Where Kansas oil data lives.

A LAS file (Log ASCII Standard) is what an analyst loads when they want to know what’s in a rock. It is the machine-readable version of the wireline log that a logging truck runs after a well is drilled. Older wells have paper logs, or no log at all. The paper is not the same as a LAS file: it cannot be regressed against, cannot be fed into a neural net, cannot be joined to a production database. For the purpose of any analysis beyond reading an individual log by eye, the paper era is functionally dark.

Kansas is the cleanest public subsurface-data pipeline in the United States. The Kansas Geological Survey publishes two things openly: a master well list (515,000 records deduping to 477,000 unique wells) and a bulk archive of LAS files. Most state regulators make you scrape per-county case-file PDFs. Kansas is a ten-minute download. That is why it is the proof-of-concept: cheap to execute, fast feedback, teaches the pipeline before we attack messier states.

Why the gap matters. Every dark offset well is pre-drill intelligence nobody extracted. A 1950 dry hole in the next section tells you what’s in your rock, if anyone digitized it. For investors and basin analysts, the first party to digitize a regional log archive has an information monopoly; in brownfield M&A, the dark-data gap is the fog of war. For policy and climate work (orphan-well remediation, groundwater contamination attribution, methane-leak source identification, carbon sequestration site selection) all of these require decades of old logs that don’t exist in machine-readable form. You cannot remediate what you cannot locate.

This is post 01 of a basin-by-basin series. Next: the Permian, Texas plus southeast New Mexico. Neither the Texas Railroad Commission nor the New Mexico OCD publishes a public LAS archive at the Kansas level of openness. The data picture there is almost certainly worse.

Dedupe. The KGS master list ships 515,000 records but the true drilled-well count is smaller. API numbers are deduplicated after normalising the format (strip dashes, strip trailing .0 introduced by Excel round-trips). Records whose status is Approved Intent, Expired Intent, or Cancelled API (i.e. “paper” wells that were permitted but never drilled) are excluded. Net: 419,777 drilled wells.

Dark/lit classification. Join the drilled-well table to the KGS LAS inventory on the normalised API key. A well is “lit” if at least one LAS file exists for it in the WebDocs archive; “dark” otherwise. This is a proxy, not a complete measure. A well can have a scanned paper log behind a paywall, or a proprietary LAS inside an operator archive, and still count as dark here. The proxy captures what a third-party analyst can see without writing a cheque, which is the quantity the series is about.

Vintage bucketing. Spud year parsed from the Spud Date field with pd.to_datetime(errors='coerce'). Real date coverage is 63.5% of drilled wells; the remaining wells go into an “unknown” bucket and are excluded from decade-level reads. The decade-level LAS coverage rate uses the decade’s denominator, not a state-wide average.

Orphan definition. A well is considered “orphan” for the Figure 03 callout if the Current Operator field is blank. This is the state’s own record: some of these are plugged-and-abandoned wells whose last operator no longer exists, some are wells where the transfer of ownership was never filed, and some are genuine unknowns. The “nobody” in the callout is whichever of those three a specific well happens to be.

Cherokee region definition. Eleven counties: Allen, Bourbon, Chautauqua, Cherokee, Crawford, Elk, Labette, Montgomery, Neosho, Wilson, Woodson. Bounding box for plotting: 36.93–38.15 N, -96.05 to -94.55 W.

Processing in Python: pandas, geopandas, shapely. Figures are matplotlib at 300 DPI. Coordinates are NAD27 as published by KGS.