Northern Prairie Wildlife Research Center
In 1988-91, we captured female mallards from mid-April through early May with decoy-hen traps (Sharp and Lokemoen 1987). We fitted each female with a 23-g harness transmitter (Dwyer 1972) and a uniquely identifiable combination of nylon nasal markers (Lokemoen and Sharp 1985). We monitored females daily to assess nesting activity (Krapu et al. 1997). At nest sites of marked females, we captured all ducklings in the brood when possible, attached web tags following a procedure modified from Haramis and Nice (1980), and attached 2-g radiotransmitters using sutures and glue to 1-4 randomly selected ducklings per brood.
In 1992-94, we located nests by systematic searching on privately owned Conservation Reserve Program (CRP) fields and WPAs. Nest searching was conducted by dragging a chain between vehicles to flush females from nests (Higgins et al. 1969). We determined developmental stages of eggs by candling (Weller 1956). Beginning about 15 days after the onset of incubation, we used modified bow traps (Salyer 1962) or walk-in traps (Dietz et al. 1994) to capture nesting females. We fitted each captured female with a 4-g anchor transmitter (Pietz et al. 1995) and a unique combination of nasal markers. After marking, we anesthesized females with methoxyflurane to reduce the risk of nest abandonment (Rotella and Ratti 1990). We web-tagged all ducklings and marked 1-4 (usually 2) ducklings per brood with 1.5-1.8-g anchor transmitters modified from Mauser and Jarvis (1991). All capture and marking procedures were approved by the Animal Care and Use Committee at Northern Prairie Wildlife Research Center and conformed to recommendations of the American Ornithologists' Union (1988).
Maximum ranges of female and duckling transmitters to ground-tracking vehicles were 2-3 km and 1.5 km, respectively. Radios were equipped with mortality sensors (mercury switches or thermistors); we attempted to retrieve carcasses as soon as possible after mortality signals were detected.
We tracked each brood from nest to wetland, then attempted to visually check broods daily to detect losses of unmarked ducklings and radio failures. If no visual sightings were obtained, we recorded the brood location and radio status using standard telemetry methods (Mech 1983). We aerially searched for missing broods (Gilmer et al. 1981) weekly.
Landscape Variables.--The National Wetland Inventory (NWI) delineated upland and wetland habitats on our study areas from high-altitude color-infrared photographs prior to our study. Land cover on each study area was documented using aerial videography beginning in May 1988 and status was updated in the database annually. We calculated percent of upland in perennial cover (PERNCOVER: native grasslands, planted cover, alfalfa hayland, woodlands, shrub lands, odd areas, and road right-of-ways) on each study area. We classified each wetland basin by the most permanent water regime assigned to part or all of that basin (temporary, seasonal, semipermanent, and lake) by the NWI (Cowardin et al. 1979, 1982). Using aerial videography (Cowardin et al. 1988), we estimated water conditions on each study area at monthly intervals from May to September. Ponds were defined as basins that contained water (Cowardin 1982) and were categorized according to basin class (e.g., water within a basin with a seasonally flooded water regime was termed a seasonal pond). We delineated inundated portions of wetland basins (i.e., pond size) using the feature-mapping process from Map and Imaging Processing System software (MIPS; Miller et al. 1990).
We evaluated the relation between water conditions and brood survival using relative abundance of seasonal ponds. We chose percent of seasonal basins with water as an explanatory variable because (1) seasonal ponds account for most annual variation in number of ponds in prairie pothole habitats (Krapu et al. 1997), (2) seasonal ponds are a preferred habitat of brood-rearing females (Talent et al. 1982), and (3) seasonal ponds can be readily monitored by managers over large areas. We calculated percent of seasonal basins containing ponds (WETSEAS) and used it to assign values to the WETSEAS variable (0 = <17%, 1 = >59%). These WETSEAS categories were used because percent of seasonal basins with ponds did not exceed 17% on study areas in 1988-90 and 1992, but was never lower than 59% on study areas in 1993 and 1994.
Unlike loss of individual ducklings, brood loss (i.e., death of all ducklings in a brood) can occur either in a single catastrophic event (Sargeant et al. 1973) or through attrition over a protracted time span and area. Consequently, we calculated landscape variables over the entire study area, and assigned a value of WETSEAS from the wetland survey date nearest to the hatch date of each brood.
Weather, Hatch Date, Brood Age, Brood Female Age, and Brood Size Variables.--We obtained daily records of precipitation and minimum air temperature for each study area from the nearest National Weather Service observation station (National Oceanic and Atmospheric Administration 1988-94). For each brood exposure day, we calculated (1) RAIN ("1" if it had rained [including values recorded as "trace"] on the current or 2 previous days and "0" otherwise); and (2) TMIN (the average of daily minimum temperatures for the current and the 2 previous days). We included any rainfall events that had occurred in the 2 previous days to allow for some lag time between rainfall and mortality (Korschgen et al. 1996) and a possible delay of ≤1 day in detecting a brood loss. The HATCHDATE was the Julian date on which the first egg of a clutch hatched. Brood age (BA) was designated as 0-7 or 8-30 days. We determined age (second-year [SY] or after-second-year [ASY]) of brood females using the greater secondary covert (Krapu et al. 1979). Brood size was the number of ducklings that hatched in each brood.
To evaluate brood survival, we used data from broods with radiomarked adult females. We used Cox (1972) proportional hazards regression (PROC PHREG; SAS Institute 1996) to test for differences in brood survival to 30 days of age in relation to TMIN (time-dependent, continuous), RAIN (time-dependent, binary), HATCHDATE (continuous), BA (time-dependent, binary), PERNCOVER (continuous), WETSEAS (binary), female age (binary), and brood size (continuous). Prior to performing this analysis, we tested for colinearity between HATCHDATE and brood size (PROC CORR; SAS Institute 1990) and found none (r = -0.06, P = 0.66). We specified BA as a time-dependent variable by resetting the time origin for each brood to 0 at day 8 of life. Our fully specified model included all main effects and 2-way interactions, except WETSEAS-by-BA, WETSEAS-by-RAIN, and all interactions with brood size. We did not include those interactions with WETSEAS because of missing cells. We used backward elimination to delete non-significant (P > 0.05) terms, beginning with the interactions. We considered all ducklings in a brood dead when no ducklings remained with the female, as determined by observations, or when erratic female movements indicated no affinity to any wetland. When the exact date of brood loss was unknown, we assigned loss at the midpoint of the interval between the last time the female was seen with a brood and the first time she was seen without a brood or exhibited erratic movements. We censored surviving broods on the date they were last observed alive or day 30 of life. To test for possible effects of duckling radiotransmitters on brood survival, we included a binary explanatory variable identifying whether or not broods contained ≥1 radiomarked ducklings in our final model.