Fourteen years of monitoring on US Highway 93 north of Missoula produced a finding that rarely survives into the citations. On the Flathead Indian Reservation between Evaro and Polson, the three mitigated study areas recorded 71.44 percent fewer large-mammal carcasses and 80.04 percent fewer reported wildlife crashes once the rise in collisions in the unmitigated control sections was taken into account. Across the corridor as a whole, the number of wildlife-vehicle collisions did not decrease at all. The report’s collision analysis covers 51.9 of the 56 rebuilt miles, between Evaro and Polson.
Both statements are in the same 2016 report to the Montana Department of Transportation. That 51.9 mile analysis corridor was mitigated along 16.8 percent of its length. The rest of it was rebuilt to modern geometry, and the report is direct about what that does: wider lanes, wider shoulders, longer sight distances and more gentle curves improve human safety in general, while wildlife-vehicle collisions are likely to increase. Corridor-wide, reported large-mammal carcasses rose 5.38 percent after reconstruction, an increase the report’s own statistical test found not significant (two-sided Mann-Whitney U, P=1.000). Reported wildlife crashes rose 83.57 percent over the same corridor, an increase the test did find significant (P=0.036), though the report says part of it may be related to the higher design speed of the rebuilt road pushing more repair estimates over the $1,000 threshold that defines a reportable crash. A reconstruction project can cut collisions substantially inside its fenced and structured sections and still leave the corridor no better off by the one measure, carcasses, that did not move.
That is the discipline this subject needs. Every published percentage is attached to a specific fenced length, a specific species, a specific span of years and a specific counting method, and those four attributes move independently of each other.
The Montana corridor is the cautionary case
The reconstruction placed wildlife crossing structures at 39 locations along the 56 mile road and put wildlife exclusion fence on both sides of approximately 8.71 miles of it. Twenty-nine of the structures were monitored with cameras, which recorded 95,274 successful crossings in total, about 22,648 a year. White-tailed deer accounted for 69 percent of them. Mule deer, and domestic dogs and cats, each accounted for roughly 5 percent. Black bear crossed successfully 1,531 times, and 20 different species of medium or large terrestrial wild mammals used the structures.
Habitat connectivity, in other words, worked. Human safety worked in the places that were fenced. The animals the fence was too short to protect were a different story. Black bear carcasses continued to be recorded after reconstruction, and the researchers found no evidence that the measures at Evaro, Ravalli Curves and Ravalli Hill reduced the number of reported black bear carcasses, which they attributed to the short fenced lengths, the design of the fence, and gaps at access roads and steep slopes. Six grizzly bear carcasses were reported between 2002 and 2015. Five of those bears were hit after reconstruction, three of them on rebuilt sections carrying only short fence segments with underpasses at selected locations.
Fence length is the variable that moves the number
The finding with the widest application on US 93 concerns the fence rather than the structures it serves. Drawing on its own data and the wider literature, the report concludes that wildlife fences were most effective at reducing collisions with large mammals, almost always better than 80 percent, where the fence and its associated measures ran for at least 3.1 miles of continuous road. Below that length the average effectiveness dropped to about 50 percent, and the report describes the effectiveness of any individual mitigated section shorter than 3.1 miles as highly unpredictable. The mechanism is fence end effects, which concentrate collisions at and near the points where the fence stops.
Wyoming’s account of its own program is consistent with the upper half of that range. The state transportation department reports that fencing combined with underpasses and overpasses can reduce collisions with large mammals an estimated 80 to 99 percent, with an average of 87 percent, and that at Nugget Canyon on US 30 the construction of seven underpasses and game-proof fencing reduced deer-vehicle collisions by 81 percent. The department’s page attaches no study or year to the 80 to 99 percent range, so it is best read as agency practice guidance rather than as a measurement of a particular corridor.
The hardware between the fence ends behaves differently by species, and the Montana monitoring put numbers on that too. Wildlife guards, which work like cattle guards at access road openings, were a near-total barrier to deer at 1.26 percent permeability for white-tailed deer and 0.45 percent for mule deer. They were close to useless against carnivores, with permeability of 94 percent for mountain lion, 73 percent for bobcat and 53 percent for black bear, which is why the researchers recommend something other than a wildlife guard wherever bears or cats are the target species. Wildlife jump-outs, the earthen escape ramps inside the fenced corridor, were used by white-tailed deer to reach the safe side only about 7 percent of the time, against about 32 percent for mule deer. One gap in the fence built so people could walk through recorded nine human crossings in three and a half years and 140 white-tailed deer crossings over the same period.
Colorado ran the cleaner experiment and published its edge
The State Highway 9 project between Kremmling and Green Mountain Reservoir rebuilt an 11 mile segment for $40 million. The project archive records average annual daily traffic of 3,100 to 3,600 for the road. The project installed five wildlife underpasses and two overpasses along with 10.3 miles of exclusion fence, 29 wildlife guards and 61 escape ramps. From November 2015 through April 2020, researchers evaluated it with 62 motion-activated cameras plus crash and carcass records.
The passage record is unusually granular: 112,678 successful mule deer passages across the seven structures at an overall success rate of 96 percent, with sixteen other species also using them. On the safety side, the mitigation helped decrease wildlife-vehicle collision crashes reported to law enforcement by 92 percent and supplementary carcass counts by 90 percent relative to preconstruction levels, which the authors convert into an average of 13 crashes and 56 mule deer mortalities prevented each year since construction finished.
The two percentages are not interchangeable, and the report does not treat them as such. Crashes reported to law enforcement and carcasses counted separately are different denominators with different biases, and a corridor can move one more than the other. The report draws its own boundary here too: collisions continued beyond the project area, south of the fence end, where ungulate activity carried on, and the authors state that the project may not fully mitigate wildlife movements across SH 9 and recommend extending the mitigation farther south to cover the whole collision hotspot.
Structure type is a species question before it is an engineering one
The overpass photographs better and the underpass is cheaper, and neither fact decides anything. What the Montana cameras recorded is a set of preferences specific to each animal. White-tailed deer used bridges, overpasses and large culverts more than expected and barely used small culverts at all. Mule deer behaved similarly. Black bear used a wider range including small culverts. Elk and moose mostly or exclusively used the wildlife overpass. Grizzly bears used large culverts exclusively, which the report qualifies by noting that large culverts are the most common structure type inside the area grizzlies use regularly. Because so many of these structures are also drainage assets, the sizing decisions described in how highway drainage systems work end up deciding which species can pass.
Two further findings should change how monitoring reports are read. There is a learning curve: deer and black bear use rose for at least five years after construction, so a report written one or two years in understates the eventual figure. And large mammal use of large underpasses varied greatly while doing so independently of fence length, which means the presence of fence and longer fence lengths did not guarantee higher use. Location relative to surrounding habitat, population density and existing movement patterns dominate. Colorado’s monitoring found both its overpass and its underpass designs effective for mule deer, which is the useful comparison: where the structure is placed matters more than which kind it is.
What the federal program actually says
The Wildlife Crossings Pilot Program is written into law at 23 U.S.C. 171, added by section 11123 of Public Law 117-58 on 15 November 2021, and the statutory findings are worth reading in the original because they are quoted loosely everywhere else. Congress found that, according to the Federal Highway Administration report titled Wildlife-Vehicle Collision Reduction Study, “there are more than 1,000,000 wildlife-vehicle collisions every year”; that collisions “represent a persistent concern that results in tens of thousands of serious injuries and hundreds of fatalities on the roadways of the United States”; and that “the total annual cost associated with wildlife-vehicle collisions has been estimated to be $8,388,000,000”.
The report Congress cited is the 2008 study to Congress prepared for FHWA by the Western Transportation Institute, whose own abstract says that, based on several national datasets, there are between one and two million such collisions annually in the United States and that the number is increasing. The statute took the bottom of that range and dropped the trend.
The money is smaller than the discussion around it. The law authorizes $60,000,000 for fiscal year 2022, $65,000,000 for 2023, $70,000,000 for 2024, $75,000,000 for 2025 and $80,000,000 for 2026, out of the Highway Trust Fund, and requires that not less than 60 percent of grant funds each year go to projects in rural areas. Set the $350 million total against Colorado’s $40 million for 11 miles and the whole five-year authorization buys roughly eight or nine comparable corridors. Congress also required a report to Congress by 31 December each year and, in a companion section, a study that updates and expands on both the 2008 report and the October 2008 best practices manual that accompanied it.
The benefit-cost test is not what justifies these projects
The mitigation measures along US 93 North did not generate monetary benefits in excess of their costs, at least not based on human safety parameters alone.
That sentence is the most useful thing in the Montana report for anyone building a case for funding. The same analysis found that costs were lowest on highway sections with long and continuous wildlife fences and underpasses but no overpass, and highest on sections that had crossing structures but only short fence segments or no fence at all. None of the configurations paid for itself in the report’s accounting: Table 16 puts monetized benefits at between 3.94 and 24.44 percent of costs across the analyzed sections, every balance negative. The authors then say why the exercise has limited force: the analysis rested almost entirely on human safety parameters, and passive use values were excluded even though those values were the principal reason the structures were built. Reducing the chance that a driver kills a grizzly bear does not have a price in the manual. The general limits of monetizing this kind of outcome are covered in highway lifecycle cost analysis, and they bite harder here than in pavement work.
Washington shows what a long build reports while it is still building
Along the 15 mile stretch of Interstate 90 between Hyak and Easton, eleven large wildlife crossing structures were complete in phases 1 and 2, with seven more planned for phases 3 and 4 and continuous wildlife fencing through most of the project area. At the end of 2023, cameras recorded the 25,000th safe wildlife crossing at completed structures since 2014, including 7,967 elk, 13,525 deer, 3,216 coyotes and six cougar. Five of those cougar crossings happened in 2023 alone, against one in the nine years before. The Keechelus Wildlife Overcrossing recorded 1,304 elk crossings in 2023, the highest single-year use of one structure by elk in the project.
On collisions, the department says only that wildlife-vehicle collisions within the completed areas dropped to a consistently low or non-existent level. That is a qualitative claim where a percentage would attract more attention, and it is the more defensible one to make while neighboring phases are still open construction sites feeding animals toward unfenced ground.
What the evidence supports, and what gets funded
Across all four programs, the transferable result is not about structures. It is about the continuous length of fence they sit inside, the treatment of the point where that fence ends, and the species-specific hardware in between. Projects are announced in structures, because a structure can be photographed and a fence end cannot, and the cheapest available improvement on many corridors is extending an existing fence past the place where animals currently walk around it. That same preference for the photographable component decides what gets built in several of the programs gathered in the environment section. The habitat that draws those animals to the shoulder is managed under a separate set of rules, set out in roadsides as habitat, and the interaction between the two is barely studied.
The three governments on US 93 agreed measures of effectiveness in advance and then published the finding that some of the human safety measures were not met. That is what accountable monitoring produces, and it is why the Montana numbers are more useful than cleaner ones from projects with no control section. The studies that report an inconvenient result are consistently the ones whose figures survive scrutiny.