Radar Is Free. Judgment Is Not.

Anyone with a laptop can now pull L-band radar imagery of a tailings dam. NASA says NISAR's provisional data were released on July 20, 2026, and the Alaska Satellite Facility says the public release began on July 20, 2026, covering observations acquired on or after June 17, 2026. Products are searchable through NASA's Earthdata Search or the ASF Vertex tool, and by script through the asf_search Python package. Sentinel-1, meanwhile, has long offered revisit times of 6 or 12 days over most of the world, according to a peer-reviewed study led by N. Rana.

The satellite pass has become a commodity. The scarce asset is the judgment to read it, and mining is about to find out how expensive that judgment is.

The pass is cheap

NASA calls NISAR a systematic mapper of changes in the planet's surface as small as a centimeter. Its pitch to miners is vegetation. JPL says the mission tracks surface change unhindered by the clouds or foliage that blind optical sensors and higher-frequency radars. NASA's Craig Ferguson said NISAR's "long wavelength L-band radar will make it possible to detect and track land subsidence in more challenging and densely vegetated regions."

Europe's free data has already shown what that looks like in practice. EUSPA, the EU space agency, says free Copernicus Sentinel data is a corner stone of making such solutions cost-efficient, transparent and scalable. In a pilot for Eramet, which EUSPA describes as the largest producer of high-grade manganese ore worldwide, Tre Altamira used Sentinel-1 and InSAR techniques to detect subtle ground shifts with millimetre precision. The movements traced to erosion, settling under stockpiles and consolidating decantation ponds. Monitoring around key tailings dams found no significant or traceable movement. Eramet data scientist Christophe Bessin said the pilot "gave us confidence to broaden its application at Eramet."

The same logic works before a mine exists. Norwegian explorer Kuniko, working with Poland's TerraEye, combined Sentinel-2 imagery with other satellite sources to find spectral signatures of mineral deposits near legacy mining zones. EUSPA says field validation later confirmed several targets. The motive is financial: Norway's winters leave a short window for on-site work, and each campaign carries high costs and risks. In EUSPA's feature, Kuniko COO Mona Schanche said: "Using space data opens new doors in terms of how we can explore our mineral deposits on a large scale."

Budgets explain why that pitch lands. The IEA's Global Critical Minerals Outlook 2026 says exploration spending declined by more than 10%, with lithium and nickel falling around 45%. Base metal prices rose by one-third between January 2025 and April 2026, with copper reaching record highs. Companies must find metal on tight budgets, and ground you need not walk is money kept. TerraEye's own claims run in that direction: it says it cut a 3,208 km² area at El Perdido in the Andes down to 47, a 98.5% reduction in ground to walk. Those are company claims, not audited results. ClaimWatch data show the same ground-hunting pressure in the US, with 6,758 new federal claims in May 2026 against 3,809 in February.

Where the free data stops

Screening tolerates false positives, because a bad target costs a drill program. Safety monitoring does not, and the radar's limits show up there.

Eos's Landslide Blog, written by Dave Petley, describes a new study of ten mining-complex failures analysed with InSAR. In every case the radar detected movement ahead of the landslide, on slopes that did not fail in the key event, and the deformation continued afterwards. Petley's reading is cautionary: only a small proportion of mining-related deformation leads to damaging failures, and reliably detecting the precursors of those that do remains a major challenge. For InSAR to be useful, he writes, analysts must separate ongoing creep from incipient failure.

The dam record shows how unsettled that separation is. The Rana study notes that at Cadia in Australia, earlier work found accelerating deformation in the two to three months before collapse. At Tonglvshan in China, the March 2017 breach showed no discernible precursor acceleration at any section. For Brazil's 2019 Feijao collapse, which killed 272 people, the InSAR studies disagree. One team found the precursor acceleration not statistically significant. Another concluded the collapse was foreseeable. Same mountain of data, opposite verdicts.

Vendors concede the point. INSAR Corporation, an Ottawa InSAR vendor, says "a stable InSAR result alone does not establish that a dam is safe." It answers "Not reliably" on whether radar can monitor a wet, active tailings beach, because fresh slurry changes between passes and the signal loses coherence. It also says satellite monitoring cannot satisfy the Global Industry Standard on Tailings Management alone and can contribute only an independent, repeated deformation record alongside on-site instruments. Its own company-claimed analysis shows two copper tailings impoundments, at Butte, Montana and near the Great Salt Lake, subsiding by up to 2,500 mm between 2016 and 2026. That is a measurement, not a verdict. Whether it signals danger depends on design, instrumentation and history, which the archive can partly reconstruct "where suitable data exist."

The main objection

The obvious rebuttal is that judgment is just another thing software eats. The IEA notes venture capital has recovered, increasingly targeting AI that could improve exploration and extraction efficiency. The Rana study itself blames a "complicated and lengthy workflow" requiring high-capacity computing for the thin archive of tailings case studies. Automate the workflow and the archive grows.

That is plausible for screening, where a model can be scored against drill assays. TerraEye says an independent assessment tested every available dataset against roughly 150,000 m of drill-hole assays and picked its Metal-Stressed Vegetation layer as the best predictor of high-grade nickel. That claim is hosted by the vendor, but at least the test is scoreable. Failure prediction has too few events to train on, and the events it has are contested. A larger archive of creeping slopes will not settle whether Feijao was foreseeable. Automation will produce more candidate alarms. Someone still has to decide which ones to act on.

What follows

Copernicus and NISAR will make baseline monitoring cheap for every miner, and that will leave a premium on the people who interpret it. NASA's David Bekaert expects that "We're going to see an influx of new discoveries from all over the world." Some of those discoveries will be new deposits. Many more will be new patches of colour on dams that are fine.

JPL's own Mexico City map carries the warning in miniature: its yellow and red areas, built on preliminary measurements, are likely residual noise expected to shrink as data accumulate. Free radar will keep passing overhead. What a red patch means is still a human call, and it is the one that costs money.