InSAR for Mining: Watching the Ground Breathe From 700km Up
A mine wall in Thar shifted 14mm last March. Nobody on the ground noticed. The satellite did.
That's the part most people don't get about InSAR. It's not magic, it's not new (the technique goes back to the 1990s), but the way we apply it to Pakistani mining sites today — that's what's changed everything for me.
Let me back up.
What InSAR Actually Does (Without the Textbook Stuff)
InSAR stands for Interferometric Synthetic Aperture Radar. Fancy name. Simple idea. A radar satellite passes over the same patch of ground twice — say, 12 days apart — and measures the difference in the radar signal's phase between those two passes. If the ground moved even a few millimeters, that movement shows up as a fringe pattern in the data.
We're talking 2-3mm precision. From 700km in space. Through clouds. At night.
For mining, this changes the math on a lot of things. Pit wall stability. Tailings dam integrity. Subsidence over underground workings. Even regional tectonic creep that affects long-term mine planning. All of it becomes measurable without sending a single person up a slope with a total station.
And honestly? I used to think InSAR was overkill for most of our clients. Small operators in Khyber Pakhtunkhwa or Balochistan don't need millimeter precision, I figured — they need to know where the gold is. But then a chromite operator near Muslim Bagh lost a wall section in 2022. Three workers injured. The deformation signal was sitting in Sentinel-1 data for weeks before it happened. We just weren't looking.
Now we look. Every site. Every pass.
The Scale Problem Nobody Talks About
Here's the thing about radar interferometry mining workflows — they're computationally brutal. A single Sentinel-1 SLC scene is around 4GB. Processing an interferogram takes coregistration, phase unwrapping, atmospheric correction, geocoding. Do that for one mine, fine. Do it for 15 mines (which is what I personally manage in Gilgit Baltistan), and you're already drowning. Do it for the 400+ sites we monitor across Pakistan and you need actual infrastructure.
This is where geomines spent most of 2023. Not building shiny dashboards. Building pipelines.
We pull Sentinel-1 data on a rolling 6-day cadence (Sentinel-1A only, since 1B died — RIP). The stack gets processed automatically. Persistent Scatterer InSAR for built-up or rocky terrain. SBAS for vegetated or low-coherence areas. The output isn't a pretty picture — it's a time series of vertical and horizontal displacement for every coherent pixel on a mining lease.
For a 50-hectare site, that's roughly 80,000 measurement points. Updated every 6-12 days. Forever.
A traditional ground survey gets you maybe 200 points. Once a quarter. If the engineer shows up.
Where InSAR Mining Monitoring Actually Pays Off
Three use cases I've seen pay for themselves within months:
1. Tailings dam monitoring. After Brumadinho in Brazil (2019, 270 dead), regulators worldwide started demanding continuous deformation monitoring. Pakistan hasn't caught up regulatorily, but smart operators aren't waiting. We tracked one tailings facility in Sindh that was creeping at 8mm/month — slow enough to miss visually, fast enough to fail eventually.
2. Open-pit wall stability. Pre-failure deformation almost always accelerates before a slope collapse. The Hoek-Brown criterion is great, but it's a model. InSAR is observation. Combining both is what actually works.
3. Subsidence over abandoned or active underground workings. Coal in Chakwal. Salt in Khewra. We've found InSAR deformation signatures over old galleries that nobody had documented properly in 40 years.
But look, it's not all clean. Decorrelation in dense vegetation, snow cover in the northern areas (try getting good coherence over a glacier-adjacent mine in Hunza in February), atmospheric phase delay in monsoon — these are real problems. We solve some of them by stacking long time series. Others we just flag and move on.
What I Got Wrong
When we started, I thought InSAR would be a premium add-on. A separate product line for big clients with deep pockets.
Wrong call.
It should've been baseline from day one. Every report we generate now includes a deformation layer by default, even if the client only asked for mineral targeting. Because finding a copper anomaly is one thing. Telling the client that the slope above their proposed pit has been creeping 5mm/year for the last decade — that's the kind of insight that changes whether they invest $40 million or walk away.
The combination matters more than any single dataset. Sentinel-2 tells us what minerals are there. ASTER refines the mineralogy. SRTM gives us structure. SAR backscatter shows surface roughness and moisture. And InSAR — InSAR tells us if the ground is staying still or quietly moving toward something bad.
None of these alone is enough. All of them together is what we mean when we say satellite intelligence for mining.
So the question I'd ask any operator reading this — when's the last time you actually measured your pit walls? Not modeled them. Measured them.
If the answer is "last quarter" or "we don't," we should probably talk.