Introduction

The intelligence beneath the forest

Every year, billions of trees are planted. Most of the decisions behind those plantings — where to plant, which species, what method — are made on intuition, historical averages, or best guesses.

Canopi replaces guesswork with prediction.

Our API returns survival probabilities for tree plantings at any site in our coverage network, accounting for species, planting method, soil composition, climate patterns, and topographic position across multiple time horizons. Think of it as the root network beneath the forest — invisible, essential, and intelligent.

Ask us: if I plant Douglas Fir by drone on this slope in the Cascades, what’s the probability those trees are alive in five years?

We’ll give you a number. And the reasoning behind it.

What Canopi does

Canopi is a predictive API, not a monitoring platform. Most forest technology looks backward — satellite imagery of existing forests, verification of current carbon stocks. Canopi looks forward. We model what happens next.

A single API call returns:

• Survival probabilities across 1-year, 3-year, and 5-year horizons
• Risk factors — the specific environmental conditions most likely to threaten survival at that site, derived from interpretable machine learning (SHAP analysis)
• Method comparison — how survival probabilities change depending on whether trees are manually planted or drone-seeded at the same location

Who it’s for

Reforestation operators — Optimize site selection. Compare planting methods before committing resources. Bid contracts with predicted outcomes, not gut estimates.

Carbon credit buyers and underwriters — Quantify reforestation risk before committing capital. Canopi predictions turn planting projects from hopes into probabilities.

Impact funds and insurers — Build portfolio risk models grounded in site-level prediction data. Underwrite reforestation outcomes with actuarial intelligence.

Conservation technologists — Integrate survival predictions into your own platform. Canopi is the intelligence layer — you build the experience.

The method question

Manual planting or drone seeding? It’s a decision worth millions of dollars annually — and until now, it’s been made on intuition.

Canopi is the only platform that models how planting method affects survival at the site level. Same location, same species, two methods, two different probability curves. The /v1/predict/compare endpoint returns both side by side.

Current coverage

Canopi currently covers Oregon and Washington — the Pacific Northwest. Our prediction network spans approximately 18,000 sites across both states, covering 20 tree species across the full ecological range from dry-side juniper woodlands to wet coastal hemlock forests.

We are expanding to additional US regions. The model architecture is designed for national coverage using USFS Forest Inventory and Analysis data available in all 50 states.

The model

Our current model is Mycel v0.2, the second generation in the Mycelium Series. It’s trained on decades of field observations from the US Forest Service, fused with soil surveys, climate records, and terrain data through gradient-boosted decision trees (XGBoost).

The Mycelium Series follows the lifecycle of the organism that holds forests together:

• Spore — genesis. The first model.
• Mycel — the network forming. Where we are now.
• Hypha — full reach. National coverage.
• Rhizo — deep roots. Multi-source, continuously learning.

Each generation deeper, wider, more connected.

Get started

The fastest path from here to your first prediction is the Quickstart — you’ll have a working API call in under two minutes.

For a deeper understanding of how predictions work, start with How Predictions Work.