Skip to main content
Reading viewAll insights →
BLOG17 min read

Karratha Has About 4 Mtpa Idle, and 82% of 894 Released Wells Around It Have Digital Logs

Karratha Has About 4 Mtpa Idle, and 82% of 894 Released Wells Around It Have Digital Logs
Tannistha Maitiby Tannistha MaitiHead of Engineering · 19 Nov 2026
Share

Woodside retired Karratha's train 2 and cut nameplate from 16.9 to 14.3 Mtpa, and on its own quarterly figures the plant's output fell by about as much: the idle room in Q2 2026, 4.27 Mtpa, is the room it had in Q3 2024 before the train went. At IEEFA's 150 MMscfd per Mtpa that is about 0.64 Bcf/d of inlet gas, for a plant now approved to run to 2070. The wells that could locate the backfill are public. We counted NOPIMS's listings for 894 released exploration and appraisal wells in the Northern Carnarvon and Browse basins: 734 (82%) hold a log item in the digital classes and 65 have scans only, 48 of them drilled before 1990. The listing text confirms a curve format for only 282. The records missing are the correlations and the confidence on net pay.

The Karratha Gas Plant gave up a train and kept its idle room. Woodside stopped production from LNG train 2 on 28 October 2024 and retired it in the second quarter of 2025 as its feedstock declined, which cut the plant's nameplate from 16.9 to 14.3 million tonnes a year [2][3][8]. On Woodside's own production figures, the plant made LNG at an annual rate of 12.65 Mtpa in the third quarter of 2024 and 10.03 Mtpa in the second quarter of 2026, so output fell by 2.62 Mtpa while capacity fell by 2.6. The idle room was 4.25 Mtpa before the retirement and 4.27 Mtpa after it. The plant is now approved to operate until 2070 [9]. The gas to fill the room has to be found in the Carnarvon and Browse basins, and the released well data from those basins is public. We counted what the national repository lists for it: of 894 released exploration and appraisal wells, 734 hold digital log data and 65 hold scans only.

The shortfall at Karratha is gas

Karratha sits on the Burrup Peninsula in the Pilbara, about 1,260 km north of Perth. It has processed gas for the domestic market since 1984 and exported LNG since 1989 [1]. It is the onshore half of the North West Shelf Project, which Woodside operates: three offshore platforms whose gas comes ashore through trunklines of about 130 km, four LNG trains since the retirement, and two domestic gas trains [1][7]. Woodside's own description of where it stands is plain. After 40 years of operations the project is entering a period of production decline, and the plant "will progressively have increased processing ullage" despite the third-party gas it already takes [7].

Woodside's quarterly reports give the plant's average daily LNG production on a 100% basis, in thousands of barrels of oil equivalent a day, for gas from the Shelf's own fields and, separately, for Pluto gas processed through the 3.2 km Pluto-KGP Interconnector that started in March 2022 [2][3][4][5][6][14]. Woodside publishes the conversion: a tonne of Karratha LNG is 8.08 boe [4]. Converting at that factor:

QuarterShelf gas, Mboe/dPluto gas, Mboe/dLNG made, MtpaNameplate, MtpaIdle, Mtpa
Q3 20242592112.6516.94.25
Q4 20242582412.7416.94.16
Q1 20252352311.6516.95.25
Q2 2025202199.9814.34.32
Q3 20252182310.8914.33.41
Q4 20252202010.8414.33.46
Q1 20262102710.7114.33.59
Q2 20262031910.0314.34.27

In the five quarters at 14.3 Mtpa the idle room ranged from 3.41 to 4.32 Mtpa and averaged 3.81. The shortfall is gas. In the fourth quarter of 2025 the Shelf's LNG reliability was 99.8% [6], and the plant still made only 10.84 Mtpa, 76% of nameplate. Some quarters carry other effects: planned maintenance in the second quarter of 2025, and more of the gas sent to the domestic market in the second quarter of 2026 [3][4]. None of them changes the order of magnitude.

The draft of this piece put the plant's whole requirement at 1.9 Bcf/d, using a rule of 130 MMscfd of feed gas per Mtpa. That rule describes the gas that leaves in the ships. On Woodside's own factors, 8.08 boe per tonne and 5,700 scf per boe [4], a tonne of Karratha LNG carries the energy of about 46,000 scf, which is 126 MMscfd for each Mtpa. The plant burns and removes gas on the way, so the inlet needs more. IEEFA puts LNG Canada's 14 Mtpa at 2.1 Bcf/d of feed gas [11], 150 MMscfd per Mtpa. At that rate a full 14.3 Mtpa plant takes about 2.15 Bcf/d at the inlet, and the room idle in Q2 2026 is about 0.64 Bcf/d of it. The exact factor depends on the gas and on the plant's fuel use, so the exhibit lets you set it.

What could fill it, and on what clock

In 2020 Wood Mackenzie warned that up to 7 Mtpa of Karratha's capacity could be idle by 2027 unless new supply was secured, and expected near-term backfill to bring new volumes into the plant from 2022 [10]. Some of it arrived. Pluto gas has been processed at Karratha since 2022 [14], and Woodside began processing Waitsia Stage 2 gas through the Shelf's facilities in the fourth quarter of 2025 [6]. Against the 16.9 Mtpa nameplate of 2020, the plant's 10.03 Mtpa in Q2 2026 leaves 6.87 Mtpa unused, close to that forecast. Retiring train 2 took 2.6 Mtpa of it off the books without adding a molecule.

The near-field option is small against that room. The Greater Western Flank Phase 4 project, a five-well subsea tie-back sanctioned in the fourth quarter of 2025, targets start-up in 2028, with 100 MMboe of gross proved plus probable reserves, and Woodside says it extends the Shelf's production "by around one year" [6]. Scarborough, Woodside's large gas field in the same basin, is going to a new train at the neighbouring Pluto plant, with first LNG targeted for the fourth quarter of 2026 [4]; its gas is lean and would have needed new nitrogen removal to enter Karratha [10].

Browse is the backfill named in the approval [9]. Its Brecknock, Calliance and Torosa fields lie about 425 km north of Broome, and the concept is two floating production facilities feeding a trunkline of about 900 km to the Shelf's existing infrastructure [12]. In July 2026 Western Australia gave the Browse to North West Shelf Project State Significant Project status [4][13]. It has no final investment decision. When Woodside exercised its pre-emption right over PetroChina's 10.67% interest in June 2026, part of the price was a US$175 million payment due only if a final investment decision on all three fields is taken on or before 30 June 2032 [4].

So the joint venture faces a backfill question with an open horizon. One tie-back buys about a year. The plant's approval runs 44 more years. Between them sits a ranked list of candidate gas zones within pipeline reach, the list a development director needs before any commercial case, and that list rests on the released wells.

The well record is public, and mostly digital

Operators in Australian Commonwealth waters submit their well data to the regulator, and after a confidentiality period it becomes open file. The National Offshore Petroleum Information Management System, NOPIMS, is the public catalogue, and the national offshore wells data collection behind it is listed under a Creative Commons Attribution 4.0 licence [15][16]. No registration is needed to search it. The listing behind the public portal can be queried directly, and for each borehole it returns every item held: its type, a description, its size and an item ID [15]. The dates in the listing show the release rule at work: a borehole rig-released on 29 June 2016 carries a basic-data date of 29 June 2018 and an interpretive date of 29 June 2021.

The draft promised three counts that tell a backfill team how fast it can answer. The first is how many candidate wells have digital curves and how many only images. That one can be made today, without anyone's permission, so we made it.

Scope. Every borehole NOPIMS flags as offshore, labelled Northern Carnarvon Basin (in any of its three spellings) or Browse Basin, drilled for exploration or appraisal: 1,094 boreholes, which fold into 894 wells once sidetracks join their parent, rig-released between 1968 and 2025. Development wells are left out; they sit inside fields already found.

Log items. We took the items whose type is a log: well log, wireline, LWD, MWD, mud log, composite and master logs, and log interpretations. Each item ID starts with a class letter. To learn what the letters mean we read the published file listing of 46 items, in a browser, across all three classes. Of 33 items in classes D and P, 28 held a LAS, DLIS, LIS or TAP file of curves. All 13 items in class I held nothing but TIF or PDF scans.

Tiers. A well counts as digital if any of its log items is in class D or P; as scans only if all of its log items are in class I; as logs in the report if it has no separate log item, only completion reports; and as unlisted if nothing is listed yet.

WellsDigitalScans onlyLogs in report onlyUnlisted
Northern Carnarvon73460956636
Browse1601259260
Both894734 (82%)65 (7%)89 (10%)6

The scan-only wells are old. Of the 65, 48 were drilled before 1990, and their median year is 1983. Before 1990, 23% of the wells have scans only; from 1990, 2.5% do. Even among the 210 wells drilled before 1990, 156 hold a digital item, and 84 of those have a class P item, which in our sample meant curves transcribed from tape. Goodwyn 1, drilled in 1971, has a single LAS file dated 2004 alongside a 1.47 GB wireline item in the image class.

The exceptions are specific. North Rankin 1, the 1971 discovery well of one of the fields that built the plant, has its logs in NOPIMS as 102 TIF files. Calliance 1, a 2005 appraisal well on one of the three Browse fields, has a single wireline item of 19.37 GB that holds 17 TIF files. Neither has a log item in the digital classes.

KARRATHA · THE ROOM AND THE WELLS4.27 MtpaIDLE IN Q2 2026, OF 14.382%WELLS WITH A DIGITAL LOG ITEMQ2 2026: 10.03 of 14.3 Mtpa made, 4.27 Mtpa idle. Filling it takes 0.64 Bcf/d of inlet gas at 150 MMscfd per Mtpa.734 of 894 released Northern Carnarvon and Browse exploration and appraisal wells hold a log item in the digital classes; 65 have scans only.KARRATHA LNG OUTPUT, MTPA00.0050.75101.50152.25Bcf/d innameplate 16.914.3 after train 24.27 idleQ3'24Q1'25Q3'25Q2'26NWS gasPluto gasQ3 2024: 12.65 Mtpa made, 4.25 idleQ4 2024: 12.74 Mtpa made, 4.16 idleQ1 2025: 11.65 Mtpa made, 5.25 idleQ2 2025: 9.98 Mtpa made, 4.32 idleQ3 2025: 10.89 Mtpa made, 3.41 idleQ4 2025: 10.84 Mtpa made, 3.46 idleQ1 2026: 10.71 Mtpa made, 3.59 idleQ2 2026: 10.03 Mtpa made, 4.27 idleEXPLORATION AND APPRAISAL WELLS PER YEAR · BOTH BASINSdigital log itemscans only, median 1983logs in report only102030401970198020002010202022.9% scans2.5% scans1990Before 1990: 48 of 210 wells scans only (22.9%). From 1990: 17 of 684 (2.5%).At 150 MMscfd per Mtpa the full 14.3 Mtpa takes 2.15 Bcf/d at the inlet. The gas carried in the LNG itself is 126 per Mtpa on Woodside's factors.Woodside quarterly reports, Q4 2024 to Q2 2026: production rates, 100% project; 8.08 boe per tonne; 5,700 scf per boe. IEEFA, Sep 2026: 2.1 Bcf/d for 14 Mtpa.NOPIMS public listings, 19 Oct 2026: 1,094 offshore exploration and appraisal boreholes, 894 wells. Digital: a log item of class D or P. Scans: class I only.
quarter:
digital means:basins:

Q2 2026: 10.03 of 14.3 Mtpa made, 4.27 Mtpa idle. Filling it takes 0.64 Bcf/d of inlet gas at 150 MMscfd per Mtpa. 734 of 894 released Northern Carnarvon and Browse exploration and appraisal wells hold a log item in the digital classes; 65 have scans only. Before 1990: 48 of 210 wells scans only (22.9%). From 1990: 17 of 684 (2.5%).

Left: Karratha Gas Plant LNG output by quarter, from Woodside's quarterly reports (average daily production, 100% project, NWS Project LNG plus Pluto gas processed through the Pluto-KGP Interconnector), converted at Woodside's 8.08 boe per tonne, against nameplate. Train 2 stopped producing on 28 October 2024 and was retired in Q2 2025. The right-hand axis turns Mtpa into inlet gas at the chosen factor. Right: released exploration and appraisal wells in the Northern Carnarvon and Browse basins, counted from the NOPIMS public listings on 19 October 2026, by first rig-release year and by what is listed for their logs. In a sample of 46 item file listings, 28 of 33 items in the digital classes held a LAS, DLIS, LIS or TAP file, and all 13 in the image class held only TIF or PDF scans.

The second toggle in the exhibit asks a different question: what does the search screen itself say? If a well counts as digital only when the description of one of its log items names a curve format, the count falls from 734 to 282, 32% of the wells. The other 452 digital wells carry log items described as "Various", "Wireline" or nothing at all. Their log items total 1.23 TB. A team that reads the catalogue's text would conclude that two-thirds of the basin needs digitising. The item classes, and the file listings that back them, say that about 7% does. Across all 894 wells the log items come to 2.26 TB, and 1.04 TB of that is in the image class: most wells have scans too, and the scans are the bulk of the bytes.

What the count does not answer

That settles the first count, and it moves the bottleneck. A digital item is not a complete, depth-matched curve set across the interval that matters; LIS and TAP files are tape images that need converting before a geologist can load them; and an item in class D or P can hold an interpretation rather than raw curves. Still, for most wells within reach of Karratha the digital data exists and can be downloaded today.

The two counts the open record cannot give are the other two in the draft. How many wells have a regional correlation that someone has checked, and how many carry picks of unknown origin? And does each candidate zone have a net pay with a confidence, or only a single number from one interpretation? Where a well found gas, it found it in particular reservoir layers. For every zone within pipeline reach of Karratha, what is the net pay, how confident are we, and is the data from those wells in a state where someone can answer in a week instead of a quarter? None of this stops a decision. It stops a fast one, because without a shared correlation every new option is reinterpreted from the logs up.

Our own demonstration of that workflow runs on released wells from Stybarrow, in the Exmouth Sub-basin. Stybarrow is an oil field. NOPIMS records its first exploration well, Stybarrow 1, as rig-released in March 2003 by BHP; the field produced from November 2007 to July 2015 as a BHP and Woodside joint venture, and Woodside has owned and operated it since the 2022 merger and is decommissioning it [17]. The demonstration loads the released logs and reports, correlates well to well, builds the geological model and ranks zones by volumetrics and chance of success. It shows the workflow on oil zones in one sub-basin. It does not show a gas backfill ranking, and the extension to the wider Carnarvon well set is still to run. The forecasting half of a backfill case, a per-field gas profile with a band summed against train capacity, is the part we are still building; today we run production forecasting for oil.

The questions worth putting to any backfill case follow from the counts. For each candidate zone, is the net pay a number with a confidence, or one number from one interpretation? How many of the wells behind it are in the digital classes, and how many are scans like North Rankin 1? And when the next well result arrives, which options does it contradict, and who is told?

Key takeaways

  1. Train 2 stopped producing on 28 October 2024 and was retired in Q2 2025, cutting Karratha's nameplate from 16.9 to 14.3 Mtpa.
  2. On Woodside's 100% figures at 8.08 boe per tonne, output fell from 12.65 Mtpa in Q3 2024 to 10.03 Mtpa in Q2 2026, so the idle room stayed at about 4.3 Mtpa; across the five quarters at 14.3 it averaged 3.81.
  3. In Q4 2025 the plant ran at 99.8% LNG reliability and made 10.84 Mtpa, 76% of nameplate: the shortfall is gas.
  4. 130 MMscfd per Mtpa is about the gas leaving as LNG (126 on Woodside's factors); at IEEFA's inlet rate of 150, the full plant takes about 2.15 Bcf/d and the Q2 2026 idle room about 0.64.
  5. GWF-4 extends the Shelf's production by about one year from 2028; Browse, about 900 km away, has State Significant status and no final investment decision; the plant is approved to 2070.
  6. Of 894 released exploration and appraisal wells in the Northern Carnarvon and Browse basins, 734 (82%) hold a digital-class log item, 65 (7%) scans only, 48 of those drilled before 1990.
  7. The NOPIMS listing text names a curve format for only 282 wells (32%): the archive is more digital than its catalogue reads.
  8. Correlations someone has checked and net pay with a confidence are not in the public record, and they are what slows a backfill decision.

Limitations

The plant figures are Woodside's production rates, not throughput measured at the inlet. The Pluto gas rows are converted at Karratha's 8.08 boe per tonne; Pluto's own factor, 8.34, would lower them by about 3%. Idle room is measured against nameplate, and no plant makes nameplate every day, so the room a backfill could fill is somewhat smaller than the table shows. Some quarters include planned maintenance or more domestic gas, which lowers LNG for reasons other than supply.

The well count reads the catalogue, not the files. The class letters are a proxy calibrated on 46 file listings: 5 of 33 digital-class items held no LAS, DLIS, LIS or TAP file, so some wells we call digital will not have usable curves. Items in class I that hold digital data would make the scan-only count an overestimate; none of the 13 we read did. Reading more listings was limited by the content delivery network's rate controls, so the sample is small.

A digital item says nothing about coverage. A well can have digital curves over one section and scans over the reservoir. Our count does not test depth range, curve suite or quality, and LIS and TAP files need conversion before use.

The scope is NOPIMS's basin labels, not a distance from Karratha. The Northern Carnarvon label covers the Exmouth Plateau and Exmouth Sub-basin as well as the Rankin Platform and Dampier Sub-basin, and some boreholes carry no sub-basin, so we did not split nearer from farther fields. Released wells are also a biased sample: they are the wells old enough to be out of confidentiality, and the ones that found something tend to carry more data.

Nothing here measures reservoir quality. A basin where every well is digital can still have no zone worth a pipeline.

References

  1. Offshore Technology. Karratha Gas Plant, Western Australia: about 1,260 km north of Perth, domestic gas from 1984, LNG from 1989, two 130 km trunklines: offshore-technology.com.
  2. Woodside Energy. Fourth quarter report for period ended 31 December 2024, ASX, 22 January 2025: train 2 cessation of production executed 28 October 2024; production rates, 100% project: announcements.asx.com.au.
  3. Woodside Energy. Second quarter report for period ended 30 June 2025, ASX, 23 July 2025: retirement of train 2, capacity from 16.9 to 14.3 Mtpa; production rates: announcements.asx.com.au.
  4. Woodside Energy. Second quarter report for period ended 30 June 2026, ASX, 29 July 2026: production rates; Karratha Gas Plant 8.08 boe per tonne, 5,700 scf per boe; Scarborough first LNG targeted Q4 2026; Browse pre-emption and the contingent payment; State Significant Project status: announcements.asx.com.au.
  5. Woodside Energy. Third quarter report for period ended 30 September 2025, ASX, 22 October 2025: production rates; planned maintenance at North Rankin and Karratha: announcements.asx.com.au. The Q1 2025 report gives the March 2025 rates: woodside.com.
  6. Woodside Energy. Fourth quarter report for period ended 31 December 2025: NWS LNG reliability 99.8%; GWF-4 final investment decision, five wells, start-up 2028, 100 MMboe gross 2P, extends production by around one year; Waitsia Stage 2 processing began: woodside.com.
  7. Woodside Energy. Half-year report for period ended 30 June 2025, ASX, 19 August 2025: three offshore platforms, four LNG and two domestic gas trains; production decline and increasing processing ullage: announcements.asx.com.au.
  8. Argus Media. Australian Woodside's output up but LNG declines in 2Q, 23 July 2025: train 2 offline from late 2024 on declining feedstock, retired April to June: argusmedia.com.
  9. Reuters via Baird Maritime. Australia approves North West Shelf gas plant to run until 2070, 12 September 2025: 48 conditions; Browse as the new supply: bairdmaritime.com.
  10. G. Thompson, Wood Mackenzie. Re-inventing the North West Shelf, opinion, 8 September 2020: up to 7 Mtpa idle by 2027; backfill volumes from 2022; Scarborough gas lean and needing nitrogen removal at the NWS: woodmac.com.
  11. IEEFA. LNG Canada Phase 2: After FID key questions remain, September 2026: Phase 1's two trains require 2.1 Bcf/d: ieefa.org.
  12. Environmental Protection Authority of Western Australia. Proposed Browse to North West Shelf Project, Commonwealth environmental impact statement consultation: fields about 425 km north of Broome, two FPSOs, a 900 km trunkline to existing NWS infrastructure: consultation.epa.wa.gov.au.
  13. Energy Connects (Bloomberg). Woodside's $34 billion Browse named significant status project, 27 July 2026: energyconnects.com.
  14. Hydrocarbon Engineering. Processing of Pluto gas starts at North West Shelf, 31 March 2022: the 3.2 km Pluto-KGP Interconnector: hydrocarbonengineering.com.
  15. NOPTA and Geoscience Australia. NOPIMS public well search and the well and report listings behind it, queried 19 October 2026: public.neats.nopta.gov.au/nopims/wells; about the system: ga.gov.au/nopims.
  16. Geoscience Australia. Australian National Offshore Wells Data Collection, Research Data Australia record: Creative Commons Attribution 4.0 International: researchdata.edu.au.
  17. Offshore Technology. Stybarrow field decommissioning, Australia: Exmouth Sub-basin oil field, first production November 2007, production ceased July 2015, Woodside owner and operator after the 2022 merger: offshore-technology.com.
  18. A neighbouring LNG supply question counted from open well files: Phase 2 Needs 2.1 Bcf/d More Montney Gas, and 91% of Its Full-Interval Logs Predate 2009.
  19. How the raster share of an archive sets how long it takes to become readable: Fully Deployed Is Not Machine-Readable.
Tannistha Maiti
Tannistha Maiti

Head of Engineering

More from EarthScan

Related research

All insights →
Mozambique LNG Spent 54 Months in Force Majeure, 2.4 Times the 22 Months It Had Worked
Insight

Mozambique LNG Spent 54 Months in Force Majeure, 2.4 Times the 22 Months It Had Worked

Phase 2 Needs 2.1 Bcf/d More Montney Gas, and 91% of Its Full-Interval Logs Predate 2009
Insight

Phase 2 Needs 2.1 Bcf/d More Montney Gas, and 91% of Its Full-Interval Logs Predate 2009

Two LNG Trains Dip Under 60% of Nameplate One Year in Three, and Seven One Year in 600
Insight

Two LNG Trains Dip Under 60% of Nameplate One Year in Three, and Seven One Year in 600

Stay ahead

EarthScan insights, in your inbox.

Field-tested research on subsurface and energy-transition AI. About twice a month. No noise.

We use your email only for this newsletter. Unsubscribe anytime Privacy.