The Conservation Fund (TCF) manages large timberland properties where accurate forest stand boundaries and stand-level attributes are critical for inventory, management, valuation, and long-term planning. However, much of this work is currently performed manually, making it time-intensive, costly, and difficult to scale consistently across properties. This project addresses that challenge by developing a reproducible, end-to-end geospatial workflow capable of automating stand delineation and generating reliable stand-level attributes from remotely sensed data. The system integrates satellite imagery and disturbance history data within a unified Python-based workflow. Segmentation is performed using satellite imagery and the Segment Anything Model (SAM), which demonstrated improved delineation performance compared to earlier methods. In parallel, preprocessing of Land Change Monitoring System (LCMS) data was standardized to handle multi-year disturbance patterns and missing observations, enabling consistent derivation of stand establishment year and stand age. These preprocessing components are integrated into a modular pipeline that combines ArcPy geoprocessing, remote sensing feature extraction, and machine learning workflows. Following stabilization of the preprocessing pipeline, the project transitioned to model development and evaluation. Machine learning models were implemented to predict stand origin and cover type using Sentinel-2 spectral features and disturbance history variables. Model performance was assessed using property-level validation strategies to ensure robustness across landscapes. Overall, this project delivers a modular, analyst-ready software workflow capable of supporting forest inventory and planning decisions. By reducing reliance on manual interpretation and enabling scalable stand-level analysis, the system provides a foundation for more efficient timberland management across diverse properties.
Mentor: Kyle Bradbury
Project poster (PDF)
