A screening tool that combines satellite imagery analysis and geospatial modeling to detect informal lead-acid battery recycling operations and industrial smelters across South Asia.
Proof-of-concept demo · NCR Delhi region · Data includes mockup detections for demonstration
Two complementary machine learning pipelines work in parallel to screen for different types of lead contamination sources.
XGBoost classifier analyzes geospatial features — battery shop density, scrap dealers, population patterns, road networks — to identify likely informal lead-acid battery recycling operations.
Deep learning CNN trained on Sentinel-2 multispectral imagery (13 bands, 10-60m resolution) to detect industrial smelter facilities through building morphology, SWIR reflectance, and land cover patterns.
Each detection receives a risk score (0.0-1.0). Sites are ranked and clustered geographically to optimize field verification routes, focusing limited resources on the highest-impact locations.
Lead poisoning is one of the largest environmental health crises affecting children worldwide.
There is no safe blood lead level. Even concentrations below the CDC reference value of 3.5 µg/dL are associated with reduced IQ, attention deficits, and behavioral problems in children.
Used lead-acid batteries are often recycled informally in residential neighborhoods without environmental controls, exposing nearby communities to toxic lead dust and fumes.
Many contamination sources operate outside regulatory oversight. Satellite imagery and geospatial data can help identify these sites at scale, where traditional monitoring cannot reach.
Machine learning enables systematic screening across entire regions, generating prioritized site lists for field teams — transforming reactive detection into proactive identification.
Explore detected sites, filter by risk level, and understand why each location was flagged.
Explore risk heatmap overlaid with detected ULAB sites, smelters, and verified locations across NCR Delhi.
Select any point to see its risk score, detection date, and the specific features that triggered the flag.
View site type distribution, risk breakdowns, and detection timelines with interactive charts.
Download filtered results as CSV for field planning or generate summary reports for stakeholders.
Built in collaboration with