AquaSolvency v2.6: basin-scale water-insolvency screening of 202 river basins from GRACE/GRACE-FO, 2002–2026

AquaSolvency screens 202 river basins for water insolvency — storage falling faster than it is replaced — from 258 months of GRACE and GRACE-FO satellite gravimetry, April 2002 to June 2026. Total water storage is read from three mascon solutions produced independently at CSR, NASA GSFC and NASA JPL, averaged over HydroBASINS level-3 outlines in an equal-area projection. The seasonal cycle is removed with harmonics; trends are estimated with Theil–Sen on annual means and tested with the Hamed–Rao modified Mann–Kendall test. An antecedent-climate model fitted to the basin's own rainfall and temperature, and scored out of sample, separates what observed weather accounts for from what it does not. Groundwater is estimated by removing modelled soil moisture, snow water equivalent and, where a record exists, lake and reservoir storage. Each basin receives one of ten status classes and a confidence grade that a failed check can lower but never raise. WHAT THIS RECORD CONTAINSThe basin-level results for all 202 basins with 86 fields, the monthly basin series, the robustness and validation tables behind them, the full reference list, a data dictionary, and the dashboard and written method as published. VALIDATIONResults are set against 41,866 in-situ wells worldwide (Spearman rho = 0.438 across 22 scored basins), 6,843 pre-monsoon observation wells in India (rho = 0.757), 32,827 km2 of land subsidence measured by radar interferometry, published studies recomputed over their own periods, and a separately produced satellite trend field estimated directly from Level-1B ranging data (correlation 0.96 across 153 assessed basins). Every basin is rerun under 288 alternative analysis settings and 48 ways of preparing the satellite record, and a perfect-model test screens 1,530 synthetic basins containing no water use at all. WHAT IT DOES NOT CLAIMThe measurement resolves about 300 km, so every result describes a whole river basin and never an aquifer or a district. It measures insolvency, not water bankruptcy: bankruptcy in the formal sense also requires irreversibility, which a storage record alone cannot establish. It is not a climate-change attribution — the model is fitted to observed weather, which carries natural and forced variability together. The three mascon solutions are not three independent measurements: they rest on the same observations and apply the same glacial isostatic adjustment model, so their spread understates the true uncertainty. Eleven months (July 2017 to May 2018) have no satellite; trends are fitted across that gap and nothing is modelled into it. Thresholds are screening thresholds, to be set locally with the people who manage a basin. The method has not been through journal peer review. ACCESSFiles are restricted while the accompanying manuscript is in preparation. The interactive dashboard is open to everyone at https://praddyk-aquasolvency.static.hf.space/ — the results can be read, filtered, charted and downloaded there without this record. Requests for the files from researchers and basin authorities who need them to check a specific result are welcome and will be answered.