Technology Radar: July 2026

Six developments, in the order we would act on them. Each is here because the primary source has been read, or because its authorship, mechanism and limitations could be established from sources that are open. Nothing is included on the strength of a headline.

Satellite data now checks what an operator reports

Williams and Gautam, with colleagues at the Environmental Defense Fund, MethaneSAT LLC, Harvard and the Universitat Politècnica de València, published basin-level methane emissions and intensities from 33 MethaneSAT scenes gathered between May 2024 and May 2025¹. The paper is peer-reviewed and open access in Atmospheric Chemistry and Physics, and it covers six producing regions across six countries and 207 counties and districts: the Permian, San Joaquin and Eagle Ford in the United States and Mexico, the Amu Darya in Turkmenistan and Uzbekistan, and the Zagros Foldbelt in Iran and Iraq.

The instrument resolves area emissions at about 4 km by 4 km. Measured emissions span an order of magnitude, from 454 tonnes an hour in the Permian to 114 in the Eagle Ford. Loss rates span nearly two orders: 2.4 per cent in the Eagle Ford, 2.6 in the Permian, 2.9 and 3.7 in the Turkmen and Uzbek Amu Darya, 12.1 in the San Joaquin and 18.6 in the Zagros Foldbelt. Every region exceeds the Oil and Gas Climate Initiative target of 0.2 per cent by between 10 and 90 times¹.

The comparison against bottom-up inventories is the part worth reading twice. Permian and Eagle Ford oil and gas emissions came in about four times the gridded EPA greenhouse gas inventory and the San Joaquin about twice. Outside the United States the gaps are wider: the Turkmen Amu Darya ran 5 to 8 times existing inventories across the full domain, and individual jurisdictions reached 15 to 62 times¹.

What it is good for: a reported emissions number is now something you can check rather than something you have to accept. The concentration and emissions datasets are public on Google Earth Engine, so a buyer can compare an operator's declared methane intensity against an independent measurement over the same acreage for the cost of an analyst's time. That matters commercially, because EU import rules and the certification schemes used to price differentiated gas rest on intensity figures that were previously self-reported.

The limits: satellite inversion and bottom-up inventory are different methods over different periods, so a discrepancy is not automatically an error in the inventory. Sectoral attribution carries wide confidence intervals and the paper is candid about it, putting the oil and gas share at 90 per cent in the Permian with a range of 64 to 100, and at 24 per cent in the San Joaquin where other sources dominate. This is one year of scenes, not a time series¹.

Temperature data you already hold can measure saturation

Guimaraes, Galvao, Horne and Tartakovsky, working across Stanford and Petrobras, show that routinely acquired downhole temperature data allows direct estimation of fluid saturations, and that this improves in-place volume estimates in reservoirs with aquifer support². The work is open access in SPE Journal.

The method treats a nonisothermal closed reservoir with impermeable boundaries at high Péclet number, builds a dimensionless formulation for an axisymmetric cylindrical domain, and derives analytical temperature expressions by the method of characteristics under pseudosteady pressure. The authors validate against commercial thermal simulators and wellbore data, then apply it to a field case with aquifer support².

What it is good for: estimating stock-tank original oil in place without assuming a water saturation value. Pressure analysis on its own cannot separate saturation from volume, so the assumed water saturation carries the answer, and that assumption is one of the most common places an in-place number quietly goes wrong. Here the input is temperature already logged and sitting in the well file. There is nothing to acquire and nothing to buy.

The limit is in the assumptions, and they are real: closed boundaries, high Péclet number, pseudosteady pressure, and one published field case. This is a strong second opinion on a volumetric estimate rather than a replacement for one, and the disagreement between the two is where the interesting work sits.

A way to test what a vendor means by autonomous

de Wardt, with Cayeux and Mihai of NORCE, Macpherson of Baker Hughes, Pirovolou of Weatherford and Annaiyappa, propose a taxonomy separating automated drilling systems from autonomous ones, drawing definitions from aerospace, control theory and other high-risk industries³. The paper is open access in SPE Journal.

The framework scores ten traits: environment complexity, problem complexity, mission, risk management, dynamic planning, situational awareness, decision-making, execution, learning from experience and human-agent interaction. Each trait carries a quantification, so a claim of autonomy can be assessed rather than accepted. The authors also argue that an autonomous system can collaborate with human operators while holding independent decision authority, which cuts against the assumption that autonomy means removing the human³.

What it is good for: procurement and diligence language. Autonomous is currently a marketing word in drilling, and a buyer evaluating a contractor or a technology has had no agreed way to test it. Ten scored traits give you a checklist to put in front of a vendor and a basis for comparing two bids that claim the same thing. It costs nothing to adopt and it is usable this quarter.

The limit: this is a framework, not a field result. It changes how you assess a claim. It does not drill anything.

Closed-loop fracturing reaches the field, with one hard constraint

Navaiz and Stark of Halliburton and Paradeis of Chevron reported what they describe as the first closed-loop completions programme at the SPE Hydraulic Fracturing Technology Conference in The Woodlands in February 2026, and JPT carried it in its June technology focus⁴. The work was done on Chevron acreage in Colorado⁵.

The mechanism is more specific than the phrase suggests. Offset-well fibre measurements, principally volume to first response, characterise how fracture growth rates vary between stages. Treatment volume is then reallocated away from underperforming stages toward better ones, holding total volume constant⁶. Halliburton attributes the execution to its ZEUS IQ platform with OCTIV auto frac for automated stage execution and Sensori for subsurface feedback, and describes the result as reacting to a localised environment rather than to a performance forecast⁵.

What it is good for: it moves completion decisions inside the job. Surface automation on a fracturing spread already controls the pumps, but it does not change the design, so the learning lands on the next well. Reallocating volume between stages while pumping puts the diagnostic to work immediately, and because total volume is held constant it is a reallocation rather than an extra spend.

The limits are real and worth knowing before anyone gets excited. ResFrac's review of the conference notes that the relationship between stage-level fluid uniformity and volume to first response is not well understood, and may be confounded by asymmetrical fracture growth, stress shadowing, stage and well sequencing, and where the observation well happens to sit. It also makes the practical objection: offset observation wells are not routinely available, and never available for every well on a pad⁶. No production or cost result has been published in any source we can read. Treat this as a method proven to run, not yet a method proven to pay.

Money goes to multilateral re-entry and SAGD ranging

Emissions Reduction Alberta awarded 37 million dollars across ten drilling technology projects worth about 179 million dollars combined. Most went to geothermal: 8 million dollars each to Eavor Technologies for superhot resource access and Rodatherm Energy for closed-loop geothermal power, and 4.4 million to Borobotics for an autonomous electric drilling robot⁷.

Two smaller awards matter more for mature oil and gas. Global Dynamics received 2.25 million dollars for multilateral well re-entry, and Weatherford Canada 1.42 million for real-time sensing and ranging in SAGD drilling. Phase Advanced Sensor Systems received 420,000 dollars for high-temperature downhole pressure sensors and Precision Drilling 3.1 million for robotic pipe handling⁷.

What it is good for: multilateral re-entry is a redevelopment tool. It reaches new drainage from a wellbore that already exists, avoiding a new surface slot and a new well cost, and that arithmetic decides whether a mature field gets a second phase or gets abandoned. High-temperature pressure sensors matter for the same reason, since thermal and deep completions are where instrumentation fails first.

The limit: this is funding, not results. Ten projects have money, none has a published field outcome, and public innovation programmes convert poorly. Treat it as a two to three year watch list.

Depleted gas reservoirs get a second life as storage

Harbour Energy's Viking CCS project has cleared front-end engineering and design and had its development funding approved. The numbers are specific enough to be useful: eight suitable depleted gas reservoirs in the Viking fields, injection at roughly 9,000 ft beneath a salt layer 600 to 1,000 ft thick, 400 million tonnes of storage capacity, an initial target of up to 4 million tonnes a year in the early 2030s, and a subsea pipeline sized for up to 30 million tonnes a year against an expansion case of 10. Investment required is put at more than 13 billion pounds⁸.

What it is good for: the reservoir engineering that makes a depleted gas field a storage site is the same work as managing it in late life. Pressure history, seal integrity, well stock and abandonment status are the inputs either way. For anyone holding a mature gas asset near industrial CO2, that is a second exit route to price alongside abandonment, and these figures give a reference point for scale.

The limit: this is a company announcement rather than a technical paper, so the capacity and containment case has not been through peer review. First operations are early 2030s, and 13 billion pounds is a plan rather than a spend.

What we would use, and where

The MethaneSAT data goes into diligence immediately, as a check on any declared methane intensity. It is free, it is peer-reviewed, and on the evidence of six basins a self-reported figure is as likely as not to be understated by a factor of two or more. We would run it on any asset where gas pricing, an import rule or a certification scheme depends on the number.

The temperature work goes into use immediately, on due diligence and redevelopment screening. It costs nothing to try, the data is already in the well files, and it bears on the number that decides the valuation. We would run it alongside a conventional volumetric and treat any disagreement between the two as the thing worth investigating.

The autonomous drilling taxonomy goes into procurement language this quarter. Ten scored traits are a better question to put to a drilling contractor than asking whether the system is autonomous, and it gives a diligence review something testable where it previously had a brochure.

Multilateral re-entry goes on the redevelopment watch list, to be revisited when the Alberta projects publish field results rather than on the strength of a grant. Viking is worth tracking as a reference case for anyone pricing a late-life gas asset, because it puts a number on what a large depleted-reservoir storage scheme looks like.