Install barrier fencing along roads

How is the evidence assessed?
  • Effectiveness
  • Certainty
  • Harms

Study locations

Key messages



  • Survival (9 studies): Three controlled studies, in the USA, Germany and Brazil, found that roadside fencing or equivalent barrier systems reduced the numbers of mammals, including wildcats and coypu, killed by vehicles on roads. Two before-and-after studies, in the USA, found that roadside fencing with one-way gates to allow escape from the road, reduced the number of collisions between vehicles and deer. A study in the USA found that a 2.7-m-high fence did not reduce road-kills of white-tailed deer compared to a 2.2-m-high fence. A controlled, before-and-after study in the USA found that barrier fencing with designated crossing points did not significantly reduce road deaths of mule deer. A replicated, controlled, before-and-after study in Canada found that electric fences, (along with an underpass beneath one highway), reduced moose-vehicle collisions. A review of fencing studies from USA, Canada and Sweden, found that longer fencing along roadsides led to a greater reduction of collisions between large mammals and cars than did shorter fence sections.


  • Behaviour change (5 studies): A controlled, before-and-after study in the USA found that 2.3-m-high fencing in good condition prevented most white-tailed deer accessing a highway. A replicated, controlled, before-and-after study in Canada found that electric fences reduced moose access to highways. Three studies (two replicated), in the USA, found that higher fences (2.4–2.7 m) prevented more white-tailed deer from entering highways than did fences that were 2.2 m high, 1.2 m high with outriggers or 1.2–1.8 m high.

About key messages

Key messages provide a descriptive index to studies we have found that test this intervention.

Studies are not directly comparable or of equal value. When making decisions based on this evidence, you should consider factors such as study size, study design, reported metrics and relevance of the study to your situation, rather than simply counting the number of studies that support a particular interpretation.

Supporting evidence from individual studies

  1. A controlled, before-and-after study in 1975 along a highway through mixed hardwood forest in Pennsylvania, USA (Falk et al. 1978) found that provided, it was in good repair, 2.3-m-high fencing prevented most white-tailed deer Odocoileus virginianus crossing a highway. Significantly fewer deer crossed the fence once it had been repaired (0–6), compared to before (77–84) and once repairs were undone (23–153), and compared to control sections (on which repairs were not carried out) during the same periods (24–247; 111–141; 53–268 crossings respectively). The 2.3-m-high fences ran either side of a four-lane highway, with a top section angled 45° away from the highway. The study site comprised two 0.8-km control sections with a 1.6-km experimental section between. Fence defects included gaps under the fence and lowered or broken top wires. Tracks in snow and sand along the fence both sides of the highway were monitored before repairs, after repairs along the experimental section and after repairs were undone. This cycle was implemented once in both winter and spring 1975 and tracks were surveyed over five days during each period.

    Study and other actions tested
  2. A before-and-after study in the 1970s along two highways in California, USA (Reed et al. 1982) found that barrier fences, including one connected to an underpass, and others to one-way gates, reduced deer-vehicle collisions by 68–87%. Fewer deer Odocoileus spp. road mortalities were recorded after construction of the six fence sections (average 2/km/year) than before (average 11/km/year). Six different lengths (1.9–7.7 km) of 2.4-m fencing were installed along Interstate 70 and Colorado Highway 82. Five of the fences were only on one side of the road, the other was on both sides and connected to an underpass. Four of the fences had one-way gates to allow deer to escape from the highway. Deer carcasses found along the road were counted in each fenced area before and after installation. Cost-benefit analysis was also undertaken using pre-fence mortality (dead deer) and fence effectiveness and estimates of cost of vehicle repair, value of deer, discount rate, cost of fence and cost of fence maintenance (see the original article for results).

    Study and other actions tested
  3. A replicated, before-and-after study in 1977–1979 along two highways in Minnesota, USA (Ludwig & Bremicker 1983) found that barrier fencing with one-way gates decreased deer-vehicle collisions. Along two fenced road sections, 1.3 and 8 deer/year were killed compared to an estimated 20/year in the pre-fence period. One fence was installed in a ditch with 1 m of water, meaning 30% of gates could not be used to escape the highway. Overall, 69% of 51 passages through gates were in the correct direction, i.e. from the highway to outside the fenced corridor. Two sections of 2.4-m-high fence with one-way gates along new highways were monitored for 18 months. Fences were 4 and 5 km long with nine and 10 pairs of gates (30 m apart), respectively. Deer were monitored crossing through gates by using baler counters and track beds. Deer-vehicle collisions were monitored for one year before (along old adjacent highway) and 18 months after installation. Cost-benefit analysis was also carried out (see the original article for further details).

    Study and other actions tested
  4. A study in 1981–1983 in forest in Pennsylvania, USA (Feldhamer et al. 1986) found that a 2.7-m-high deer-proof fence reduced the number of white-tailed deer Odocoileus virginianus on the highway compared to a 2.2-m-high fence, but did not reduce road-kills. A total of 240 groups of deer were observed on the highway alongside 23 km of 2.7-m-high fence compared to 465 alongside 18 km of 2.2-m-high fence. Overall, 1,687 deer (82% of all sightings) were on highway verges. In 1981–1983, one hundred deer died on the highway (1.2 deer/km/year) and numbers did not differ between fence types. Deer were monitored along a 41-km section of a 4–6-lane highway, 23 km of which had a 2.7-m-high mesh fence and the remainder a 2.2-m-high fence with an overhang. Thirty-six spotlight surveys were undertaken along the highway from January 1981 to January 1983.

    Study and other actions tested
  5. A controlled, before-and-after study in 1991–1995 along two highways in Utah, USA (Lehnert & Bissonette 1997) found that barrier fencing with designated crossing points and warning signs did not reduce road deaths of mule deer Odocoileus hemionus. Deaths fell on both fenced and unfenced sections but the rate of fall was not significantly higher on fenced road sections (after: 36–46; before: 111–148) than on unfenced sections (after: 34–63; before: 75–123). The number of deer on road verges fell by 34–55% following fence installation. In September 1994, four and five crossing points were installed along a two- and a four-lane highway respectively. Fencing, 2.3 m high, restricted access to roadsides and directed deer towards crossing points. At these points, deer could jump a 1-m-high fence into funnel shaped fencing (2.3 m high) with a narrow opening to the road. One-way gates allowed deer trapped along the road to escape. Three warning signs, spaced 152 m apart, and painted lines across the road at crossings, indicated to drivers that it was a crossing point. Road deaths (weekly) and behaviour were monitored along fenced and nearby unfenced roads before and after installation, from October 1991 to November 1995. Spotlight count surveys were undertaken twice/month.

    Study and other actions tested
  6. A controlled, before-and-after study in 1998–2002 along a highway in Florida, USA (Dodd et al. 2004) found that a barrier wall-culvert system reduced mammal road-kills. After construction, 33 mammals of ≥12 species were recorded dead on the 2.8-km section of road with the barrier (2.8 km) compared to 50 mammals on a 400-m section without barriers. Of those killed along the barrier, 17 were rice rats Orozomys palustris, which climbed adjacent vegetation to get over the barrier. In 2000–2001, a 1-m-high concrete wall with 15-cm overhanging lip was constructed along a 2.8-km section of a highway. Eight concrete culverts were spaced 200–500 m apart below the wall. Roadkills were monitored on three days/week before (August 1998–1999) and after (March 2001-March 2002) barrier wall construction.

    Study and other actions tested
  7. A replicated, before-and-after study in 2003–2005 along two highways in Québec, Canada (Leblond et al. 2007) found that electric fences, along with an underpass beneath one highway, reduced moose Alces alces access to highways and moose-vehicle collisions. There were fewer moose-vehicle collisions after fence construction (zero) than before (1–5/year) and moose tracks on the road decreased by 76–84%. Only 33% (of 53) of moose tracks on the road were from moose that had crossed a fence; most entered through vehicle access routes (31%) or at fence ends (7%). Fences prevented 78% (7/9) of radio-collared moose from crossing the highway. Electric fences (1.5 m high, cables 0.3 m apart) were installed along both sides of a 5-km section of Highway 175 in 2002 and a 10-km section of Highway 169 in 2004 (both two-lane). Moose were monitored along fenced and adjacent equal-length unfenced road sections using weekly track surveys in May–August of 2003–2005. GPS collars were fitted to 47 moose and locations recorded every 2–3 hours for 1–3 years. An underpass was constructed along one highway (23 m long, 16 m wide, 7 m high) and a fence opening on the other (that triggered dynamic warning road signs).

    Study and other actions tested
  8. A controlled study in 2001–2005 along a motorway through forest and agricultural land in Germany (Klar et al. 2009) found that installing roadside fencing designed to keep wildcats Felis silvestris off the road reduced road-related wildcat mortality. Wildcat mortality was lower where wildcat fencing was installed (0.07 deaths/km/year) than in areas with other types of fencing (0.41–0.44 deaths/km/year). This difference was not tested for statistical significance. In 2002, two-metre-high wildcat fencing, with 5 × 5 cm mesh, a 50-cm-wide metal sheet overhang and a board down to 30 cm below ground, was installed along 6.4 km of road. Fine-meshed fence (same specifications as the wildcat fence, but without the overhang) was installed along 4 km of road. Standard wildlife fencing was installed on 7 km of road. Wildcat mortality data collected by researchers was supplemented by reports from motorway authorities and members of the public.

    Study and other actions tested
  9. A replicated, before-and-after study in 2009–2010 along a university campus road in Georgia, USA (Gulsby et al. 2011) found that a 2.4-m-high fence was more successful at preventing white-tailed deer Odocoileus virginianus accessing the road than was a 1.2-m-high fence with outriggers attached to the top. Fewer deer crossed the road in a section with 2.4-m-high fencing (<0.01 crossings/day) than in a section with 1.2-m-fence with 0.6-m outriggers (0.05 crossings/day). Before fence construction, deer made 0.3–1.0 crossings/day. In May–June 2009, a vertical wire fence (1.6 km long, 2.4-m-high) and an outrigger fence (1.6 km long, 1.2 m high with a 0.6-m-long outriggers at 45°, attached to the top and threaded with five wires) were erected. Between January 2009 and March 2010, movements of eight adult female deer were monitored using GPS collars. Four deer had home ranges that overlapping the 2.4-m-high fence and four overlapped the 1.2-m-high fence with outriggers.

    Study and other actions tested
  10. A replicated, controlled study in 2008 in fields in Georgia, USA (Stull et al. 2011) found that white-tailed deer Odocoelus virginianus did not jump 2.4-m-high barrier fencing, at 1.8 m fewer jumped if fencing was opaque and 1.2-m-high fences with outriggers angled towards deer were jumped less than those angled away. Among deer that jumped the 1.2-m control fence, fewer jumped each subsequently taller fence (1.5 m: 92%; 1.8 m: 75%; 2.1 m: 42%; 2.4 m: 0%). In opaque fence trails, 90% jumped 1.2 and 1.5-m fences and 50% jumped the 1.8-m fence. With an outrigger, fewer jumped when this was angled towards deer (60%) than away (90%). Three treatment areas (0.1–0.2 ha) were bisected with a test fence. Designs were woven-wire fencing either alone (1.5, 1.8, 2.1 and 2.4 m high), covered with opaque fabric (fence 1.2, 1.5 and 1.8 m high), 1.2 m high with a 0.6-m 50% opaque plastic outrigger angled at 45°, or a 1.2-m-high control fence. Ten adult female deer were each tested with each design in each treatment area. After 48 hours habituation and limited food, deer were enclosed the opposite side of test fences from food. Deer were videoed throughout each 25-hour trial.

    Study and other actions tested
  11. A controlled, before-and-after study in 1995–2002 along a highway through a wetland in Rio Grande do Sul, Brazil (Bager & Fontoura 2013) found that roadside fencing and underpasses reduced the number of road-kills of coypu Myocastor coypus. Fewer coypu were killed by cars after fencing was installed (3.6 coypu/100 km/day) than before (8.3 coypu/100 km/day). The total number of animal road-kills (including all mammals, birds and reptiles) after fencing was installed (10.3 animals/100 km/day) was smaller than before fencing (15.3 animals/100 km/day) (this result was not tested for statistical significance). Road-kill rates fell in fenced sections but increased in the unfenced section (see paper for details). Two sections of a two-lane highway, totalling 10.2 km long, were fenced in 1998. The fence was 50–100-mm mesh, 1.10 m high. Between these sections was a 5.5-km-long unfenced section. Nineteen underpasses in total were also installed along these three road sections. Road-kills were counted from a car from July 1995 to June 2002. Monitoring was conducted at an average speed of 50 km/h, by 2–4 observers, along 15.7 km of highway. A total of 619 monitoring runs were made before fence installation (July 1995 to September 1998) and 571 afterwards (October 1998 to June 2002).

    Study and other actions tested
  12. A 2016 review of fencing studies from USA, Canada and Sweden (Huijser et al. 2016), found that longer fencing along roadsides led to a greater reduction of collisions between large mammals and cars than did shorter fence sections. Results were not tested for statistical significance. Fences reduced collisions between large mammals and cars more in road sections fenced along >5 km (average 84% reduction in relation to before fencing) than in sections fenced along <5 km (average 53% reduction). The review identified 21 fenced road sections (18 from the USA, two from Canada and one from Sweden). Fences were 0.6–33.8 km long and 2.1–2.5 m high. Large mammals targeted by surveys included white-tailed deer Odocoileus virginianus, moose Alces alces, roe deer Capreolus capreolus, mule deer Odocoileus hemionus, elk Cervus canadensis and bighorn sheep Ovis canadensis.

    Study and other actions tested
Please cite as:

Littlewood, N.A., Rocha, R., Smith, R.K., Martin, P.A., Lockhart, S.L., Schoonover, R.F., Wilman, E., Bladon, A.J., Sainsbury, K.A., Pimm S. and Sutherland, W.J. (2020) Terrestrial Mammal Conservation: Global Evidence for the Effects of Interventions for terrestrial mammals excluding bats and primates. Synopses of Conservation Evidence Series. University of Cambridge, Cambridge, UK.

Where has this evidence come from?

List of journals searched by synopsis

All the journals searched for all synopses

Terrestrial Mammal Conservation

This Action forms part of the Action Synopsis:

Terrestrial Mammal Conservation
Terrestrial Mammal Conservation

Terrestrial Mammal Conservation - Published 2020

Terrestrial Mammal Conservation

What Works 2021 cover

What Works in Conservation

What Works in Conservation provides expert assessments of the effectiveness of actions, based on summarised evidence, in synopses. Subjects covered so far include amphibians, birds, mammals, forests, peatland and control of freshwater invasive species. More are in progress.

More about What Works in Conservation

Download free PDF or purchase
The Conservation Evidence Journal

The Conservation Evidence Journal

An online, free to publish in, open-access journal publishing results from research and projects that test the effectiveness of conservation actions.

Read the latest volume: Volume 21

Go to the CE Journal

Discover more on our blog

Our blog contains the latest news and updates from the Conservation Evidence team, the Conservation Evidence Journal, and our global partners in evidence-based conservation.

Who uses Conservation Evidence?

Meet some of the evidence champions

Endangered Landscape ProgrammeRed List Champion - Arc Kent Wildlife Trust The Rufford Foundation Save the Frogs - Ghana Mauritian Wildlife Supporting Conservation Leaders
Sustainability Dashboard National Biodiversity Network Frog Life The international journey of Conservation - Oryx Cool Farm Alliance UNEP AWFA Bat Conservation InternationalPeople trust for endangered species Vincet Wildlife Trust