Wildlife crossing structures (WCS) are proposed as the primary way to improve wildlife connectivity across transportation. Barrier fencing is used independently to reduce wildlife-vehicle collisions (WVC). WCS effectiveness depends on understanding the interacting influences of wildlife and human activity. There are no published, science-based, and objective methods for decision-support for WCS siting at US state extents. To inform WCS planning on 8 highways throughout CA, supported by ~$28 million from the Wildlife Conservation Board, locations were identified: 1) where new over OR under-crossings are ecologically needed and feasible to construct, and 2) where existing structures provide connectivity. The following spatial datasets were used: 1) available wildlife occurrences and movement, 2) WVC, 3) habitat suitability models, 4) topography, 5) human disturbance and development, 6) ownership, and 7) infrastructure data. I developed a logic-based model in GIS to associate the disparate data types in a decision-support framework. The result was identification of places along each highway of types of actions critical to reduce wildlife-vehicle conflict and improve connectivity. For each highway, one to three locations are being or have been evaluated by engineers and plans developed for construction of 13 WCS and fence alignments, with an eventual cost of over $200 million.

Transportation infrastructure can restrict wildlife movement and increase wildlife-vehicle collisions (WVCs) that kill over 48,000 deer in California annually. Medians, which separate opposing traffic lanes, vary in type (e.g., concrete barriers, metal guardrails, vegetated strips) and may influence WVC rates. At Caltrans' request, we assessed the impact of median types on WVC rates along California highways. Using Google Street View, we categorized median types at 1,069 mule deer (Odocoileus hemionus) WVC sites in Caltrans District 9 and 332 western gray squirrel (Sciurus griseus) WVC sites in Caltrans District 2 between January 2015 and April 2024. Chi-square tests showed significant differences between WVC site medians and randomly generated site medians, suggesting that median type affects WVC density for both species. Additionally, we analyzed 73 paired highway transects where the median transitions from one type to another, comparing WVC rates along 1-mile stretches. Wilcoxon tests revealed WVC rates were 1.95 times higher along metal guardrails than vegetated strips (p=0.045), though other pair comparisons were not significant. These findings highlight the influence of median treatments on WVCs and suggest that transportation planning should consider wildlife connectivity impacts to reduce collisions.

The California Roadkill Observation System (CROS) was launched in August 2009 as a novel US-state scale wildlife-vehicle conflict (WVC) reporting system. Since its inception, CROS has collected over 212,800 observations making it the largest WVC reporting system in the US. WVC observations have come from several sources, including amateur observers; state, federal, and local agencies; private entities; and highway patrol officers. Spatial accuracy of observations is <13m (median error) and species identification is >97% correct, with no difference between professional and amateur observers. Multiple uses have been made of the data: 1) Planning dozens of wildlife fencing/crossing projects throughout the state; 2) To model habitat suitability and species distributions; 3) To locate high-density and statistically-significant clusters of WVC (hotspots); and 4) To estimate the economic costs of these crashes and compare to costs of building fencing. Because of its longevity and size, CROS has served as a global standard for other more recent large-extent systems. It also seems to be motivating state policy and expenditures, evidenced by citation of the system in legislative language. There are many lessons learned from developing, managing and using a system like this, not least of which is the near-impossibility of getting funding.

Wildlife-vehicle collision (WVC) causes direct mortality and reduces population connectivity. Although the idea of WVC is well-characterized, its total impact on wildlife populations remains poorly understood. Here, we describe a method for estimating the impact of WVC at the US state scale, with a case study on California mule deer (Odocoileus hemionus). We used GenEst (USGS) (Dalthrop et al., 2018) to fit models, which incorporates rates of carcass observation, persistence, and effort, taken from our data collections systems, and search efficiency, from the literature. Our models were constrained by habitat suitability and split into categories by: traffic volume, bioregion, and year and season. Rates of deer-vehicle collisions (DVC) were estimated for each category using 43,021 observations of DVC between 2016-2023. We find region-variable rates, between 0-1.4 DVC/km (South Coast) and 0.4-5 DVC/km (Sierra Nevada/Modoc). We estimated that 34,000 to 53,000 mule deer were killed per year. This represents about 10% of the estimated statewide population, suggesting WVC may play an important role in its steady decline. Our approach is readily applicable to other populations with sufficient data collection, which we are carrying out for other CA species.

Interstate 5 (I-5) is a high-traffic highway that potentially acts as a continental barrier to wildlife. We propose the definition that “a continental barrier is any divide across a continent that restricts wildlife movement and creates habitat connectivity issues and/or behavioral changes across ecoregions along its length”. We conducted a meta-analysis of literature using an observed (in literature) to expected (from habitat suitability models) ratio. We divided I-5 into seven ecoregions and used habitat maps to create lists of ‘ecologically expected’ (EE) native mammalian species for each region. Using Google Scholar, we collected literature that addresses the effects of I-5 on any regional EE species (OI-5), or any regional species interactions with highways (OA). Ungulates had the highest OI-5/EE and OA/EE ratios (0.7 and 0.83), while soricomorphs had the lowest (0.09 and 0.2). For all species, Southern California had the highest OI-5/EE and OA/EE ratios (0.67 and 0.79), while Northwestern WA had the lowest (0.12 and 0.21). Of the 119 pieces of literature collected, analysis thus far shows 84% support I-5 being a continental barrier for a limited subset of EE species.

Bobcats (Lynx rufus) have had success persisting in urban areas, but vehicle collisions are a large source of mortality. Identifying how bobcats alter behavior near roads can help wildlife managers develop management strategies to reduce mortality from vehicles. To determine how roads affect bobcat movement, we analyzed GPS data from bobcats collared by the Bobcats in Tucson Research Project using continuous-time movement analyses. Our study focused on three questions regarding bobcat movement near roads: 1. Do bobcats avoid crossing roads? 2. Do bobcats use culverts and underpasses to cross roads? 3. Does bobcat behavior change when road density increases? We found that bobcats crossed roads 11% less frequently than expected from random chance, but we found no evidence that bobcats use culverts or underpasses to cross roads or that bobcat movement behavior (i.e., speed or home range size) varies with road density. Our results suggest that managers interested in reducing bobcat mortality from vehicle collisions need to do more than simply providing crossing structures. Fences to funnel bobcats toward crossing structures, rumble strips to scare bobcats from roads, reduced speed limits, and wildlife warning signs for drivers may be effective tools to reduce bobcat mortality from vehicle collisions.