Methods

We tracked annual changes in caribou habitat by quantifying the amount of disturbed area within woodland caribou ranges in British Columbia and Alberta.

Changes in caribou habitat are tracked by quantifying the amount of disturbed area within caribou ranges, a key component of the federal definition of critical habitat for woodland caribou^1,2,3. For each caribou range, the amount of disturbed area is estimated annually from 1961 to the most recent publicly available updates, ensuring an up-to-date assessment of habitat change. Caribou range boundaries are delineated by provincial jurisdictions.

Disturbance data

Disturbances within caribou range are modelled using Geographic Information System (GIS) data representing various disturbance types, including: wildfire, forest harvest, roads, geophysical exploration lines (e.g., seismic lines), clearings related to oil and gas resource development (e.g., pipelines and well pads), forests killed by mountain pine beetle outbreaks, and other anthropogenic disturbances such as mines, agriculture, railways, and urban areas.

The datasets used represent the best available data for the given jurisdiction (i.e., Alberta or BC) and disturbance type. Note that each dataset will vary in the degree to which it accurately and precisely represents the true disturbance footprint, and thus outputs from all analyses should be viewed as best estimates of disturbance impacts within caribou ranges. See Data Sources for more information on each dataset used to quantify disturbance.

Most datasets contain information on the year of disturbance creation (i.e., timestamped data), which is necessary to track annual changes in disturbed area. However, datasets differ in their timespans. For example, in British Columbia, the timespan for wildfire and forest harvest data extends to 2024 whereas all other disturbance types have data sets that extend only to 2021. Similarly, wildfire data sets in Alberta extend to 2024, but the timespan of other disturbance data extends only to 2022. Metrics of cumulative disturbance over time, such as net total disturbance, are therefore calculated up to the most current year where data are available for all disturbance types.

Accuracy for dates of disturbance will vary based on the source of the date. Some reference datasets are updated annually, while others (e.g., provincial aerial imagery sets in Alberta) may only be updated sporadically, with gaps potentially spanning several decades.

All datasets are clipped to caribou range boundaries. For each caribou range, we report on the Current Current Habitat Condition, the Change in Disturbance Over Time, and Buffered Disturbance.

Calculating the change in disturbance over time

For a subset of disturbance types—forest harvest and wildfire— disturbances are allowed to recover back to undisturbed habitat. Forest harvest and wildfire are thought to influence caribou population declines by increasing leafy, deciduous forage for other ungulates (e.g., moose and deer), causing their populations to increase as well as those of their generalist predators (e.g., wolves), which incidentally prey on caribou at unsustainable rates. Because this process is initiated by an increase in forage for moose and deer, an intuitive criterion for recovery is when forage production in disturbed areas returns to pre-disturbance levels (Appendix A)⁴.

Forage production for moose and deer was modelled by the annual amplitude of change in the Enhanced Vegetation Index (ΔEVI)^4,5,6. The ΔEVI index is sensitive to changes in deciduous and herbaceous vegetation, which comprise a high proportion of moose and deer diet^7,8,9. The ΔEVI index has also shown positive correlations with moose and deer densities^6,10 and moose occupancy⁵.

Generalized additive mixed-effects models (GAMMs) were used to evaluate the response of ΔEVI pre- and post-disturbance in harvested and burned areas (Appendix A). Recovery was identified as the time taken (in years) for ΔEVI to return to its pre-disturbance geometric mean. Users can vary the time for recovery within the 95% confidence interval for a given recovery estimate using the Range Disturbance and Recovery Tool.

Because the recovery process can vary by climate and other site conditions¹¹, recovery criteria were estimated for different biogeoclimatic zones in British Columbia and natural subregions in Alberta (Appendix A). Both classification systems partition large geographic areas into units (zones or subregions) with relatively uniform climatic conditions and distinct vegetation communities. 

At this time, recovery for other disturbance types are not considered. Some disturbances are effectively permanent, such as agriculture and urban areas, and are unlikely to transition back to mature forest. Other polygonal disturbances such as abandoned oil well pads and borrow pits are expected to recover eventually, but most have not shown significant recovery within the timespan of available data, particularly when assessed using ΔEVI-based analyses.

For linear features (e.g., seismic lines and trails), which represent an important and widespread disturbance type, recovery is highly variable and often constrained by site-specific factors such as soil compaction, altered hydrology, repeated human use, and the persistence of early-successional vegetation^12,13. For these features, analyses using ΔEVI are not appropriate because the widths of these features are often much smaller than the resolution of the ΔEVI data (30-m), and the primary mechanisms by which linear features negatively impact caribou stem from changes in predator behavior more so than by these features creating additional forage for moose and deer^14,15. Variation in site-limiting factors means that vegetation regeneration on linear features is frequently slow or stalled, and recovery is not strongly linked to disturbance age as it is for wildfire or forest harvest. Consequently, tracking recovery over time is challenging.

In future iterations of the CHIP, the current state of vegetation regeneration on seismic lines and trails will be evaluated using light detection and ranging (lidar) data.

Calculating buffered disturbance

To calculate annual and total buffered disturbance, all unrecovered human-caused features were buffered by 500 m, which is also done in the Federal Recovery Strategy for boreal caribou¹. This buffered human footprint was combined with unbuffered natural disturbances then overlaps were dissolved to create a single layer. Note that this buffered metric will differ from federal estimates because of differences in the disturbance data used and differences in recovery criteria^{1, 2}.

Buffered disturbance values may differ from other estimates^1,16 because of differences in mapping resolution, how recovery is tracked for various disturbances, and disturbance definitions vary across reporting systems.

• CHIP: Maps disturbances using high-resolution publicly available datasets. These data are updated frequently, including annual mapping when possible (fire and harvest), and recovery of wildfire and timber harvest is based on when they are no longer expected to contribute to increased moose and deer forage.
• Federal Recovery Strategy: Maps all visible disturbances using LANDSAT imagery at a scale of 1:50 000. These data are updated infrequently, use coarse resolution data, and recovery of disturbances is based on visual interpretation from imagery.

The division-by-three adjustment applied to unbuffered linear-feature area for the southern mountain caribou (see Data Sources) was not applied when calculating 500 m buffered disturbance. Applying the adjustment before buffering would have very little effect on the resulting buffered disturbance estimate because the buffer, rather than the width of the original feature, accounts for most of the mapped area. Retaining the original 100-m pixels also maintains consistency with the linear-feature data used by Environment and Climate Change Canada for buffered disturbance estimates.