Step 00 / 16
North Dakota Montana
Dark · no public log
Lit · North Dakota
Lit · Montana
Dark: spud/completed before 2000 AND not a known horizontal. See step 05.
Loading 45,921 wells
Project 03 · Subsurface · Walk 03

Walking the Williston.

One structural basin, two states, one clipping polygon drawn by the USGS, not by a legislature. 45,921 drilled wells inside the Bakken and Three Forks Total Petroleum System. We went looking for what the Bakken horizontal era lit, and what it left dark.

01 · The question

Does the Bakken era buy a lit basin?

Kansas was 94% dark: old plays, shallow oil, nobody ever had the incentive to publish. Permian was 78% dark: wealth, horizontal drilling, still mostly hidden. The Williston is the cleanest natural experiment in between. A fifteen-year horizontal campaign sitting directly on top of seventy years of conventional vertical drilling.

The hypothesis we went in with: the post-2007 Bakken is lit. The pre-2000 legacy is not. The basin-wide average lands somewhere in between.

02 · The basin

A boundary geology drew.

We clip to the USGS Bakken and Three Forks Total Petroleum System polygon. Sixteen Assessment Units unioned into one shape, published in the 2021 National Assessment. Operators use this boundary to book Bakken reserves. So do we.

The polygon crosses North Dakota, Montana, and a sliver each of South Dakota and Saskatchewan. We publish North Dakota and Montana. South Dakota has under five hundred wells inside the polygon. Saskatchewan is a different regulator, a different currency, and a different unit system. Both are flagged on the methodology page. Neither is in this cut.

The Bakken does not care where North Dakota ends and Montana begins. The data does.

03 · North Dakota arrives

43,445 wells. One shapefile. Spud dates attached.

NDIC's Oil and Gas Division publishes the state's entire well inventory as a single ArcGIS shapefile, available as a direct HTTPS download from the DMR GIS portal. Each surface hole ships with API, operator, field, spud date, well status, and coordinates in NAD83.

A second shapefile, OGD_Horizontals, ships the directional survey points. 5.4 million rows across 22,258 unique APIs: the full horizontal population of the state's subsurface. We use that to flag every ND well that has a horizontal leg on record.

The ND data-delivery system is the cleanest per-state well inventory in the United States. It is also the reason the ND side of this story is the lit side.

04 · Montana arrives

41,943 wells. No spud date. No horizontal flag.

MT BOGC publishes the state's well inventory as WellSurface.zip on a directory-listed file share. The shapefile ships API, operator, field, current status, completion date, and coordinates. It does not ship spud date. It does not ship a horizontal indicator.

We use the completion-date year as a vintage proxy. Completion typically trails spud by thirty to ninety days in conventional wells, so the vintage shift is small. We cannot distinguish vertical from horizontal in Montana. Both the 2000s Elm Coulee horizontal pulse and the 1970s Madison vertical field count as "not horizontal" in our schema.

This biases the Montana dark share upward by a few points relative to North Dakota. The bias is documented, not flattened.

05 · What "dark" means

Two fields, same floor.

ND dark = spud year is before 2000 AND the well is not flagged as horizontal in OGD_Horizontals. NDIC's digital filing era begins around 2000; anything older relies on paper scans that are not systematically public.

MT dark = completion year is before 2000. Montana has no horizontal flag we can credibly apply, so the horizontal exclusion drops.

The asymmetry is the finding, not the problem. Every basin in this series has its own rule for what counts as lit. We publish the rule and publish the numbers it produces.

06 · The basin clip
45,921

Wells inside the TPS polygon.

85,388 wells across the two states fall inside their own borders. Only 45,921 fall inside the Bakken and Three Forks polygon: 40,656 in North Dakota and 5,265 in Montana. Everything outside the polygon is a different petroleum system, drilled for different rock, and does not belong in this cut.

North Dakota's share of the basin is 89%, which matches every Bakken production accounting we cross-checked against. Montana's share is 11%, almost all of it in Richland, Roosevelt, and Sheridan counties.

07 · The reveal
45.7%

Williston, dark.

21,003 of 45,921 wells. Kansas was 94% dark. Permian was 78% dark. Williston lands at 46%: the lowest dark share in the series so far. The Bakken horizontal campaign carried the basin to the majority-lit side. Barely.

The headline number hides the structural split that comes next. Nothing about a basin-wide average holds once you cross the state line.

08 · The state gradient

The basin split, at 104°W.

North DakotaND
43.3% MontanaMT
64.9%

The same Bakken rock runs under both states. The ND side gets a 2008+ horizontal campaign that wrote its logs digitally from day one. The MT side mostly pre-dates that era. The dark share inverts at the state line.

"The Bakken does not care about borders" is the geological claim. "The Bakken's data record does" is the claim this chart makes.

09 · Vintage, basin-wide

The basin is a 2010s well.

pre-1980
7,909 1980s
5,699 1990s
1,608 2000s
3,101 2010+
21,225 unknown
6,379

46% of the wells inside the Bakken and Three Forks polygon were spudded in 2010 or later. 21,225 of them, almost all North Dakota, almost all horizontal. Under 4% of the basin was drilled in the 1990s: the decade the vertical Williston ran out of rate.

The 2010+ bar is the entire lit signal. Remove it and the basin looks like Kansas.

10 · The horizontal wedge
22,258

North Dakota horizontals, with survey tables.

Every horizontal well that has ever been drilled in North Dakota ships its directional survey through NDIC's OGD_Horizontals shapefile. 22,258 unique APIs. 21,484 of them inside the TPS polygon.

The horizontal population is where the Bakken went from an oil-rich mudstone to a commercial play. It is also where the data went from paper to digital. These are not two separate stories.

If we ignored the horizontal flag and classified pre-2000 as dark on spud year alone, the ND dark share rises by seven points. Choosing to count horizontals as lit is a defensible call, not a free one.

11 · Elm Coulee

Montana's 2000s pulse.

Elm Coulee field, Richland County, Montana, is the reason the industry believed the middle Bakken could be drilled horizontally at scale. Burlington Resources and Halliburton proved the completion design here between 2000 and 2005, three years before the ND horizontal boom.

You see the pulse in the Montana vintage curve: a 2000s bar that stands alone, with most of the rest either pre-1980 verticals or a thin post-2010 residual. What you cannot see, and we cannot show, is which of those Elm Coulee wells are horizontal. The BOGC shapefile does not ship the flag.

If a future MT BOGC data update publishes a horizontal indicator, the MT dark share in this basin will fall. Until then, Elm Coulee is undercounted here.

12 · The unknown bucket
6,379

Wells with no parseable vintage.

14% of the basin has no usable spud or completion year. The ND unknown wells are mostly legacy pre-NDIC-digital-filing inventory. The MT unknown wells are mostly string values the parser cannot handle. Both are classified dark by default.

Kansas had a similar bucket. So did New Mexico. Every basin in this series carries a dark floor made of wells that were drilled before anyone standardized how to write the date down.

13 · The operators

Ten names drilled the lit basin.

Continental
3,069 Hess Bakken
2,874 Whiting
2,511 Devon Williston
2,319 Oasis Petroleum
2,192

The top five operators are the named Bakken majors, each with a dark share between 5% and 22%. The horizontal-era operators run a separate data regime: they log their wells digitally because the next pad's completion design depends on it.

Below the top ten, dark shares rise. Below the top fifty, the median operator looks more like Kansas than like EOG.

14 · Where the lit wells live

Parshall, Sanish, Bakken core.

The lit wells concentrate in a tight arc across Mountrail, McKenzie, Williams, and Dunn counties: the Bakken horizontal core. Every pad on that arc carries a full modern log suite and a digital completion report. The vertical wells on the state's eastern margin and the southern shelf are not lit, and never will be.

Zoom into the core and the violet dominates. Zoom out to the basin and the ghost dominates. Both are true at once. The map asks you which scale you trust.

15 · The tell

Dark data is a stratigraphy of time.

Kansas was structurally dark: old plays, shallow oil, no pressure. Permian was selectively dark: wealth on top, dark history underneath. Williston is the cleanest version of the same rule. The horizontal era lit this basin. Everything before it is still paper, still unindexed, still not in anyone's public stack.

The question the basin poses is not whether the industry could digitize its pre-2000 record. It is why, a decade and a half into the digital era, it still hasn't.

The Williston is the good state.

Three basins in, the series has a curve. Kansas sits at 94% dark. Permian at 78%. Williston at 46%. Not because the Williston is older or shallower or poorer, but because it is dominated by a single fifteen-year drilling campaign that happened to coincide with the digital-filing era. Strip that campaign and the basin snaps back into Kansas territory. The lit basin is the horizontal Bakken. Everything else is the floor the industry still hasn't swept.

Where this goes

46% dark is a floor, not a ceiling. The pre-2000 Williston is still there, and still mostly unreachable.

  • Operators & modellers. Bakken type-curve work leans hard on the horizontal log suite. Vertical Madison and Mission Canyon wells that ring the basin still hold the stratigraphic context the petroleum system was built on, and most of them are paper.
  • Acquirers & underwriters. Vertical-era well files carry plug-and-abandon liability that has to be priced against a log record that does not exist publicly. 6,379 wells with no parseable vintage are exactly the inventory nobody wants to inherit.
  • Regulators & policy teams. Orphan inventories, CCS screening, seismicity review, and long-interval pressure modelling all depend on the legacy wells being findable. "The basin is 46% lit" is an accurate headline and a misleading policy statement.

I build the pipelines, basin-clip analyses, and AI-assisted subsurface QC that close this kind of gap, across basins, regulators, and the geological detail that still matters. If you work any side of this, let's talk.

Sources: NDIC Oil & Gas Division GIS (OGD_Wells, OGD_Horizontals) · Montana BOGC (WellSurface shapefile) · USGS ScienceBase item 618e90c8: Bakken and Three Forks Assessment Unit boundaries, 2021 National Assessment. Wells spatial-joined in EPSG:5070 (US Albers). Coordinates in EPSG:4269 (NAD83). Built with deck.gl and scrollama. Analysis: @salamituns.