The conversion of rural lands to suburban communities is having profound effects on the state of Maryland. Beginning in the 1960’s and continuing into the present, land use and land cover changes in previously rural areas have left behind a patchwork of habitat that has proven ideal for the white-tailed deer (Odocoileus virginianus). The resulting unchecked increase in deer numbers around an expanding road network with increasing traffic speed and volume have brought to our attention the phenomenon of deer-vehicle collisions.

In Maryland, reported deer-vehicle collisions have doubled in the last nine years (1988-1996), currently up to 3200 and causing nearly $10 million in property damage (MD-DNR Press Release, 1996). The number of collisions is estimated to be 726,000 per year for the entire United States at a cost of $2 billion each year (Conover 1997a). Such alarming numbers have warranted further analysis of deer-vehicle collision trends to determine which conditions (structural or environmental) may bring about high accident rates or serve to decrease them.

Various studies have been performed analyzing temporal trends (Puglisi et al. 1974, Allen and McCullough 1976), vehicle speed (Pojar et al. 1975), animal movement and behavior (Carbaugh et al. 1975, Feldhamer et al. 1986, Putman 1997), and land use/land cover in the vicinity of highway rights-of-way (Bellis and Graves 1971, Puglisi et al. 1974, Carbaugh et al. 1975, Allen and McCullough 1976, Bashore et al. 1985). However, these investigations were conducted without the benefit of computerized spatial analysis technology. The goal of this paper is to perform a more spatially accurate analysis of accident distribution patterns and determine which combination of environmental and structural conditions strongly influence deer-vehicle collision rates in Howard County, Maryland. It is assumed the utilization of GIS can provide new insights into factors influencing deer movement and behavior along interstate highways.

This study was conducted on a 20 mile stretch of Interstate 70 which lies completely within Howard County, Maryland (Figure 1). The interstate enters Howard County in the east via a bridge over the Patapsco River and exits at the Carroll County line in the west. The section is a controlled access, 4-6 lane divided highway built in the early to mid 1960’s. There are 7 interchanges within the county, 4 of which are cloverleafs. Fencing is in place (~6ft in height) along both sides of the interstate for the entire length of the study area with breaks occurring at interchanges. The speed limit ranges from 55 in the east to 65 in the west and mean daily traffic volume for the past 5 years (1995-1999) has been 82,666 vehicles/day in the eastern county and decreasing to 37,294 vehicle /day in the western county.

Howard County has historically been a rural area. However, the rapid expansion of human populations into the area has forever changed this image. Approximately 23% (37,300 acres) of the county’s total area (162,177 acres) is classified as urban or built-up land. Forested areas (~33% of total area) occur mainly in and around residential areas in the eastern county with deciduous forest accounting for 26% of that amount. Small patches of trees exist at the fenceline and cloverleafs of the interstate throughout the study area.

The rural image still holds true for the western half where a large portion of the county's 33% agricultural land can be found.

There are no strong estimates of deer herd size or density within Howard County but the existing habitat and annual deer harvest numbers have led to high assumptions.

The most critical data obtained for this study were Maryland State Police animal-related accident reports for I-70. These reports, provided by the State Highway Administration Office of Traffic and Safety, contained the spatial and temporal data required for a preliminary GIS analysis of deer-vehicle collisions.

The distance (feet or miles) from an accident to the nearest interchange was measured and included in each report. Given this information, along with 1-meter digital orthophotographs, each accident was plotted using ArcView GIS (Esri Redlands, California) to within approximately .01 mile of its reported location although accuracy is unchecked. Next, a 3,000ft buffer was placed around each accident to help enlarge the study area scale and allow for more thorough analysis. Spatial trends including density of accident locations (½ mile search radius) were calculated using Spatial Analyst 1.0 extension (Esri Redlands, California) for ArcView GIS.

Information regarding the movement and behavior of white-tailed deer along I-70 could be gained through analysis of supplemental data provided in accident reports. Temporal trends (time of day, month of year) were available along with weather conditions (clear, rain, fog, snow) and visibility (daylight, dawn or dusk, dark:streetlights on, dark:no streetlights). Diagrams drawn of accident scenes displayed the movement of deer before and after an accident which made possible the determination of how frequently deer entered the right-of-way from the shoulder or median.

Traffic volume data for I-70 was obtained from the State Highway Administration for the years of 1995-1999. The data consisted of hardcopy maps which divided I-70 into 7 segments using interchange locations as boundary lines. Each segment was labeled with the average vehicles/day using it in the given year. The volume was a combination of traffic in both directions (east-west). These values were used to investigate possible relationships between traffic volume and deer-vehicle collisions while also providing a glimpse into urbanization of Howard County.

An analysis of land use and land cover within Howard County was also undertaken to determine if any significant trends were present. Using 1-meter digital orthophotographs and ArcView GIS, forest patches along the I-70 corridor were digitized and overlaid with accident locations. The proximity of tree cover to these locations could then be determined. On a much smaller scale, a countywide land use/land cover dataset was obtained from Howard County GIS. It is important to note this data was classified using an Anderson Level II Scheme (Anderson et al. 1976) therefore land use and land cover were combined within a single theme. Despite this setback, using ArcView GIS, each land use/land cover type was merged and total area calculated. Next, the 3,000ft buffer of accident locations was used to intersect the land use/land cover theme to create a local dataset. Afterwards the numbers of accidents within each land use/land cover type were counted to help determine accident frequency trends.

A total of 97 deer-vehicle collisions were reported for the 20-mile section of I-70 from 1995-1999. The yearly totals were 17, 19, 19, 18, and 24 respectively.

The plotted accident locations yielded some interesting spatial relationships. Visual inspection alone revealed a clustering around the interchanges with cloverleafs. This trend was reinforced by the density analysis (Figure 2) in which the areas of highest frequency (22 max. within ½ mile radius) overlaid almost perfectly with these cloverleafs. Route 97 was the site of 22 accidents with 18 accidents located at Marriottsville Rd., 17 accidents at Rt. 94, 12 at Rt. 32, and 9 at Rt. 29.

Several temporal trends (Figure 3&4) were evident after analysis of the accident database. October and November were the peak of accident activity (30 and 24 respectively) accounting for 56% of the total. May and June also hosted an increase in accidents (12 and 9 respectively). These spring and fall cycles account for 77% of all reported accidents within the study area.

Daily cycles were also present with 58 of 97 accidents (60%) occurring between 6pm and 6am. Within this timeframe, 9 accidents took place during the 6, 8, and 9pm hours and 7 during the 10pm and 12am hours.

The amount of lighting present at the time of each accident was also documented. Not surprisingly, 51 of 97 accidents took place at night on sections of I-70 with no streetlights. In contrast, 37 accidents were reported during daylight hours. Similar in nature, adverse weather conditions and the visibility associated with them influenced accident frequency. All but 2 accidents occurred during clear conditions.

The most significant findings involved deer movement shortly before an accident. 1995-1997 accident reports (55 total) showed 32 accidents were the result of deer entering the roadway from the median. Of these, 20 took place between 6pm and 6am. On a broader temporal scale, 21 of 32 median-originated accidents occurred in October and November with 8 more in May-June for a total of 29 (91%) during the fall-spring cycles.

Traffic volume on I-70 increased from 1995 to 1996 then dropped off roughly 5,000 vehicles/day for 1997. Volume then increased approximately 2,000 vehicles/day from 1997-1999. The marked decrease in traffic flow from 1996 to 1997 was not reflected in the accident data since the yearly totals were the same (19). Overall, average daily traffic volume for the study area decreased in a westward direction (82,666-37,294) from 1995-1999 yet the number of accidents increased slightly.

Spatial analysis of accidents within the proximity of vegetation (tree patches) yielded very significant trends involving cloverleaf areas. There were 9,387 acres of forest having at least one border on I-70 in Howard County (cloverleafs accounted for 166 acres). Subsequently, 64 of 97 accidents occurred within 50 ft and 81 of 97 within 100ft of these forest patches. However, the influence of cloverleafs was hidden within these numbers. Cloverleafs were removed from the vegetation theme to reveal only 39 accidents (a decrease of 26%) occurring within 50 ft and 52 (30% decrease) within 100ft of roadside forest patches. Cloverleafs only constitute 2% of the total forest area which bordered I-70 yet have a clearly significant effect on accident frequency.

Using the county land use/land cover data intersected with the 3,000ft buffer of accidents, it was found 36.5% of the total buffer area was designated cropland and pasture, 24% deciduous forest, and 18% residential. Within these delineations, 40 accidents occurred near agricultural lands, 39 near deciduous forest, and 15 near residential areas. Findings based on this data however were not considered significant due to the coarse resolution (10m), lack of knowledge concerning data creation and integrity, and the combination of land use with land cover into a single theme.

The temporal trends exhibited in this study were well documented in previous literature. Accidents occurring during nighttime hours were seen in Michigan (Allen and McCullough 1976) and Pennsylvania (Carbaugh et al. 1975). The movement of deer to foraging areas at these times was considered the most significant reason (Carbaugh et al. 1975). However, data from Michigan was obtained through spotlight surveys which cannot be considered representative of the entire population, only those visible to a passing automobile at a given time. Pennsylvania data was based on police reports which could be more reliable but only concerns reported accidents and therefore assumes this number was representative of the entire population.

Accidents occurring in spring and fall cycles were also evident in previous studies (Allen and McCullough 1976, Bellis and Graves 1971, Carbaugh et al. 1975, and Puglisi et al. 1974). Three reasons are offered to explain this trend although they are general assumptions made from changes in environmental or human-induced behavior, not results from scientific studies. Peaks in accident activity in the fall are often attributed to the deer mating season (Carbaugh et al. 1975, Allen and McCullough 1976, Puglisi et al. 1974). The corresponding hunting pressure during the fall has also been offered as a possible influence on increased deer strikes (Allen and McCullough 1976, Puglisi et al. 1974). However, an interesting question raised by Allen and McCullough stated accidents should occur during the daytime hunting hours instead of the observed nighttime incidences.

A final theory put forth to explain seasonal variation was forage availability on the right-of-way (Bellis and Graves 1971, Carbaugh et al. 1975, Feldhamer et al. 1986). This trend could be an explanation for the large proportion of median-originated accidents seen on I-70. Vegetation grew uncontrolled on the highway side of the fencing which has led to increased accident frequency (Puglisi et al. 1974). Also of considerable note was the existence of wooded cloverleafs at I-70 interchanges with high densities of deer-vehicle collisions. It is quite conceivable deer utilize these areas (>1 acre each) for cover and forage. Deer are believed to access the right-of-way in search of forage during the fall and spring months due to an insufficiency within woodland areas. Allen and McCullough disputed this assumption by noting sex ratio of kills should then be similar if both sexes were in the right-of-way feeding. However, the sex ratio was biased towards males and females during different seasons in their Michigan study. Further research with longer, more accurate observations is needed in this area to determine the movement and behavior of deer on a daily and seasonal basis.

Density of deer-vehicle collisions was not unexpected due to the fencing in place along the entire length of I-70. The clustering around interchange areas has been found in Pennsylvania (Feldhamer et al. 1986) while overall aggregation of accidents was reported by Puglisi et al (1974) and Bashore et al. (1985). The trend with regard to fencing can possibly be explained by gaps at interchanges which may have a bottleneck effect on deer movement and allow access to the right-of-way. Deer may also exploit holes under or through sections of fencing. More general clustering can be a combination of factors such as vegetation (Carbaugh et al. 1975), topography (Bellis and Graves 1971), or in-line visibility (Bashore et al. 1985).

Traffic volume on I-70 displayed a seemingly inverse relationship to accident frequency. Accidents increased in a westward direction (slightly) even though average daily traffic volume decreased by more than half (82,666-37,294). This trend does not follow conventional wisdom which would suggest more cars relate to more accidents. However, Carbaugh et al (1975) concluded no relationship existed between traffic volume and accident frequency while Allen and McCullough (1976) speculated increased traffic volume during crepuscular hours may have caused higher accident frequencies. The relationship between traffic volume and deer strikes obviously can be quite ambiguous and should be used cautiously as an indicator in predictive models.

Given the current trends on I-70 in Howard County, several mitigation options are available which can reduce the attractiveness of the right-of-way and prohibit entrance to it. Fencing seems the most effective method for deterring deer however it can be extremely expensive. Considering I-70 already has fencing in place, consistent maintenance to fix gaps (caused by humans, fallen trees, etc.) and prevent erosion is required for its effectiveness to be assured. Mitigating gaps in fencing at interchanges can be accomplished through the installation of cattle gates or similar structures embedded within the road surface which deter deer movement. The combination of fencing and cattle gates can reduce the number of deer on the right-of-way, however they do not reduce the incentive. Highway planting and maintenance programs that embellish interstates with flowers and leave areas of edge vegetation along the fenceline must be changed. Aesthetically pleasing highways are a good reflection on the state of Maryland but must be weighed against the $10 million a year spent on deer-vehicle collisions. A properly maintained right-of-way with grass mowed up to and even beyond fencing along with reduced amounts of forage (in cloverleafs and ROW) can make highways less attractive to wildlife. Administrators in transportation must consider the cost and benefit of these mitigation options and decide whether beauty or safety is more important.

One trend evident in previous deer-vehicle collision research is the lack of knowledge concerning deer movement and behavior on and around highway rights-of-way in urban/suburban environments. Feldhamer et al (1986) tracked movements using radio telemetry and found no general patterns relative to highways. This data was collected daily for two years and hourly on 5 occasions for 22 total deer (7 male, 15 female). However, telemetry is field-intensive work, which requires little to no animal movement between readings, and produces fairly course location data. It is not feasible for the continuous tracking of deer which is needed to more accurately assess movement in highway rights-of-way as well as residency time.

The advent of GPS technology and its dissemination into the private sector has spawned new and innovative approaches in animal tracking. The switch off of selective availability only serves to augment these. Accurate locations (x, y, and z) along with a continuous time component bring a powerful analytical tool into the hands of wildlife managers. With the proper configuration of software and hardware, data can be continuously collected from a remote location without animal movement-induced errors. Applications for tracking deer-vehicle collisions are numerous. Use of GPS collars on selected at-risk deer can help determine if deer simply use highways as travel corridors or as foraging areas. On I-70, usage and residency time within cloverleafs could be established as well as routes of access to the right-of-way. Passages under or through fencing which are being exploited by deer would be discernible with GPS data. The home range of individual deer within certain land use and land cover can be calculated as well as dispersal related to changes in these. The newly available data can then be used to fill in knowledge gaps related to deer behavior, which hinder predictive models.

In addition to deer behavior and movement, GPS data can improve and assess the accuracy of reported accident locations thereby increasing confidence in analysis of deer-vehicle accidents. Investigations would be possible at more local scales providing insight into trends around a single interchange. Current methods of locating accidents include obtaining police reports and animal control information which is not always gathered with spatial analysis in mind and therefore accuracy is suspect. More effort among transportation administrators, police, animal control agencies, insurance groups, and natural resource managers is necessary to assemble accurate and complete statewide accident databases, which is the first step to solving a complex problem.

I would like to personally thank Dr. Tim Foresman whose advice and guidance as both a professor and friend are appreciated. Also included is Mr. Sam Walker who has helped this project in many ways by devoting his personal time and efforts.

Manu G. Shah, PE of the Maryland State Highway Administration Office of Traffic and Safety must be congratulated for his efforts in obtaining and providing accident data as well as expressing personal interest in my efforts. The Research Division of State Highway, especially Dr. Richard Woo and Mr. William Branch, have also expressed interest in mitigating deer-vehicle collisions and are currently considering a proposal to fund research proposed in this paper.

Finally, I would like to thank the Center for Conservation Research and Technology (www.ccrt.org) for recognizing the potential of this project and offering their wide range of innovative animal tracking technology and expertise. I look forward to doing some applied research with them in the near future.

Allen, R.E., and D.R. McCullough. 1976. Deer-car accidents in southern Michigan.

Journal of Wildlife Management 40:317-325.

Anderson J.R., E.E. Hardy, J.T. Roach, and R.E. Witmer. 1976. A land use and land

cover classification system for use with remote sensor data. U.S. Geol. Surv. Prof.

Pap. 964p.

Bashore, T.L., W.M. Tzilkowski, and E.D. Bellis. 1985. Analysis of deer-vehicle

collision sites in Pennsylvania. Journal of Wildlife Management 49:769-774.

Bellis, E.D., and H.B. Graves. 1971. Deer mortality on a Pennsylvania Interstate. Journal

of Wildlife Management 35:232-237.

Carbaugh, B., J.P. Vaughan, E.D. Bellis, and H.B. Graves. 1975. Distribution and activity

of white-tailed deer along an interstate highway. Journal of Wildlife Management

39:570-581.

Conover, M.R. 1997a. Monetary and intangible valuation of deer in the United States.

Wildlife Society Bulletin 25:298-305.

Feldhamer, G.A., J.E. Gates, D.M. Harman, A.J. Loranger, and K.R. Dixon. 1986.

Effects of interstate highway fencing on white-tailed deer activity. Journal of

Wildlife Management 50:497-503.

Pojar, T.M., R.A. Prosence, D.F. Reed, and R.H. Woodward. 1975. Effectiveness of a

lighted, animated deer crossing sign. Journal of Wildlife Management 39:87-91.

Puglisi, M.J., J.S. Lindsey, and E.D. Bellis. 1974. Factors associated with highway

mortality of white-tailed deer. Journal of Wildlife Management 38:799-807.

Putman, R.J. 1997. Deer and road traffic accidents: options for management. Journal of

Wildlife Management 51:43-57.

Kevin M. Armstrong

GIS/Remote Sensing Analyst

UMBC Department of Geography and Environmental Systems

1000 Hilltop Circle

Room 007 Social Sciences Building

Baltimore, MD 21061

(410) 455-2900

kmarm@hotmail.com