Mapping the Jagersfontein tailings flow five days after the dam broke

A tailings dam at the Jagersfontein diamond workings, in the Free State, gave way early on 11 September 2022. Roughly five million cubic metres of saturated waste ran through the edge of the town and out across the grazing land beyond. Homes were destroyed and people died.

The flow ran about seven kilometres over dry ground before it reached the stream network. Tailings were traced a further fifty odd kilometres down the Prosesspruit and the Riet to the Kalkfontein reservoir inside a day. Farms were buried, and the people who worked them needed to know where and how much.

Three years later there is still no downloadable polygon of that footprint.

The analysis covers a box of about 24 by 19 kilometres holding the impoundment, the town and the full overland runout, on five archive passes: 27 August and 6 September before the failure, then 16 September, 6 October and 5 November after it.

Ground counts as buried where the vegetation index falls by 0.10 or more between a pre-event scene and a post-event scene. That threshold was fixed in advance, applied to every pair of dates in the study, and never adjusted to bring the answer closer to anyone's published figure. Areas are the sum of connected patches of half a hectare and larger.

The same test also runs across the two scenes from before the failure. Those bracket ten days in which the dam was still standing, so whatever they return is what the method produces from sun angle, atmospheric correction, grazing and passing cloud shadow. Every result below is quoted against it.

Scoring is restricted to ground that carried some vegetation beforehand, a restriction set before any post-event scene was read. The Karoo in September is dry and sparse, so on this terrain the restriction moves the five day figure by about one per cent. On wetter ground it does considerably more.

This is a footprint of living ground that was buried, about 12 square kilometres of it. The published estimates of 16 and more than 26 square kilometres describe damaged land in a broader sense, and a good deal of the runout across the Karoo was bare before the tailings ever arrived. The three numbers measure three different things. Ours is the one that comes with a threshold, a control and a boundary you can load into a GIS.

The evidence was already sitting in the archive on 16 September 2022, five days after the wall came down and a day before the first Charter product was delivered. Nothing had to be tasked and nobody had to go to site.

Timing changes the answer. Between the 16 September pass and the 5 November pass the measured footprint falls by 37 per cent, so a single map made two months late understates what was buried. Liability on an event like this should be assessed against the earliest clean scene rather than the most convenient one, and for a site with a live impoundment the same test run on a schedule is a monitoring programme rather than a forensic one.

The control is what makes any of it arguable in front of a regulator. Nineteen hectares from ten quiet days is the floor this method produces on this terrain, and it is the fact that turns 910 hectares from an assertion into a measurement.