A national inventory of well lithology and hydraulics for the conterminous United States (v1.1)

Version 1.1.0 (2026-08-16) — units correction. In v1.0.0 the columns transmiss_ft2_d and hyd_cond_ft_d were declared in ft²/day and ft/day but carried the agency value unconverted. Michigan's EGLE Wellogic publishes transmissivity in gal/day/ft and conductivity in gal/day/ft² and states this nowhere in its download, so all 775,182 Michigan rows — effectively the entire transmissivity column (775,367 rows total) — were overstated by a factor of 7.48052. Montana's values state their own units in the source text and were correct as published. v1.1.0 harmonizes these columns to the declared units and adds three columns so the conversion is auditable and lossless: transmiss_agency and hyd_cond_agency hold the agency value verbatim, and hyd_units_agency records which convention each row was published on (gpd_per_ft or ft2_per_d). Michigan's median transmissivity becomes 187 ft²/day (was 1,400); Montana's 70 ft²/day is unchanged. No rows were added, removed or re-flagged — row counts, state coverage, per-column non-null counts and all QA flags are identical to v1.0.0; only values changed. Users of v1.0.0 transmissivity or conductivity should re-download or divide the Michigan values by 7.48052. Harmonized, openly-harvested inventory of per-well lithology, hydraulics, and locations for the 48 conterminous United States, accompanying the manuscript “A national inventory of well lithology and hydraulics for the conterminous United States” (Rafiei, 2026). Compiled entirely from primary state-agency sources through public endpoints (no privileged access); every value is as published by the source agency, with no interpolation, gap-filling, or modeled substitution. Per-row provenance (source_id, and for lithology a source_type separating driller/geologist records, vision-language-recovered samples, and interpreted framework contacts) plus a documented QA layer (qa_flag; sentinel fillers nulled, implausible values flagged, nothing imputed). Contents (well tables join on (state, well_id)): gw_lithology.parquet — 28,518,082 depth-resolved lithology intervals across 5,165,761 wells in 34 states, with verbatim descriptions and a derived normalized material class; gw_hydraulics.parquet — per-well structured hydraulics for 6,895,791 wells (static water level, depth, yield, drawdown, specific capacity, screened interval, surface elevation, transmissivity/conductivity with provenance, pump-test duration); gw_well_locations.parquet — coordinates (EPSG:4326) for 7,796,626 wells with an agency-derived precision qualifier; gw_framework_lake_michigan.parquet — the USGS Lake Michigan Basin under-lake hydrogeologic framework (30,290 grid cells), kept separate from the well tables. Depths in feet; records use each state's own primary key. Released under CC-BY-4.0; underlying records are public US state well data; no well-owner personal information is included.