Mining footprint detection with multi-modal satellite data
Regulators need accurate mining boundaries to check compliance and plan land rehabilitation. In Australia, GIS analysts took months to map each region’s active and historical sites by hand. Therefore, the cost made continuous national monitoring impractical.
Sentinel-2 images alone give unreliable mining boundaries. Clouds interrupt images over time, while disturbed land can resemble bare soil or dry plants. Moreover, optical sensors miss work underground or below tree cover. Sentinel-1 radar and elevation data provide further evidence. However, inputs differ in size and timing, and sites vary in ore type and biome. The model must combine these inputs while preserving those differences.
We built a pipeline combining images, radar, and terrain data. Its inputs included 13-band Sentinel-2 images, Sentinel-1 VV/VH radar, and elevation with slope. Separate encoders extracted and aligned features from each source before combining them. Moreover, we fine-tuned NASA and IBM’s Prithvi model to mark mining areas. Prithvi is a geospatial foundation model, trained broadly on satellite data for later specialised tasks. This was the first published use of a geospatial foundation model for mining segmentation.
I built pipelines to ingest and prepare more than 500 GB of labelled multispectral images. These images came from Google Earth Engine and recorded several light wavelengths. Moreover, I coordinated annotation with geographers at the University of Queensland, UCL, and the University of Nottingham. Agreement checks maintained label quality across ore types and biomes.
We compared image stacks spanning several years to separate active mines from rehabilitated or historically disturbed land. This distinction has direct legal and financial importance under Australian regulation.
The pipeline replaces months of manual GIS analysis with automated maps after each satellite update. Therefore, it supports compliance and rehabilitation work across large mining regions. Uses include Australian state regulation and international mining companies’ environmental, social, and governance (ESG) reporting.
- Monash University, Primary host (A/Prof Risqi U. Saputra, Prof Alex Lechner)
- University of Queensland, Remote sensing science and annotation
- UCL, Annotation and domain expertise
- University of Nottingham, Annotation and domain expertise