13 places are sinking fast, and no tested explanation fits.

The unexplained thirteen

Every patch below was tested against all seven layers and none applied. Coordinates included so anyone can check.

A candidate present, case not closed

These have only weak evidence nearby — topo-map pit symbols (on Delmarva, mostly sand and gravel pits) or carbonate bedrock with no mapped sinkhole hotspot. A suspect named is not a case closed, so they are neither explained nor counted in the headline.

The explained thirty — including all the famous ones

Houston's groundwater bowl, the Beverly Hills and Wilmington oil fields, Phoenix, Las Vegas, Fresh Kills landfill, the Georgia tidal marshes, Fort Worth over the Barnett Shale wells. The method proving it can explain the known cases is what makes its “unexplained” worth taking seriously. Wells are an association, not a proven mechanism — the table says wells are present, not that extraction did it.

Method — so the result can be checked

1. Cluster: every ~1 km cell holding 30+ measurements at -5 mm/yr or below, merged with its neighbours; clusters under 3 cells dropped as speckle.
2. Name: each cluster reverse-geocoded to county and place via the US Census geocoder.
3. Subtract, layer by layer, each with three outcomes — hit, pass, or not-tested, never two: USGS-documented groundwater subsidence (cited per county), state oil & gas well databases (California and Texas), the USGS karst map (bedrock and sinkhole hotspots separately), USGS topo mine symbols, Superfund proximity, SSURGO soils sampled at the sinking points, and attribution in the source papers themselves.
4. Controls: six known cases must classify correctly before any verdict counts — the wells layer must find both LA oil fields, the groundwater layer must find Houston, Phoenix and Las Vegas, and the literature layer must find Fresh Kills. A layer that returns zero everywhere is tested against a place where the answer must be nonzero before its zeros are believed.

Ground motion: Ohenhen et al. (2025) Nature Cities 28-metro VLM; Ohenhen et al. Atlantic-coast VLM; Ohenhen (2025) NYC–Long Island VLM — peer-reviewed, open-licensed. We did not produce the measurements; we produced the join. Analysis dated 2026-08-05. Download the full table: every cluster with its verdict and evidence (CSV).

Is your ground moving?

The same archive answers for a single US address, free.

Does "unexplained" mean nobody can explain it?

No — it means none of the seven causes we could test applies: documented groundwater subsidence, oil and gas wells, karst bedrock and mapped sinkhole hotspots, recorded mines and pits, Superfund sites, and compressible wetland soils checked at the sinking points themselves. Two candidates our tests could not check remain open: glacial isostatic adjustment — the slow collapse of land that bulged up around the ice-age ice sheet — and agricultural groundwater withdrawal, which has no public site-level dataset. Regional isostatic adjustment runs about 1 to 2 mm per year; these patches sink at 5 to 8, so if that is the answer it is not the whole answer.

Why is most of the unexplained list on the Delmarva Peninsula?

That is the finding. Eleven of the thirteen unexplained patches sit on the peninsula shared by Virginia, Maryland and Delaware — on upland farmland, not marsh. The marshes were tested and cleared separately: our soils check samples the actual sinking points, and where they sit on tidal marsh or muck the cluster is classified as explained by soil compaction, as happened on the Georgia coast. The Delmarva patches passed that test on dry ground.

Is this everywhere in America?

No. The source datasets cover 28 major metros, the Atlantic coast, and the New York area — peer-reviewed vertical land motion measurements published by Ohenhen and colleagues. A fast-sinking place outside that coverage would not appear here, so the honest claim is "in the areas these datasets cover", never "in America".

What about Houston, Phoenix, and the famous sinking places?

They are in the explained column, which is how it should be. The fastest sinking in the covered US — the Katy area of Houston at around 19 mm per year — is also the best documented: groundwater withdrawal, managed by a subsidence district created in 1975. A method that flagged Houston as a mystery would be broken, and that check is built in: the analysis carries six positive controls that must pass before any verdict counts.