Citation
Forcey, G. M., K. Gilland, and C. Denny. 2016. A GIS-based model to predict electrocutions of bald eagle and wood stork in Florida. Jean Doucet, editor. Environmental Concerns in Rights-of-Way Management 11th International Symposium, Halifax, Nova Scotia, Canada.
Background
According to APLIC, wildlife electrocutions can cause injury or mortality, reduce system reliability, and increase utility costs. Bird electrocutions occur on utility wires when there is contact between two phases of wires or phase to ground. Larger birds are more susceptible to electrocutions because of the increased likelihood of contacting multiple contact points. Most electrocutions occur on low-voltage lines due to the smaller spacing between energized components. Adding pole equipment (e.g., transformers and capacitors) increases the electrocution likelihood because of the additional energized surface area.
We developed a habitat and structural model to predict bald eagle (Haliaeetus leucocephalus) and wood stork (Mycteria americana) electrocution risk on distribution poles in Florida. We identified areas where bald eagles (Figure 1) and wood storks (Figure 2) would be most like to occur based on the surrounding habitat. The habitat model output was overlaid with pole equipment, the number of phases (lines), and pole frame to estimate a pole-specific relative risk throughout the study area.


Methods
Habitat Model
Based on what is known about the habitat preferences of bald eagles and wood storks, we fit a series of hierarchical linear-mixed habitat models in an information-theoretic framework to understand how large-scale habitat variables affect relative abundance. We used eBird data and FNAI landcover data for bird and habitat data, respectively. We also considered water management district pump stations, landfills, and nest locations as additional predictors that could influence abundance. After fitting habitat models, we translated the statistical model back to a mapped model to facilitate visualization and interpretability. Each variable in the mapped model was weighted based on the variable’s coefficients in the statistical model. Relative risk values were ranked on a 1 to 10 scale.
Structural Model
The second model component used data on pole equipment, the number of phases, framing, and a bird mortality database to understand the specific combinations of pole characteristics that influence electrocutions. In the mortality database, all recorded large bird mortalities were used as surrogates for bald eagle and wood stork electrocutions because there were insufficient numbers of mortalities of the latter two species to draw any meaningful conclusions. We calculated the electrocution risk for each combination of equipment, number of phases, and framing on a pole. This was calculated by dividing the number of incident poles (number of poles where an electrocution occurred) of each type by the total number of poles of the same type within the system. The final electrocution probabilities of all combinations of equipment, number of phases, and framing configurations were sorted and ranked according to the quantile classification from 1 to 10.
Cumulative Risk Model
We used results from both the habitat and structure risk models to calculate a cumulative risk value for each distribution pole within the system. The cumulative model considered both habitat suitability and pole equipment, phases, and framing. Habitat and structural risk were weighted equally in the final risk ranking. Each pole-specific risk value is relative to other pole-specific risk predictions and is not an absolute measure of risk. Relative risk predictions are still useful for understanding the factors that influence electrocution risk and help prioritize management and retrofitting strategies that are optimally beneficial.
Model Evaluation
Two different levels of model evaluation were done:
- We compared the ability of the habitat model to predict bald eagle and wood stork abundance outside the study region
- Compared the habitat model predictions to locations where mortality has been observed, assessing their association with a Spearman’s rank correlation.
Results
Habitat Model
Bald eagles were most influenced by open-water herbaceous vegetation interspersed with other habitats, the number of bald eagle nests, and the number of landfills. Wood stork abundance was most affected by open-water herbaceous vegetation, the number of wood stork nests, and the number of landfills. Other environmental variables we examined had a weaker influence on bald eagle and wood stork abundance.
The model predicted a higher abundance than surrounding areas for bald eagles in small, localized regions, which was driven by the presence of nests and landfills. Wood stork abundance was similar to bald eagle abundance predictions, with the addition of higher predicted abundance in the Everglades National Park.
Structural Model
Bald eagles and wood storks were electrocuted at a rate of 0.16 and 0.02 incidents per 10,000 poles, respectively, over the seven years the data was collected. Poles with switches had the highest electrocution risk for large birds, followed by poles with both transformers and switches. In general, poles with equipment were between four and seven times riskier than poles without equipment. Double-phase poles were the riskiest for electrocutions, compared to one and three-phase poles. Crossarm framing had three times the risk as vertical framing, which was the next highest risk framing type.
Cumulative Risk Model
Poles with a cumulative risk rank between 2–4 (on a 1–10 scale) were the most numerous across the study region. Generally, poles with a risk rank ≥ 8 were associated with a landfill or nesting colony as equipment along would not generally give a pole that risk rank (Figure 3).

Model Evaluation
The bald eagle and wood stork habitat models had moderate to poor fit predicting absolute abundance, respectively. Despite this limitation, the models were better at predicting electrocution incidents of raptors and wading birds (ρ = 0.55) which was the study’s primary goal (Figure 4).

Conclusions
Overall predicted abundance for bald eagles and wood storks was low throughout Florida, with locally higher abundance predictions, mainly due to the scattered presence of nests and landfills. In addition, wood storks had a higher abundance around the Everglades National Park and surrounding agricultural area, likely due to a large amount of open water herbaceous habitat in this area.
Although two-phase poles were the riskiest, this finding was likely because of the low number of poles in the system, which increased the number of incidents per pole. Electrocutions were four times more frequent on crossarm poles than on other framing configurations. This is likely because crossarm framing provides ample perches between phases, and there is more opportunity for birds to come into contact with those phases, according to the APLIC electrocution manual.
The goal of this model was to assign a relative measure of electrocution risk to each distribution pole in the system to prioritize management activities for retrofitting poles to bird-safe designs. Prioritizing the riskiest poles for retrofitting first, ensures that labor and monies for these activities are used optimally to increase system reliability and reduce electrocutions to the greatest extent possible.
Limitations
- Sample size limitations of the mortality database precluded us from evaluating all possible equipment combinations, number of phases, and framing on the poles. Many combinations had no incidents or were very rare in the system. This was addressed by examining equipment, number of phases, and framing independently, though these factors are not likely independent in the real world.
- The structural model was built and validated on mortalities of raptors and wading birds in the study region (not just bald eagles and wood storks). This was done to increase sample size and provide a greater ability to evaluate the electrocution risk of pole equipment, number of phases, and framing.
References
(APLIC) Avian Power Line Interaction Committee. 2006. Suggested practices for avian protection on power lines: the state of the art in 2006. Edison Electric Institute, APLIC, and California Energy Commission.
Forcey, G. M., K. Gilland, and C. Denny. 2016. A GIS-based model to predict electrocutions of bald eagle and wood stork in Florida. Jean Doucet, editor. Environmental Concerns in Rights-of-Way Management 11th International Symposium, Halifax, Nova Scotia, Canada.