Exploring drivers of fire-caused forest structural changes using modeling and digital aerial photogrammetry

Abstract:

A warming climate and a century of fire exclusion in western North America have contributed to large contiguous patches of homogeneous forest. Consequently, there are questions about forests’ resilience (ability to recover structure and function after disturbance), particularly in the face of increased fire activity. There is a need to better characterize and understand burn mosaic patterns created by current fire regimes, as well as how these patterns affect forest resilience. We investigated how current fire regimes create forest structures characteristic of resilient forests (heterogeneity in gaps and openings). Specifically, we ask: for a given pre-fire forest structure and burn severity classification, what are the dominant pathways of post-fire forest structure change? How do these pathways vary with biophysical setting and forest type?

We used 40 cm resolution stereo Digital Aerial Photogrammetry (DAP) data from one month pre- and two and four years post-fire for three 2015 fires in Colville National Forest (Washington State) to analyze changes in forest canopy patterns across a gradient of burn severities. We created pre- and post-fire digital surface models (DSMs) from DAP data, then used DSMs to calculate canopy cover, canopy opening, and canopy fragmentation indices. We used a hierarchical classification combined with a machine learning algorithm to produce forest structure classes from the DAP canopy indices. We also addressed biophysical conditions including forest type, pre-fire basal area, topography, and climate data associated with the above dominant pathways using random forest modeling. Most of the forests in the study area contained dense continuous canopy cover pre-fire, and we found multiple pathways to post-fire forest structure per burn severity class. Low- and moderate-severity fire created fine and meso-scale patterns of clumps and openings of various sizes. High-severity fire areas transitioned from mostly continuous canopy to mostly open, fragmented canopies. This novel pre- and post-fire photogrammetry-derived dataset allows for a unique opportunity to use low-cost, high-fidelity data to assess fire-caused change to forest structure across broad spatial scales.