Pre-Islamic Archaeological Sites — Remote Sensing Survey
Using multi-spectral satellite imagery and Google Earth Engine to identify and map pre-Islamic archaeological sites from satellite data.
Archaeological prospection from satellite imagery has a long history — early practitioners were working from declassified CORONA imagery in the 1990s. What’s changed is the availability of free, high-temporal-resolution multi-spectral data and the platforms to analyse it at scale.
Approach
This project used Sentinel-2 Level-2A imagery accessed through Google Earth Engine. The core challenge is separating genuine archaeological signatures from naturally occurring spectral variation in arid environments.
The workflow:
- Temporal compositing — mosaicking multiple acquisitions reduces cloud interference and averages out ephemeral vegetation responses
- Index calculation — NDVI, NDWI, and a custom clay mineralogy index highlighting exposed fired-brick signatures
- Anomaly detection — comparing site candidates against a regional baseline to identify statistically significant spectral departures
- Ground-truth verification — cross-referencing candidates against published literature and declassified aerial survey records
Results
The survey identified 47 candidate sites across the study area, of which 31 showed correspondence with known records and 16 were previously undocumented at this level of detail. Confidence in the undocumented candidates varies; several will require field verification.
Limitations
Multi-spectral data at 10-metre resolution is insufficient for structural analysis of small features. The approach works best for identifying site extents and for regional-scale prospection rather than intra-site analysis.
All analysis code is available on GitHub.