Modeling the Regeneration of Aspen Using Remote Sensing, Machine Learning, and Field Validation in Post-Fire Ecosystems.

Abstract

Post-fire aspen recovery in mixed-conifer systems remains poorly understood, particularly where conifer encroachment has suppressed the clonal root networks that drive resprouting. Aspen (Populus tremuloides) is a keystone species across western North America, underpinning disproportionate levels of biodiversity and ecosystem service provision relative to its spatial extent. The Monroe Canyon Fire (summer 2025), having burned over 29,834 ha (73,721 ac) of encroached aspen and mixed-conifer forest in central Utah, presents a critical opportunity to examine how fire interacts with legacy vegetation composition to shape post-fire recovery trajectories. This study models post-fire landscape regeneration across the Fishlake National Forest, investigating how fuels, treatments, and burn severity interact to drive post-fire aspen regeneration, with ecology and management application at the forefront. To do so, satellite imagery and ground data from past operational crown fires were integrated to identify the strongest predictors of regeneration, using ramet density (stems/ha) as the response variable alongside topographic, treatment, severity, and proximity metrics. FASMEE plot data from three operational high-severity burn units served as training data for the model. Burn severity emerged as the strongest predictor of aspen regeneration success, with secondary contributions from pre-fire aspen presence and topography, supporting the hypothesis that high-severity fire promotes regeneration by reducing competing overstory and stimulating root sprouting, while legacy vegetation composition and aspect-driven moisture gradients moderate recovery outcomes. Integrating remote sensing spectral indices within a Random Forest modeling framework, a spatial predictive model of post-fire aspen recovery potential was developed from FASMEE training data and applied across the Monroe Canyon Fire landscape. Outputs distinguish areas of high regeneration potential (≥5,000 stems/ha) from those likely requiring active management intervention. The model identified over 880 ha (2,175 ac) of high-certainty regeneration and an additional 4,955 ha (12,244 ac) with greater-than-chance likelihood of recovery. Results are intended to support post-fire restoration prioritization. An improved understanding of how fuels, treatments, and fire severity shape recovery has broad implications for forest and fire management across the western United States.