Northern Prairie Wildlife Research Center
Both the number of May ponds and the index of conserved soil moisture varied dramatically during the 25-year period of study, but the CSM index less so (Fig. 1). Drought was especially pronounced during 1977, 1980-1981, and 1989. May ponds and CSM were strongly correlated (r = 0.67, P < 0.001; Table 1). The number of May ponds was higher in years with cool April and May temperatures and wet Aprils; the CSM index was related to weat manner, but less strongly (Table I).
Average temperatures during April and May tends study period (Fig. 2). Temperatures in the two months were highly correlated (r = 0.67, P<0.001; Table 1),indicating the persistence of seasonal temperature patterns within years. On the other hand, monthly precipitation totals were very erratic (Fig. 3). Unlike average temperatures, precipitation totals in April and in May were not significantly correlated, indicating that precipitation patterns were less enduring than temperature patterns.
Seven dabbling duck species, three diving duck species, and one stiff-tailed duck were common in the Woodworth Study Area (Table II). In decreasing order of abundance, these were blue-winged teal (making up an average of 4396 of the total waterfowl population), gadwall (12.9%), mallard (10.6%), northern shoveler (6.2%), lesser scaup (6.0%), northern pintail (5.6%), redhead (5.2%), ruddy duck (5.0%), canvasback (2.0%), American wigeon (1.9%), and greenwinged teal (1.6%). Gadwall and mallard populations were the most consistent in size, as measured by the coefficient of variation; most variable in numbers were the canvasback, ruddy duck, and redhead (Table 2).
Interactions among the species
Among dabbling ducks in the Woodworth Study Area, populations of the different species generally positively correlated with each other (11 of 21 pairwise comparisons significant; Table 3). Numbers of gadwalls and northern pintails especially were associated with other dabbling duck species, while at the other extreme, numbers of mallards were significantly related to those of only one other species, the gadwall. The three diving ducks plus ruddy ducks also tended to covary together (5/6 pairwise comparisons significant; Table III). Numbers of dabbling ducks and diving ducks tended to vary inversely, but just 3 pairwise comparisons were significant (with one positive), and x2 = 2.33, P > 0.05 that the number of negative correlations (14/21) was higher than that expected by chance (10.5/21). It seems that the two groups varied independently.
Correlations between residuals from the best-fitting model for each species and counts of the other species were more frequently positive than negative (Table 4). Significant positive correlations (P < 0.10) with at least one other species were calculated for all modeled species except blue-winged teal. Some species pairs were symmetric in their significance; for example, residuals from the model for mallards were positively related to gadwall numbers, and residuals from the gadwall model were positively related to mallard counts. Other such species pairs with symmetric positive correlations were gadwall-American wigeon, gadwall-northern pintail, American wigeon-northern pintail, northern shoveler-redhead, northem shoveler-canvasback, redhead-ruddy duck, and canvasback-lesser scaup.
The only species pair with symmetrical negative correlation coefficients involved the blue-winged teal and lesser scaup. Other negative correlations were scattered, except that residuals from the model for canvasback were negatively related to counts of three species: gadwall, American wigeon, and green-winged teal.
Species relations to explanatory variables
The populations of these duck species have fluctuated at the continent-wide level during the past quarter-century (Figs. 4-14), presumably in response to a variety of factors that include varying conditions of the wetland habitat, changes in land use, predator populations, hunting pressure, and possibly others. Correlations between numbers and calendar year indicate more-or-less linear trends over the period 1965-1989 that are negative at both local (Woodworth) and continental levels for all seven species of dabbling ducks except green-winged teal at the continental scale. Of these trends, decreases were significant at the local scale for green-winged teal, American wigeon, northern pintail, and gadwall(Table 2). For mallard, the trend was significant at only the continental scale, whereas northern pintail declined significantly at both scales (Table 2).
No diving duck had a significant continental trend (although two species showed a tendency to decline), whereas at the local scale all three species showed significant increases over the quarter-century. The stiff-tail, ruddy duck, increased significantly during this period at both local and continental scales.
Thus the population trends of ducks in the Woodworth Study Area did not always track continental changes closely, either in direction or in level of significance. Local counts are compared to the continental indexes in Figs. 414. Although correlations between counts at Woodworth and continental indexes were positive for virtually all species, they were significant for only ruddy duck, northern pintail, and blue-winged teal(Table 4, last column).
The number of ducks in the study area was positively correlated with May ponds for each of the 11 species, but significantly so for only blue-winged teal (r = 0.71) and northern shoveler (r = 0.58), with northern pintail and ruddy duck correlations indicating 0.10 > P > 0.05(Table 5). Correlations with the Conserved Soil Moisture index, which serves as a proxy for wetland conditions, were generally similar to, but weaker than, those with May ponds (Table 5)).
Among the seven dabbling duck species, correlations with proximate weather variables, namely, April and May mean temperatures and precipitation, showed a general pattern of negative correlations with temperature in both months and generally positive correlations with precipitation, especially in April (Table 5). These results are not surprising, as cooler and wetter conditions favor the maintenance of water in wetland basins. Neither of these relationships was expressed for the three diving ducks or ruddy duck, however.
The possibility that counts in any one year show carryover effects from the previous year is tested next. For two dabbling-duck species (American wigeon and green-winged teal), counts in successive years were weakly positively correlated (Table 5), but not significantly. The mallard counts, inexplicably, were negatively correlated with those of the previous year. In contrast, counts of each diving duck species and the ruddy duck were strongly and positively related to those in the previous year.
Regression models provided different results for the various species. No model yielded a significant multiple correlation coefficient for the mallard, and R2 reached only 22.7% for the best model. This result was consistent with the general lack of significance among simple correlations with explanatory variables (Table 5). Similarly, models for the American wigeon and green-winged teal failed to produce significant R2 values; for the former species, the R2 value for the best model was only 14.0%, and in the latter species no best model was defined. Like the mallard, numbers of these two species were not strongly correlated with individual explanatory variables; their relatively low densities at Woodworth may explain the lack of identified relationships.
Regression models typically explained about a third of the variance in numbers of the gadwall. The "best" model, explaining about a fourth of the variance, included April temperature and May precipitation, both with negative coefficients. Stronger relationships were apparent in the blue-winged teal, with population numbers and proximate habitat variables giving R2 = 57 3%; here the best model explained two-thirds of the variation in numbers and included effects of May ponds, continental population, and the previous year's count. This last variable was included in the model despite the fact that its simple correlation coefficient with the species' numbers was O (Table 5). A similar dependence of numbers on proximate habitat variables was shown in northern shoveler, where the best model included only May ponds and accounted for a third of the variation. Counts of the northern pintail were related both to proximate habitat features and to the continental index. The best model, explaining nearly half of the variation in numbers, included only the continental index, however.
Of the diving ducks, significant R2 values for the redhead resulted from models including proximate habitat features and the previous year's count. The best model, with R2 = 44.6%, included the previous year's count, May precipitation, and May ponds. Numbers of canvasbacks were not related to proximate habitat variables, but were closely (and only) related to numbers in the previous year, with the best model explaining 43.8% of the variation. For the lesser scaup, models produced significant R2 values when the previous year's count was included; the best model, accounting for half of the variation, included the previous year's count, May precipitation, May temperature, and May ponds. The ruddy duck yielded models with significant multiple correlation coefficients whenever the continental population index was included. For this species the best model, incorporating only that variable, accounted for nearly half of the overall variance in census numbers.