Exposing Greenwashing: Satellite ML for Carbon Credit Verification
Speaker
Neeraj Pandey
Neeraj is the co-founder of Vivid Climate, a climate management and DMRV platform. Neeraj is a polyglot. Over the years, he has worked on a variety of full-stack software and data-science applications, as well as computational arts, and likes the challenge of creating new tools and applications, and is an active international speaker with talks and tutorials presented at multiple conferences.
Abstract
The carbon market is set to reach 1T dollars by 2030, yet 84% of offsets fail to deliver real climate benefits. Verification still relies on sparse site visits and self-reported data. This poster shows a Python workflow that audits carbon projects using satellite imagery and ML, detecting over-crediting and leakage in REDD+ sites. With open data and open-source tools, anyone can compare claimed versus observed forest outcomes and verify what projects actually deliver.
Description
The global carbon market is projected to reach $1 trillion by 2030, yet it remains plagued by a fundamental credibility crisis - studies show 84% of offset projects fail to deliver promised environmental benefits (that's potentially $840 billion in phantom climate action). Traditional verification relies on infrequent site visits and self-reported data, creating blind spots where fraud thrives undetected. This poster presents a Python-based framework for independently auditing carbon credit projects using satellite imagery, drone data and machine learning, exposing risks that conventional methods miss.
Our pipeline processes Landsat-8 and Sentinel-2 imagery using rasterio and geemap, calculating vegetation indices (NDVI/SAVI) to monitor forest health across project boundaries. Using scikit-learn's Isolation Forest, we detect anomalous claims without needing labeled fraud examples - the algorithm identifies carbon stock reports that deviate from satellite-observed patterns, flagging outliers for investigation rather than requiring us to define 'fraud' in advance. Time-series analysis with the BFAST algorithm identifies sudden deforestation events and leakage patterns - where "protected" forests simply shift destruction to neighboring areas.
Applied to certified REDD+ projects, the framework identifies over-crediting risks where claimed sequestration exceeds observed vegetation changes by 2-3x. It detects leakage signatures where deforestation increases in buffer zones despite reported project success. By comparing reported stocks against satellite-derived indices while monitoring adjacent areas, the pipeline addresses key integrity challenges.
Visitors can explore an interactive demo comparing claimed vs. observed forest cover for real carbon projects, see anomaly detection results on live satellite feeds, and access the full pipeline for their own investigations.