https://cpw.cvlcollections.org/files/original/6c823a011247a49b4c60a39eaaf02efd.pdf f2361b4247bea54790c2e8d9cb457744 PDF Text Text The research in this publication was partially or fully funded by Colorado Parks and Wildlife. Dan Prenzlow, Director, Colorado Parks and Wildlife • Parks and Wildlife Commission: Marvin McDaniel, Chair • Carrie Besnette Hauser, Vice-Chair Marie Haskett, Secretary • Taishya Adams • Betsy Blecha • Charles Garcia • Dallas May • Duke Phillips, IV • Luke B. Schafer • James Jay Tutchton • Eden Vardy �The Journal of Wildlife Management 79(1):60–68; 2015; DOI: 10.1002/jwmg.801 Research Article Habitat and Herbivore Density: Response of Mule Deer to Habitat Management ERIC J. BERGMAN,1 Colorado Parks and Wildlife, 317 West Prospect Avenue, Fort Collins, CO 80526, USA PAUL F. DOHERTY JR., Department of Fish, Wildlife and Conservation Biology, Colorado State University, Fort Collins, CO 80523, USA GARY C. WHITE, Department of Fish, Wildlife and Conservation Biology, Colorado State University, Fort Collins, CO 80523, USA DAVID J. FREDDY, Colorado Parks and Wildlife, 317 West Prospect Avenue, Fort Collins, CO 80526, USA ABSTRACT The suite of demands competing for wildlife management funds necessitates direct assessment of management decisions, especially when these decisions have direct costs, as well as tangible opportunity costs. We conducted a mark–resight study that estimated mule deer (Odocoileus hemionus) density across multiple study units in southwest Colorado that had been exposed to different intensities of habitat treatments. Our treatments were comprised of common habitat management techniques including hydro-axe and roller-chopper disturbances as well chemical control of weeds and reseeding with desirable mule deer browse species. Reference study units received no habitat management treatments. Total deer densities varied between 20–84 deer/km2 in southern study units and 4–12 deer/km2 in northern study units. We did not observe a consistent pattern of higher deer density on advanced treatment study units despite it being the primary hypothesis of the study. We observed a wide range of variation in deer density among years. Resighting probabilities (range 0.070–0.567) were best modeled as an interactive function of study unit and year, although sampling method was also influential. We recommend that if population density is to be used as a population response variable, it be used in tandem with other, possibly more sensitive parameters such as overwinter survival or late winter body condition. Ó 2014 The Wildlife Society. KEY WORDS abundance, Colorado, environmental change, habitat management, mark–resight, mule deer, Odocoileus hemionus, resighting probabilities. As wildlife agencies implement management actions, assessing the impact of those actions on populations is necessary to improve the management process. Similarly, environmental change due to habitat succession, climate change, or human development also affects wildlife populations. Knowing which population parameters will be most affected by habitat change and determining if a response is biologically significant can be difficult to discern. An example can be found in habitat management. Management of landscapes for the benefit of mule deer has been a consideration for many decades (Wallmo 1981, Watkins et al. 2007). When assessment of these actions has occurred, it has focused on indirect response variables such as plant abundance, plant diversity, or changes to mule deer distribution in treated areas (Anderson 1969, Kufeld 1983, Barnitz et al. 1990, Long et al. 2008) with population impacts implied. However, the suite of demands competing for funds necessitates a more direct assessment of habitat management actions, as well as other sources of environmental change. Some wildlife managers speculate that the quantity of mule deer habitat has declined throughout western North America since the 1960s. In response, managers and biologists have Received: 26 April 2014; Accepted: 15 August 2014 Published: 12 November 2014 1 E-mail: eric.bergman@state.co.us 60 interest in maximizing the quality of remaining habitat via habitat treatments. Historically, treatments to improve mule deer habitat included burning and chaining (Kufeld 1983) but more recently have focused on hydro-axe and roller-chopper techniques (Watkins et al. 2007). However, implementation of habitat treatments is usually expensive. In the face of limited financial resources, delivery of habitat treatments also comes with the lost opportunity of conserving additional lands. To help inform decisions pertaining to this trade-off, an evaluation of the impact of habitat management techniques on mule deer habitat is warranted. An intuitive parameter for managers to describe population change is density, although estimating density for large ungulates can be cost prohibitive. Measurement of key vital rates such as sex ratios, age ratios, or survival is often more practical and less costly. Advancements tied to the merger of new analytical techniques with more efficient computing have made density estimation more tenable during the past 3 decades. With respect to large mammals, mark–resight models (McClintock and White 2007, Keech et al. 2011), sightability models (Samuel et al. 1987, Ackerman 1988, Anderson and Lindzey 1996, Walsh et al. 2009), population quadrat sampling (Gill 1969, Kufeld et al. 1980), and distance sampling (Schmidt et al. 2012) have struck a balance between cost of implementation and the desired level of precision. Within this, mark–resight and other capture– mark–recapture methodologies have proven to be particuThe Journal of Wildlife Management � 79(1) �Table 1. Number of marked mule deer (with number of resight flights) that we used to estimate abundance and density in each of 8 study units on the Uncompahgre Plateau in southwest Colorado. Study units are identified by name and location. The number of marked animals in each unit could decrease between flights because of mortality. The number of available marks increased between observation surveys because additional animals were captured as part of a separate research project. The area surveyed (km2) for each study unit varied among years. Unit (location) Sowbelly (North) Area surveyed Peach (North) Area surveyed Transfer (North) Area surveyed Shavano (South) Area surveyed Colona (South) Area surveyed McKenzie (South) Area surveyed Buckhorn (South) Area surveyed BCSWA (South) Area surveyed 2006 2007 2008 2009 10 and 40 (5)a 67 km2 20 (5) 52 km2 43 (6) 85 km2 24 (6) 51 km2 20 (6) 47 km2 22 (6) 84 km2 28 (5) 94 km2 19 and 18 (5)b 82 km2 18 (4) 49 km2 23 (3) 27 km2 64 (5) 15 km2 18 (3) 20 km2 24 and 23 (3)c 23 km2 47 (5) 19 km2 68 and 65 (5)d 19 km2 20 (5) 31 km2 61 (5) 19 km2 76 and 75 (5)e 19 km2 24 (4) 27 km2 54 (4) 24 km2 a 10 marks available on flights 1 through 3, 40 marks available on flights 4 and 5. 19 marks were available on flight 1, 18 marks available on flights 2 through 5. c 24 marks available on flight 1, 23 marks available on flights 2 and 3. d 68 marks were available on flights 1 through 3, 65 marks available on flights 4 and 5. e 76 marks were available on flight 1, 75 marks available on flights 2 through 5. b larly useful when the opportunity to have a meaningful proportion of marked individuals in the population is possible because these approaches do not hinge on assumptions of perfect detection or constant sightability (Williams et al. 2001). In an attempt to provide an assessment of habitat management actions on a mule deer population, we conducted a 4-year study that estimated mule deer density on multiple study units that had been exposed to different intensities of habitat treatments. One approach to evaluating habitat treatments is the measurement of vegetation metrics. However, many habitat treatments in Colorado are delivered with mule deer as the primary intended beneficiary. Thus, our objective was to use mule deer abundance as a tool to evaluate habitat treatments. Habitat treatments ranged in intensity from repeated treatment efforts on some units to no treatments on other units. Our hypotheses were largely congruent with results from other research on the same system during the same 4-year period. A 1.14� magnitude increase in fawn survival was documented on our treated study units (Bergman et al. 2014a). Likewise, a 1.06� magnitude increase in percent ingesta-free body fat for adult females was also documented (Bergman et al. 2014b). We hypothesized that these increases in fawn survival and body condition on study units that received habitat treatments would translate to higher deer densities, as compared to reference units. STUDY AREA We conducted this research on 8 study units (Table 1) on the Uncompahgre Plateau in southwest Colorado (Fig. 1). Study units fell between 388 150 N and 388 490 N latitudes and between 1078 410 W and 1088 280 W longitudes and elevations ranged between 1,670 m and 2,380 m. Study units were comprised of pinyon pine (Pinus edulis)-Utah juniper Bergman et al. � Habitat and Herbivore Density (Juniperus osteosperma) forests interspersed with open meadows. Forest openings typically held browse and grass species such as sagebrush (Artemesia spp.), cliffrose (Purshia mexicana), antelope bitterbrush (Purshia tridentate), mountain mahogany (Cercocarpus spp.), rabbittbrush (Ericameria spp.), western wheatgrass (Pascopyrum smithii), green needlegrass (Nassella viridula), Indian ricegrass (Achnatherum hymenoides), and bluegrass (Poa spp.). Closed canopy forests were late-seral stage with little or no understory vegetation. Depending on habitat treatment history (discussed below), vegetation in open canopy settings varied between late-seral stage, browse-dominated habitats to early-seral stage, browse-grass-forb communities. All study units were centered on public lands (U.S. Bureau of Land Management and Colorado State Wildlife Areas) although most study units had privately owned land at lower elevations. METHODS We selected study units based on habitat treatment history. Reference study units had no history of habitat management. Traditional treatment units received a single habitat treatment effort, and advanced treatment units received a traditional habitat treatment as well as a follow-up treatment. Traditional treatments were comprised of mechanical rollerchopper and hydro-axe treatments. Rollerchopper treatments were delivered by pulling a large drum affixed with perpendicular blades behind a tracked bulldozer. The bulldozer uprooted trees and other vegetation and subsequently pulled the drum over the newly downed vegetation, breaking it into smaller pieces. Rollerchopper treatments were effective at treating large areas (Watkins et al. 2007). Hydroaxe treatments were delivered by a boom-mounted mulching blade affixed to a reticulated, rubber-tired tractor. A hydroaxe mulched individual trees to ground level and was capable 61 �Figure 1. Map depicting Colorado Parks and Wildlife Data Analysis Unit (DAU) boundaries and the greater study area located on the Uncompahgre Plateau and neighboring valleys in the San Juan Mountains in southwest Colorado. The greater study area (solid gray DAUs), which encompassed the 8 study units (white polygons), is shown in relation to the surrounding communities of Delta and Montrose, Colorado (black circles). From north to south, study units included Sowbelly, Peach, Transfer, Shavano, Colona, Buckhorn, McKenzie, and Billy Creek State Wildlife Area (BCSWA). of a more refined approach to treating the landscape than rollerchopping (Watkins et al. 2007). Advanced habitat treatment efforts included vegetative reseeding and chemical control of weeds on the same piece of ground that had received a traditional habitat treatment. The seed mixes were comprised of key browse species for mule deer, including bitterbrush, cliffrose (Purshia mexicana), sagebrush, serviceberry (Amelanchier alnifolia), and four-wing saltbush (Atriplex canscens). Herbicide treatments were comprised of Plateau1 (imazpic; BASF, Research Triangle Park, NC), Milestone1 (aminopyralid: Dow AgroSciences, Indianapolis, IN), and glyphosate, and primarily targeted cheatgrass (Bromus tectorum) and jointed goatgrass (Aegilops cylindria). Mechanical habitat treatments were delivered within the previous 2–8 years (Bergman et al. 2014a,b). We did not incorporate potential study units that had received mechanical treatments within the previous 1–2 years because of uncertainty about time lags in vegetation response following mechanical disturbance. This time lag safeguarded against the potential for habitat quality to decline immediately following treatment, until browse species had opportunity to establish and grow under less competitive constraints (Young et al. 1985, Bates et al. 1998, Bates et al. 2000, Miller et al. 62 2000). For advanced treatment units, follow-up treatments were implemented at the same time as the deer population monitoring efforts in this study. To assess the effect of habitat treatment effort, and to control for spatial variation, advanced treatment units were spatially paired with reference units. Prior knowledge regarding lower annual precipitation at northern latitudes in the greater study area influenced this pairing (Fig. 1). Two study units were paired in the northern portion of the greater study area. The northernmost of these, Sowbelly, was a reference unit and Peach Orchard Point (Peach) was an advanced treatment study unit. A third study unit, Transfer Road (Transfer), was also located in the northern portion of the greater study area (Fig. 1) and was a traditional treatment study unit. We chose the location of all study units such that each was far enough isolated that deer exposure to other, non-research related habitat treatments was unlikely. All remaining study units were located in the southern portion of the greater study area. Shavano Valley (Shavano), Colona Tract (Colona), and McKenzie Buttes (McKenzie) were all traditional treatment study units (Fig. 1). The southernmost study unit was Billy Creek State Wildlife Area (BCSWA; Fig. 1), which was the southern advanced treatment study unit for which Buckhorn The Journal of Wildlife Management � 79(1) �Mountain (Buckhorn) was the paired reference study unit. We estimated mule deer density in Sowbelly, Peach, BCSWA, and Buckhorn during 4 consecutive years (2006–2009). We estimated density on the Shavano, Colona, McKenzie, and Transfer study units during 2006, 2007, 2008, and 2009, respectively. We collected winter severity data from the National Water and Climate Center. We used mean daily snow depth data for the month of March from 2 sites, Columbine Pass and Red Mountain Pass, as relative indicators of annual winter severity on our study units. Both weather stations were located within the same drainage but at higher elevations than our study units. We used snow data from the site located at Columbine Pass as an indicator for our 3 northern study units, and data from Red Mountain Pass as an indicator for our 5 southern study sites. Field Methods Mule deer capture and marking.—We captured mule deer and marked them as part of a larger research project (Bergman et al. 2014a,b). We captured all deer either by helicopter netgunning (Webb et al. 2008, Jacques et al. 2009) or baited drop-nets (Ramsey 1968, Schmidt et al. 1978, White and Bartmann 1994). Capture, handling and radio-collaring procedures for all aspects of this study were approved by the Institutional Animal Care and Use Committees at Colorado Parks and Wildlife (protocol #10-2005) and Colorado State University (protocol #08-2006A). The marked sample of deer available for our resighting flights included 6-month old fawns and adult females. All marked deer came from 1 of 3 different subsets of animals. The first subset was comprised of 6-month old fawns. At the onset of each winter (late Nov through early Dec), we captured 25 fawns on each study unit and fitted fawns with very high frequency (VHF) radio collars (Lotek, Inc., Newmarket, ON, Canada; Bergman et al. 2014a). All radio collars were constructed with tan canvas belting. To enhance visual detection by observers, we sewed either white or yellow rubber neckband material to the sides and top of each radio collar. The second subset of marked animals was comprised of adult female mule deer that were captured in early March of each winter for body condition scoring purposes (Bergman et al. 2014b). We captured 30 adult females on each of 2 study units each winter. During the first winter (2006), we captured deer on the Sowbelly and BCSWA study units. During the final 3 winters (2007–2009), we captured deer on the Buckhorn and BCSWA study units. During the winters of 2006 and 2007, we fitted adult females with VHF radio collars similar to those deployed on mule deer fawns. During the winters of 2008 and 2009, we fitted adult females with temporary neckbands that were either yellow (2008) or blue (2009) in color. We constructed temporary neckbands using surgical tubing that degraded when exposed to ultraviolet sunlight, making it unlikely that temporary collars would be retained in the next winter. We did not observe yellow neckbands (deployed in 2008) in 2009. The final subset of marked animals was comprised of surviving animals from earlier research projects (Bishop et al. 2009, Lukacs et al. 2009). Bergman et al. � Habitat and Herbivore Density This final subset of animals was unique to the Colona and Shavano study units. Some animals that had been fitted with VHF collars as part of these completed projects were still alive and relocated on these study units prior to our density reflights (<1 week). Although animals wearing non-functioning VHF collars possibly were on study units but not accounted for as marked deer available to be seen, this deflation in collar count would have been negligible. Regardless of which subset animals came from, and regardless of which study unit animals were located on, marks could not be individually identified by observation during resighting flights. Thus, we classified all deer that we encountered on flights as either marked or unmarked, based on visual observation of a collar. To address population closure concerns, we determined the number of VHF-marked deer that were available to be seen on each study unit during each aerial survey for each day that a survey occurred. For adult females fitted with neckbands, we could not make a similar determination so we assumed that these deer were available to be seen during each survey. Based on published survival rates of adult female deer in this area (Bishop et al. 2009, Lukacs et al. 2009), the daily survival rate of these animals was likely high (S^ ¼ 0.999), lending support to this assumption. Likewise, the mid-winter immigration and emigration rates of deer on these study units, based on the radio-collared subset of deer, was negligible (0.2%; 2014a), supporting the assumption that departure from the study units did not occur following capture. Aerial surveys.—Each year, we conducted aerial surveys during the last 2 weeks of March using a Bell 47-Soloy helicopter (Bell Helicopter, Hurst, TX). The same pilot and main observer conducted surveys on all flights. We surveyed study units 3–6 times each year. We flew all surveys at 55– 80 km/hr and 15–45 m above the ground. We typically conducted surveys on consecutive days, although because of weather delays and other conflicts this was not always possible. We defined the sampling boundaries for each study unit on a yearly basis by relocation data collected from radio-collared deer. Helicopter surveys followed topographical strip contours (Kufeld et al. 1980, Freddy et al. 2004). During flights, the pilot and observers used topographic features such as open ridges and washes to partition deer groups into observed and not observed categories. By incrementally following topographical contour lines, the pilot was able to partition previously unclassified deer groups and herd them into areas that contained previously observed deer, as they were being classified by the observers. This flight pattern minimized the potential for observed and not observed deer groups to mix, and minimized the distance that any single group of deer was pursued by the helicopter. If we needed less than 1 hr of flight time to survey an entire study unit, then we sampled the entire unit on each flight. Four of our study units—Sowbelly, Peach, Transfer, and McKenzie—were too large to be completely surveyed within 1 hour. For these units, we generated a random flight path within the study unit. We created random flight paths by overlaying each study unit with a grid composed of 1-km2 cells. We randomly selected 10 cells, without replacement, and created a flight path that efficiently incorporated each of these cells. Because the entire lengths of 63 �random flight paths were within the defined sampling area, we classified all observed deer. We used random flight paths for 1 flight and then replaced them with new random flight paths for each subsequent flight. Because of the process used to generate random flight paths, all marked animals had an equal probability of being available under each flight path. Resighting probabilities were not immediately comparable among study units. For smaller study units, every marked animal had the opportunity to be observed during each flight. For larger study units that we sampled using random flight paths, resighting probabilities were the product of the probability of being available for detection under a flight path and the probability of being detected. Statistical Analysis We analyzed all data using mark–resight models (McClintock and White 2011) in Program MARK (White and Burnham 1999). Mark–resight models largely replicate the results of the joint hypergeometric maximum likelihood estimator used in program NOREMARK (White 1996) but provide the ability to model heterogeneity in resighting probabilities and the ability to compare model results using model selection theory (Burnham and Anderson 2002). We used logit-normal models for all analyses because the exact number of marked animals on each study unit was assumed to be known with a high degree of confidence. However, marked animals could not be individually identified during aerial surveys. Because marked deer could not be individually identified, we fixed the parameter used to estimate the variance of individual heterogeneity (s2) in resighting probability (P) at 0 for all models. We calculated a single abundance estimate for each study unit during each winter. We divided annual abundance estimates by the size of each study unit for that year to generate density estimates. Logit-normal models allow for a high degree of flexibility in the estimation of resighting probabilities. To explore factors that influenced our resighting probabilities, we built a model set composed of 19 models. The simplest model generated a single estimate for resighting probability (i.e., no spatial or temporal patterns incorporated). More complex models allowed resighting probabilities to vary among years, study units, and flights within a single year. We also structured models to allow resighting probabilities to vary based on sampling method (i.e., smaller study units were completely surveyed during each flight, whereas larger study units were only partially surveyed using random flight paths). We compared the relative importance of each model by evaluating individual model weights and differences in Akaike’s Information Criterion values that were corrected for small sample size (DAICc; Burnham and Anderson 2002). To make comparisons about the relative importance of the different factors that influenced resighting probability on our flights, we built all possible combinations of additive models and multiplicative interaction models, as suggested by Doherty et al. (2012). RESULTS Mean daily snow depth during March at Columbine Pass, for the period on record (2004–2009), was 1.30 m (SD ¼ 0.28 m). For the 2006–2009 time period, March snow depths 64 Figure 2. Deviation from long-term mean daily snow depths during the month of March from 2 high elevation sites near study units in southwest Colorado. Data for the southern sites (black columns) were collected between 2000 and 2009. Data for the northern sites (white columns) were collected between 2004 and 2009. were below average every year except for 2008, during which the mean daily snow depth was 0.22 m greater than the 6year average (Fig. 2). The 10-year (2000–2009) mean snow depth at Red Mountain Pass was 1.64 m (SD ¼ 0.29 m). As was the case with Columbine Pass, mean daily snow depths at Red Mountain Pass were below the long-term average for 2006–2009 except for 2008 when the snow depth was 0.53 m greater than the 10-year mean (Fig. 2). Over the course of 4 years, we calculated late winter abundance and density estimates for the 8 study units. During the first winter of the study, sampling effort (i.e., the number of helicopter resight flights) was less than during subsequent years (Table 1). Based on estimates from the first winter, we increased sampling effort in all study units to improve the precision of estimates (Table 1). The area surveyed, as dictated by deer relocation data collected each winter, showed moderate variation among years (Table 1). The area surveyed for our southern study unit pairing (BCSWA and Buckhorn) varied between 19 km2 and 27 km2, whereas our northern study unit pairing (Peach and Sowbelly) varied between 47 km2 and 94 km2. Comparison among years for the remaining study units was not possible because we surveyed these units only during a single year (Table 1). Based on the above average snowpack during 2008, we expected a reduction in area sampled in all sites, especially the northern sites, but the only site where a reduction occurred was Sowbelly. Total deer densities in our southern study units varied between 20 deer/km2 and 84 deer/km2 (Fig. 3). Despite southern study units tending to have more marked deer available for resighting, the large number of deer in each southern unit resulted in a relatively low proportion of marked individuals. The mean estimate of the proportion of marked individuals varied between 0.016 and 0.146 for our southern study units. The coefficient of variation for density estimates in our southern study units ranged between 0.053 and 0.294. In our southern study units, we observed a consistent trend of higher mule deer density in our reference study units than in our advanced treatment and traditional treatment study units (Fig. 3). The Journal of Wildlife Management � 79(1) �Table 2. Model selection results for the 7 best mark–resight abundance estimation models for mule deer (Odocoileus hemionus) from 8 study units on the Uncompahgre Plateau and neighboring drainages in southwest Colorado. Model evaluation is based on Akaike’s Information Criterion that has been corrected for small sample size (AICc). We present only the model structure related to the estimation of resighting probabilities (P). The year covariate allowed for annual variation in P, sampling method accounted for different sampling strategies, and study unit and flight covariates allowed P to spatially and temporally vary, respectively. The total number of parameters in each model is also provided (K). Model Study unit � year Sampling method � year � flight Sampling method � year Study unit þ year þ flight Study unit þ year Sampling method þ year þ flight Sampling method þ year a b Model AICc DAICca weight likelihood Kb 0.00 0.02 4.46 7.94 10.54 14.09 17.15 0.471 0.466 0.051 0.009 0.002 0.001 0.000 1.000 0.988 0.108 0.019 0.005 0.001 0.000 40 59 28 36 31 30 25 AICc for the top model was 4,208.33. Parameter count for each model includes 20 parameters dedicated to abundance, although abundance estimates did not influence estimation of resighting probabilities. Figure 3. Mule deer (Odocoileus hemionus) density estimates, with 95% CIs, for 8 study units on the Uncompahgre Plateau in southwest Colorado. Northern study units are depicted in panel A. Southern study units are depicted in panel B. Reference study units are depicted by white bars, advanced treatment study units are depicted by light gray bars, and traditional treatment units are depicted by dark gray bars. Note the difference in scale between panels. Deer densities on our northern study units varied between 4 deer/km2 and 12 deer/km2. The mean estimate of the proportion of marked individuals in the northern study units varied between 0.044 and 0.096. The coefficient of variation for density estimates in our northern study units ranged between 0.107 and 0.417. Density estimates for the traditional treatment study units, geographically located between the northern and southern unit pairings varied between 5.77 deer/km2 and 36.95 deer/km2 (Fig. 3). The trend of higher deer density in our southern reference study unit was not evident in our northern units (Fig. 3). Based on our data, our top 2 models of resighting probabilities combined held the majority of support (AICc weight ¼ 0.471 and AICc weight ¼ 0.466, respectively; Table 2). However, the second of these models had 19 additional parameters (Table 2). The first model modeled resighting probability as an interactive function of study unit and year. The second model modeled resighting probability as an interactive function of sampling method, year, and flight (Table 2). When we considered the complete model set, year, sampling method, study unit, and flight had cumulative AICc weights of 1, 0.517, 0.483, and 0.475, respectively. Model-averaged resighting probabilities showed a high level of variation among units and years (range: 0.070– Bergman et al. � Habitat and Herbivore Density Figure 4. Model-averaged mark–resight resighting probabilities for mule deer (Odocoileus hemionus), with 95% CIs, from study units on the Uncompaghre Plateau and in neighboring drainages of the San Juan mountains in southwest Colorado. Northern study units are depicted in panel A. Southern study units are depicted in panel B. Reference study units are depicted by white bars, advanced treatment study units are depicted by light gray bars, and traditional treatment units are depicted by dark gray bars. 0.567; Fig. 4). Resighting probabilities tended to be lower in northern study units (0.070–0.310) than in our southern study units (0.151–0.567; Fig. 4). However, this result was largely driven by the different sampling methods used in 65 �large and small study sites. Based on model results that only accounted for sampling method, resighting probabilities for large study units (P ¼ 0.203, SE ¼ 0.012) were less than those from smaller study units (P ¼ 0.414, SE ¼ 0.011). If the resighting probability from our smaller study units is considered to be an applicable resighting estimate for large study estimates, we can derive the probability that deer would fall within the random resight paths that we used to survey large units. Under this assumption, the probability of a deer being included in the flight path is P ¼ 0.490. Finally, models built with multiplicative interaction structures consistently outperformed models with only additive effects (Table 2). DISCUSSION Based on patterns of deer density in our study units, density of mule deer on reference units appeared to be more dynamic than on units that had been exposed to habitat treatment efforts. The stable winter range density estimates on treated study units may have stemmed from habitat management efforts. However, we did not observe a consistent pattern of higher deer density on advanced treatment study units, as we had hypothesized. A general trend in deer density among our northern study units was difficult to infer (Fig. 3). During the 2007 and 2009 winters, no apparent difference in mule deer density occurred between our reference and advanced treatment study units (Fig. 3). We observed an increase in density on our reference study unit during 2008, as compared to other years and other study units (Fig. 3), but this was likely related to winter severity (Fig. 2). Potential explanations could include a reduction in winter range availability caused by the above average snowpack, or behavioral shifts in deer as they sought late seral stage forests (i.e., untreated areas) for thermal cover. We observed a more distinct pattern in density in our southern study units (Fig. 3). However, this pattern did not follow the prediction that we would observe higher densities in our advanced treatment study units (Fig. 3). As was the case in our northern study units, we believe the above average snowpack during 2008 caused a spike in deer density on our southern reference unit. If this data point is viewed as a stochastic outlier, a declining trend in deer density on our southern reference unit can be discerned. Our hypothesis was that habitat treatment management actions would result in greater densities of deer on treated landscapes. This hypothesis largely paralleled the predictions and results of simultaneous research conducted by Bergman et al. (2014a). The work of Bergman et al. (2014a), which was conducted on the same study units during the same time frame as our mark–resight density estimation flights, showed a 1.14� magnitude increase in overwinter survival rates of mule deer fawns on advanced treatment study units. This documented increase in survival was directly linked to the advanced habitat treatment management actions and was expected to result in higher densities of deer. However, the specific management actions that defined advanced treatment units were implemented as part of this research project only during the summer of 2006 in 66 Peach and during the summers of 2006 and 2007 in BCSWA. If increases in overall density were to occur, they likely would have been easier to detect after the increased survival of fawns had been allowed to compound for a longer period of time. For example, research in other parts of Colorado has shown that when using similar helicopter mark–resight surveys on mule deer, detecting a 30–40% change in density may take 5–6 years (C. Anderson, Colorado Parks and Wildlife, unpublished data). Alternatively, if the downward trend in density observed on the Buckhorn reference study unit during 2006, 2007, and 2009 (Fig. 3) reflects a true population trajectory, the lack of a similar trend on BCSWA may be important. If habitat management can lead to a local stabilization in density in light of concurrent declines in neighboring areas, this has important management implications. Unfortunately, stochastic annual variation and relatively large estimates of variance challenge the robustness of such conclusions. As expected, resighting probabilities differed between large and small study units. For large study units, the estimated resighting probabilities were the product of the probability of being available to be observed and the probability of being observed (i.e., the probability that an animal was present under a random flight path and was then seen). As study unit size transitions from large to small, the probability of an animal being under the randomly generated flight path should increase. This pattern can generally be observed in McKenzie and Transfer, which had larger areas surveyed (Table 1) and smaller resighting probabilities (Fig. 4) when compared to smaller study units. Based on published visibility results, our resighting probabilities aligned with rates reported for smaller group sizes of animals and higher percentage of vegetation cover (Samuel et al. 1987). Ultimately, the large and stochastic annual variation in density estimates on our individual study units demonstrates that large fluctuations in density do occur among years. We believe that these fluctuations were primarily driven by changes in the amount of area used by deer, as opposed to the total number of deer using the areas. However, this variation was dampened in advanced treatment study units (Fig. 3). This phenomenon may be attributable to habitat management efforts. Annual variability in winter severity likely explains the spike in density observed on our reference units during 2008. Although the spike in density on Buckhorn during 2008 resulted in an abnormally high estimate, we feel this spike was an artifact of annual variation and we do not believe such a density could be sustainable for any extended period of time. The relationship between increasing snow depth at higher elevations and the concentration of animals on limited, lower elevation habitat has been established for several species (Gilbert et al. 1970, Bruggeman et al. 2009). Presumably, as conditions at higher elevations were less accommodating during the 2008 winter, overall density should have increased on all study units as deer moved to lower elevations and concentrated in smaller areas. Our data indicate this did occur on our reference study units but not on our advanced treatment study units. We speculate that deer that traditionally spent winter on study units that had The Journal of Wildlife Management � 79(1) �experienced habitat improvement efforts did not face the same food limitation stress as deer that wintered on reference units and thus moved onto winter range regardless of winter severity. A similar phenomenon with spring migration has been observed in that various species follow the spring plant phenological progression of green-up (Mysterud et al. 2001). We believe deer are migrating to lower elevations because of a building snowpack that forces animals off summer and transition range. Specifically, deer that overwinter on habitat that offers poorer quality, and less abundant forage, are more reluctant to move down in elevation (i.e., leave summer and transition range) and only do so when confronted with extreme conditions. Individuals that anticipate high quality habitat and abundant forage will move prior to being forced off of summer range. However, these ecological scenarios do not explain the lower density of deer observed on our treatment areas. The most viable explanation for this counterintuitive result is that our treatment study units, and particularly BCSWA in the south, served as winter range for a smaller portion of summer range than the reference unit. Unfortunately, we did not have the ability to validate this scenario by capturing deer on summer range and tracking their movements onto winter range. Ultimately, the use of density as a population response parameter for habitat assessment, as our results demonstrate, may not be ideal. Similar conclusions have been reached in the past (Van Horne 1983, Hobbs and Hanley 1990). In particular, simulation models presented by Hobbs and Hanley (1990) provide an explanation that aligns well with our results as well as from other research in our study system (Bergman et al. 2014a). Hobbs and Hanley (1990) demonstrated that animals living in habitat with lower resource quality might subsequently experience reduced reproductive output. For populations to be maintained under these conditions, a higher density of productive animals is necessary. As demonstrated by Bergman et al. (2014a), overwinter survival of 6-month old mule deer fawns was lower in our reference study units. When lower juvenile survival rates and higher overall densities are considered in tandem, they can be viewed as evidence that our reference study units may have been near their local carrying capacity. This conclusion may be further validated if the potential downward trend in density on our southern reference study unit is viewed as an actual population trajectory and not a spurious effect. Evaluation of habitat management actions, especially when financial resources are limited, is important to judge the effectiveness of actions and dollars invested. Within this evaluation, a simultaneous assessment of different population parameters is prudent. Despite a relatively high level of effort, as well as a relatively high financial cost, the density and abundance estimation procedures we employed were not particularly sensitive. Overall, research from this system suggests that response variables more closely aligned with fitness traits (e.g., survival of young as well as reproductive success combined with body condition of adult females) may be most appropriate for assessing the wildlife population– habitat relationship versus density estimation. The intensity Bergman et al. � Habitat and Herbivore Density of landscape treatments was high in our study. Similar intensities are unlikely to occur during short time intervals when mediated by climate change or habitat succession. Likewise, during our study the intensity of population monitoring exceeded that of routine population management. This combination of factors exacerbates the difficulty in relying on density as a response variable. Ultimately, if population abundance and density are to be used as demographic response variables, we encourage their use in tandem with other fine-scale population response variables such as survival rates and body condition. MANAGEMENT IMPLICATIONS Despite a relatively high level of effort, as well as a relatively high financial cost, the density estimation procedures we employed were not particularly sensitive. Our results also stand as cautionary evidence in the use of density estimates to evaluate the impacts of other environmental change. Our habitat treatments created a large change on the landscape; naturally induced change can be expected to be more subtle. Interpreting changes in density amidst wide annual variation should be done with caution. Ultimately, if population abundance and density are to be used as demographic response variables, we encourage their use in tandem with other fine-scale population response variables such as survival rates and body condition. ACKNOWLEDGMENTS Financial support for this research was provided by Colorado Federal Aid Wildlife Restoration Project Funding, the Colorado Division of Parks and Wildlife Habitat Partnership Program, the Mule Deer Foundation, and the Colorado Division of Parks and Wildlife Big Game Auction and Raffle Grant program. Delivery and funding of mechanical treatments were largely coordinated by the Uncompahgre Plateau Project, a non-profit group composed of many agencies, organizations, and individuals who saw the value in working cooperatively to improve wildlife habitat. We are indebted to A. Cline, C. Harty, B. Lamont, R. Lockwood, D. Lucchesi, J. McMillan, C. Santana, C. Tucker, and K. Yeager for their assistance with fieldwork. Fixed wing pilots S. Waters, L. Gepfert, and D. Felix provided assistance with aerial telemetry flights and helicopter pilots R. Swisher and M. Shelton helped with capture. D. Felix piloted many hours of helicopter flight time during resight surveys. His demeanor, attitude, and attention to animal welfare made resight flights an enjoyable and safe experience. We are especially appreciative of the willingness of C. Harty, R. Lockwood, D. Coven, K. Yeager, and C. Santana to serve as the second observer on resighting flights. The time and energy provided by Colorado Division of Parks and Wildlife Area 18 personnel were instrumental to this research. Valuable insight, discussion, and support were also provided by A. Holland, B. Banulis, B. Watkins, C. Anderson, and R. Kahn. Preliminary reviews of this paper were provided by C. Anderson and H. Johnson. 67 �LITERATURE CITED Ackerman, B. B. 1988. Visibility bias of mule deer aerial census procedures in southeast Idaho. Dissertation. University of Idaho, Moscow, USA. Anderson, A. E. 1969. 2,4-D, sagebrush, and mule deer–cattle use of upper winter range. Colorado Division of Wildlife, Special Report No. 21, Fort Collins, USA. Anderson, C. R. Jr., and F. G. Lindzey. 1996. A sightability model for moose developed from helicopter surveys. Wildlife Society Bulletin 24:247–259. Barnitz, J. A., V. W. Howard, and G. M. Southward. 1990. Mule deer and rabbit use of areas of pinyon-juniper woodland treated by two-way cabling. New Mexico Agricultural Experiment Station, Bulletin 752. New Mexico State University, Las Cruces, USA. Bates, J. D., R. F. Miller, and T. Svejcar. 1998. Understory patterns in cut western juniper (Juniperus occidentalis spp.) woodlands. Great Basin Naturalist 58:363–374. Bates, J. D., R. F. Miller, and T. J. Svejcar. 2000. Understory dynamics in cut and uncut western juniper woodlands. Journal of Range Management 53:119–126. Bergman, E. J., C. J. Bishop, D. J. Freddy, G. C. White, and P. F. Doherty. Jr., 2014a. Habitat management influences overwinter survival of mule deer fawns in Colorado. Journal of Wildlife Management 78:448–455. Bergman, E. J., P. F. Doherty, Jr., C. J. Bishop, L. L. Wolfe, and B. A. Banulis. 2014b. Herbivore body condition response in altered environments: mule deer and habitat management. PLoS ONE 9:e106374. Bishop, C. J., G. C. White, D. J. Freddy, B. E. Watkins, and T. R. Stephenson. 2009. Effect of enhanced nutrition on mule deer population rate of change. Wildlife Monographs 172. Bruggeman, J. E., R. A. Garrott, P. J. White, F. G. R. Watson, and R. W. Wallen. 2009. Effects of snow and landscape attributes on bison winter travel patterns and habitat use. Pages 623–647 in R. A., Garrott P. J. White and F. G. R. Watson, editors. The ecology of large mammals in central Yellowstone: sixteen years of integrated field studies. Academic Press, Boston, Massachusetts, USA. Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Second edition. Springer, New York, New York, USA. Doherty, P. F., G. C. White, and K. P. Burnham. 2012. Comparison of model building and selection strategies. Journal of Ornithology 152:317– 323. Freddy, D. J., G. C. White, M. C. Kneeland, R. H. Kahn, J. W. Unsworth, W. J. DeVergie, V. K. Graham, J. H. Ellenberger, and C. H. Wagner. 2004. How many mule deer are there? Challenges of credibility in Colorado. Wildlife Society Bulletin 32:916–927. Gilbert, P. F., O. C. Wallmo, and R. B. Gill. 1970. Effect of snow depth on mule deer in Middle Park, Colorado. Journal of Wildlife Management 34:15–22. Gill, R. B. 1969. A quadrat count system for estimating game populations. Colorado Game, Fish and Parks Game Information Leaflet 76, Fort Collins, USA. Hobbs, N. T., and T. A. Hanley. 1990. Habitat evaluation: do use/ availability data reflect carrying capacity? Journal of Wildlife Management 54:545–522. Jacques, C. N., J. A. Jenks, C. S. Deperno, J. D. Sievers, T. W. Grovenburg, T. J. Brinkman, C. C. Swanson, and B. A. Stillings. 2009. Evaluating ungulate mortality associated with helicopter net-gun captures in the Northern Great Plains. Journal of Wildlife Management 73:1282–1291. Keech, M. A., M. S. Lindberg, R. D. Boertje, P. Valkenburg, B. D. Taras, T. A. Boudreau, and K. B. Beckmen. 2011. Effects of predator treatments, individual traits, and environment on moose survival in Alaska. Journal of Wildlife Management 75:1361–1380. Kufeld, R. 1983. Responses of elk, mule deer, cattle and vegetation to burning, spraying, and chaining of Gambel oak rangeland. Colorado Division of Wildlife, Technical Publication No. 34, Ft. Collins, USA. 68 Kufeld, R. C., J. H. Oterman, and D. C. Bowden. 1980. A helicopter quadrat census for mule deer on the Uncompahgre Plateau, Colorado. Journal of Wildlife Management 44:632–639. Long, R. A., J. L. Rachlow, and J. G. Kie. 2008. Effects of season and scale on response of elk and mule deer to habitat manipulation. Journal of Wildlife Management 72:1133–1142. Lukacs, P. M., G. C. White, B. E. Watkins, R. H. Kahn, B. A. Banulis, D. J. Finley, A. A. Holland, J. A. Martens, and J. Vayhinger. 2009. Separating components of variation in survival of mule deer in Colorado. Journal of Wildlife Management 73:817–826. McClintock, B. T., and G. C. White. 2007. Bighorn sheep abundance following a suspected pneumonia epidemic in Rocky Mountain National Park. Journal of Wildlife Management 71:183–189. McClintock, B. T., and G. C. White. 2011. From NOREMARK to MARK: software for estimating demographic parameters using mark– resight methodology. Journal of Ornithology 152:S641–S650. Miller, R. F., T. J. Svejcar, and J. A. Rose. 2000. Impacts of western juniper on plant community composition and structure. Journal of Range Management 53:574–585. Mysterud, A., R. Langvatn, N. G. Yoccoz, and N. C. Stenseth. 2001. Plant phenology, migration and geographical variation in body weight of a large herbivore: the effect of variable topography. Journal of Animal Ecology 70:915–923. Ramsey, C. W. 1968. A drop-net deer trap. Journal of Wildlife Management 32:187–190. Samuel, M. D., E. O. Garton, M. W. Schlegel, and R. G. Carson. 1987. Visibility bias during aerial surveys of elk in northcentral Idaho. Journal of Wildlife Management 51:622–630. Schmidt, J. H., K. L. Rattenbury, J. P. Lawler, and M. C. Maccluskie. 2012. Using distance sampling and hierarchical models to improve estimates of Dall’s sheep abundance. Journal of Wildlife Management 76:317–327. Schmidt, R. L., W. H. Rutherford, and F. M. Bodenham. 1978. Colorado bighorn sheep-trapping techniques. Wildlife Society Bulletin 6:159–163. Van Horne, B. 1983. Density as a misleading indicator of habitat quality. Journal of Wildlife Management 47:893–901. Wallmo, O. C. 1981. Mule and black-tailed deer of North America. University of Nebraska Press, Lincoln, USA. Walsh, D. P., C. F. Page, H. Campa, S. R. Winterstein, and D. E. Beyer. 2009. Incorporating estimates of group size in sightability models for wildlife. Journal of Wildlife Management 73:136–143. Watkins, B. E., C. J. Bishop, E. J. Bergman, B. Hale, B. F. Wakeling, A. Bronson, L. H. Carpenter, and D. W. Lutz. 2007. Habitat guidelines for mule deer: Colorado Plateau shrubland and forest ecoregion. Mule Deer Working Group, Western Association of Fish and Wildlife agencies, Cheyenne, Wyoming, USA. Webb, S. L., J. S. Lewis, D. G. Hewitt, M. W. Hellickson, and F. C. Bryant. 2008. Assessing the helicopter and net gun as a capture technique for white-tailed deer. Journal of Wildlife Management 72:310–314. White, G. C. 1996. NOREMARK: population estimation from mark– resighting surveys. Wildlife Society Bulletin 24:50–52. White, G. C., and R. M. Bartmann. 1994. Drop nets versus helicopter net guns for capturing mule deer fawns. Wildlife Society Bulletin 22:248–252. White, G. C., and K. P. Burnham. 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46:120–139. Williams, B. K., J. D. Nichols, and M. J. Conroy. 2001. Analysis and management of animal populations. Academic Press, San Diego, California, USA. Young, J. A., R. A. Evans, and C. Rimbey. 1985. Weed control and revegetation following western juniper (Juniperus occidentalis) control. Weed Science 33:513–517. Associate Editor: David Euler. The Journal of Wildlife Management � 79(1) � Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Journal Articles Description An account of the resource CPW peer-reviewed journal publications Text A resource consisting primarily of words for reading. Examples include books, letters, dissertations, poems, newspapers, articles, archives of mailing lists. Note that facsimiles or images of texts are still of the genre Text. Dublin Core The Dublin Core metadata element set is common to all Omeka records, including items, files, and collections. For more information see, http://dublincore.org/documents/dces/. Title A name given to the resource Habitat and herbivore density: response of mule deer to habitat management Description An account of the resource <span>The suite of demands competing for wildlife management funds necessitates direct assessment of management decisions, especially when these decisions have direct costs, as well as tangible opportunity costs. We conducted a mark–resight study that estimated mule deer (</span><i>Odocoileus hemionus</i><span>) density across multiple study units in southwest Colorado that had been exposed to different intensities of habitat treatments. Our treatments were comprised of common habitat management techniques including hydro-axe and roller-chopper disturbances as well chemical control of weeds and reseeding with desirable mule deer browse species. Reference study units received no habitat management treatments. Total deer densities varied between 20–84 deer/km</span>²<span> in southern study units and 4–12 deer/km</span>²<span> in northern study units. We did not observe a consistent pattern of higher deer density on advanced treatment study units despite it being the primary hypothesis of the study. We observed a wide range of variation in deer density among years. Resighting probabilities (range 0.070–0.567) were best modeled as an interactive function of study unit and year, although sampling method was also influential. We recommend that if population density is to be used as a population response variable, it be used in tandem with other, possibly more sensitive parameters such as overwinter survival or late winter body condition.</span> Bibliographic Citation A bibliographic reference for the resource. Recommended practice is to include sufficient bibliographic detail to identify the resource as unambiguously as possible. Bergman, E. J., P. F. Doherty, G. C. White, and D. J. Freddy. 2014. Habitat and herbivore density: response of mule deer to habitat management. The Journal of Wildlife Management 79:60-68. <a href="https://doi.org/10.1002/jwmg.801" target="_blank" rel="noreferrer noopener">https://doi.org/10.1002/jwmg.801</a> Creator An entity primarily responsible for making the resource Bergman, Eric J. Doherty Jr, Paul F. White, Gary C. Freddy, David J. Subject The topic of the resource Abundance Colorado Environmental change Habitat management Mark–resight Mule deer <em>Odocoileus hemionus</em> Resighting probabilities Extent The size or duration of the resource. 9 pages Date Created Date of creation of the resource. 2014-11-12 Rights Information about rights held in and over the resource <a href="http://rightsstatements.org/vocab/InC-NC/1.0/" target="_blank" rel="noreferrer noopener">In Copyright - Non-Commercial Use Permitted</a> Format The file format, physical medium, or dimensions of the resource application/pdf Language A language of the resource English Is Part Of A related resource in which the described resource is physically or logically included. The Journal of Wildlife Management Type The nature or genre of the resource Article