Abstracts

Presented by Michael G. Anderson - Email: m_anderson@ducks.ca

Many waterfowl surveys in the prairie pothole region were established along grid roads, generally spaced a mile apart, where indicated pairs were counted for some set distance on either side of the roadway. Often this comprised a strip 0.25 miles wide. Depending on the survey, these estimates were made by observers on foot, in vehicles, or low-flying aircraft. A necessary assumption when extrapolating such estimates to larger landscapes is that pair densities encountered along these strips are representative of entire landscapes. In perched moraines like the area near Minnedosa, MB, where drains usually cannot convey water off farms, drainage instead tends to consolidate surface water along edges of fields (e.g., road ditches) or in larger interior wetlands. We used annually replicated walking surveys for canvasbacks and redheads designed to view every single wetland, on two 6-square-mile study areas from 1983-1990, to assess whether proximity to roads affected estimates of pair densities. Other variables evaluated were a wetness index, species, study block, overall pair densities and plausible interactions. Year was included as a random effect. Poisson mixed effects models were used for modeling density of indicated pairs. The best approximating model included location (near-road vs. interior), pair density, species and location x wetness and location x density interactions. Model-estimated pair counts were higher in near-road locations for all circumstances in our study. These differences were greatest when pair density and wetness index or both were lower. An important next step would be to assess temporal changes in the distribution of wetlands relative to roads, and therefore the possibility of directional temporal bias. For Minnedosa, there are aerial photos of some sections from 1948 through the 1990s, but additional sites should be sought across the Canadian and U.S. prairies.