Modeling Elk Nutrition and Habitat Use in Western Oregon and Washington

Studies of habitat selection and use by wildlife, especially large herbivores, are foundational for understanding their ecology and management, especially if predictors of use represent habitat requirements that can be related to demography or fitness. Many ungulate species serve societal needs as game animals or subsistence foods, and also can affect native vegetation and agricultural crops because of their large body size, diet choices, and widespread distributions. Understanding nutritional resources and habitat use of large herbivores like elk ( Cervus canadensis ) can benefit their management across different land ownerships and management regimes. Distributions of elk in much of the western United States have shifted from public to private lands, leading to reduced hunting and viewing opportunities on the former and increased crop damage and other undesired effects on the latter. These shifts may be caused by increasing human disturbance (e. g., roads and traffic) and declines of early‐seral vegetation, which provides abundant forage for elk and other wildlife on public lands. Managers can benefit from tools that predict how nutritional resources, other environmental characteristics, elk productivity and performance, and elk distributions respond to management actions. We present a large‐scale effort to develop regional elk nutrition and habitat‐use models for summer ranges spanning 11 million ha in western Oregon and Washington, USA (hereafter Westside). We chose summer because nutritional limitations on elk condition (e. g., body fat levels) and reproduction in this season are evident across much of the western United States. Our overarching hypothesis was that elk habitat use during summer is driven by a suite of interacting covariates related to energy balance: acquisition (e g., nutritional resources, juxtaposition of cover and foraging areas), and loss (e g., proximity to open roads, topography). We predicted that female elk consistently select areas of higher summer nutrition, resulting in better animal performance in more nutritionally rich landscapes. We also predicted that factors of human disturbance, vegetation, and topography would affect elk use of landscapes and available nutrition during summer, and specifically predicted that elk would avoid open roads and areas far from cover‐forage edges because of their preference for foraging sites with secure patches of cover nearby. Our work had 2 primary objectives: 1) to develop and evaluate a nutrition model that estimates regional nutritional conditions for elk on summer ranges, using predictors that reflect elk nutritional ecology; and 2) to develop a summer habitat‐use model that integrates the nutrition model predictions with other covariates to estimate relative probability of use by elk, accounting for ecological processes that drive use. To meet our objectives, we used 25 previously collected data sets on elk nutrition, performance, and distributions from 12 study areas. We demonstrated the management utility of our regional‐scale models via application in 2 landscapes in Washington. The elk nutrition model predicts levels of digestible energy in elk diets (DDE; kcal DE/g of consumed forage) during summer. Model input data were from foraging experiments using captive female elk and field measurements of site characteristics at fine scales (∼0.5 ha). The nutrition model included a set of equations that predicted forage biomass as a function of site characteristics and a second set that predicted DDE primarily as a function of forage biomass. We used the nutrition model to develop a DDE map across the Westside. We then evaluated performance of the model by comparing predicted DDE to nutritional resource selection by elk and to population‐level estimates of autumn body fat and pregnancy rates of lactating elk. To model elk habitat use, we compiled 13 unique telemetry data sets from female elk ( n  = 173) in 7 study areas (data collected June–August 1991–2009). We used a generalized linear model with 5 of the data sets, coupled with ecologically relevant covariates characterizing nutrition, human disturbance, vegetation, and physical conditions, to estimate intensity of use with the negative binomial model. We evaluated model performance by mapping predicted habitat use with the regional model and comparing predictions with counts of elk locations using 8 independent telemetry data sets. The nutrition model explained a reasonably high amount of variation in forage biomass ( r 2  = 0.46–0.72) and included covariates of overstory canopy cover, proportion of hardwoods in the canopy, potential natural vegetation (PNV) zone, and study area. Dietary DE equations in the model explained about 50% of the variation in DDE ( r 2  = 0.39–0.57) as a function of forage biomass by PNV zone and study area. Broad‐scale application of the nutrition model in the Westside region illustrated the predominance of landscapes that failed to meet nutritional needs of lactating females (≤2.58 kcal/g) and their calves, especially at moderate elevations in closed‐canopy forests in both the Coast Range and the southern Cascades. Areas providing DDE at (>2.58–2.75 kcal/g) or in excess (>2.75 kcal/g) of the basic requirement of lactating females were uncommon (<15% of area) or rare (<5% of area), respectively, and primarily occurred in early‐seral communities, particularly at higher elevations. Wild elk avoided areas with DDE below basic requirement and selected for areas with DDE >2.60 kcal/g. Percentage of elk ranges providing DDE levels near or above basic requirement was highly correlated with pregnancy rates of lactating females. Autumn body fat levels were highly correlated with percentage of elk ranges providing DDE levels above basic requirement. The regional model of elk habitat use with greatest support in the empirical data included 4 covariates: DDE, distance to nearest road open to motorized use by the public, distance to cover‐forage edge, and slope. Elk preferred habitats that were relatively high in DDE, far from roads, close to cover‐forage edges, and on gentle slopes. Based on standardized coefficients, changes in slope (−0.949) were most important in predicting habitat use, followed by DDE (0.656), distance to edge (−0.305), and distance to open road (0.300). Use ratios for the regional model indicated these changes in relative probability of use by elk: a 111.2% increase in use for each 0.1‐unit increase in DDE; a 22.7% increase in use for each kilometer away from an open road; an 8.1% decrease in use for each 100‐m increase in distance to edge; and a 5.3% decrease in use for each percent increase in slope. The regional model validated well overall, with high correlation between predicted use and observed values for the 4 Washington sites ( r s  ≥ 0.96) but lower correlation in southwestern Oregon sites ( r s  = 0.32–0.87). Our results demonstrated that nutrition data collected at fine scales with captive elk can be used to predict nutritional resources at large scales, and that these predictions directly relate to habitat use and performance of free‐ranging elk across the Westside region. These results also highlight the importance of including summer nutrition in habitat evaluation and landscape planning for Westside elk. The models can inform management strategies to achieve objectives for elk across land ownerships. The regional model provides a useful tool to understand and document spatially explicit habitat requirements and distributions of elk in current or future landscapes. The 2 examples of management application demonstrated how effects of management on elk nutrition and habitat use can be evaluated at landscape scales, and in turn how animal performance and distribution are affected. Results further illustrated the importance of managing for nutrition in combination with other covariates (i e., roads, slope, cover‐forage edges) that affect elk use of nutritional resources to achieve desired distributions of elk. Our meta‐analysis approach to habitat modeling provides a useful framework for research and management of wildlife species with coarse‐scale habitat requirements by identifying commonalities in habitat‐use patterns that are robust across multiple modeling areas and a large geographic range. Use of such methods in future modeling, including application in monitoring programs and adaptive management, will continue to advance ecological knowledge and management of wildlife species like elk. © 2018 The Authors. Wildlife Monographs published by Wiley on behalf of The Wildlife Society.