Independent Territorial Audit for ESG Due Diligence Using Open Satellite Infrastructure

Independent verification of corporate environmental impact claims remains inaccessible to many researchers, civil society organizations, and communities in the Global South due to the cost and infrastructure requirements of conventional remote sensing workflows. We present GhostWorks-ESG, a zero-cost framework for independent territorial audit that combines open satellite data, semantic foundation model embeddings, and free-tier cloud infrastructure. The framework integrates Synthetic Aperture Radar (Sentinel-1), vegetation monitoring (Sentinel-2 NDVI), 64-dimensional AlphaEarth semantic embeddings, and nighttime light activity (VIIRS) to identify and characterize territorial change around industrial facilities. Applied to a 5 km radius around the Landulpho Alves Refinery in Bahia, Brazil, the system detected 21 territorial hotspots between 2019 and 2024, including four locations with critical composite risk scores. Detected signals include a +7.19 dB SAR anomaly near the Joanes River, substantial vegetation suppression (NDVI −0.741), and a 13.4× increase in nighttime activity. All evidence is geolocated, date-stamped, physically verifiable through open imagery, and translated into automatically generated audit questions. The complete pipeline requires no institutional affiliation, licensed datasets, GPU infrastructure, or paid computing resources. Beyond the specific ESG application, this case demonstrates how recent releases of open satellite foundation model embeddings shift the limiting factor for environmental AI research from access to computational infrastructure toward workflow design. We argue that this shift creates new possibilities for independent researchers and organizations in the Global South to develop, deploy, and independently verify AI-enabled environmental intelligence.