Discussion
This study is the first attempt to account for unevenly distributed wildlife populations when assessing strikes along a railroad; its information was used to help
understand the spatial pattern of train strikes along the CPR. Hotspots were segments
of the CPR that had strike rates significantly higher than the overall mean, and were
identified for elk and deer, but not bears. Hotspots were generally associated with
higher relative abundance for elk and deer at the analysis scale (3 miles). Although
there was disagreement between bears’ relative abundance and strike rates at this
scale (Fig. 9.3c, d), general correlation was apparent at a larger scale. For example,
bear strikes and relative abundance were relatively higher in the western half of the
study area (Fig. 9.3d segments 0–3), compared to the eastern half (Fig. 9.3d segments 14–27). High-risk segments had significantly higher kill rates than expected,
based on estimates of relative wildlife abundance, and they indicated potential
problem areas that may have been overlooked using strike data alone.
There was a significant relationship for elk, deer and bears with at least one train
or railroad design variable. These relationships indicated higher abundance and
train speeds (elk and deer), and larger right-of-way widths (deer) were associated
with increased strike rates. For bears, the number of barriers and presence of bridges
was also positively correlated with strike rates. These results supported the
hypothesis that there are at least three general variables that affect the spatial pattern
of train strikes: 1) the relative abundance of wildlife either on the rail bed or on the
adjacent landscape; 2) train speed; and 3) railroad designs such as highway overpasses, rock-cuts, tunnels, snow sheds and bridges. Other railroad studies have
shown similar associations with bears. In Slovenia, Eurasian brown bears
(U. arctos) were struck at rock cuts and on bridges (Kaczensky et al. 2003). At least
one road study has noted the effect of bridges on strikes with bears (Van Why and
Chamberlain 2003). Huber et al. (1998) studied locations where brown bears were
struck by trains in Croatia, and used variables similar to those of this study. They
found no difference in verge slope or longitudinal or perpendicular visibility but did
detect a difference in the presence of bear foods at strike locations compared to
random locations. They noted a slight difference in the distance at which trains were
first audible at strike locations (Huber et al. 1998). Data on train noise was not
collected in this study, but field observations suggested that train volume varied
depending on train direction within the study area, and the direction in which the
trains were moving when each animal was struck was not recorded.
There are at least four possible explanations for the lack of correlation between
bear strikes and relative abundance found in this study. The first is based on the
strength of non-constant risk phenomena, and three others deal with limitations of
the methods and analysis. First, railroad design or other variables may have strongly
affected strike probabilities. If these variables had strong causal or probabilistic
effects, the spatial pattern of strikes would be a function of the spatial pattern of
these variables and not the bears’ relative abundance. However, it is unclear
whether this is the case because the sample size is relatively low (n = 80) and no
150
B.P. Dorsey et al.
This study is the first attempt to account for unevenly distributed wildlife populations when assessing strikes along a railroad; its information was used to help
understand the spatial pattern of train strikes along the CPR. Hotspots were segments
of the CPR that had strike rates significantly higher than the overall mean, and were
identified for elk and deer, but not bears. Hotspots were generally associated with
higher relative abundance for elk and deer at the analysis scale (3 miles). Although
there was disagreement between bears’ relative abundance and strike rates at this
scale (Fig. 9.3c, d), general correlation was apparent at a larger scale. For example,
bear strikes and relative abundance were relatively higher in the western half of the
study area (Fig. 9.3d segments 0–3), compared to the eastern half (Fig. 9.3d segments 14–27). High-risk segments had significantly higher kill rates than expected,
based on estimates of relative wildlife abundance, and they indicated potential
problem areas that may have been overlooked using strike data alone.
There was a significant relationship for elk, deer and bears with at least one train
or railroad design variable. These relationships indicated higher abundance and
train speeds (elk and deer), and larger right-of-way widths (deer) were associated
with increased strike rates. For bears, the number of barriers and presence of bridges
was also positively correlated with strike rates. These results supported the
hypothesis that there are at least three general variables that affect the spatial pattern
of train strikes: 1) the relative abundance of wildlife either on the rail bed or on the
adjacent landscape; 2) train speed; and 3) railroad designs such as highway overpasses, rock-cuts, tunnels, snow sheds and bridges. Other railroad studies have
shown similar associations with bears. In Slovenia, Eurasian brown bears
(U. arctos) were struck at rock cuts and on bridges (Kaczensky et al. 2003). At least
one road study has noted the effect of bridges on strikes with bears (Van Why and
Chamberlain 2003). Huber et al. (1998) studied locations where brown bears were
struck by trains in Croatia, and used variables similar to those of this study. They
found no difference in verge slope or longitudinal or perpendicular visibility but did
detect a difference in the presence of bear foods at strike locations compared to
random locations. They noted a slight difference in the distance at which trains were
first audible at strike locations (Huber et al. 1998). Data on train noise was not
collected in this study, but field observations suggested that train volume varied
depending on train direction within the study area, and the direction in which the
trains were moving when each animal was struck was not recorded.
There are at least four possible explanations for the lack of correlation between
bear strikes and relative abundance found in this study. The first is based on the
strength of non-constant risk phenomena, and three others deal with limitations of
the methods and analysis. First, railroad design or other variables may have strongly
affected strike probabilities. If these variables had strong causal or probabilistic
effects, the spatial pattern of strikes would be a function of the spatial pattern of
these variables and not the bears’ relative abundance. However, it is unclear
whether this is the case because the sample size is relatively low (n = 80) and no
150
B.P. Dorsey et al.