---

Introduction

Estimates of annual survival are vital to understanding changes in regional and continental waterfowl populations (Anderson 1975, Johnson et al. 1988). However, a thorough understanding of waterfowl population dynamics requires age-, sex-, and cause-specific estimates of seasonal mortality rates (Anderson and Burnham 1976, Blohm et al. 1987) and knowledge of their local variation. Only a few such estimates are available for mallards (Hestbeck et al. 1989, Dugger et al. 1994, Reynolds et al. 1995), but without them, managers cannot effectively manipulate waterfowl habitats or hunting seasons. Management strategies that target demographic groups, specific time periods, areas where losses are significant, and causes that show greatest potential for modification are most likely to succeed (Nichols 1989).

Survival of young from fledging to autumn migration remains among the least-understood components of waterfowl recruitment. Historically, investigators postulated that survival rates were near 1.0 for the postfledging-prehunting period (e.g., Jahn and Hunt 1964:45-46). However, Nichols and Hines (1987) and Johnson et al. (1992) noted substantial differences between survival rates of young and adult mallards banded before hunting season and inferred greater prehunting mortality of fledged juveniles because age-related differences in survival are seldom evident by late winter (Hopper et al. 1978, Rakestraw 1981, Nichols and Hines 1987; but see Reinecke et al. 1987). Causes and timing of mortality during the period from fledging to migration remain largely unknown (Sargeant and Raveling 1992).

Objectives of our study were (1) determine causes and timing of mortality for fledged juvenile mallards in northcentral Minnesota, (2) compare local survival rates to estimates for a larger geographic area encompassing our study area, and (3) determine from these survival rates whether reducing natural mortality in the vicinity of natal marshes could have produced measurable gains in recruitment to migration. Our results are based on data collected during 1972-74 by the Northern Prairie Wildlife Research Center of the U.S. Fish and Wildlife Service (now of the Biological Resources Division, U.S. Geological Survey) and the Department of Ecology and Behavioral Biology (now the Department of Ecology, Evolution, and Behavior) at the University of Minnesota. Other aspects of these data have been described by Gilmer et al. (1977), Kirby et al. (1981, 1983, 1989) and Kirby and Cowardin (1986).

Methods

Data Collection

Data were collected within a 932-km² portion of Beltrami, Cass, and Itasca counties, Minnesota, and largely within Chippewa National Forest. Waterfowl habitat consisted of large sand-bottomed lakes, meandering streams, rivers, isolated wetlands, and beaver (Castor canadensis) flowages. Cowardin and Johnson (1973) and Gilmer et al. (1977) presented detailed descriptions of the study area.

We captured 36 Class III ducklings (Gollop and Marshall 1954) and 59 newly fledged (immature) mallards were captured in late July, early August, and mid-September of 1972-74 by nightlighting on brood-rearing lakes. Each duckling was fitted with a breast-mounted radiotransmitter (Gilmer et al. 1974) and a U.S. Fish and Wildlife Service leg band. Mean transmitter mass ranged from 22.0 to 24.8 g (<3% of the body weight of fledged mallards). Gilmer et al. (1974) reported effects of these radiotransmitters on mallards and wood ducks (Aix sponsa): birds returned to normal behavior within 1 week of marking, and most birds (84%) recovered by hunters were in good to excellent condition.

We located radiomarked mallards from aircraft an average of 5 times per week and at least once daily on the ground or water by triangulation. Locations were confirmed visually, when possible to do so without flushing birds. If a bird did not appear to move during the interval between 2 successive searches and visual confirmation of status was impossible, a ground search was conducted, usually within 12 hr, to determine if the bird was alive. Dead birds were necropsied to determine cause of death (predation, hunting, unknown). Dates of death were obtained by correspondence with hunters who reported recoveries.

Statistical Analyses

We constructed survival records from relocations of radiomarked birds, dates when investigators sought but failed to locate radiomarked birds, dates of death reported by hunters, and field notes describing causes of deaths. For each year, we used opening day of hunting season (1 Oct in 1972 and 1973, 2 Oct in 1974) as the origin of our time scale. We excluded birds from our analysis if they were killed by predators prior to fledging. We attributed deaths of crippled birds subsequently killed by predators to hunting mortality incurred on the date of crippling.

We expected radiotransmitters to operate for 75-90 days. Chances of a radiotransmitter being rendered inoperative by a predator or shotgun pellets were slight (Kirby et al. 1981), and observers noted changes in signal characteristics that preceded premature failures. Thus, we were able to distinguish radio failures from deaths and departures. If a duck carrying a radiotransmitter with a satisfactory signal that had operated <75 days was not located during a thorough search of the study area and neighboring wetlands, we concluded the bird had flown elsewhere. If a bird departed from the study area and did not return, or if its radiotransmitter failed during the interval between 2 searches, we right-censored its survival record at the midpoint between searches.

We used Cox proportional hazards models to determine whether survival rates varied among sexes, age classes at capture, or years of study. Our objective was to determine whether effects ruled out pooling data from different groups; hence, we used both forward selection and backward elimination to test effects and selected a liberal criterion of P ≤ 0.20 for statistical significance.

Our full model for comparing sexes and age classes at capture was

where S was the survival rate, YEAR identified 3 strata, SEX distinguished males and females, AGE described age at first capture, and interactions (SEX x SEAS, AGE x SEAS) allowed AGE and SEX to have different effects prior to hunting season. Conditional on nonsignificant effects of age class and sex, we subsequently fit a second set of Cox models to test for differences between years. Our full model for this subsequent analysis was

with YEAR treated as a covariate instead of as a stratifying variable. Conditional on nonsignificant differences among years, we pooled data across years and used the Kaplan-Meier (K-M) staggered-entry estimator (Pollock et al. 1989) to calculate survival rates and associated 90% confidence intervals. To prevent K-M survival estimates of zero, which resulted when the last remaining member of a cohort was killed after other members migrated, we right-censored the record for the last bird in each cohort. We implemented survival analysis with SAS statistical software (PROC PHREG; SAS Institute 1997).

Results

We equipped 95 juvenile mallards with radiotransmitters. One transmitter failed the night it was attached. Another bird became entangled in its antenna, and 3 were killed by predators prior to fledging. These events reduced our sample to 90 birds (Table 1).

Table 1. Kaplan-Meier estimates of survival rates for fledged, radiomarked juvenile mallards in northcentral Minnesota from 23 July to 20 October 1972-74.

Monitoring interval	Cohort	na	Exposure days	Deaths	Survival	SE
Preseason	Females	48	1,636	5	0.84	0.063
	Males	42	1,050	2	0.88	0.077
	Class IIIb	31	1,339	4	0.85	0.071
	Immaturesb	59	1,347	3	0.86	0.072
	All birds	90	2,686	7	0.86	0.047
Hunting season	Females	28	231	10	0.40	0.141
	Malesc	21	67	9	0.37	0.148
	Class IIIb	16	123	6	0.30	0.166
	Immaturesb	33	175	13	0.29	0.129
	All birds	49	298	19	0.29	0.107
Fledging to migration	Females	48	1,867	15	0.34	0.121
	Malesc	42	1,117	11	0.33	0.134
	Class IIIb	31	1,462	10	0.25	0.142
	Immaturesb	59	1,522	16	0.25	0.113
	All birds	90	2,984	26	0.25	0.094
aMaximum number at risk during the monitoring interval. bAge class at capture. cCensored last survivor on 12 October.

Of the 90 birds included in our survival analysis, we could determine fates with certainty for 41. Thirty-one were killed on the study area before 20 October, the latest date a surviving marked bird was observed. Seven of these deaths occurred prior to the hunting season and were attributed to predation: 5 to mink (Mustela vison), 1 to a red fox (Vulpes vulpes), and 1 to an avian predator (probably a great horned owl [Bubo virginianus]). We attributed 24 deaths to hunting on the study area; 5 of these, however, occurred after we lost radio contact. Of the 19 hunting deaths that befell birds with active radios, 9 occurred on opening day of hunting season. Hunters reported killing 10 additional birds that left our study area on migration.

When estimating survival rates, we censored survival records of 64 birds when we lost radio contact due to departure or radio failure. These instances of censoring included 5 of 24 birds killed on our study area by hunters. The median time elapsed from last observation to censoring was 1.2 days.

Figure 1. Survival distribution (solid line) for radiomarked juvenile mallards from northcentral Minnesota, 23 July to 20 October 1972-74, with upper and lower 90% confidence limits (dashed lines).

Our estimates of survival rates (Table 1, Fig. 1) do not rule out the possibility of important sex- or age-related differences in survival rates, but we found no statistical evidence (P > 0.20) that such differences affected results we obtained by pooling data across years and sexes or age classes. Estimated survival rates were similar for males and females as well as for birds marked before and after fledging (Table 1).

Discussion

To increase local recruitment by reducing mortality, waterfowl managers first need to identify mortality factors that (1) account for a substantial portion of total annual mortality, and (2) show potential for modification (Nichols 1989). In northcentral Minnesota, eliminating natural mortality would have increased survival to migration only slightly (from 0.25 to 0.29). Hence, management of natural mortality had limited potential for increasing recruitment of mallards to the fall flight.

Management of natural mortality could have been more beneficial if we underestimated its effect and overestimated the effect of hunting. However, as the following example shows, our findings are robust to such errors. A conservatively low estimate for survival prior to hunting season is 0.77, the lower 90% confidence limit for our estimate. A conservatively high estimate for survival during hunting season is 0.50, our upper 90% confidence limit. Even if these estimates were correct, overall survival was 0.39. Eliminating natural mortality entirely would have increased overall survival by only 11 percentage points.

Although the preceding example supports our conclusion, it is unrealistic. Our estimates of survival to hunting season are more likely low than high, due to effects of capture, handling, and radiotransmitters (Gilmer et al. 1974, Reinecke et al. 1992). Substantial negative bias seems unlikely, however, because our estimates of survival to hunting season are not far from 1 and are comparable to (Hestbeck et al. 1989, Longcore et al. 1991) or higher than (Parker 1991) other published estimates for mallards and for American black ducks (Anas rubripes).

Our estimates of survival during hunting season were susceptible to both positive and negative biases. Some marked birds may have been killed by hunters and removed from the study area without our knowledge. Conversely, radiotransmitters may have predisposed some birds to hunting mortality (Reinecke et al. 1992). The rapid behavioral adjustment of birds to our transmitters (Gilmer et al. 1974) and high survival rate prior to hunting season suggest the latter effect was not serious; however, our estimate of 29% survival during hunting season seems low in comparison with other published estimates. For example, Kirby and Cowardin (1986) reported an annual survival rate of 45.5% for young mallards banded during 1967-74 in a much larger region that included our study area.

We anticipated the disparity between our results and those of Kirby and Cowardin (1986) because we estimated survival rates for birds living within a comparatively small area subjected to heavy hunting pressure. Heavy hunting pressure caused the departure of many marked birds from our study area (Kirby et al. 1989); hence, it seems likely that birds departed to areas with lower hunting pressure and experienced lower mortality rates as a result. Birds that departed and experienced higher survival removed themselves from our sample, but not from the band recovery analysis of Kirby and Cowardin (1986).

Management Implications

Factors other than hunting caused few mortalities of fledged juvenile mallards and showed little potential for modification in forested habitats of northcentral Minnesota during 1972-74. Strategies that increased nest success, reduced brood losses, or decreased hunting mortality would more likely have produced meaningful gains in recruitment because these factors caused much greater losses of mallards from our study area (Jessen 1970, Ball et al. 1975, Gilmer et al. 1977, Cowardin and Johnson 1979, Kirby et al. 1989).

Although potential benefits of reducing natural mortality were limited, small gains in recruitment may be worth pursuing when they can be achieved easily. In this case, however, reducing natural mortality would not have been feasible. Deaths were attributed to a variety of predator species and distributed thoughout a broad geographic area devoted to a variety of uses, not just to mallard production. Devising acceptable, cost-effective means of reducing predation is especially difficult under such circumstances (Greenwood and Sovoda 1996). The ability of managers to achieve effective control is an important consideration in predator management (Sargeant et al. 1995), but one that is often overlooked.

Since 1974, changes in waterfowl habitats on our study area probably have not increased rates of mortality due to predation of fledged juvenile mallards. Lakefront property has been developed, but development probably provides mallards with a refuge from some species of predators; highest concentrations of mallard broods are now found in such areas. Conversely, rates of mortality due to hunting are probably still relatively high. Bag limits are more liberal and seasons are longer now than they were in the 1970's, and the area remains a favorite with hunters. Thus, for mallard populations in northcentral Minnesota, the relative importance of nonhunting mortality of fledged birds is probably not substantially greater now than when our data were collected.

---

NPWRC Home | Site Map | About Us | Staff | Search | Contact | Web Help | Copyright