Understanding which forest structural attributes affect insect infestations at fine scale is challenging. In this paper, we investigated the use of a bitemporal airborne laser scanning dataset to extract forest structural characteristics linked to spruce budworm (Choristoneura fumeraria) infestations in the boreal forests of Quebec, Canada using unsupervised machine learning. Canopy height and cover best characterized the effect of different infestation severity on the landscape.

Landscape configuration is known to affect insect infestation movement and severity, but a spatially-explicit characterization of which properties of the landscape configuration affect these dynamics is still missing. In this paper, we leveraged 13 years of Landsat time series surface reflectance composites and Random Forest to assess the role of forest fragmentation on the severity of the infestations in Quebec, Canada. We found that increased fragmentation may increase the likelihood of more severe infestations based on past observations, potentially as a result of landscape saturation effects.

Forest resilience is a central aspect in forest management, but frameworks to quantifying it in a comparable way across studies and environments remain theoretical. In this paper, we operationalized these frameworks by leveraging three decades of Landsat time series surface reflectance composites, in combination with a two-stage clustering approach to quantify forest resilience to insect infestations in a spatially-explicit fashion. To do so, we calculated spectral impact and recovery rate for infested pixels in the boreal forests of Quebec, Canada via three key spectral indexes describing greenness, moisture, and structure. Our results show that boreal forestes are more resilience in greenness and structure than moisture, potentially owning to post-infestation foliage recovery dynamics.