NASA Space Apps 2026 · Cambridge, MA
Where is the ground sinking? We measured 29 places from space.
Using the first data from NASA and ISRO's new NISAR radar satellite, we built one open method that turns radar echoes into centimetres of ground sinking, ran it on 29 basins over the same 93 days, and checked every number.
Rates in cm/100d (centimetres per 100 days; 1 in = 2.54 cm) along satellite LOS (line of sight: toward or away from the radar), from NASA-JPL GUNW (Geocoded Unwrapped Interferogram: paired radar images processed to show ground movement). NISAR data are provisional. Calibration and validation are carried out by the NISAR project (NASA-JPL and ISRO), not by our team.
How we know it's right
A satellite number is only useful if you can trust it. We tested ours three ways.
Ten permanent GPS stations in California measured the same sinking as the satellite. Typical difference: about 1 cm (0.4 in).
5 caught We hunt for our own mistakesRadar can jump by exactly one wave cycle (12.1 cm). An automatic test checks every basin for these jumps and bad dates; flagged basins are not ranked.
15 of 29 We only rank what passesBasins with thin coverage or failed checks stay "indicative". Even Konya, our strongest signal, stays indicative after our own review.
NISAR data are provisional. Calibration and validation are carried out by the NISAR project (NASA-JPL and ISRO), not by our team. How it works, step by step →
What we built · in one look
Nine things we made. Each one is a section below.
- 01Global atlas29 basins, same 93 days
- 02Self-checkcaught our own 12.1 cm error
- 03GPS proofr = 0.92 at 10 stations
- 04People exposed1.17 M where ground sank >5 cm
- 05Water lost0.32 km³/yr gone for good
- 06Infrastructure81 of 645 sections flagged
- 07Bulletins29 one-page reports
- 08Simulatorlearn the physics in 30 s
- 09Open engineone command, any region
A global subsidence atlas
What it is: ground motion maps for 29 basins on three continents, all made the same way from NISAR's first radar data and ranked on the same 93 days.
Why it matters: for the first time you can compare sinking in Türkiye, Iran, Mexico, California and Asia side by side, with the same ruler.
A check that catches our own errors
What it is: an automatic test on every basin's time series. It looks for jumps of exactly one radar phase cycle (12.1 cm / 4.8 in) and for single bad dates.
Why it matters: it caught our first number one. Semarang's −12.1 cm (−4.8 in) was one processing jump; re-processed, the real peak on the coast is −6.2 cm (−2.4 in).
Proof on the ground
What it is: a comparison with 10 permanent GPS stations in California's Central Valley, where the ground moved up to 7.8 cm (3.1 in) away from the satellite in 108 days.
Why it matters: GPS stations are the reference that engineers trust. NISAR and GPS tell the same story, station by station.
We also compared with NASA JPL's OPERA product: over Mexico City the pattern matches at 10 points (r = 0.91), although OPERA covers winter and NISAR summer.
Who lives on sinking ground
What it is: each atlas map overlaid with a population map (GHSL 2025), counting only sinking larger than twice the measurement noise.
Why it matters: it turns centimetres into people, the number a city or ministry needs.
it sank more than 5 cm (2 inches) in one summer
more than 5 cm (2 inches) in one summer
Water storage lost for good
What it is: eleven months of NISAR data split into ground that bounces back each winter and ground that does not. The part that does not come back is squeezed clay: aquifer space lost forever.
Why it matters: water managers count it in cubic kilometres, so we did too.
Canals and railways at risk
What it is: every canal, road and railway line in the Central Valley checked for uneven sinking, with a probability that includes our measurement error.
Why it matters: uneven sinking cracks canals and tilts tracks; engineers can see which section to inspect first.
A one-page bulletin for every basin
What it is: 29 printable bulletins in the style of NASA disaster products: rate, worst areas, people affected, confidence, revision history.
Why it matters: a water manager does not read papers; one page with the number and its reliability is what they use.
Open the Konya Plain bulletin →A simulator anyone can play
What it is: run a farming valley for 24 months. Pump more and earn more, but watch the ground sink and part of it never come back.
Why it matters: the model reproduces what NISAR measured, so students and farmers learn from real data, in 30 seconds.
Play the simulator → Play Aquifer Run → For kids: Water Hero →
An open engine anyone can re-run
What it is: NISAR Vision, the pipeline behind every number here. One command searches NASA's archive, downloads, measures and checks a new region.
Why it matters: any agency or student can repeat our atlas for their own city, and check our work.
Get the code and data →# measure a new region from public NASA data python -m pipeline.nisarvision run \ --name konya --bbox 32.4,37.4,33.9,38.1 \ --start 2026-06-17 --end 2026-10-31 # rebuild and self-check the atlas python pipeline/pc10/build_atlas.py