Forests are vital for our planet. They help fight climate change by absorbing a lot of carbon dioxide from the air, acting as major carbon sinks. They store large amounts of carbon in biomass and soil, estimated to absorb about 30% of human-caused CO2 emissions annually worldwide.
However, scientists and project managers must track forest health. They need to know how much carbon forests store. This helps ensure that efforts to protect or grow forests are effective. This is called measuring, monitoring, reporting, and verifying forest carbon, often shortened to MMRV.
Eyes in the Sky: How Remote Sensing Sees Forests Differently
Measuring carbon in forests is tricky and expensive. Usually, people go out into the forest and measure trees by hand, which takes a lot of time and effort. It’s hard to do this over large areas, especially in dense or remote forests.
This is where remote sensing comes in.
Remote sensing is a way to gather information about forests without going there in person. It uses satellites, airplanes, or drones equipped with cameras and sensors. This technology can take pictures and collect data. It helps scientists learn how tall trees are, how dense the forest is, and how much carbon it might store.
There are different kinds of remote sensing data:
- Optical imagery: like normal photos taken from space or planes, showing the tops of trees and land features.
- Radar: which uses radio waves and can see through clouds and work at night.
- Lidar: which uses lasers to map the exact height and shape of trees in 3D.
The Challenge with Remote Sensing Data
Each data type has strengths and weaknesses. Optical images are good and widely available, but they can’t see through clouds and only show forest surfaces. Radar can see through clouds but has trouble measuring details in dense forests. Lidar is very accurate but expensive and covers less area.
To get the best info, scientists combine different types of data using artificial intelligence (AI) and machine learning techniques. Machine learning helps computers find patterns in huge amounts of data to make better estimates.
Meta’s Canopy Height Map: AI-Powered Forest Intelligence
Meta developed a unique AI model that merges high-resolution satellite images with lidar data. This model maps tree canopy heights globally with great detail—less than one meter per pixel. This means it can see individual trees in many places.
The map and the AI model are open-source and freely available, so anyone can use them to help forest projects. They enable better planning, monitoring, and verification of forest carbon projects. Reza Rastegar, Senior Manager of Research Science at Meta, stated:
“When applied thoughtfully, we believe AI research and remote-sensing tools, particularly those that are open source, have the potential to revolutionize the transparency and accessibility of the carbon market.”
Meta’s model has been validated with mean absolute errors of 2.8 meters in U.S. forests and 5.1 meters in Brazil. This reflects a promising improvement in estimating canopy height at fine scales. These advanced datasets and models are helping to track natural regeneration, selective logging, and forest degradation more accurately, which is vital for credible MMRV of carbon credits.
What’s special about this model?
- It works globally with very fine detail.
- It can help identify important areas to protect or restore.
- It can make new maps for different times if good images are available.
- It helps detect small changes in forests, like selective logging (cutting some trees but not all).
- It supports methods from carbon credit standards. This is important for those who need dynamic baselining or updating project baselines with real data from nature.
RELATED: Meta and Microsoft Secured Long-Term Carbon Credit Deals to Support Olympic Rainforest
From Pixels to Carbon Credits: Turning Data into Climate Action
Forest carbon projects use different official methods to create and verify forest carbon credits. The three main methods Meta focuses on are:
- Verra VM0045 – for improved forest management (IFM).
- Verra VM0047 – for afforestation, reforestation, and revegetation (ARR).
- American Carbon Registry (ACR) IFM – a US-based improved forest management method.
Here’s how Meta’s canopy height map and AI model fit into these methods:
- In project planning, the map helps find good parcels of forest to include, determine project boundaries, and understand forest structure.
- For dynamic baselining, especially in ARR and ACR’s IFM methods, the AI model can help update baselines based on real forest growth or loss over time.
- For reversals monitoring (tracking if carbon gains are lost, e.g., due to fire or logging), the map gives better details to detect forest disturbances.
The Fine Print: What Meta’s Model Gets Right—and Where It Struggles
Many traditional satellite products can’t reliably measure forest height or biomass in dense forests or small areas. Meta’s model, because it uses very high-resolution images, helps overcome this.
Monitoring small or fragmented forests, river corridors, or areas with selective logging is crucial. These places are difficult to track using low-resolution data.
Meta’s canopy height model is a powerful tool for estimating forest structure, but it comes with limitations. It works best with high-quality imagery at 0.5–1 meter resolution. The global canopy height map uses images from 2009 to 2020. This means it might not show current forest conditions. So, there’s a need for updated maps.
Accuracy may also drop in underrepresented forest types, so local validation with field or lidar data is advised. Using the model requires significant computing power and technical expertise, which may limit adoption.
For forest carbon projects, remote sensing offers great promise but faces barriers. There is no universal agreement among registries, buyers, and developers on acceptable methods or datasets.
In addition, technical skills, computational capacity, and access to affordable, high-quality datasets remain limited. Uncertainty around accuracy—and lack of consensus on acceptable error levels—make trust and comparability difficult.
For the identified barriers, the report authors recommend the following:
Experts want clearer standards for how datasets can be used. They also seek better reporting on uncertainty and clearer rules for issuing carbon credits. A global benchmarking database with verified data and a central portal for quality datasets could help boost adoption.
Moreover, easier AI tools would make this process smoother. Integrating advanced models like Meta’s into accessible platforms, alongside collaborative standard-setting, will be crucial to scaling reliable forest carbon monitoring and verification.
Examples of New and Exciting Uses of Meta’s Model
- Counting trees in agroforestry projects to monitor performance.
- Mapping old-growth forests and biodiversity hotspots.
- Detecting subtle forest degradation, like selective logging.
- Monitoring reversals (losses of carbon stored) with greater accuracy.
- Supporting more accurate estimates of above-ground biomass.
Forests are vital to fighting climate change by storing carbon, but measuring how much carbon they hold and how this changes over time is tough. New technologies like remote sensing are making this easier, faster, and cheaper.
Meta’s AI-powered canopy height map is a cutting-edge tool offering very detailed, global forest height data that can help in planning, monitoring, and verifying forest carbon projects.
