Meta and WRI release Canopy Height Maps v2

Meta and the World Resources Institute released Canopy Height Maps v2 on March 10, updating their model and global forest mapping data. The announcement describes a switch to a DINOv3 backbone and more geographically varied lidar training data.

A revised foundation for forest maps

The release includes an open-source model and maps that estimate canopy height from imagery. Meta reports improved agreement with its evaluation measurements. The announcement also describes limitations involving geographic coverage, image geometry and when observations were collected.

Those boundaries matter for anyone combining the release with another geographic dataset. A canopy estimate describes a modeled property of an observed area; using it to estimate carbon, habitat quality or commercial timber requires additional assumptions and validation.

Validate the map where decisions happen

Our analysis: a useful first evaluation is regional. Select areas resembling the intended deployment, compare predictions with independent measurements where available, and inspect errors by terrain and forest type. An average across a global benchmark can conceal a troublesome local pattern.

Preserve acquisition dates, coordinate reference systems and processing versions when joining these maps to property boundaries or field surveys. A clean geographic overlay can still compare measurements taken at different times. Apparent change between releases also needs investigation: it may reflect a revised model rather than a changed forest.

Teams building public dashboards should expose uncertainty and provenance alongside the visual layer. A sharp image can suggest more certainty than the underlying estimate supports. The practical opportunity is a reproducible starting point for analysis, with local checks determining which decisions the new maps can responsibly inform.