ThinkingEarth Talks: UC2B

What is the main purpose of the Forest Biomass Monitoring and Carbon Sink Assessment tool?

The primary purpose of the tool is to provide accurate, scalable, and high-resolution monitoring of forest Above-Ground Biomass (AGB) and carbon stocks using Earth Observation data and AI. It is designed to support Measurement, Reporting and Verification (MRV) for REDD+, improve carbon accounting, assess carbon sink performance, and enable transparent evaluation of carbon credit projects over large areas such as the Amazon Basin.

What challenges in the carbon credit industry does this tool aim to address?

The tool addresses major challenges such as lack of transparent carbon stock assessments; labour-intensive, expensive field-based verification; limited scalability of traditional MRV methods; insufficient spatial resolution and monitoring frequency of existing solutions. By automating biomass estimation from satellite data, the tool aims to provide faster, more consistent and scientifically validated assessments.

How does the tool use satellite imagery and AI to estimate forest biomass and carbon stocks?

The tool integrates multiple Earth Observation datasets into a deep learning framework built around the Copernicus Foundation Model to generate high-resolution Above-Ground Biomass (AGB) maps. Earth Observation datasets, including Sentinel-1 SAR imagery, Sentinel-2 optical imagery, and Digital Elevation Model (DEM) data, are used as input. The trained model predicts spatially continuous AGB maps, which are then integrated into the web platform to estimate forest carbon stocks, and support visualization, analysis, and reporting for users.

What are the main data sources used by the tool, and why were they chosen?

The principal datasets include:

• Sentinel-1 SAR

• Provides forest structural information.
• Works in all weather conditions and through cloud cover.

• Sentinel-2 optical imagery

• Captures vegetation reflectance useful for biomass estimation.

• GEDI LiDAR

• Provides highly accurate forest height and biomass reference measurements for model training and validation.

• ESA Climate Change Initiative (CCI) Biomass products

• Used as benchmark biomass datasets.

• Digital Elevation Models (DEM), meteorological variables and land-cover datasets

• Improve model performance by accounting for environmental conditions influencing biomass.

These complementary datasets provide both horizontal and vertical forest information, leading to more reliable biomass estimation.

How can this tool improve the transparency and credibility of carbon credit projects?

The tool improves transparency by producing independently derived biomass estimates; providing continuous satellite-based monitoring rather than occasional field visits, reporting validation metrics; incorporating Explainable AI (xAI) to improve model interpretability, supporting repeatable and traceable MRV workflows. This allows investors, project developers and regulators to make more informed decisions regarding carbon credits.

Who are the primary users of this tool, and how do they benefit from its outputs?

The intended users include: Carbon project developers, Forestry project managers, Financial institutions, Carbon credit investors and traders, Government agencies, Environmental organisations, Researchers.

The tool provides: high-resolution biomass maps, carbon stock estimates, carbon sink assessment, decision support for forest management, independent evaluation of carbon assets, reporting for MRV compliance.

How is the accuracy of the biomass estimates validated against real-world measurements?

The biomass estimates are validated by comparing the predicted values with trusted reference measurements, including GEDI LiDAR observations and, ESA-CCI. The agreement between predictions and reference data is assessed using standard accuracy metrics such as RMSE, MAE, and R².

What are the biggest limitations or challenges when monitoring forest biomass using Earth Observation data?

Some important challenges include: cloud cover affecting optical imagery; spatial variability across different forest ecosystems; limited availability of high-quality ground truth data; need for consistent annual monitoring over very large regions; high computational requirements for processing multi-source satellite data; maintaining model accuracy across diverse geographical regions.

Would this tool have to be adapted to different forests depending on size and density?

Yes. Different forest ecosystems differ in: tree species composition; canopy density; forest height; biomass distribution; climate; topography.

To maintain high accuracy, AI models may require calibration or fine-tuning using representative training data from different forest types. The project also highlights improving model transferability across diverse forest ecosystems through foundation models and domain adaptation techniques.

What future improvements or additional features are planned to enhance the tool and its applications?

Future enhancements, include: expansion to annual monitoring across large geographical regions; improved transferability across diverse forest ecosystems; independent carbon scoring methodologies; and transform the platform from a biomass mapping tool into a comprehensive decision-support system for carbon accounting, climate policy, and sustainable forest management.