Measuring boreal forest fragmentation after fire: Which configuration metrics are best?

Studying changes to the shape, size, and arrangement of patches of forest habitat remains a challenge in the field of landscape ecology. A major issue is that most landscape pattern metrics measure both the amount of habitat as well as habitat configuration. To obtain independent measures of habitat configuration, the established approach is a detrending analysis using regression residuals between configuration metrics and habitat abundance. We compared this detrending approach with a new set of three normalized configuration metrics and evaluated their suitability to detect changes to forest fragmentation in the Canadian boreal forest as a result of fire disturbance. We found that the combination of two of the three normalized configuration metrics responds well to habitat configuration dynamics after fire, whereas the classical approach provides an inferior measure of changes to habitat configuration. Our second objective was to examine whether spatial configuration metrics can be directly predicted from non-spatial surrogates that describe the initial habitat structure and the disturbance regime. This has practical value, because many models that guide forest management and conservation are non-spatial. We found that normalized configuration metrics were predicted with moderate accuracy (average adjusted r2 = 0.53), while detrended metrics could not be predicted (r2 = 0.16).

► Habitat fragmentation is an important cause of population decline or extirpation. ► To better monitor habitat fragmentation we developed and tested a new set of metrics. ► Metrics were largely independent of total habitat, allowing for flexible comparisons. ► In a case study of fire disturbance, trends in fragmentation were accurately captured. ► The metrics could also be well predicted into the future based on disturbance regimes.