Spatial forest patterns in Canada are largely driven by disturbances like wildfires, harvesting, road building, and energy development (e.g., mines, oil and gas). These spatial patterns of forests impact habitat use for a wide range of organisms in Canada. To better understand how these disturbances contribute to the evolution of landscape forest spatial patterns over time, we analyzed thousands of landscapes across Canada using Landsat satellite image time series. We stratified Landsat "scenes" across major ecozones of the Canadian boreal forest and downloaded all available Landsat imagery, which were transformed into annual "best available pixel" composites (read more). Pixels in these annual Landsat satellite images were classified into forest, non-forest, and water using a time series algorithm that detected forest disturbances (i.e., conversion from forest to non-forest) and forest recovery (i.e., conversion from non-forest to forest) in order to track trends through time.
We further subdivided the Landsat images into smaller 5,000 ha landscapes and used Fragstats to calculate four simple landscape pattern metrics: forest cover (% of total landscape area), largest forest patch (the largest contiguous forest patch expressed as % of total landscape area), forest edge density (m/ha), and core forest cover (total forest cover nibbled by 90 m, expressed as % of total landscape area). Forest cover is an indicator of forest dominance in a particular landscape, largest forest patch indicates how intact the forest is, edge density indicates abundance of forest edge habitat (important for ecological processes that occur at forest edges), and core forest cover relates to the relative abundance of forest cover available for habitat of some organisms.
Forest cover declined slightly over the 25-year time series, which also reduced the largest forest patch, but these trends were amplified for the other fragmentation metrics. Forest edge density increased significantly for densities between 20 and 70 m per ha. Landscapes that had low edge density at the beginning of the time series declined by the end of the time series. Most landscapes had core forest cover above 60% in 1985, but by 2010 the trend had shifted between 20 and 60%.
Read the article in PLOS ONE.