STEMS: AI-Driven Precision Inventory

The Challenge

Precision forest management demands frequent, high-resolution structural data to localize problem areas and orchestrate targeted, high-value interventions. Traditional intensive ground surveys are prohibitively expensive, logistically slow, and provide only a sampled, static "snapshot" rather than a complete portrait of a dynamic stand. Compounding this challenge, piece-size distribution varies wildly across mixedwood tracks; without tree-level visibility, industrial operators face unpredictable harvest rate models and massive inefficiencies in supply-chain logistics.

The Solution: Automated Intelligence

STEMS (Single TrEe Metrics and Stand assessment) replaces manual workflows with an automated pipeline. By converting drone imagery into high-fidelity structural data, it delivers a new class of inventory that is accurate, repeatable, and deployable across complex canopy conditions.

Results

Automated stem detection: STEMS isolates individual stems with over 90% detection accuracy.

Species classification: STEMS classifies 35 species with 86 - 96% precision.

Biophysical modeling: Height estimated with LiDAR equivalent accuracy (RMSE: 1.18m)

Aggregation at various scales

Impact

• 40% reduction in field costs
• $0.41/m³ savings
• $280,000 annual savings
• Operationally ready as a webapp

Conducted during our lead advisor's tenure at FPInnovations. The analytical methods, software workflows, and trademarks remain the intellectual property of FPInnovations.