A national inventory of well lithology and hydraulics for the conterminous United States

Abstract. Continental and coupled groundwater models require a three-dimensional picture of the subsurface – the vertical sequence of geologic materials and their hydraulic properties with depth – yet no harmonized national compilation of this information has existed for the United States, where it is generated as a regulatory byproduct in millions of water-well driller’s logs held by individual state agencies in heterogeneous, partly-digitized forms. We present an open, harmonized inventory of well lithology and hydraulics for the 48 conterminous states, compiled entirely from primary state-agency sources through public endpoints with no privileged access. The inventory brings together ∼28.5 million depth-resolved lithology intervals for ∼5.2 million wells across 34 states, harmonized to a single canonical schema (well identifier, interval top and bottom depth, verbatim material description, a derived normalized material class, and per-row source provenance); structured hydraulics – static water level, total depth, yield, drawdown, specific capacity, screened interval, and, where reported, transmissivity – for ∼6.9 million wells in 41 states; and coordinates for ∼7.8 million wells spanning all 48 conterminous states, with the agency’s own precision qualifier attached. Every value is reported as published by the source agency, with no interpolation, gap-filling, or modeled substitution: release quality assurance nulls only documented agency filler codes, and flags – never alters – implausible or internally inconsistent values. Every record is traceable to its agency source. We document per-state coverage and the exact access pattern for every source, so the compilation is fully reproducible and extensible; a companion availability census classifies each of the ∼13.3 million registered wells by how – or whether – its lithology is published. Where a state exposes lithology only as scanned driller’s-log images, we recover it with a reproducible, cost-bounded vision-language procedure validated against a state agency’s hand-transcriptions (91.5 % material-class accuracy, Cohen’s κ = 0.90; 96.0 % coarse/fine texture, κ = 0.92). The inventory is released openly on Zenodo and is browsable through a live web platform. It supplies the depth-resolved subsurface-structure and hydraulics layer that national well-location databases omit, ready for use in groundwater modeling, aquifer characterization, and data-driven hydrogeology.