Northern Prairie Wildlife Research Center
Fox families were accurately censused on six townships (559 km2) within the three-county study area (Fig. 1) by repeated aerial flights during May and June each year from 1969 to 1973 (Sergeant et al. 1975). Evidence indicated that fox densities on the townships were representative of those throughout the reference area. Thus, estimates of fox family densities for each year 1969-73 were made by expanding the township totals by 168.46 ( = 94,245 km2 in study area / 559 km2 in six townships). We estimated fox densities in the other years either by using mid-April rural mail carrier sightings (RMC) of adult foxes gathered by the North Dakota Game and Fish Department or by using the number of foxes killed at strychnine-baited draw stations (SDS) maintained by the U.S. Fish and Wildlife Service.
Allen and Sargeant (1975) found a close relationship between the township census (TWP) data and RMC data obtained for the same region. The rural mail carrier counts (live adult foxes seen per 1,609 km driven) in each of six strata were weighted by the proportion of that stratum in our reference area (Table 4). The agreement between these index values and the TWP data was excellent, as demonstrated by the regression equation
and a correlation coefficient of r = 0.998 (3 df) for 1969-73, the period for which both sets of data were available. By multiplying the calculated value of TWP by 168.46 (the expansion factor converting TWP to reference area data), we obtained estimates of spring fox family populations for the reference area extending back to 1964.
The only available fox population data for 1963 were the SDS reports, which correlated well with the RMC data; the regression equation was
and r = 0.958 (1 df), based upon the overlapping period 1964-66. By using equations 1 and 2 and multiplying the result by 168.46, we obtained an estimate of the 1963 fox population.
The average estimated spring fox population was 9,905 families (SD = 4,019). The trend for the study period described in Table 4 was in general agreement with observations of hunters and trappers who were active throughout the period.
Our estimates of spring mallard densities were taken from the annual Waterfowl Breeding Pair Surveys conducted by the U.S. Fish and Wildlife Service. Pospahala et al. (in preparation) gave procedural details of the surveys, which involved aerial counts by species and sex of all waterfowl observed along prescribed transects. The surveys were generally conducted during the latter half of May. Intensive ground searches of certain surveyed areas were made to estimate the proportion of birds overlooked during the aerial count, so that the estimates obtained from the flights could be adjusted for this visibility rate.
The combined strata 45 and 46 of the aerial surveys nearly coincide with our reference area. The area sampled within the strata varied from 1,445 to 1,503 km2, which represented a sampling fraction of 1.4%.
For groups of one to five drakes, each male was counted as a pair, under the assumption that the female was either with the male or on the nest. Groups of more than five males were presumed to be nonbreeders and were tallied as the number of individuals seen. Our model only required the number of breeding males, and because the number of flocked males counted in strata 45 and 46 was negligible (R. S. Pospahala, personal communication), we halved the estimated total breeding population for these strata to arrive at the number of breeding males. Because our reference area was 11.0% smaller than that of strata 45 and 46, a corresponding reduction in the male mallard population was made (Table 5).
The fox denning season in North Dakota begins about 1 April, lasts approximately 10-15 weeks, and coincides with the principal waterfowl nesting season (Sergeant 1972). During this season, prey are brought to dens for use by the pups and food remains accumulate in and around the dens. A convenient and seemingly accurate index of fox predation rates on mallards was the average number of mallards discernible in food remains found above ground at fox rearing dens. Most ducks brought to fox dens were almost completely consumed by the foxes. Thus, in determining the number of mallards at individual dens, we interpreted the presence of any identifying part such as a wing or a few body feathers as representing one mallard taken by foxes. Duplicate parts were needed before additional birds of a sex were assigned to a den. The species of some ducks brought to dens could not be determined and those unidentified ducks were excluded from our calculations. The difficulty of assigning mallards to a den increased with the number of mallards present because the chance of finding sufficient distinguishing parts decreased.
Estimates of the average number of mallards taken per fox family were obtained by expanding the surface food remains index to account for: (1) mallards at dens but not detectable in food remains found above ground, i.e., additional birds identified by remains below ground, (2) mallards taken by foxes but not brought to dens, and (3) the number of dens used by foxes to rear their pups. Estimates of each factor were available.
Incidence of mallards on surface of dens-- From 0.22 to 0.36 mallard per den was identified in food remains found above ground at fox rearing dens during 1969-73 (Table 6). The annual samples of dens were from all major habitat types in the reference area and included dens at which no food remains were found and dens in areas of poor mallard habitat. The dens were found and visited with nearly equal intensity throughout May and June each year and the data are believed to depict the frequency of mallards in surface remains at dens throughout the Prairie Pothole Region of North Dakota.
Mallards found below ground-- The relationship between the number of mallards identified in food remains found above ground and the total number above and below ground was determined from a sample of 75 completely excavated dens having the remains of one or more mallards (Table 7). Fifty-nine (49%) of 121 mallards discernible in the food remains at these dens were found above ground; the remainder could be identified only by excavation of the dens. This figure was consistent among years and thus by multiplying the number of above-ground mallards by 2.0 ( = 1.0/0.49), we estimated the total number of mallards at dens, and used this expansion factor as a constant in the model.
Mallards not brought to dens-- Not all food acquired by foxes was brought to dens. The percentage of mallards acquired by foxes that were brought to dens and detectable in the food remains at those dens was estimated in a study conducted in Stutsman County during 1969-70 (G. L. Rohde, unpublished data). Live mallards, each equipped with a radio transmitter on one wing and serially numbered tags on extremities, were tethered throughout the denning season in areas occupied by five families of red foxes. The disposition of birds taken by foxes was determined by frequent locations of the radio transmitter. Dens were excavated after they were abandoned by the foxes and the food remains were examined for the presence of the tethered ducks. From a sample of 285 mallards taken by the foxes during April, May, and June, totals of 18%, 47% and 16%, respectively, were detectable in the combined above and below-ground food remains at the dens (Table 8). The number of mallards at each den was determined by the same criteria used to determine fox predation rates on mallards in the three-county study area. Thus, duplicate parts were needed before an additional mallard was added to the list of birds at a den. For the model we relied on the figure for May, when we believe most wild mallards were taken by foxes. This estimate, 47%, was the most conservative, but still represented an expansion factor of 2.1 (1.0/0.47). It was used as a constant in our model because sufficient data were not available to assess annual variation.
Number of dens used-- Many workers have shown that a fox family typically uses several dens to rear its pups and that changes of dens occur frequently, often following disturbances such as those caused by agricultural practices (Table 9). All data in Table 9 were from agricultural regions where disturbances to dens were frequent, thus paralleling conditions in our reference area.
Dens were very difficult to locate before pups were 4-5 weeks old because pups remained underground and few prey were brought to them. One den of each of those tallied per family in Table 9 was believed to be from this early period, but few of these early dens were included in our sample of dens in Table 6. We accordingly subtracted one den from the average in Table 9 to estimate the average number of dens of the type we sampled. The average of 3.8 dens per fox family was used as a constant in the model.
Sex ratios of mallards at fox dens-- Data on the sex ratio of mallards brought to rearing dens were available for 1969-73 (Table 10). Most of these mallards were found at dens in the three-county study area, although some were from elsewhere within the reference area. The preponderance of females each year was minimal because the distinct coloration of males made them more easily identifiable than females in food remains at dens.
Based upon the expansion factors discussed and the sex ratios of mallards found at fox dens, we estimated the predation rates for male and female mallards for each year 1969-73 (Table 11). On the average, each fox family took 0.959 male and 3.601 female mallards.
Estimates of annual survival rates for male and female mallards were obtained for 1963-67 from bandings of adult mallards in Minor Reference Area 131 as defined by Anderson and Henny (1972). This area is composed of the Prairie Pothole Region and the Agassiz Lake Plain of North Dakota. Because few if any bandings were done in the Agassiz Lake Plain, the data were considered to be representative of our reference area.
Survival estimates were obtained from recoveries of birds banded in late summer and early autumn and analyzed by Anderson (1975), who used efficient statistical methods recently developed by Seber (1970) and Robson and Youngs (1971). These methods allowed determination of survival estimates for 12-month intervals from one banding period to the next. The data were sufficient to estimate survival rates for both adult males and adult females only for 1963-67 (Table 12), but we assumed no appreciable change occurred after this period. Examination of survival estimates from nearby areas (western Minnesota, southwestern Manitoba, and southeastern Saskatchewan) did not contradict this assumption.
Direct estimates of hunting mortality rates were not available, but a useful index to these rates existed in the annual recovery rates of banded birds. The rate for the ith year, denoted Pi, is defined as the number of banded birds shot that year and reported to the USFWS Bird Banding Laboratory, divided by the total number of banded birds alive at the beginning of the hunting season. By adjusting these rates to account for nonreporting of bands and for the loss of birds that were shot but not retrieved by hunters, it was possible to estimate the hunting mortality rate.
Annual recovery rates for 1963-67 were estimated (Anderson 1975) by applying Seber's (1970) method to the same banding data used for survival rates. Table 12 shows estimates of the annual recovery rates of adults, and Table 13 gives the summary statistics for annual male and female survival and recovery rates.
To account for banded birds taken by hunters who did not report the bands, we needed an estimate of the rate of band reporting. A study conducted in 1972-73 (Henny and Burnham 1976) compared the recovery rates of normally banded mallards and mallards banded at the same time and place but carrying an additional band offering a reward of $10 to the hunter who submits the band. By assuming that all such reward bands are turned in, the authors estimated the reporting rate. The estimate would be biased upward, and our estimated hunting mortality downward, if some reward bands went undetected by hunters or were knowingly not submitted. The reward band study indicated that the average band reporting rate for mallards in North America in 1972 was 44% (C. J. Henny and K. P. Burnham, personal communication), although it varied both geographically and according to distance of the recovery from the banding site. (Results from 1973 bandings were not available at the time of our analysis.) Recovery rates were thus multiplied by 2.27 ( =1.0/0.44) to account for nonreporting of bands.
In addition to birds recovered by hunters, some were downed but not retrieved. Some of these birds died immediately; others succumbed later from the gunshot wounds or because of heightened susceptibility to predators, starvation, or freezing. Presumably, few of the hunter-downed birds recovered. Kimball (1973) provided an estimate of 17.9% unretrieved mallards, based upon observations of 2,564 birds downed by hunters throughout the United States in the 1965-69 hunting seasons. These were "regular" seasons, in which regulations did not discourage the shooting of female mallards as opposed to male mallards.
Thus, by multiplying a recovery rate by 2.27 and adding 17.9% to account for birds not retrieved, we estimated the hunting mortality rate for a given year. The average hunting mortality rates were 15.8% for adult males and 10.2% for adult females.