonymous 1995). Two examples of using the pair density map for management decisions are given below.

Example 1. On the map in Figure 1, area "A" in Stutsman County, North Dakota is characterized by a large group of high-priority units, mostly near the center of the circle. A visit to the area showed that the land cover in this area is dominated by native grassland used for cattle production. Because grassland provides high nest security compared with cropland (Klett et al. 1988), treatments designed to prevent the conversion of this land to cropland are appropriate. One such treatment is a grassland easement under a program currently administered by the U.S. Fish and Wildlife Service (FWS). Under the terms of the easement agreement, the landowner receives a one-time payment that is determined as a percentage of the land value. In exchange, the landowner agrees to an easement encumbrance on the land with the principal condition that the land must remain as grassland for the duration of the contract period (in many cases, perpetuity). Another treatment, currently being supported by the FWS and Ducks Unlimited, Inc. in the PPJV, is a program to assist landowners in establishing rotational grazing systems on grassland that previously was grazed continuously during the growing season. Rotational grazing systems are intended to improve the structure of the grass cover, thus providing increased forage for cattle and nesting cover for ducks. Additionally, by increasing the profitability of cattle production, grazing systems may help prevent conversion of grassland to alternative uses such as grain production.

Example 2. Land use in area "B" (Figure 1), Stutsman County, North Dakota, is characterized by cropland used intensively for small grain production. Cropland is the least productive of all the major cover types for ducks nesting in the area (Cowardin et al. 1985, Klett et al. 1988). Since 1985, numerous cropped fields in this area have been converted to undisturbed perennial cover through the U.S. Department of Agriculture's Conservation Reserve Program (CRP). Some of these CRP fields occur in the high-priority units shown in Areas "B." Recent studies have demonstrated that CRP cover is attractive to nesting ducks and provides relative security from nest predators (Kantrud 1993, Reynolds et al. 1994). CRP fields that occur on high-priority units could receive priority consideration for extending existing CRP contracts and accepting additional conservation bids under future farm programs. Other programs, such as predator enclosures or predator removal, also would be appropriate in landscapes such as area "B," especially if grass cover is limited.

Discussion

Low duck recruitment in the prairie pothole region of the United States and Canada is the result of excessive predation associated with extensive landscape changes due to agriculture and other development. Improving this landscape to benefit duck production is a large, complex and costly task. Managers charged with this task in the prairie pothole region are faced with the difficult problem of applying a variety of management strategies to an almost unlimited number of situations affected by political, social, economic, geographic and ecological constraints. Not all the landscape can be treated. Most often, opportunity (a favorable juncture of circumstances) dictates the type and location of treatment application. Such an approach can lead to less than maximum benefits from the treatment. Many of the strategies needed to improve duck production in developed agricultural areas are intensive in nature. That is, they require considerable amounts of money and manpower to develop and maintain (e.g., predator barriers). Therefore, to maximize cost effectiveness, appropriate management practices should be targeted to areas where they can benefit the most breeding ducks.

We offer a tool that waterfowl managers can use to ensure that habitat treatments are applied to the most appropriate part of the landscape. We link breeding duck populations to the landscape by applying duck pair/wetland relationships and breeding duck home range characteristics to digital wetland data using GIS technology. By identifying areas of different breeding duck densities, managers can apply treatments to areas where they will potentially benefit the largest numbers of breeding ducks. For example, guidelines for electric fence enclosures (designed to reduce predation on nests and nesting hens) call for placing this treatment in areas with relatively high breeding duck densities and low amounts of competing nesting cover outside of the fence. Such guidelines are intended to increase the probability that the relatively expensive tract of fenced cover is used by high numbers of nesting hens. Our maps allow managers to find ureas of relatively high duck densities easily and then, through field visits or other habitat information, select specific sites with little competing nesting cover. Our maps also provide a very useful mechanism for targeting both wetland and grassland protection (but not restoration) efforts to provide the greatest benefits to breeding ducks. Maps currently are available for all of the prairie pothole region of South Dakota and North Dakota. Within six months, we intend to complete the mapping for North and South Dakota. When digital wetland data become available from NWI for eastern Montana, we will apply the process there as well.

The described process only considers ducks in the Prairie Pothole Joint Venture area. We believe landscape changes affecting breeding ducks in this area also affect many other grassland and wetland-dependent species. With adequate data, similar processes could be used to target management for other species. GIS modeling of densities of other species would help managers to pool resources from multiple sources to maximize benefits to a variety of species throughout the PPR. We invite other agencies and organizations to use our maps to help target their programs. In the future, we plan to develop a digital coverage of upland habitats in the prairie pothole region. Such an upland habitat layer, when combined with the pair density map presented here, would provide an even more valuable tool for prioritizing landscapes for protection and enhancement.

Acknowledgments

We thank M.F. Murphy and L.J. Shaffer, National Wetland Inventory, for providing digital data for wetlands. We thank R.R. Johnson, South Dakota State University; K.F. Higgins, National Biological Service; and M.R. Luther, North Dakota Geological Survey, for assistance in processing digital wetland data. Partial funding for processing digital wetland data was provided by S.A. Peterson, Environmental Protection Agency. T.L. Shaffer and L.M. Cowardin assisted with model development. Special thanks is extended to M.D. Koneff for processing the data used to model our study area. A.C. Schaff typed the manuscript. We thank L.L. Strong, T.L. Shaffer, K.R. Willis and R.A. Warhurst for reviewing the manuscript.