Mapping the World's Forests with Greater Precision: Introducing Canopy Height Maps v2

March 10, 2026

TL;DR

Canopy Height Maps v2 (CHMv2) is an open-source model and world-scale map set for forest analysis.
It utilizes Meta's DINOv3 self-supervised vision model for high-resolution global forest mapping.
CHMv2 offers substantial improvements in accuracy and detail compared to previous versions.
The model enhances capabilities for monitoring forest health, tracking restoration, detecting degradation, and estimating carbon storage.
It has seen a significant increase in its R² score from 0.53 to 0.86, indicating higher prediction accuracy.
CHMv2's training dataset was expanded and improved with more diverse lidar examples and specialized tools.
The technology is being used by organizations like Forest Research in the UK and the European Commission's Joint Research Centre.
It supports urban planning initiatives such as 'Cities for Smart Surfaces' in the US and 'Cool Cities Lab' by WRI.
Future work aims to improve predictions in data-sparse regions, address viewing-geometry effects, and extend temporal coverage for change detection.

Continue reading the original article