The DrillerDB Research Program is open. Faculty, graduate students, agencies, and nonprofit research organizations can apply for free record-level access to 18 million+ water well records across the United States and Canada: well depth, static water level, yield, construction details, drill dates, and locations, compiled from state and provincial well-completion reports. Applications are a short form and a one-page data use agreement; exports arrive as plain CSV scoped to your study area.
The layer researchers cannot get anywhere else
Every time a water well is drilled, the driller records what the bit passed through - clay, sand, gravel, shale - and files that log with a state agency. Those logs are the only direct observational record of the shallow subsurface that exists at continental scale. But state portals publish them the way they were filed: scanned forms and free-text descriptions whose wording changes from county to county, and a log written for Wisconsin's system was never meant to be read next to one from Kansas.
DrillerDB processed more than 10 million of those logs - including hundreds of thousands that survive only as scanned handwriting - and parsed more than 53 million individual depth intervals, assigning 51.3 million of them a lithology from a single controlled vocabulary, with color and grain size captured as separate fields where the driller recorded them. The result, the Underground Lithology Grid, is the largest standardized, record-level compilation of driller-described subsurface lithology ever assembled for North America - roughly 33 times the size of the largest previously published compilation, the U.S. Geological Survey's 1.57-million-record glacial aquifer study (Bayless et al., 2017) - and the first to apply one lithologic vocabulary across state and provincial boundaries at continental scale.
We taught ten million drillers' logs to speak the same language - and kept every original word next to every coded layer so a researcher can check our work.
What the records show
An analysis of nearly 1.4 million domestic wells in the dataset shows how much well construction has changed in fifty years. In the 2010s, the average new domestic water well drilled in Colorado reached 360 feet - 152 feet deeper than the average well drilled there in the 1970s. In California, up 135 feet; in Maine, up 139. In Michigan and Wisconsin, meanwhile, well depths barely moved. The pattern holds inside individual counties: in California, every one of the 38 counties with enough records in both decades shows deeper new wells. (Well depth measures how wells are built, not the depth of the water table itself.)
How accurate is the coding
Standardization is overwhelmingly deterministic: 96.5 percent of intervals are coded by a curated, versioned dictionary rather than a language model, with under 1 percent AI-assigned and 1.7 percent left explicitly uncoded rather than guessed. In an independent audit of 1,000 randomly sampled coded intervals, the assigned lithology matched the driller's original description in 96 percent of cases, was judged partially correct in 3 percent, and outright wrong in fewer than 1 percent. The original driller's wording is preserved alongside more than 99 percent of intervals, so every coding decision can be checked against the source.
We are equally explicit about what the dataset is not. Driller descriptions are made from cuttings at the rig, so depths carry uncertainty; terminology is not standardized to a laboratory scale; description detail concentrates in water-bearing zones; and wells are drilled where water is wanted, not on a scientific sampling grid.
What it supports
- Hydrogeology and aquifer characterization at a record density observation networks do not reach
- Groundwater depletion studies pairing static water levels with drill dates across decades
- Water scarcity, drought, and climate work joining county-level water table behavior to precipitation and climate models
- Well depth and construction trends, like the fifty-year deepening analysis above
- Water quality and public health research on the more than 40 million Americans who rely on private wells
- Geothermal siting and shallow-subsurface characterization
The terms, plainly
Access is free for non-commercial academic and institutional research. We ask for three things: use it non-commercially, cite DrillerDB in anything you publish, and do not redistribute the raw data. Well locations are delivered exactly as the state source of record publishes them. These are public records, and we do not reduce their precision. The underlying logs were filed under state law by working drillers and preserved by the agencies that collect them; DrillerDB claims no ownership of any driller's log. The citation is the price of admission - it is how the program stays free.
The research program page has the field list, coverage notes, and the application template. We read every application ourselves and respond within five business days.
- Apply to the Research Program
- Browse the records on the public well map
- Read the press release
- Press kit: logos, maps, fact sheet
Coverage
The announcement went out on the wire August 4, 2026, was carried by AP News, and was syndicated to more than 45 local news stations nationwide, from Chicago to Denver to New York.