David C. Bowden

An improved model for spatially correlated binary responses

In this paper, we use covariates and an indication of sampling effort in an autologistic model to improve predictions of probability of presence for lattice data. The model is applied to sampled data where only a small proportion of the available sites have been observed. We adopt a Bayesian set-up and develop a Gibbs sampling estimation procedure. In four examples based on simulated data, we show that the autologistic model with covariates improves predictions compared with the simple logistic regression model and the basic autologistic model (without covariates). Software to implement the methodology is available at no cost from StatLib.

Generalized mark-sight population size estimation applied to Colorado moose

A new procedure for constructing confidence interval estimates of population size in mark-sight experiments is presented. The method of selecting animals for marking must be equivalent to a simple random sample without replacement. Marked animals are required to be individually identifiable. The number of times animals are sighted must be independent of their mark status. The sighting process does not need to be composed of independent sighting trials or even decomposable into separate trials. Sighting probabilities can vary among individuals and can depend on such factors as group size and vegetational cover. Other methods of constructing confidence intervals in mark-sight experiments given these latter conditions have failed to achieve their stated nominal confidence level. Our confidence interval procedures are shown by simulation to have actual confidence levels close to nominal under conditions encountered in an application to a Colorado moose (Alces alces shirasi) population. For this population with 29 radio-collared moose and 5 helicopter sighting flights, the 90% confidence interval for the moose population size on the 1,400-km 2 area was 382-505.

Dynamics of interacting Elk populations within and adjacent to Rocky Mountain National Park

We studied population subdivision and density-dependent and independent factors influencing population processes between 1965 and 2001 for elk (Cervus elaphus) inhabiting Rocky Mountain National Park (park) and the adjacent Estes Valley (town), Colorado, USA. Elk numbers within the park were held relatively constant by management controls until 1967, after which time they were allowed to increase without human interference. Radiotelemetry of 73 elk indicated limited exchange between the subpopulations; combined with clear distinctions in population dynamics, this suggests that these subpopulations are relatively independent despite the absence of physical barriers between them. The elk subpopulation within the park initially increased at 6.5%/year between 1968 and 1970, then growth gradually slowed-exhibiting density-dependent reductions both in calf survival and recruitment with increasing population size-and approached an estimated carrying capacity of 1,069 ± 55 (?±SE). Since 1991, this subpopulation has remained within ±5% of this equilibrium. The adjacent Estes Valley subpopulation grew at an estimated maximum 5-year average rate of 11.0% from 1979 to 1983 and is still increasing at 5.2%/year (1991-2001 average). Estimated town population currently is about 70% of our projected carrying capacity of 2,869 ±415 elk based on projection of observed calf recruitment decline with increasing population. Both carrying-capacity estimates are consistent with independent estimates based on forage biomass and energy considerations. Adult cow survival rate was not found to differ between park and town, and we estimated a constant rate of 0.913 [95% CI = 0.911, 0.915]. Bull survival rates increased in the park from 0.52 to 0.79 between 1965 and 2001, but remained constant at 0.42 [0.35, 0.47] in the Estes Valley. Colder winter temperatures were correlated with reduced calf recruitment (calves:cow at age 0.5 yr) and with reduced calf survival (between age 0.5 and 1.5 yr) in town. Recruitment of town elk also increased with warmer summer temperatures and greater summer precipitation. No weather covariates were significantly correlated with calf recruitment or survival in the park. Declining calf recruitment has been nearly linear and similar in both the park and town. In the park, calf survival responded little to density when the population was well below carrying capacity, but responded at an increasing rate as the population neared carrying capacity. This pattern may explain why calf survival response to density has not yet been detected in town. We estimated current combined population size of 3,049 [2759, 3369] elk in 2001. Elk in the town sector currently outnumber elk in the adjacent national park by almost 2:1 and are projected to increase by 46% before being nutritionally limited, suggesting that human-elk conflicts likely will increase in the absence of active management intervention.

Field Exposure of Seven Species or Subspecies of Salmonids to Myxobolus cerebralis in the Colorado River, Middle Park, Colorado

Abstract Recent failures in recruitment of rainbow trout Oncorhynchus mykiss in the Colorado River in Middle Park, Colorado, USA, led to studies of the effect of the myxosporean parasite Myxobolus cerebralis, the causative agent of whirling disease, on the wild trout fishery in the river. During 1995–1996 and 1996–1997, we conducted field exposures of sentinel fish to examine the vulnerability of seven species or subspecies of salmonids exposed to the parasite in the Colorado River. During 1995–1996, brook trout Salvelinus fontinalis and Colorado River cutthroat trout O. clarki pleuriticus experienced 85% or higher mortality within 4 months of exposure. Groups of fish introduced at greater mean weight or later in the summer tended to survive better than others, but mortality often was severe among these groups during the second summer. By April 1996, dead fish or sacrificed fish representing seven groups were shown by histology to contain mature M. cerebralis spores. During 1996–1997, three groups of cutt...

Effects of fertility control on populations of ungulates : General, stage-structured models

Regulating the abundance of ungulate populations using hunting can prevent populations from reaching levels that cause harm to natural and human dominated systems. However, there are an increasing number of cases where hunting is infeasible, and in such cases, fertility control has been widely advocated as an alternative means for controlling populations. Here, we develop simple analytical models offering general insight into the feasibility of using fertility control to regulate the abundance of ungulates. The models are structured in stages to represent variation in the duration of effect of fertility control agents. Analysis of these models offers several predictions, amenable to testing in field studies. (1) More than 50% of fertile females will need to be maintained infertile to achieve meaningful reductions in ungulate numbers even when fertility rates are low. (2) The relationship between the proportion of females maintained infertile and the steady state density is highly nonlinear. This means that small errors in estimating levels of infertility can lead to large errors in achieved density. It also means that managers should expect to see little change in steady-state density across a broad range of delivery rates. (3) The efficacy of fertility control as a management technique depends strongly on the persistence of the effect of the fertility control agent and the ability of managers to recognize previously treated animals. (4) Fertility control using long-lived agents can be more efficient than culling in regulating ungulate numbers. (5) Treating small populations with irreversible agents magnifies the likelihood of population extinction relative to treatment by culling. As with all techniques, managing population fertility must extend from a sound understanding of the influence of management actions on the state and dynamics of the population.

Evaluation of the lek‐count index for greater sage‐grouse

Abstract Counts of birds attending leks traditionally have been used as an index to the population size of greater sage-grouse (Centrocercus urophasianus) and, more recently, as a means to estimate population size. The relationship between this index and the actual population has not been studied. We used intensive counts of individually marked and unmarked greater sage-grouse on leks to evaluate how sex and age of birds, time of day, and time of season impact lek-attendance patterns and lek counts. These within-season sources of variation need to be considered when estimating detection probability of birds on leks and ultimately adjusting the lek-count index to estimate true population parameters. On average, 42% of marked adult males, 4% of marked hens, and 19% of yearling males were observed on leks per sighting occasion with all 15 known leks being intensively counted. We discovered that lek counts as currently conducted may be useful as an index to greater sage-grouse populations, but standardization of protocols is needed to allow for better spatial and temporal comparisons of lek-count data. Also the probability of detecting birds on leks must be estimated in order to relate lek counts to population parameters. Lastly, we evaluated use of the bounded-count methodology for correcting lek-count data. We showed large biases associated with this technique and below-nominal coverage of confidence intervals even at large numbers of counts, demonstrating the unreliability of the bounded-count method to correct lek-count data.

NEONATAL MULE DEER FAWN SURVIVAL IN WEST-CENTRAL COLORADO

Abstract Declining mule deer (Odocoileus hemionus) populations resulting from apparent low recruitment brought management and political focus on neonatal fawn survival. We captured mule deer fawns on the Uncompahgre Plateau (5,957 km2) in west-central Colorado, USA, at a mean age of 3 days (range = newborn to 6 days), and we radiomarked them with mortality-sensing, drop-off radiocollars. We radiomarked 230 fawns with samples of 50 in 1999, 88 in 2000, and 92 in 2001. Designated neonate survival period was from capture to 14 December. Survival was different among years (χ22 = 6.160, P = 0.046) with 3-year mean survival of 0.501. Cause-specific mortality ordered from highest to lowest was sick/starve, coyote, unknown, trauma, bear, and feline. Neither all predation combined (coyote, bear, feline; P = 0.379) nor coyote predation alone (P > 0.989) differed among years. By 31 July, 74% of the sick/starve mortality and 75% of the predation mortality had taken place, with 76% of mortality from all sources occurring by this date. Mean fawn weights at capture were different among years (P = 0.044). We also found a difference in hind-foot length among years (P = 0.002). Weight and hind-foot means were different between 2000 and 2001 (P > 0.017), with 1999 not different from either 2000 or 2001 (P < 0.017). Mean capture date was 19 June (SD = 4.83 days) and median capture date was 19 June (range = 9 Jun–6 Jul), with 94.8% of all captures occurring between 13 and 30 June. This implies that most does were bred during their first estrus cycle. Neonatal survival through 14 December did not completely account for observed low fawn:doe (f:d) ratios. We hypothesized that fetus mortality during late pregnancy or mortality of fawns at birth (before they could be detected for capture) as potential causes of poor recruitment.

Confidence Bands of Uniform and Proportional Width for Linear Models

Abstract In this paper confidence bands are given for simple linear models. The confidence bands that are given are straight lines rather than curves and are either (1) parallel or (2) trapezoidal. The confidence bands are over a finite length interval.

Influence of hunting on movements of female mule deer.

Hunting is a fundamentally important tool for wildlife managers. We examined the null hypothesis that hunting does not influence deer movement and their use of habitat types. Seventeen radio-collared, adult, female Rocky Mountain mule deer (Odocoileus hemionus hemionus) were located 1 day before the 1983 first Colorado deer season, and during day 2 of the first and day 3 of the second deer seasons in the foothills west of Fort Collins, Colorado. Distance from the preseason location to each location during hunting seasons were calculated for each deer. There were no differences between mean distance from pre-hunting season location to hunting season location for 10 deer that had all 3 locations in the area closed to hunting, and 4 deer that had 3 locations in the area open to hunting (P = 0.34 and 0.52). All 17 deer had all 3 locations in the interior of their minimum convex polygon home ranges. Those home ranges had a mean size of 226 ha and range of 117 to 323 ha. However, deer in the section open to hunting generally moved to vegetation types with increasingly better escape cover as the hunting seasons progressed. We conclude that hunting pressure did not cause deer movement in terms of distance or cause them to leave their normal home ranges, but did cause deer to move into more adequate cover.

MAXIMUM LIKELIHOOD ESTIMATION FOR MIXTURES OF TWO NORMAL DISTRIBUTIONS

This paper is primarily concerned with estimation of the parameters 1, al2 /-2 a22 and p in the mixture of two normal distributions when independent sample information is available from one of the populations. The solution to the maximum likelihood (ML) equations was obtained using Newtons iterative method. Some interesting results for the moment estimates were obtained for the case when independent sample observations are available from one population. Extensive Monte Carlo simulation was employed to obtain the sample variances of the estimates as well as the estimated asymptotic variances. The variances of the estimates are influenced by the separation of the two means with respect to the variances, the mixture proportion (p), and, of course, the size of the sample. When the number of observations is small and the means are not well separated, the sample variance of the estimates can be as much as three times greater than the estimated asymptotic variances.