A habitat-based framework for grizzly bear conservation in Alberta

Abstract

Grizzly bear (Ursus arctos L.) populations in Alberta are threatened by habitat loss and high rates of human-caused mortality. Spatial depictions of fitness would greatly improve management and conservation action. We are currently challenged, however, in our ability to parameterize demographic rates necessary for describing fitness, especially across gradients of human disturbance and for land cover types. Alternative approaches are therefore needed. We describe here a method of estimating relative habitat states and conditions as surrogates of fitness using models of occupancy and mortality risk. By combining occurrence and risk models into a two-dimensional habitat framework, we identified indices of attractive sinks and safe harbour habitats, as well as five habitat states: non-critical habitats, secondary habitats (low-quality and secure), primary habitats (high-quality and secure), secondary sinks (low-quality, but high risk), and primary sinks (high-quality and high risk). Primary sink or high attractive sink situations were evident in the foothills where bears were using forest edges associated with forestry and oil and gas activities on Crown lands, while primary habitats or safe harbour sites were most common to protected alpine/sub-alpine sites. We suggest that habitat states and indices be used for setting baseline conditions for management and comparison of habitat conditions over time and identification of grizzly bear conservation reserves. A no net loss policy of critical habitats could be used to maintain existing habitat conditions for landscapes threatened by human development. Under such a policy, conversions of primary habitat would require restoration of equivalent amounts of primary sinks through decommissioning of roads.

Introduction

Understanding the distribution and abundance of species in space and time is the primary definition of ecology (Krebs, 1985). With the recent advent of geographic information systems (GIS), together with widespread availability of digital geo-spatial data, predicting species occurrence and/or abundance has become commonplace (Boyce and McDonald, 1999, Scott et al., 2001). Applications of such models include climate-change assessments (Tellez-Valdes and Davila-Aranda, 2003), restoration or range expansion (Mladenoff et al., 1995, Boyce and Waller, 2003), ecological risk assessment (McDonald and McDonald, 2002), and conservation gaps or reserve design (Flather et al., 1998, Yip et al., 2004). Ultimately, understanding large-scale patterns and temporal changes to rare, threatened or endangered species helps focus conservation needs (Dobson et al., 1996, Mattson and Merrill, 2002).

Describing species occurrence, or even that of abundance, however, does not necessarily parallel habitat relationships for populations, as occurrence and abundance can be poor surrogates for demographic performance (Van Horne, 1983, Hobbs and Hanley, 1990, Tyre et al., 2001). Relating life history traits to habitats is critical for understanding habitat processes and ultimately the management of species of conservation concern (Franklin et al., 2000, Breininger and Carter, 2003). Without understanding such functions, one risks assuming that animal occurrence or abundance relates directly to habitat quality, something that is not always the case. For instance, some sites considered high in habitat quality from an occupancy standpoint may be low in survival and/or final recruitment. These ‘attractive’ habitat patches can produce local population sinks, and therefore have been called attractive sinks (Delibes et al., 2001, Naves et al., 2003) or ecological traps (Dwernychuk and Boag, 1972, Ratti and Reese, 1988, Donovan and Thompson, 2001). Recognizing this phenomenon within conservation habitat models and resulting planning maps is therefore crucial for fully representing habitat quality. For many species, however, we lack the necessary data to formulate habitat-specific demographic parameters and waiting for such data to be collected for long-lived species with low reproductive rates might simply result in documenting the decline rather than providing an initial recommendation for the conservation problem. No doubt, collection of long-term life history information needs to be gathered, but exploiting existing data sources also is necessary for short-term conservation management. Commonly, what is available is information on animal occupancy from aerial surveys or radiotelemetry studies and sometimes a distribution of mortality locations from government management databases (e.g., hunting, problem wildlife, vehicle-wildlife collisions, etc.). Identifying attractive sink habitats, as well as some form of source or secure habitats from these data, would prove useful for conservation planning and wildlife management.

One species ideally suited for exploring conservation habitat modelling from an occupancy and survival framework is grizzly bears Ursus arctos L. Grizzly bears are an important keystone species (Tardiff and Stanford, 1998) that have declined substantially throughout much of North America in the past century (McLellan, 1998, Mattson and Merrill, 2002), largely due to vulnerability from late maturation, low density, low reproductive rates and a high trophic level (Russell et al., 1998, Purvis et al., 2000a, Purvis et al., 2000b, Woodruffe, 2000, Garshelis et al., 2005). Current efforts to identify grizzly bear habitats have relied on radiotelemetry analyses of habitat selection (Waller and Mace, 1997, Mace et al., 1996, Mace et al., 1999, McLellan and Hovey, 2001, Nielsen et al., 2002, Nielsen et al., 2003) or assessments of occupancy based on field surveys and published records (Naves et al., 2003, Posillico et al., 2004). Resulting habitat models provide assessments of animal occurrence or use, but cannot suggest overall habitat quality based on demographic performance. Even spatial models that predict grizzly bear abundance (Apps et al., 2004), although adding additional information over and above occupancy and use, still lack an explicit mechanism to identify conservation actions. What is needed is an approach that merges habitat-related occurrence or animal abundance models with critical life history parameters.

For grizzly bears, it is widely accepted that survival, especially that of females, is the most sensitive parameter for population growth (Knight and Eberhardt, 1985, Mattson et al., 1996, Wiegand et al., 1998, Boyce et al., 2001, Wielgus et al., 2001). Most grizzly bear mortalities are human-caused (McLellan et al., 1999, Benn and Herrero, 2002) and related to human access (Nielsen et al., 2004a). Incorporating some form of survival within habitat maps would therefore be helpful. Although population-level estimates of survival have been estimated for grizzly bears (e.g., McLellan et al., 1999), few have attempted to define or index these in a spatial manner necessary for targeting on-the-ground management (see however, Nielsen et al., 2004a, Johnson et al., 2004). Recently, Naves et al. (2003) used a spatial framework for defining brown bear habitats in northern Spain that incorporated both survival and reproduction simultaneously. Such modelling and mapping approaches are attractive management tools for identifying conservation needs because they record attractive sinks where animals are likely to be present, but suffer high mortality rates, and source or secure habitats where animals are present and enjoy high survival. Both habitat states provide managers with 2 separate conservation strategies: (1) preservation and protection of existing source and secure areas to impede habitat degradation; and (2) mitigation of sites where habitat conditions are excellent, but risk of mortality is high and manageable.

Here, we develop a framework for identifying attractive sink and source-like habitats for grizzly bears in west-central Alberta, Canada. Such an approach is especially warranted for this region, given the recommendation of threatened status by Alberta’s Endangered Species Conservation Committee (Stenhouse et al., 2003). With any such change in status, effective habitat maps will be necessary for appropriate management and conservation planning. Despite the recognition of population declines and the importance of secure habitats, current management is largely based on a 1988 assessment of land cover and human disturbance (Stenhouse et al., 2003). We propose to update this assessment and redefine grizzly bear habitat using empirical models of animal occurrence and risk of human-caused mortality specific to current conditions in the east slopes of the Alberta Rocky Mountains. By placing occupancy and mortality risk models in a two-dimensional framework, we define indices of attractive sink and safe-harbour (source-like or secure) habitats as well as a classification of 5 habitat states including, non-critical habitat, secondary sink, primary sink (similar to high index values of attractive sink habitats), secondary habitat, and primary habitat (similar to high index values safe harbour habitats). Although these habitat indices and states are not based directly on demographic parameters, they have value in tracking temporal changes in habitats and ranking areas for conservation action.