A Comparison of Two Modeling Approaches for Evaluating Wildlife-Habitat Relationships

ABSTRACT Studies of resource selection form the basis for much of our understanding of wildlife habitat requirements, and resource selection functions (RSFs), which predict relative probability of use, have been proposed as a unifying concept for analysis and interpretation of wildlife habitat data. Logistic regression that contrasts used and available or unused resource units is one of the most common analyses for developing RSFs. Recently, resource utilization functions (RUFs) have been developed, which also predict probability of use. Unlike RSFs, however, RUFs are based on a continuous metric of space use summarized by a utilization distribution. Although both RSFs and RUFs predict space use, a direct comparison of these 2 modeling approaches is lacking. We compared performance of RSFs and RUFs by applying both approaches to location data for 75 Rocky Mountain elk (Cervus elaphus) and 39 mule deer (Odocoileus hemionus) collected at the Starkey Experimental Forest and Range in northeastern Oregon, USA. We evaluated differences in maps of predicted probability of use, relative ranking of habitat variables, and predictive power between the 2 models. For elk, 3 habitat variables were statistically significant (P < 0.05) in the RSF, whereas 7 variables were significant in the RUF. Maps of predicted probability of use differed substantially between the 2 models for elk, as did the relative ranking of habitat variables. For mule deer, 4 variables were significant in the RSF, whereas 6 were significant in the RUF, and maps of predicted probability of use were similar between models. In addition, distance to water was the top-ranked variable in both models for mule deer. Although space use by both species was predicted most accurately by the RSF based on cross-validation, differences in predictive power between models were more substantial for elk than mule deer. To maximize accuracy and utility of predictive wildlife-habitat models, managers must be aware of the relative strengths and weaknesses of different modeling techniques. We conclude that although RUFs represent a substantial advance in resource selection theory, techniques available for generating RUFs remain underdeveloped and, as a result, RUFs sometimes predict less accurately than models derived using more conventional techniques.     

The effect of terrain and female density on survival of neonatal white‐tailed deer and mule deer fawns

Juvenile survival is a highly variable life‐history trait that is critical to population growth. Antipredator tactics, including an animal's use of its physical and social environment, are critical to juvenile survival. Here, we tested the hypothesis that habitat and social characteristics influence coyote (Canis latrans) predation on white‐tailed deer (Odocoileus virginianus) and mule deer (O. hemionus) fawns in similar ways during the neonatal period. This would contrast to winter when the habitat and social characteristics that provide the most safety for each species differ. We monitored seven cohorts of white‐tailed deer and mule deer fawns at a grassland study site in Alberta, Canada. We used logistic regression and a model selection procedure to determine how habitat characteristics, climatic conditions, and female density influenced fawn survival during the first 8 weeks of life. Fawn survival improved after springs with productive vegetation (high integrated Normalized Difference Vegetation Index values). Fawns that used steeper terrain were more likely to survive. Fawns of both species had improved survival in years with higher densities of mule deer females, but not with higher densities of white‐tailed deer females, as predicted if they benefit from protection by mule deer. Our results suggest that topographical variation is a critical resource for neonates of many ungulate species, even species like white‐tailed deer that use more gentle terrain when older. Further, our results raise the possibility that neonatal white‐tailed fawns may benefit from associating with mule deer females, which may contribute to the expansion of white‐tailed deer into areas occupied by mule deer.     

Scale for resource selection functions

Resource selection functions (RSFs) are statistical models defined to be proportional to the probability of use of a resource unit. My objective with this review is to identify how RSFs can be used to unravel the influence of scale in habitat selection. In wildlife habitat studies, including radiotelemetry, RSFs can be estimated using a variety of statistical methods, all of which can be used to explore the role of scale. All RSFs are bounded by the resolution of data and the spatial extent of the study area, but also allow predictor covariates to be measured at a variety of scales. Conditional logistic regression permits designs (e.g. matched case) that relate the process of habitat selection to a limited domain of resource units that might better characterize what is truly ‘available’ to the animal. Scale influences the process of habitat selection, e.g. food resources are often selected at fine spatial scales, whereas landscape patterns at much larger scales typically influence the location of home ranges. Scale also influences appropriate sampling in many ways: (1) heterogeneity might be obliterated (transmutation) if resolution or grain size is too large, (2) variance of habitat characteristics might be undersampled if extent or domain is too small, (3) timing and duration of observations can influence RSF models, and (d) both spatial and temporal autocorrelations can vary directly with the intensity of sampling. Using RSFs, researchers can examine habitat selection at multiple scales, and predictive models that bridge scales can be estimated. Using Geographical Information Systems, predictor covariates in RSF models can be measured at different scales easily so that the predictive ability of models at alternative spatial and temporal domains can be explored by the investigator. Identification of the scale that best explains the data can be evaluated by comparing alternative models using information‐theoretic metrics such as Akaike Information Criteria, and predictive capability of the models can be assessed using k‐fold cross validation.     

Influence of Well Pad Activity on Winter Habitat Selection Patterns of Mule Deer

ABSTRACT Conversion of native winter range into producing gas fields can affect the habitat selection and distribution patterns of mule deer (Odocoileus hemionus). Understanding how levels of human activity influence mule deer is necessary to evaluate mitigation measures and reduce indirect habitat loss to mule deer on winter ranges with natural gas development. We examined how 3 types of well pads with varying levels of vehicle traffic influenced mule deer habitat selection in western Wyoming during the winters of 2005–2006 and 2006–2007. Well pad types included producing wells without a liquids gathering system (LGS), producing wells with a LGS, and well pads with active directional drilling. We used 36,699 Global Positioning System locations collected from a sample (n = 31) of adult (>1.5-yr-old) female mule deer to model probability of use as a function of traffic level and other habitat covariates. We treated each deer as the experimental unit and developed a population-level resource selection function for each winter by averaging coefficients among models for individual deer. Model coefficients and predictive maps for both winters suggested that mule deer avoided all types of well pads and selected areas further from well pads with high levels of traffic. Accordingly, impacts to mule deer could probably be reduced through technology and planning that minimizes the number of well pads and amount of human activity associated with them. Our results suggested that indirect habitat loss may be reduced by approximately 38–63% when condensate and produced water are collected in LGS pipelines rather than stored at well pads and removed via tanker trucks. The LGS seemed to reduce long-term (i.e., production phase) indirect habitat loss to wintering mule deer, whereas drilling in crucial winter range created a short-term (i.e., drilling phase) increase in deer disturbance and indirect habitat loss. Recognizing how mule deer respond to different types of well pads and traffic regimes may improve the ability of agencies and industry to estimate cumulative effects and quantify indirect habitat losses associated with different development scenarios.     

Mule deer and energy development—Long‐term trends of habituation and abundance

As the extent and intensity of energy development in North America increases, so do disturbances to wildlife and the habitats they rely upon. Impacts to mule deer are of particular concern because some of the largest gas fields in the USA overlap critical winter ranges. Short‐term studies of 2–3 years have shown that mule deer and other ungulates avoid energy infrastructure; however, there remains a common perception that ungulates habituate to energy development, and thus, the potential for a demographic effect is low. We used telemetry data from 187 individual deer across a 17‐year period, including 2 years predevelopment and 15 years during development, to determine whether mule deer habituated to natural gas development and if their response to disturbance varied with winter severity. Concurrently, we measured abundance of mule deer to indirectly link behavior with demography. Mule deer consistently avoided energy infrastructure through the 15‐year period of development and used habitats that were an average of 913 m further from well pads compared with predevelopment patterns of habitat use. Even during the last 3 years of study, when most wells were in production and reclamation efforts underway, mule deer remained >1 km away from well pads. The magnitude of avoidance behavior, however, was mediated by winter severity, where aversion to well pads decreased as winter severity increased. Mule deer abundance declined by 36% during the development period, despite aggressive onsite mitigation efforts (e.g. directional drilling and liquid gathering systems) and a 45% reduction in deer harvest. Our results indicate behavioral effects of energy development on mule deer are long term and may affect population abundance by displacing animals and thereby functionally reducing the amount of available habitat.     

Increases in residential and energy development are associated with reductions in recruitment for a large ungulate

Land‐use change due to anthropogenic development is pervasive across the globe and commonly associated with negative consequences for biodiversity. While land‐use change has been linked to shifts in the behavior and habitat‐use patterns of wildlife species, little is known about its influence on animal population dynamics, despite the relevance of such information for conservation. We conducted the first broad‐scale investigation correlating temporal patterns of land‐use change with the demographic rates of mule deer, an iconic species in the western United States experiencing wide‐scale population declines. We employed a unique combination of long‐term (1980–2010) data on residential and energy development across western Colorado, in conjunction with congruent data on deer recruitment, to quantify annual changes in land‐use and correlate those changes with annual indices of demographic performance. We also examined annual variation in weather conditions, which are well recognized to influence ungulate productivity, and provided a basis for comparing the relative strength of different covariates in their association with deer recruitment. Using linear mixed models, we found that increasing residential and energy development within deer habitat were correlated with declining recruitment rates, particularly within seasonal winter ranges. Residential housing had two times the magnitude of effect of any other factor we investigated, and energy development had an effect size similar to key weather variables known to be important to ungulate dynamics. This analysis is the first to correlate a demographic response in mule deer with residential and energy development at large spatial extents relevant to population performance, suggesting that further increases in these development types on deer ranges are not compatible with the goal of maintaining highly productive deer populations. Our results underscore the significance of expanding residential development on mule deer populations, a factor that has received little research attention in recent years, despite its rapidly increasing footprint across the landscape.     

Seasonal shifts in habitat selection of a large herbivore and the influence of human activity

Failure to recognize factors contributing to variation in habitat models like resource selection functions (RSFs) can affect their application for projecting probabilities of occurrence, and thereby limit their relevance for conservation and management. We compared seasonal RSFs (2006–2008) for 16 adult female moose (Alces alces) with home ranges located in western Algonquin Provincial Park (APP), Ontario, Canada, to those of 14 adult females located in provincial Wildlife Management Unit (WMU) 49, 40 km west of the protected area. Wildlife and habitat management practices differed between regions: hunting was higher in WMU 49 compared to APP, and APP preserved large tracts of old growth forest rarely found in WMU 49. Seasonal RSFs projected expected similarities in moose resource use between regions (e.g., responses to wetlands and stands of eastern hemlock, Tsuga canadensis [in winter]); however, we also observed differences consistent with the hypothesis that animals, through effects of hunting, would shift habitat use seasonally and in response to roads. We further observed evidence of functional responses in habitat selection due to underlying differences in forestry practices (e.g., responses to stands of old-growth hemlock forest). Given the close proximity and shared biogeographic region between study areas, we believe that observed spatial dynamics in RSFs were ultimately reflective of divergent management strategies between areas and ensuing differences in predation and hunting mortality risk, and functional habitat.     

Are polar bear habitat resource selection functions developed from 1985–1995 data still useful?

相似文献   

Does small-perimeter fencing inhibit mule deer or pronghorn use of water developments?

Wildlife water development can be an important habitat management strategy in western North America for many species, including both pronghorn (Antilocapra americana) and mule deer (Odocoileus hemionus). In many areas, water developments are fenced (often with small-perimeter fencing) to exclude domestic livestock and feral horses. Small-perimeter exclosures could limit wild ungulate use of fenced water sources, as exclosures present a barrier pronghorn and mule deer must negotiate to gain access to fenced drinking water. To evaluate the hypothesis that exclosures limit wild ungulate access to water sources, we compared use (photo counts) of fenced versus unfenced water sources for both pronghorn and mule deer between June and October 2002–2008 in western Utah. We used model selection to identify an adequate distribution and best approximating model. We selected a zero-inflated negative binomial distribution for both pronghorn and mule deer photo counts. Both pronghorn and mule deer photo counts were positively associated with sampling time and average daily maximum temperature in top models. A fence effect was present in top models for both pronghorn and mule deer, but mule deer response to small-perimeter fencing was much more pronounced than pronghorn response. For mule deer, we estimated that presence of a fence around water developments reduced photo counts by a factor of 0.25. We suggest eliminating fencing of water developments whenever possible or fencing a big enough area around water sources to avoid inhibiting mule deer. More generally, our results provide additional evidence that water development design and placement influence wildlife use. Failure to account for species-specific preferences will limit effectiveness of management actions and could compromise research results. © 2011 The Wildlife Society.     

Migratory Disturbance Thresholds with Mule Deer and Energy Development

Fine-scale movement data has transformed our knowledge of ungulate migration ecology and now provides accurate, spatially explicit maps of migratory routes that can inform planning and management at local, state, and federal levels. Among the most challenging land use planning issues has been developing energy resources on public lands that overlap with important ungulate habitat, including the migratory routes of mule deer (Odocoileus hemionus). We generally know that less development is better for minimizing negative effects and maintaining habitat function, but we lack information on the amount of disturbance that animals can tolerate before reducing use of or abandoning migratory habitat. We used global positioning system data from 56 deer across 15 years to evaluate how surface disturbance from natural gas well pads and access roads in western Wyoming, USA, affected habitat selection of mule deer during migration and whether any disturbance threshold(s) existed beyond which use of migratory habitat declined. We used resource and step selection functions to examine disturbance thresholds at 3 different spatial scales. Overall, migratory use by mule deer declined as surface disturbance increased. Based on the weight of evidence from our 3 independent but complementary metrics, declines in migratory use related to surface disturbance were non-linear, where migratory use sharply declined when surface disturbance from energy development exceeded 3%. Disturbance thresholds may vary across regions, species, or migratory habitats (e.g., stopover sites). Such information can help with management and land use decisions related to mineral leasing and energy development that overlap with the migratory routes of ungulates. © 2020 The Wildlife Society.     

Embracing Complexity and Context to Improve Science Communication

Members of the public play a primary role in successful implementation of wildlife management plans, making communication between scientists and the public a vital component of wildlife management. Although there is substantial public interest in the health of ungulate populations, stakeholder perspectives can vary widely, rendering a single approach to communication ineffective. To improve science communication, we characterized perspectives regarding issues negatively affecting mule deer (Odocoileus hemionus) in Wyoming, USA. We used Q methodology, a mixed quantitative-qualitative approach where participants ranked a series of statements followed by semi-structured interviews, to identify shared perspectives. We interviewed individuals (n = 37) representing prominent stakeholder groups (e.g., ranchers, hunters, conservation non-profits) in Wyoming. We identified 3 perspectives (52% of variance explained) that captured shared views regarding what factors are negatively affecting mule deer: bottom-up (n = 17 participants; 26% variance), human contributions (n = 9; 14% variance), and top-down (n = 8; 12% variance) perspectives. Most participants shared the idea that mule deer are being negatively affected, but participants diverged in views as to the primary issues. Perspectives ranged from being focused on bottom-up factors (e.g., habitat fragmentation, condition of winter ranges) to top-down factors (e.g., predation, disease) to factors focused on human contributions (e.g., human activity, public and political interests). Based on how participants diverged in perspectives and their interest in mule deer management, we discuss opportunities for scientists to improve communication by incorporating ecological complexity and nuance, moving towards a 2-way dialogue of communication, and sharing their own first-hand experiences in future communications with stakeholders. © 2021 The Authors. The Journal of Wildlife Management published by Wiley Periodicals LLC on behalf of The Wildlife Society.     

Machine learning-based risk factor analysis and prevalence prediction of intestinal parasitic infections using epidemiological survey data

BackgroundPrevious epidemiological studies have examined the prevalence and risk factors for a variety of parasitic illnesses, including protozoan and soil-transmitted helminth (STH, e.g., hookworms and roundworms) infections. Despite advancements in machine learning for data analysis, the majority of these studies use traditional logistic regression to identify significant risk factors.MethodsIn this study, we used data from a survey of 54 risk factors for intestinal parasitosis in 954 Ethiopian school children. We investigated whether machine learning approaches can supplement traditional logistic regression in identifying intestinal parasite infection risk factors. We used feature selection methods such as InfoGain (IG), ReliefF (ReF), Joint Mutual Information (JMI), and Minimum Redundancy Maximum Relevance (MRMR). Additionally, we predicted children’s parasitic infection status using classifiers such as Logistic Regression (LR), Support Vector Machines (SVM), Random Forests (RF) and XGBoost (XGB), and compared their accuracy and area under the receiver operating characteristic curve (AUROC) scores. For optimal model training, we performed tenfold cross-validation and tuned the classifier hyperparameters. We balanced our dataset using the Synthetic Minority Oversampling (SMOTE) method. Additionally, we used association rule learning to establish a link between risk factors and parasitic infections.Key findingsOur study demonstrated that machine learning could be used in conjunction with logistic regression. Using machine learning, we developed models that accurately predicted four parasitic infections: any parasitic infection at 79.9% accuracy, helminth infection at 84.9%, any STH infection at 95.9%, and protozoan infection at 94.2%. The Random Forests (RF) and Support Vector Machines (SVM) classifiers achieved the highest accuracy when top 20 risk factors were considered using Joint Mutual Information (JMI) or all features were used. The best predictors of infection were socioeconomic, demographic, and hematological characteristics.ConclusionsWe demonstrated that feature selection and association rule learning are useful strategies for detecting risk factors for parasite infection. Additionally, we showed that advanced classifiers might be utilized to predict children’s parasitic infection status. When combined with standard logistic regression models, machine learning techniques can identify novel risk factors and predict infection risk.     

Teaching animal habitat selection using wildlife tracking equipment

We present a hands-on outdoor activity coupled with classroom discussion to teach students about wildlife habitat selection, the process by which animals choose where to live. By selecting locations or habitats with many benefits (e.g., food, shelter, mates) and few costs (e.g., predators), animals improve their ability to survive and reproduce. Biologists track animal movement using radio telemetry technology to study habitat selection so they can better provide species with habitats that promote population growth. We present a curriculum in which students locate “animals” (transmitters) using radio telemetry equipment and apply math skills (use of fractions and percentages) to assess their “animal's” habitat selection by comparing the availability of habitat types with the proportion of “animals” they find in each habitat type.     

Influence of interspecific competition on mule deer birthing and rearing site selection

Ungulates often alter behavior and space use in response to interspecific competition. Despite observable changes in behavior caused by competitive interactions, research describing the effects of competition on survival or growth is lacking. We used spatial modeling to determine if habitat use by female mule deer (Odocoileus hemionus) was affected by other ungulate species prior to, during, and after parturition. We conducted our study in the Book Cliffs region of eastern Utah, USA, during 2019 and 2020. We used resource selection function (RSF) analysis to model space use of 4 ungulate species that potentially competed with mule deer: bison (Bos bison), cattle, elk (Cervus canadensis), and feral horses. We incorporated RSF models for competing species into a random forest analysis to determine if space use by mule deer was influenced by these other ungulate species. We used survival and growth data from neonate mule deer to directly assess potential negative effects of other ungulates. Habitat use by elk was an important variable in predicting use locations of mule deer during birthing and rearing. The relationship was positive, suggesting interference competition was not occurring. Survival of neonate mule deer increased as the probability of use by elk increased (hazard ratio = 0.185 ± 0.497 [SE]). Further, probability of use by elk in rearing habitat had no influence on growth of neonate mule deer from birth to 6 months of age, suggesting that exploitative competition was not occurring.     

Seasonal patterns of habitat selection in the insectivorous bat‐eared fox

Understanding how animals utilize their habitat provides insights about their ecological needs and is of importance for both theoretical and applied ecology. As changing seasons impact prey habitat selection and vegetation itself, it is important to understand how seasonality impacts microhabitat choice in optimal foragers and their prey. We followed habituated bat‐eared foxes (Otocyon megalotis) in the Kalahari, South Africa, to study their seasonal habitat selection patterns and relate them to the habitat preferences of their main prey, termites (Hodotermes mossambicus). We used Resource Selection Functions (RSFs) to study bat‐eared foxes’ 3rd‐ and 4th‐order habitat selection by comparing used locations to random ones within their home ranges. Third‐order habitat selection for habitat type and composition was weak and varied little between seasons. We found that patterns of fox habitat selection did not mirror habitat selection of Hodotermes (quantified using RSFs), even when feeding on them (4th‐order). Taken together, these results might indicate that bat‐eared foxes’ food resources are homogenously distributed across habitats and that prey other than Hodotermes play an important role in bat‐eared foxes’ space use.     

Spatial Partitioning of Predation Risk in a Multiple Predator-Multiple Prey System

ABSTRACT Minimizing risk of predation from multiple predators can be difficult, particularly when the risk effects of one predator species may influence vulnerability to a second predator species. We decomposed spatial risk of predation in a 2-predator, 2-prey system into relative risk of encounter and, given an encounter, conditional relative risk of being killed. Then, we generated spatially explicit functions of total risk of predation for each prey species (elk [Cervus elaphus] and mule deer [Odocoileus hemionus]) by combining risks of encounter and kill. For both mule deer and elk, topographic and vegetation type effects, along with resource selection by their primary predator (cougars [Puma concolor] and wolves [Canis lupus], respectively), strongly influenced risk of encounter. Following an encounter, topographic and vegetation type effects altered the risk of predation for both ungulates. For mule deer, risk of direct predation was largely a function of cougar resource selection. However, for elk, risk of direct predation was not only a function of wolf occurrence, but also of habitat attributes that increased elk vulnerability to predation following an encounter. Our analysis of stage-based (i.e., encounter and kill) predation indicates that the risk effect of elk shifting to structurally complex habitat may ameliorate risk of direct predation by wolves but exacerbate risk of direct predation by cougars. Information on spatiotemporal patterns of predation will be become increasingly important as state agencies in the western United States face pressure to integrate predator and prey management.     

Can We Accurately Characterize Wildlife Resource Use When Telemetry Data Are Imprecise?

ABSTRACT Telemetry data have been widely used to quantify wildlife habitat relationships despite the fact that these data are inherently imprecise. All telemetry data have positional error, and failure to account for that error can lead to incorrect predictions of wildlife resource use. Several techniques have been used to account for positional error in wildlife studies. These techniques have been described in the literature, but their ability to accurately characterize wildlife resource use has never been tested. We evaluated the performance of techniques commonly used for incorporating telemetry error into studies of wildlife resource use. Our evaluation was based on imprecise telemetry data (mean telemetry error = 174 m, SD = 130 m) typical of field-based studies. We tested 5 techniques in 10 virtual environments and in one real-world environment for categorical (i.e., habitat types) and continuous (i.e., distances or elevations) rasters. Technique accuracy varied by patch size for the categorical rasters, with higher accuracy as patch size increased. At the smallest patch size (1 ha), the technique that ignores error performed best on categorical data (0.31 and 0.30 accuracy for virtual and real data, respectively); however, as patch size increased the bivariate-weighted technique performed better (0.56 accuracy at patch sizes >31 ha) and achieved complete accuracy (i.e., 1.00 accuracy) at smaller patch sizes (472 ha and 1,522 ha for virtual and real data, respectively) than any other technique. We quantified the accuracy of the continuous covariates using the mean absolute difference (MAD) in covariate value between true and estimated locations. We found that average MAD varied between 104 m (ignore telemetry error) and 140 m (rescale the covariate data) for our continuous covariate surfaces across virtual and real data sets. Techniques that rescale continuous covariate data or use a zonal mean on values within a telemetry error polygon were significantly less accurate than other techniques. Although the technique that ignored telemetry error performed best on categorical rasters with smaller average patch sizes (i.e., ≤31 ha) and on continuous rasters in our study, accuracy was so low that the utility of using point-based approaches for quantifying resource use is questionable when telemetry data are imprecise, particularly for small-patch habitat relationships.     

Behavioral and Demographic Responses of Mule Deer to Energy Development on Winter Range

Anthropogenic habitat modification is a major driver of global biodiversity loss. In North America, one of the primary sources of habitat modification over the last 2 decades has been exploration for and production of oil and natural gas (hydrocarbon development), which has led to demographic and behavioral impacts to numerous wildlife species. Developing effective measures to mitigate these impacts has become a critical task for wildlife managers and conservation practitioners. However, this task has been hindered by the difficulties involved in identifying and isolating factors driving population responses. Current research on responses of wildlife to development predominantly quantifies behavior, but it is not always clear how these responses scale to demography and population dynamics. Concomitant assessments of behavior and population-level processes are needed to gain the mechanistic understanding required to develop effective mitigation approaches. We simultaneously assessed the demographic and behavioral responses of a mule deer population to natural gas development on winter range in the Piceance Basin of Colorado, USA, from 2008 to 2015. Notably, this was the period when development declined from high levels of active drilling to only production phase activity (i.e., no drilling). We focused our data collection on 2 contiguous mule deer winter range study areas that experienced starkly different levels of hydrocarbon development within the Piceance Basin. We assessed mule deer behavioral responses to a range of development features with varying levels of associated human activity by examining habitat selection patterns of nearly 400 individual adult female mule deer. Concurrently, we assessed the demographic and physiological effects of natural gas development by comparing annual adult female and overwinter fawn (6-month-old animals) survival, December fawn mass, adult female late and early winter body fat, age, pregnancy rates, fetal counts, and lactation rates in December between the 2 study areas. Strong differences in habitat selection between the 2 study areas were apparent. Deer in the less-developed study area avoided development during the day and night, and selected habitat presumed to be used for foraging. Deer in the heavily developed study area selected habitat presumed to be used for thermal and security cover to a greater degree. Deer faced with higher densities of development avoided areas with more well pads during the day and responded neutrally or selected for these areas at night. Deer in both study areas showed a strong reduction in use of areas around well pads that were being drilled, which is the phase of energy development associated with the greatest amount of human presence, vehicle traffic, noise, and artificial light. Despite divergent habitat selection patterns, we found no effects of development on individual condition or reproduction and found no differences in any of the physiological or vital rate parameters measured at the population level. However, deer density and annual increases in density were higher in the low-development area. Thus, the recorded behavioral alterations did not appear to be associated with demographic or physiological costs measured at the individual level, possibly because populations are below winter range carrying capacity. Differences in population density between the 2 areas may be a result of a population decline prior to our study (when development was initiated) or area-specific differences in habitat quality, juvenile dispersal, or neonatal or juvenile survival; however, we lack the required data to contrast evidence for these mechanisms. Given our results, it appears that deer can adjust to relatively high densities of well pads in the production phase (the period with markedly lower human activity on the landscape), provided there is sufficient vegetative and topographic cover afforded to them and populations are below carrying capacity. The strong reaction to wells in the drilling phase of development suggests mitigation efforts should focus on this activity and stage of development. Many of the wells in this area were directionally drilled from multiple-well pads, leading to a reduced footprint of disturbance, but were still related to strong behavioral responses. Our results also indicate the likely value of mitigation efforts focusing on reducing human activity (i.e., vehicle traffic, light, and noise). In combination, these findings indicate that attention should be paid to the spatial configuration of the final development footprint to ensure adequate cover. In our study system, minimizing the road network through landscape-level development planning would be valuable (i.e., exploring a maximum road density criteria). Lastly, our study highlights the importance of concomitant assessments of behavior and demography to provide a comprehensive understanding of how wildlife respond to habitat modification. © 2021 The Wildlife Society.