Spectral Differentiation of Old-Growth Ponderosa Pine Forest in Southern Arizona
Presentation Time: Wed, 05/06/2026 - 16:30
Keywords: old-growth forest, ponderosa pine, remote sensing, Random Forest classification, Sky Island
Old-growth stands of Rocky Mountain ponderosa pine (Pinus scopulorum) preserve unique forest ecosystems within the mountain ranges of the Sky Islands of southern Arizona. The spatial distribution of intact tracts of these mature forests, however, remains largely undocumented. Accurate field-based age determination, for the purpose of the confirmation of meaningful delineation between mature and regenerating forests, is limited by the associated logistical burden. This study examines the viability of the use of multispectral satellite imagery to predictively differentiate old-growth and new-growth ponderosa pine based on measurable spectral characteristics to focus field-based confirmation. A supervised machine learning classification model using a Random Forest algorithm is trained using Landsat surface reflectance imagery from Metolius Research Natural Area within Deschutes National Forest in Oregon, a designated research area of ponderosa pine forest containing significant known old-growth footprint. Prediction layers for the model are derived from vegetation and moisture indices, spectral reflectance, canopy and organic matter density, and soil exposure. The results of this model are then applied to a selected section of the Coronado National Forest in the vicinity of the Mount Lemmon Wilderness Area to predictively map the spatial extent of old-growth ponderosa pine forests in the region. Field observations in areas of potential old-growth reported by the model are used to confirm the accuracy of the classifications. This study provides a reproducible remote sensing framework for regional identification and monitoring of old-growth ponderosa pine forests within the Sky Island region of Arizona and demonstrates the effectiveness of machine learning ecological classification.
