A freight line does not fail all at once. It settles — a millimetre a year on soft soil, quietly, under axle loads that never let up. For a century the only way to feel that was to send a survey train down my rails and measure. A team at the University of Twente has now read the same subsidence from roughly 700 km up.

In Sensors (2020), Ling Chang, Nikhil Sakpal and Sander Oude Elberink of Twente, with Haoyu Wang of Fugro, stacked 170 Sentinel-1a/b radar scenes — C-band, VV, six-day revisit, January 2017 to December 2019 — over ~13 km of the Betuwe corridor, the dual-track freight spine running from the Port of Rotterdam toward central Europe. That is my kind of asset: heavy, relentless, and never idle long enough to close for a proper look.

The trouble with Persistent Scatterer Interferometry is an old one. It gives you deformation to the millimetre in line-of-sight, but it only knows where each scatterer sits to the nearest few metres. Millimetre motion of an object you cannot name is a rumour, not a diagnosis. Which rail? Which pole? Which embankment? The Twente method answers by fusing the radar with AHN3, the national Dutch LiDAR (18 points/m², 5 cm height accuracy), as a real-object datum. Each scatterer is given a 3-D error ellipsoid — elongated badly in the cross-range and up directions — and a nearest-neighbour link in whitened space, so “nearest” means statistically nearest, not merely closest. That sorted 3259 of 3328 scatterers (98%) onto actual objects: 1552 on the rail track, 1030 on embankment, 677 on the surroundings.

Then the part that matters on an asset manager’s desk. Decompose the line-of-sight into the track’s own normal direction — transversal and longitudinal motion assumed near zero, confirmed by the rail owner — and localized differential settlement between adjacent scatterers becomes readable. On the validated A–B segment, rail carried on ~435 m of fixed structures (bridges) held stable, while the ~975 m riding on soft soil subsided by an average of ~30 mm — right up against the EN 13848-5 stability threshold of 27 mm for 70–100 km/h track. The satellite pattern (2017–2019) broadly matched Fugro’s train-borne RILA survey (2013–2015). The soft ground compacts under heavy axle loads; the bridges do not. Radar saw the difference without touching my ballast.

←TODAY: 170 free Sentinel-1 scenes plus one LiDAR pass flag the subsiding 975 m before a survey train is dispatched. →3012: every century-scale corridor carries a standing orbital baseline — settlement is a monitored quantity, not a surprise dig. Fulcrum: the millimetre only becomes a maintenance decision once you know exactly which metre of track it belongs to.

Atelier: For an office designing stations, platforms or anything founded along a live corridor, this reframes the geotechnical brief — differential settlement between a bridge abutment and its soft-soil approach is now a measurable, historic time series, not a lump-sum assumption. Your Monday move: pull the free Sentinel-1 archive for your site’s footprint (Copernicus Open Access Hub or the ASF vertex) and ask your geotech consultant for the InSAR velocity field over the last three years before you fix the foundation transition detail. The data already exists; someone just has to open it.

Hack: Size the noise floor of a differential-settlement flag before you trust it. Adjacent scatterers each carry a normal-direction uncertainty, and the paper’s rule is that those standard deviations add along the pair — so the difference you flag is noisier than either point alone. Run the arithmetic on your own numbers:

sigma_a, sigma_b = 1.22, 1.25      # per-scatterer normal-direction sigma, mm
sigma_pair = sigma_a + sigma_b     # differential settlement noise, paper's rule
print(round(sigma_pair, 2))        # 2.47 mm  -> ~1.5x a 1 mm LOS error

Anything below ~2.5 mm of apparent differential movement is inside the noise; the 30 mm on the soft-soil segment is not. Know which side of that line your alarm sits on.

The lesson for anyone building a line to last: the failure my generation regrets is never the inspected bridge — it is the soft-soil segment nobody was watching because watching it meant a closure. Free radar removes that excuse. Open the archive for your corridor, and put a settlement baseline on paper before the first heavy axle rolls.