Accounting for Habitat Distribution in Caribou Zone of Influence: A New Estimation using Generalized Additive Mixed Modelling MethodsThursday, November 21, 2019 - 11:20am to 11:40am Theatre Two
Barren-ground caribou are a culturally and ecologically important species in Northern Canada. Recent population estimates suggest that the barren-ground Bathurst caribou herd, has declined by as much as 98% from peak numbers experienced in the 1980s and early 1990s. Research elsewhere in Canada and in Scandinavia on other caribou and reindeer populations, together with Traditional Knowledge in the North indicates that caribou may avoid certain anthropogenic activities, leading to theories that barren-ground caribou may respond to industrial development by changing their distribution. The term “Zone of Influence (ZOI)”, was coined to refer to the distance at which caribou change their behaviour, habitat selection and distribution relative to an anthropogenic disturbance (e.g., a mine or road).
In 2012, Boulanger et al. published a first analysis using aerial survey data to determine if Bathurst caribou were avoiding the Ekati-Diavik mine complex, situated within the herd’s summer range, in a ZOI. The authors reported an 11-14 km ZOI surrounding mine infrastructure and hypothesized fugitive dust from the mine complex to be a possible mechanism for this ZOI. Subsequently, a 14 km ZOI around active mines has been assumed by land managers; however, the underlying mechanisms influencing the size of ZOI remain poorly understood.
The current study re-examined the ZOI at the Ekati-Diavik mine complex using an alternative analysis approach that incorporated two major differences from the Boulanger et al. 2012 estimation of ZOI:
1) the underlying assumption that caribou distribution occurs in a complex, non-linear fashion, particularly in areas of higher quality habitat (i.e., heath tundra), as determined by Resource Selection Function studies conducted in the Ekati-Diavik mine complex.
2) the use of a flexible spline model for distance that allowed for an unequal distribution of occupancy that is not accounted for by habitat alone, rather than a piecewise linear model. The current model predicted caribou occurrence peaks at several distances from the mine complex: 8 km, 16 km, and 27 km. Statistical analysis of the results showed that the base habitat model is the primary control (an Area Under the Curve of 0.7897); when distance to infrastructure was added to the model, the model fit was not significantly improved (i.e., an increased AUC score of only 0.0056). This suggests that the distribution of caribou at the mine complex can be accounted for by habitat variables rather than a mine-related ZOI.
It is hypothesized that previous studies on ZOI at Ekati-Diavik identified a ZOI because the authors used straight line fits that constrained the log odds occurrence of caribou to be linear with distance from infrastructure and interpreted the natural variation in the distribution of caribou and the relationship with better quality habitat as a meaningful ZOI. This study demonstrates that the underlying nature of habitat quality and caribou occurrence needs to be carefully considered when examining caribou habitat use. It is recommended that land managers consider using ZOI models that allow the distance effect to differ across direction from human disturbances or alternatively, use a 3D spatial approach.