Performance indices to identify attributes of highway crossing structures facilitating movement of large mammals

Abstract

Studies assessing the efficacy of wildlife crossing structures often lead to spurious results because of their failure to address masking effects of confounding variables. Confounding variables include variation in human activity, density of crossing structures along the highway corridor, and equality of species' perceived access to each crossing structure. We investigated these issues for wide-ranging large carnivores and their prey species in Banff National Park, Alberta, using data obtained from systematic, year-round monitoring of 13 newly constructed crossing structures for wildlife (underpasses and overpasses) for 34 months post-construction. We standardized the first confounding variable by selecting crossing structures remote from areas of human activity. The second confounding variable we standardized by developing probability models of crossing structure usage assuming habitat homogeneity. We standardized the third confounding variable by developing species-specific, performance indices of crossing structures (=observed through passage usage–expected through passage usage). We regressed the species performance indices against 13 crossing structure variables encompassing structural, landscape, and human activity. Our results suggest that in absence of high human activity structural attributes best explained the performance indices for both large predator and prey species, while landscape and human-related factors were of secondary importance. Crossing structures that were high, wide and short in length strongly influenced passage by grizzly bears Ursus arctos, wolves Canis lupus, elk Cervus elaphus, and deer Odocoileus sp. More constricted crossing structures were favoured by black bears Ursus americanus and cougars Puma concolor. Distance to cover was the most important crossing structure landscape attribute for cougars (negative correlation) and was a significant factor determining passage for grizzly bears, elk and deer (all positive correlations). Our findings underscore the importance of: (a) integrating temporal and spatial variability a priori when addressing the efficacy of crossing structures, and; (b) demonstrate that species respond differently to crossing structure features. In light of these results, we suggest that to maximize connectivity across roads for multiple large mammal species, road construction schemes should include a diversity of crossing structures of mixed size classes. Mitigation planning in a multiple-species ecosystem is likely to be a challenging endeavour and long-term research will aid in the decision-making process.

Introduction

Over the last decade, federal land management and transportation agencies have become increasingly aware of the effects that roads have on wildlife (Bennett, 1991; Canters, 1997; Transportation Research Board, 1997). Significant advances in understanding these impacts have been made; however, the means to adequately mitigate these impacts are slower in coming (Evink, 2002; Transportation Research Board, 2002a).

In order to mitigate the effects of roads, crossing structures for wildlife are being designed and incorporated into road construction and improvement projects (Keller and Pfister, 1997; Spellerberg, 2002; Forman et al., 2003; Cain et al., 2003). Although federal land management and state transportation agencies are building costly structures for wildlife connectivity, long-term monitoring to determine the most effective approaches has not taken place (Evink, 2002). Currently there is limited knowledge of effective and affordable crossing structure designs for most wildlife species (Romin and Bissonette, 1996; Underhill and Angold, 2000; Transportation Research Board, 2002b).

One reason for the lack of available information is because few mitigation programs have implemented monitoring programs incorporating sufficient experimental design into pre- and post- construction evaluation. Thus, results obtained from most studies remain observational at best. Furthermore, those studies that collected data in more robust manners generally failed to address the need for wildlife habituation to such large-scale landscape change (Opdam, 1997). Habituation periods may take several years depending on the species as species experience, learn, and adjust their own behaviours to the wildlife structures (Clevenger et al., 2002a). The short monitoring periods frequently implemented are simply insufficient to draw reliable conclusions from (Forman et al., 2003).

Further, many earlier studies focused primarily on crossing structure relationships of single species, paying limited attention to multispecies or community level responses (see Forman et al., 2003 for review). Because poor crossing structure designs have the potential to de-couple ecosystem level processes, for example, in the formation of prey-refuge zones in predator-prey relations (Woods et al., 1996; Clevenger and Waltho, 2000), most crossing structure designs are “selectively permeable”. The apparent success of monitoring programs aimed at single species may fail to recognize the barrier effects imposed on other non-target species. Thus, systems can be severely compromised if land managers and transportation planners rely on simple extrapolation from data on individual species. To date we are unaware of any monitoring program that addresses this issue specifically.

Information deficiencies may also be due to the masking effects of confounding variables not considered in study designs (Underwood, 1997). Confounding variables are sources of variation that may bias or even mask the efficacy of one crossing structure design over another. Such variables include the variation in human use of the crossings (Clevenger and Waltho, 2000), density of crossing structures along the highway corridor, and the equality of species' perceived access to each crossing structure. If, for example, a species perceives crossing structure A as good, but not accessible, then it may choose crossing structure B (whose design is not favoured) for accessibility reasons alone. To control for these factors a robust experimental design requires a sufficiently large number of crossing structures - much larger than is realistically feasible for a public works project to finance.

In this paper, we investigate these issues using data obtained from systematic, year-round monitoring of 13 newly constructed crossing structures (underpasses and overpasses) for 34 months post-construction. These new crossing structures are sufficiently remote from centres of human activity (e.g., the town of Banff) that human use is significantly reduced and therefore not expected to be a dominant factor (Clevenger and Waltho, 2000). We standardized against the remaining confounding variables by developing species-specific, performance indices, and then tested for significant correlations against each of the crossing structure attributes. We then rank-ordered the significant coefficient of determinations and assumed that the higher the coefficient the greater importance that attribute had in influencing species passage (positive or negative). A multivariate analysis of this type allowed us to explore the extent and influence of numerous attributes associated with the crossing structures independent of confounding variables.

Our design allowed us to address relevant and current questions concerning the efficacy of crossing structures; specifically: (i) How to ascertain the strengths and weaknesses of design characteristics for a multiple large mammal species? and, (ii) What are the requirements for effective crossing structures designed for wide-ranging large carnivores and their prey species? Information on the effectiveness of mitigation measures in reducing barrier effects will provide critical information needed for future mitigation planning in the Bow Valley transportation corridor in Banff National Park.