USGS Home
Contact USGS
Search USGS

Sites ranged from intensively cropped to having substantial acreage enrolled in CRP or managed for wildlife (Table 3). Water levels in wetlands were low in 1990 following several years of drought. Wetland conditions improved somewhat on all study sites in 1991 and more so in 1992 when precipitation was near the long-term average. We considered 1990 to have "dry" and 1991 and 1992 to have "average" pond conditions when executing the default model and as a basis for the pond conditions in the customized model (Appendix C). Land use was relatively stable during the study. Minor changes in habitat availability that occurred on 2 sites were incorporated into both the default and customized model executions.

Field Data

Mallard hens were trapped from 23 April to 1 June 1990, 13 April to 15 May 1991, and 12 April to 13 May 1992 (Appendix E). From 16 to 26 hens were captured on each study site each year (Table 4). Radiotransmitters were placed on 66 to 75 hens each year. There were fewer resident hens on the study sites than the total marked each year due to hen mortalities within the first 3 days after capture and emigration. Totals of 38, 46, and 55 hens were considered residents of the combined study sites and potentially capable of nesting on the sites in 1990, 1991, and 1992 respectively.

Body mass of resident hens ranged from 900 to 1350 g (Table 5). There was little overlap between adult and yearling weights with adults averaging 33 g more than yearling hens. Twenty-nine hens could not be aged because of indeterminate DF scores or damage to the GSC2 that was collected. Weights of indeterminate-age hens were intermediate to adults and yearlings and likely included birds from both age groups.

Table 5. Body mass (g) of resident radiomarked hen mallards when they were captured on 3 study sites in southwestern Minnesota, 1990-92.
Age	n		(90% Cl)	Range
Adult	87	1092	(1082-1103)	900-1350
Yearling	22	1059	(1034-1084)	900-1220
Unknown	28a	1072	(1040-1105)	975-1250
aAn unknown age hen was released in 1990 before being weighed and was not included in the body mass analysis.

Resident hens initiated 18, 62, and 58 nests on the study sites in 1990 to 1992, respectively (Table 6). Nests were found in six of the 14 habitat types in 1990, eight in 1991, and six in 1992. An additional 55 nests could not be used to examine habitat-specific nest occurrence; 14 were from unmarked hens and 41 were from radiomarked hens nesting off the study sites. The 3 habitat types used most by radiomarked hens, ranked from highest to lowest, were CRP, unmanaged grass, and ROW in 1990; CRP, ROW, and hayland in 1991; and CRP, wildlife planted cover, and ROW in 1992.

We detected no interactions among year, study site, and habitat type that influenced age at which nests were found (all Ps > 0.37). The study site effect was significant (F = 5.62; df = 2, 148; P = 0.004) as was the habitat type effect (F = 7.06; df = 2, 148; P = 0.001) (Table 7). There was also a marginally significant effect associated with year (F= 3.14; df = 1, 148; P = 0.079).

Table 7. Mean age at which mallard nests were found using radiotelemetry on 3 study sites in southwestern Minnesota, 1990-92.
	Study site
	Ivanhoe		Slayton		Windom
Habitat category	n	a	n	a	n	a
Grasslands	42	5.1Aa	28	8.0Ba	36	8.3Ba
Road right-of-ways	5	1.8Ab	12	4.4Bb	8	5.4Bb
Otherb	5	6.0Aa	9	10.3Ba	9	5.7Ba
ªMeans followed by the same uppercase letter are similar within habitat types; those followed by the same lowercase letter are similar within study sites.
bWetland vegetation, odd areas, and cultivated fields.

Field Data and Mallard Model Comparisons

Mortality at the nest and away from the nest are treated separately in the model when predicting summer hen mortality (Table 8). Conditional hen mortality rate is an input parameter and not a predicted output of the model. We estimated the conditional hen mortality from 79 hens experiencing 103 nest destructions. The input value for conditional hen mortality used in the default and customized models was within the 90% confidence interval of our field estimate.

The daily probability of hen mortality away from the nest is also an input parameter and not a predicted output. Daily mortality of hens away from the nest combined with conditional hen mortality, yields an output prediction of summer hen mortality. Predictions of summer hen mortality from both the default and customized models were far outside the 90% confidence interval estimated from field data. Summer hen mortality predicted by both models was less than the lower confidence limit by 34% of the observed mean.

No single site-year stratum contained all 14 habitat types. Thus, we modeled mallard productivity from 11 to 13 habitat types depending on the stratum. In the field, nests were observed in 8 habitat types (Table 9). Proportional nest occurrence among habitats predicted by the default model fell within the confidence limits of our field estimates for five of the 8 habitats. For the customized model, nests predictions for 6 habitats were within the confidence limits. We observed the most nests in CRP, followed by ROW and unmanaged grass habitats. The default model ranked CRP as most important, but predicted other habitat and wildlife planted cover as the second and third most frequently used habitats. By comparison, the customized model ranked CRP first followed by ROW with other habitat third.

The deviation between the observed distribution of nests among habitats and the predicted proportional nest occurrence varied among the 3 most commonly used habitats. Predicted proportional occurrence of nests in CRP for both models exceeded the upper limit of the interval by 12% of the observed mean. Although the CRP prediction from both models fell outside the 90% confidence interval, it did not deviate from the observed value as far as predictions for other habitats. The default prediction for ROW fell outside the confidence interval by 42% of the observed mean while the customized model prediction was within the confidence interval. In contrast, both models severely underpredicted the occurrence of nests in unmanaged grass. The default prediction for unmanaged grass was lower than the lower limit of the observed mean by 41% while the prediction from the customized model was less than the lower confidence limit by 35% of the observed mean.

Habitat-specific estimates of nest success ranged from 3.3% in unmanaged grass to 18.9% in wetland vegetation (Table 10). Confidence intervals were large because we found relatively few nests in most habitats. Predicted habitat-specific nest success from both the default and customized models were within the confidence intervals for all habitats. However, when habitats were ranked by predicted nest success, the order of the top three differed from that produced by ranking the field estimates. Observed nest success was greatest in wetland vegetation followed by CRP and other habitat (i.e., odd areas). The default model predicted the highest success in CRP, wetland vegetation, and wildlife planted cover, respectively. Predictions from our customized model ranked CRP first, wildlife planted cover second, and wetland vegetation third.

Both models predicted values that were greater than the upper 90% confidence limit for four of 5 reproductive parameters (Table 11). Only the predicted hatch rates were within the field estimate confidence interval. Although clutch size predicted by both models exceeded the upper confidence limit, the predictions were outside the confidence interval by <5% of the observed mean. In contrast, predictions for hen success and number of hatched nests from either model were above their respective upper confidence limits by >35% of their observed means.