Cover Image

Not for Sale



View/Hide Left Panel
Click for next page ( 60


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 59
59 Table 25. Logistic regression analyses for modeling factors contributing to UVCs. Variable Spatially Accurate Mile-Marker Habitat Water 1 Coniferous forest 4 Deciduous forest 5 Open forest mix 2 Distance to drainage 3 Barrier-guardrail N/A+ Road width N/A+ Barrier length N/A 1 Open water N/A HosmerLemeshow test 0.764 0.512 Model development & validation 81.8 76.9 64.4 63.3 accuracies (%) N/A = standard deviation in the logistic regression output was equal to 0 areas (dry meadows, clearings). Further, distance to drainage barrier installation, road widening and improvements, had a significant negative correlation with the occurrence of lighting), thus confounding analysis and resulting in possi- UVCs in the GPS model. The distance to barrier-guardrail and ble spurious results. the length of the barriers within the buffer both showed a Previous explanations for the clustering of WVCs included negative correlation with UVCs. In the mile-marker model, parameters such as animal distribution, abundance, and dis- barrier length showed a significant negative correlation with persal and road-related factors including local topography, UVCs. vegetation, vehicle speed, and fence location or type.190,4,47 In Table 25, results are presented from the logistic regression Few studies have demonstrated that WVCs were correlated analyses for modeling the factors contributing to UVCs using with traffic volume.160,4,47,124 The majority of WVCs in the two datasets. They include a spatially accurate dataset (n = 499 analysis took place in the provincial section of the TCH locations; 391 high- and 108 low-density points) and a mile- followed by Highway 93 South in Kootenay National Park. marker dataset (n = 120; 63 high- and 57 low-density points). However, when the roadkill frequencies were standardized by Also shown are their associated ranking of significant (P < highway length in the study area, the rate of roadkill was 0.10) standardized estimate coefficients and their sign. Num- found to correlate positively with traffic volume. bers indicate the rank of importance of the variable. The sign Factors in addition to traffic volume may influence colli- indicates the influence the variable or variable level has on the sion rates, but may be masked if a more detailed and rigor- probability of a roadkill occurring [() negative correlation or ous analysis is not conducted. Previous research in the same (+) positive correlation]. Hosmer-Lemeshow goodness-of-fit Canadian study area found that elkvehicle collision rates test and overall cross-validation accuracies are included; the were significantly different between road types and declined term N/A means that the standard deviation in the logistic over time on the TCH in Banff and Yoho National Parks, regression output was equal to 0. and Highway 93 South.53 In this analysis, when the effects of traffic volume and elk abundance on elkvehicle collision rates were isolated, the latter was particularly important.53 Interpretation, Appraisal, and Applications Significant interactions indicated that road type influenced these effects and greater elk abundance led to increased Summary of UVC Data elkvehicle collisions. For this analysis, the research team For this analysis, the research team used the largest data- did not include elk abundance as an independent variable base of its kind with spatially accurate information on the because the elk abundance data available for analysis was occurrence and specific carcass location of WVCs. The traf- not at the same spatial resolution as the site-specific loca- fic mortality database is also unique in that it spans a tions in the accurate UVC model. Of the five highways in- relatively short time period (19992005), whereas other cluded in this study, the relative abundance of ungulates is databases, regardless of their spatial accuracy, often contain highest in the provincial section of the TCH and Kootenay roadkill information from a decade or more. The short time River Valley along Highway 93 South. The other highways span used in this analysis is important because over long (TCH-Banff, TCH-Yoho, and Highway 40) are situated at time periods, environmental variables may change (e.g., higher elevations and have lower ungulate densities. Few roadside vegetation and motorist visibility, habitat quality), studies investigating factors influencing WVCs have as can road-related variables (e.g., guardrail and Jersey included data on animal abundance.20,190,53

OCR for page 59
60 Models of UngulateVehicle Collisions Factors that explain collisions. The spatially accurate model indicated that adjacent habitat type was the most im- Spatial distribution and aggregation. The spatial distribu- portant variable in explaining UVCs. The proximity to open tion of UVCs on all five highways in the study area was not habitat increased the likelihood of UVCs as opposed to habi- random. The most notable aggregation was along the 24 km tats characterized by open water, deciduous forest, closed stretch of highway on 93 South. This segment of highway bi- coniferous forest, and open forest mix. Gunther et al.109 re- sects key ungulate ranges in the valley bottoms of the mon- ported that elk were involved in collisions significantly more tane region, with elevation less than 1240 m.188 often than expected in non-forested cover types. Many Several environmental and road-related variables had high deervehicle collisions in Pennsylvania were concentrated explanatory power in describing UVCs on all highways, and around woodland-field interfaces in predominantly open these variables were dependent on the spatial accuracy of the habitat.15 On the other hand, some studies have not found dataset. Results of the univariate analysis demonstrated that this association between habitat type and UVCs.4,25 Wildlife the GPS dataset had substantially more significant variables tends to be associated with specific habitats that provide (n = 10 variables) explaining the factors associated with UVCs resources and environmental conditions that promote occu- than the mile-marker dataset (n = 3 variables). pancy and survival.176 Thus, the spatial distribution of habi- tat types adjacent to or bisected by a highway transportation Predictive ability of datasets. Univariate tests and logis- corridor would likely influence the extent, severity, and loca- tic regression analysis were used to determine the predictive tions of vehicle collisions with wildlife. ability of the two datasets. Landscape variables other than habitat and topography may also be important attributes determining UVCs. For example, Univariate tests. Among the field-based variables, only distance to nearest drainage was significantly and negatively two were identified in the mile-marker dataset as being sig- correlated with the occurrence of UVCs in the spatially accu- nificant in detecting differences between high- and low-kill rate model. Ungulates had a greater tendency to be involved in UVC zones. The same variables were also identified among traffic collisions close to drainage systems. Drainage systems the six significant variables in the GPS dataset. Two of the are known travel routes for wildlife, particularly in narrow variables from the distance-to-landscape features and GIS- glacial valleys such as Banff's Bow Valley.51 Furthermore, generated buffer variables were significant from the spatially research has shown that topography, particularly road align- accurate dataset, whereas the mile-marker dataset had none. ment with major drainages, strongly influences the movement Univariate tests are often used as a preliminary step to of ungulates toward roadways and across them.20,45,159,86 identify one or more variables that are most likely good pre- The proximity to potential barriers such as Jersey barriers dictors of responses to include in an a priori logistic regres- and guardrails was an important predictor of UVCs in the sion analysis.123 The results of the univariate tests of signifi- study area. The same result was found when measuring the cance provide an interesting comparison of how well each length of Jersey barrier or guardrail within the 800 m buffer in dataset is able to describe the relationship between predictor high- and low-kill UVC zones. UVCs were found to occur variables and the location of UVCs. Of the 22 variables used nearer to Jersey barriers and guardrails, which may be because in the initial univariate test to identify variables that differed animals are funneled to the ends of the barriers and cross the significantly between high- and low-kill UVC zones, 10 highway at this point. Furthermore, fewer animals were killed (roughly half) of the spatially accurate variables compared to when the length of barriers within the 800 m buffer decreased. only 3 (< 10%) of the mile-marker variables were statistically These results suggest that the barrier is obstructing animal significant (see Table 24). movement and funneling animals to barrier ends, or particu- lar features in the landscape associated with barriers such as Logistic regression analysis. Results of the logistic re- lakes and steep topography are deterring animals from ap- gression analysis to predict the likelihood of UVCs for the two proaching the highway at these locations. Barnum 14 found datasets analyzed in this study showed the GPS model was that animals crossed more frequently at culverts, bridges, and statistically significant, however, the mile-marker model was at-grade crossings with no guardrail or median barrier. The not. Further, both of the models differed considerably in how only study modeling AVCs that included guardrails in the well they predicted the likelihood of UVCs. Strong support of analysis also found that animals tended to avoid highway the predictive ability of the GPS model compared to the mile- sections with these potential barriers; i.e., collisions were less marker model was found with the higher cross-validation likely to occur where barriers were present.158 scores. These results provide strong evidence that the GPS- The results have important ecological implications because collected data is more likely to be informative in explaining they suggest that median barriers and guardrails may obstruct WVCs than the mile-marker data. animal movement across highways. Further, the results have

OCR for page 59
61 important management implications because state trans- or highway segments with animalvehicle collisions (hotspots) portation agencies are constructing highway median barriers without knowledge of the inherent spatial error,20,15,89,25,208 with virtually no information on how they affect wildlife (2) referencing to a highway mile-marker system,124 (3) refer- movement and mortality. Despite these potential impacts, encing to a 0.1-mile-marker (or 0.1-km) system,190,158,206,125 or the 2003 AASHTO Roadside Design Guide does not address (4) using spatially accurate UTM locations (< _ 10 m error) ob- the impact of median barrier installation. Resource managers tained by a GPS unit at the collision location.53,192,193 and transportation biologists have identified this lack as a The previous review of published studies illustrates that severe shortcoming that needs immediate attention. A recent many studies that modeled animalvehicle collisions typically Transportation Research Board report highlighted the urgent have used data with a significant amount of spatial error, in- need to better understand how wildlife respond to and are troduced by relying on a mile-marker system or an equally potentially impacted by highway barriers.233 flawed approach of not being able to verify the degree of spa- tial error associated with the collision data. One study that Spatial accuracy and interpretation of results. In the rigorously measured the reporting error in the Canadian mile-marker dataset, few landscape variables were significant. Rocky Mountain parks using GPS locations found the error For example, level or gentle topography due to flat terrain is was on average 516 808 m, and ranged from 332 446 m to bisected by the TCH in the province of Alberta. Further, road 618 993m.53 width was a significant explanatory variable due to the width Plotting animalvehicle collisions on maps using grid co- and number of lanes of traffic on the TCH in the province of ordinates may not improve spatial accuracy in reporting. In Alberta. Both of these variables are not as dependent on spatial the previously mentioned study, the average distance report- accuracy, because they were broad-scale measurements with ing error associated with roadkill records (based on occur- low variability occurring on large sections of the highway. rence reports and mortality cards from the mountain None of the distance-to-feature variables showed signifi- national parks) was 969 1,322 m.53 The work presented in cance in the mile-marker dataset. These types of variables are this report is the first to the research team's knowledge to test strongly dependent on spatial accuracy of reporting UVCs. the value of low-resolution spatial data by comparing model For example, if a UVC location has an error up to 800 m, it performance results with a high-resolution spatially accurate will be evident in the measurement of these variables. dataset. Besides learning about the parameters that con- The GIS-generated buffer variables could be used to meas- tribute to UVCs in the study area, the research team discov- ure factors associated with UVCs in a mile-marker dataset.158 ered that spatially accurate data does make a difference in the The buffer encompasses the entire area in which the UVCs ability of models to provide not just statistically significant re- would have occurred, thus the factors associated with that sults, but more important, biologically meaningful results for roadkill are incorporated into the measurement of the transportation and resource managers responsible for reduc- variables. Barrier length was a significant explanatory variable ing UVCs and improving motorist safety. in both datasets and area of open water was marginally signif- These results have important implications for transporta- icant in the mile-marker dataset. These variables would have tion agencies that may be analyzing data that is referenced to to be a broad-scale landscape feature such as the area of a a mile-marker system and is spatially inaccurate. These impli- feature within the entire buffer. cations are equally important for statewide analyses or even smaller districts. Spatially inaccurate data would be suitable Dataset comparison. The primary result of the analyses for coarse-scale analysis to identify UVC hotspots, but for was that the GPS UVC model identified more factors that may fine-scale needs (project or district level), greater accuracy in contribute to UVCs than the mile-marker model. This result data will be essential for a rigorous analysis and development lends strong support to a categorical distinction between high- of sound mitigation recommendations. kill versus low-kill UVC zones (or where they are less likely to A joint U.S.Canada-wide standard for the recording of occur) when modeling is performed with high-resolution animalvehicle collisions would not only stimulate transporta- spatially accurate UVC data. tion departments and other organizations to collect more spa- Animalvehicle collisions have been modeled at a range of tially accurate roadkill data, but it would also allow for better in- spatial scales, from local to state and nationwide analy- tegration and analyses of the data. Some transportation agencies ses.124,183,158,206,192 Previous studies have used readily available are also beginning to use personal data assistants (PDAs) in data (carcass or collision statistics) to identify variables that in- combination with a GPS for routine highway maintenance ac- fluence the risk of animalvehicle collisions and have tivities (e.g., Washington State).126 These two initiatives can help recommended measures to reduce the number of fatalities. agencies collect more spatially accurate and standardized data These studies have largely relied on referencing collision data that will eventually lead to more informed analyses for trans- several ways: (1) accepting and using location data (point data) portation decision making.