The decline of many woodland caribou (Rangifer tarandus caribou) populations is thought to be linked with habitat disturbances resulting from industrial development, including timber harvesting and its network of haul roads. Defining a disturbance-abundance relationship offers a tool to assess and potentially manage for the influence of disturbance on caribou abundance. Defining this relationship is challenged by limited historical land use and abundance data, the choice of a disturbance measure, and variability in the relationship between subpopulations and across different habitat types (core vs. matrix habitat). For 12 subpopulations of woodland caribou within the southern mountain population, we linked longitudinal caribou abundance data with historical forestry disturbances simulated from forest harvest data. We compared disturbance measures estimating the proportion of forested area commercially harvested with even-aged, regeneration treatments (cutblocks) and converted to roads for transporting timber within subpopulation-specific core and matrix habitats as predictors of caribou abundance. Non-linear mixed models provided evidence that disturbances in matrix habitats negatively influenced caribou abundance, with the effects in core habitat being variable between subpopulations. Of the disturbance types evaluated, the best predictors included roads buffered by 50 m (R50), cutblocks ≤80 years old, and the cumulation of cutblocks ≤80 years old plus roads buffered by 50 m. The top-ranked model was composed of R50 present in core and in matrix habitats. This model predicted a 42% (95% CI = 33–51%) reduction in caribou abundance for every 1% increase in matrix R50 (holding core R50 constant). We recommend that in addition to existing protections of core habitat, few if any new roads should be built in core habitat, and that timber harvest in matrix habitat should be designed to minimize the establishment of associated roads.