Automated Detection of Ancient Copper Slag, Arava Valley

Machine-learning detection of ancient copper-smelting slag from drone and national orthophoto imagery, with Prof. Adi Eliyahu Behar (Ariel University), Prof. Naama Yahalom-Mack (The Hebrew University of Jerusalem), Dr Uzi Avner (Arava Institute) and the Geological Survey of Israel.

Location: Arava Valley, Eilat Mountains and Timna
Status: Research / Demonstration
Goal: Test whether slag can be detected automatically, so archaeological survey can scale from single sites to the whole Arava Valley.
Data: Centimetre-level drone imagery to country-wide orthophoto
Environment: Remote desert wadis and slopes with scattered slag heaps
Duration: 2026 – ongoing

The challenge

Hundreds of copper-slag sites are scattered across hundreds of square kilometres of remote desert, and dark slag is easily confused with natural rock. Resurveying the valley on foot is impractical.

The DynamicSense intervention

We flew high-resolution drone surveys over selected smelting sites, built an expert-annotated training set and trained a machine-learning model that was tested on both drone imagery and the national orthophoto archive. A follow-up field experiment in Timna Park tested how well thermal imaging picks out slag of different sizes, including dusty and partly buried fragments.

Outcomes & decision support

A first automated workflow for slag detection in Israel, and a path to valley-wide archaeological screening at a fraction of the cost of field survey. Results of the thermal experiment are being analysed.

Structured outputs