Estimated reading time: 21 minutes
Ukraine has now struck most of Russia’s largest oil refineries. Casey Biggerstaff argues that the case for attacking what remains grows stronger as the campaign destroys the facilities that supplied Russian forces — that infrastructure previously devoted to commercial purposes may increasingly qualify as a military objective by purpose because it could be converted to military use. That reasoning deserves examination, and it does not survive it in the form in which it is usually put.
The question matters beyond this war. Purpose is the one criterion in Article 52(2) AP I that looks forward, and it is the criterion the harder targets in any long strike campaign will turn on. It is also the criterion on which the manuals do not agree — a divergence that has attracted far less attention than the war-sustaining debate.
My earlier post on What is a Lawful Target set out what the four criteria of Article 52(2) require, including that purpose turns on intentions known to guide the adversary rather than on worst-case assumptions. A second post, on data as a military objective, asked whether the same provision reaches intangible objects. This post takes the purpose criterion into a campaign long enough to exhaust its easy cases.
One thing this post does not do. It reaches no conclusion on the lawfulness of any individual Ukrainian strike, or of the campaign as a whole. The evidence on which qualification under Article 52(2) turns is what a commander knew about a particular facility at the time, and that evidence is not public. Where the analysis touches named facilities it does so conditionally. The strikes are here because they pose the question, not because I am in a position to answer it about any of them.
The scope is qualification alone. Whether an attack that strikes a lawful target is nonetheless disproportionate, and what weight environmental damage carries in that judgment, are addressed by Davit Khachatryan and not here.
I. The Campaign and the Question It Poses
During 2026, Ukraine’s long-range drone strikes on Russian oil refining became methodical rather than occasional. Eight of the ten largest Russian refineries had been hit by the end of June, measured by processing capacity. Analysts cited by Khachatryan put around a third of national refining capacity out of action, with rationing appearing across a major energy producer. The Moscow Oil Refinery was hit twice in a single week in mid-June. Two years earlier the picture was far smaller: Markus Bernhart recorded roughly a tenth of capacity affected as at March 2024.
The target set is also wider than “refineries” suggests. Biggerstaff records strikes reaching the Baltic export terminals at Primorsk and Ust-Luga, the Krasny Yar pumping station and the Grushevaya storage and transshipment facility; Adamski and his co-authors add the Unecha and Nikolskoye pumping stations on the Druzhba pipeline.
Ukraine has justified the strikes in two registers. Its General Staff has named particular facilities as supplying Russian forces, among them the Poltavska depot, the Ryazan and Samara refineries and the Koltsevoy pipeline. Kyiv’s public framing, as Khachatryan reports it, has been broader, resting on the contribution these facilities make to sustaining the Russian war effort by the revenues they generate. Both justifications are on the record, and they are not the same claim. Readers of my earlier post on lawful targets will rightly conclude that the first can be based on Article 52 (2) as it concerns the actual use of a refinery to supply fuel to armed forces. The second is a different matter because it rests on the view adopted in Section 5.6.6.2 of the U.S. DoD Law of War Manual that a war sustaining capability of an object is sufficient to qualify it as a military objective. As will be shown immediately below, this view has remained controversial.
The clearest demonstration that rejecting the war sustaining thesis does not necessarily prevent an object from becoming a military objective sits inside a single article. Yoram Dinstein rejects the war-sustaining limb as too broad and requires a proximate nexus to military action. His illustration is the American position’s own: Union forces destroyed Confederate raw cotton because its sale funded almost all Southern arms and ammunition. If that connection suffices, Dinstein reasons, almost any civilian activity could be said to sustain a war effort — especially in a long conflict.
Dinstein´s Two Positions
Ten pages later, and without contradiction, he treats oil installations as the infrastructure of the military industry, and endorses — quoting Green, and putting it as something that can well be argued — the view that oil installations of every kind are open to destruction.
Whether or not that view is sound, it shows the two questions travel separately. Someone can take an expansive line on oil targeting while rejecting the war-sustaining extension outright — and conflating the two has obscured which one is doing the work here..
Does Inaction Change the Law?
Wolff Heintschel von Heinegg has argued that States are increasingly prepared to accept the war-sustaining approach, once dismissed as an isolated American position, and that they have neither protested nor doubted the Ukrainian attacks on Russian oil export infrastructure.
The inference is weaker than it looks. Himanil Raina answers it from the ILC’s conclusions on customary international law, under which inaction counts as practice only in certain circumstances, and failure to react evidences acceptance as law only where States were in a position to react and the circumstances called for some reaction. Neither condition can be assumed here. Hathaway, Khan and Revkin press a parallel argument about a different expansion of the military-objective definition, and the reasoning carries across: where endorsement and practice remain confined to a small group of States who look to one another, the views of the majority go unconsidered, and custom does not form from the agreement of the few.
There is a further difficulty, and it lies on the acting State’s side. As Section I recorded, Ukraine has justified the campaign in two registers. A campaign whose own belligerent reaches for military supply as often as for revenue supplies thin material for endorsing a doctrine about war-sustaining revenue.
The contested targets are argued on purpose, not on revenue.
III. Military Objective by Purpose: What the Authorities Say
The Formulation Most Authorities Use
Purpose is the criterion that looks forward. The ICRC Commentary draws the line plainly: purpose concerns an object’s intended future use, while use concerns its present function. The same paragraph of The ICRC Commentary attaches the doubt rule in Article 52(3) to that analysis, so the presumption is not confined to what an object is doing now — a point I examined in my earlier post on lawful targets and do not reopen here. The UK Manual and the Australian Manual state the criterion as the intended future use of an object, and the UK cites the ICRC Commentary for it. Canada puts it as an object the enemy uses or intends to use, and anchors that to its reservations on ratifying Additional Protocol I. Four authorities, one formulation.
Where the DoD Law of War Manual Differs
The DoD Law of War Manual states it differently. At § 5.6.6.1 it provides that “purpose means the intended or possible use in the future.” Its footnote points the reader to the Australian and UK manuals — the two that just say intended. The signal is illustrative rather than one of full support, and manuals footnote more loosely than law reviews do. But the extra word possible arrives without authority behind it.
The same provision defines nature, location and use as well, and each of those three definitions tracks the source it cites. Purpose is the one that adds a word its sources do not contain.
The illustrations show what the word possible does. Runways at a civilian airport may qualify as military objectives because they could be needed at once if military runways are put out of action. Bridges and tunnels may qualify because of the possibility that they would be used — and, the Manual adds, even though nothing is using them at that moment.
A reader might expect the DoD Law of War Manual to constrain this elsewhere, and two provisions look as though they might. At § 5.4.3.2 it forbids attacking an object on mere speculation about its possible current status. At § 5.5.3 it lists feasible precautions for verifying that a target is a military objective.
The precautions ask present-tense questions. Is this the object I meant to strike? What is it doing now? Does it respond to a summons or a warning? Useful questions, all of them — but none can test whether an object will be used for something it is not being used for yet.
The speculation rule is more interesting, because it does reach purpose. An object qualifying by purpose is a military objective now, so a decision to attack it rests on its current status, and speculation about that status is forbidden. But the same Manual defines purpose as intended or possible use. Possibility is much nearer to speculation than intention is. The Manual prohibits the reasoning at one section and admits it at another, and says nothing about how to tell them apart.
Two smaller signs point the same way. When the Annotated Supplement’s editors restated the test in their own words, they wrote it as the “intended or likely use in the future.” Likely is not possible. And Lattimer, writing in May, set the Manual’s phrase beside the accepted understanding of purpose as future intended use, then moved on to proportionality without pausing on the gap.
What an Attacker Must Establish
So what must an attacker actually establish? Three expert bodies answer, and they agree more than they appear to.
The ILA Study Group starts from the consequence: if the mere possibility that an object might be converted sufficed, almost no limits on target selection would survive. It then requires two things — clear indications that the enemy will use the object, and information objective enough for a reasonable commander to conclude that a specific object will contribute. It records a suggestion that intentions must rest on reasonable certainty rather than on worst-case contingency planning, and two sentences later adopts that standard itself, as “far more than mere speculation or conjecture.”
Oeter reaches the same requirement from the commentary side. Purpose, he writes, “involves more than the mere potential for use by a belligerent,” and there must be clear indications of concrete planning or intention. (Fleck, ed., The Handbook of International Humanitarian Law, 4th edn, at p. 178.)
The Tallinn Manual is often read as the permissive one. Its commentary to Rule 100 does say the law fixes no particular standard of likelihood, and asks instead what a reasonable attacker would decide. But the paragraph before requires the purpose to have crystallised, and its two examples are a party about to buy particular hardware for military use, and a party that has announced it will take over civilian satellite transponders. Named objects. Stated intentions.
Michael Schmitt presses the question of certainty further. He rejects the idea of a single threshold: the greater the risk of civilian harm if the attacker has misidentified the object, the less reasonable his determination becomes. That is a sliding scale, not a fixed line — and it has to be met rather than worked around.
A sliding scale governs how sure an attacker must be that an object will be put to military use. It does not change what he must be sure of. Growing more confident that some of Russia’s remaining refineries will be pressed into military supply never becomes confidence about the refinery in front of him. Ukraine’s General Staff naming the Ryazan refinery as supplying jet fuel for military action is a statement about one specific object; the other is a statement about a category.
In operational terms, purpose is a real criterion and it does real work — but the work is answerable rather than open-ended. An attacker relying on it should be able to say three things: which object will be used, which military function is moving to it, and why that function is moving to that object rather than to another. A fourth follows from the first three: the object is not yet in that use, because once it is, use has already done the job and purpose is unnecessary. Those questions have answers when the enemy has requisitioned a facility, when it has designated one as an alternate, or when it has said what it intends. They have no answers when the only thing the attacker can point to is that the alternatives are gone.
IV. The Limits of Convertibility
I use convertibility as shorthand for a particular piece of reasoning: that an object qualifies as a military objective because it could be put to military use if the need arose. The word is mine, not anyone else’s, and the reasoning is worth isolating because it is what the contested targets in the Ukrainian campaign actually turn on.
Start with what Casey Biggerstaff requires, because it is close to what I would require. Article 52(2), he writes, calls for a fact-intensive inquiry into whether a particular object makes an effective contribution, and objects qualify to the extent they bear an identifiable nexus to military operations. He is explicit that dependence does not automatically convert objects into military objectives as a matter of law: the question is one of evidentiary proof and common sense.
His argument about revenue follows that standard faithfully. The more heavily an adversary depends on a revenue stream, he reasons, the more likely it becomes that some of the infrastructure generating it can be shown to contribute effectively to military action. That is a claim about the probability that evidence exists. It does not dispense with evidence, and nothing in this post is directed at it.
His observations about purpose are built differently. Infrastructure previously devoted to commercial purposes, he writes, may increasingly qualify as a military objective by purpose “because it can be readily converted to military use.”
Biggerstaff’s two arguments thus differ in what they require. The revenue argument asks whether a facility can be shown to contribute effectively to military action, which calls for evidence. The purpose argument asks whether a civilian object can be readily converted to military use, which calls only for capacity. Capacity is what § 5.6.6.1 of the DoD Law of War Manual admits when it treats possible future use as sufficient.
A third argument looks at the sequence of the campaign itself. Because Russia’s finite, interconnected facilities suffer strategic bottlenecks, Biggerstaff suggests that the case for treating the remaining ones as objectives by purpose grows stronger as the campaign destroys the facilities that do supply Russian forces. That argument reasons from the shape of Russia’s oil network to the status of the facilities inside it. Article 52(2) AP I, by contrast, asks about the purpose of each facility.
That distinction is not new, and the sharpest statement of it comes from someone objecting to it. W. Hays Parks, then the U.S. Army’s senior law-of-war lawyer, complained in 1990 that Article 52 fails to appreciate the necessity of attacking target systems, and assumes more certainty about the connection between a segment of a target system and military operations than an attacker normally possesses. He was arguing about effective contribution to military action rather than the military purpose of an object, so the point does not transfer directly. But he read the provision as demanding object-level certainty, and he disliked it for precisely that reason.
Why the difference between the shape of a network and the status of an object residing in it matters in practice: a rising likelihood that some object in a class qualifies as a military objective never identifies which one. It cannot ground an attack on any particular member of the class. And purpose is predicated on the enemy’s intentions, so reasoning about what it would be rational for the enemy to do is not evidence of what the enemy intends.
Russia’s actual behaviour makes the point. Its response to mounting refinery losses, set out above, was to ration and to import. None of it was the conversion of a commercial refinery to military supply. That does not show conversion will not happen. It shows that it is contingent — and a purpose finding cannot rest on a contingency.
When a Campaign Expands Its Own Target Set
There is also a structural oddity in the sequencing argument. On that reasoning, destroying the facilities that supply Russian forces is what makes the remaining commercial ones the next candidates. A campaign’s success then enlarges the set of objects it may lawfully attack — unless what remains can still be shown to contribute to military action. So the question is what supplies the limit. If it is the evidence, the sequencing adds nothing, because the evidence would qualify the object to have a military purpose whenever it appeared. If it is the sequencing itself, there is no limit at all.
What the Purpose Criterion Can Support
None of this makes purpose an empty criterion. It works whenever the attacker can point to something about the facility itself, of the kinds set out in Section III. All of these are close to what Biggerstaff himself requires.
V. Sorting the Target Classes
The classes below sort facilities by the evidence available about them, not by what they nominally do. None of them carries a presumption in either direction. It is the only way of grouping these facilities that survives the object-by-object rule.
Start with why the obvious sorting fails. Refined product is fungible, so what a refinery produces does not by itself show who consumes it. A national refining system cannot be divided into military plants and civilian plants by looking at the output. Article 52(2) does not ask for that. It operates on objects, not molecules, and the question is always what is known about the facility in front of you.
Facilities with an Identified Military Supply Function
Where evidence ties a facility to supplying the armed forces, it qualifies by use, and the purpose criterion is not needed. Schmitt states it directly: a refinery supplying fuel to enemy forces is a military objective by use, whatever revenue it also generates. Ukraine has named facilities in these terms — the Ryazan and Samara refineries, the Poltavska depot, the Koltsevoy pipeline.
The ILA Study Group works the same distinction through its own example: an oil production facility that both generates revenue for the war effort and provides fuel for the armed forces. It is the second function that makes the object a military objective.
Facilities with No Facility-Specific Evidence
That is the class this post is about. Where nothing is known about the particular facility beyond its place in a national system, qualification depends on purpose, and the standard set out in Section III applies. Nothing in the class label decides it.
Export Terminals and Pipelines
Crude export terminals and pipeline infrastructure raise the revenue question, on which the prevailing view is against qualification, as Section II set out. Ust-Luga, Primorsk and the Druzhba pumping stations sit here.
But the label does not settle the case. Biggerstaff observes that limited redundancy in Russia’s Soviet-era infrastructure may mean some export facilities also handle fuel destined for military consumption. Where that can be shown, the facility belongs in the first class and qualifies by use. The class is a starting point for asking what evidence exists, not an answer.
Feedstock for Munitions Production
Facilities supplying feedstock to weapons manufacture raise the nature and use questions rather than the purpose question. The Annotated Supplement lists munitions factories among military objectives. The route matters less here, because the evidentiary link to armaments production is usually direct.
Why the Classes Cannot Become Categories
There is no third category between military objective and civilian object. The DoD Law of War Manual is explicit: objects used by both the armed forces and the civilian population, the sort usually called “dual-use”, are either military objectives or they are not, and there is no intermediate legal status. The shorthand is convenient and it flattens the very distinction the classes above are drawn to preserve.
Michael Schmitt presses the same point from the other side. What must be assessed is the individual target, not sectors, systems, industries or networks; Ukraine cannot conclude that Russia’s energy sector is a military objective. And qualifying one bridge does not relieve an attacker of assessing the next. The same holds for the twelfth refinery.
Not everyone agrees. Dinstein argues it would be imprudent to disregard the possibility that civilian airports and ports become hubs of military operations, noting that ports can readily be adapted to naval use, and citing both the Eritrea-Ethiopia Claims Commission on a civilian airport and the Cultural Property Protocol’s generic reference to an aerodrome or a port. That is qualification by category, and it is the strongest case for it.
Two things answer it. The first is register: Dinstein is describing what belligerents do and predicting what they will do, not stating what the rule requires. The second comes from the American side of the argument. Biggerstaff closes his own study by observing that targeting law does not operate through rigid taxonomies or categorial presumptions and is better applied contextually, case by case. Oeter makes the same point from treaty design, noting that Article 52(2)’s abstract definition has a distinct advantage over the enumerated lists proposed before it.
How wide the range of respectable opinion still is can be seen from where the two ends sit. Dinstein endorses the view — quoting Green, and putting it as something that can well be argued — that oil installations of every kind are open to destruction, with only filling stations in residential areas away from major highways exempt. Hathaway, Khan and Revkin propose, as reform, that industries contributing only indirectly should not be targeted at all, naming the oil industry save for fuel used specifically for military purposes. Two serious treatments, one class of facility, opposite ends of the range.
VI. What This Means for the Next Target
What Qualifies the Next Target
In operational terms, what qualifies an object under the purpose criterion is what is known about that object. Not the stage the campaign has reached, and not how much of the alternative capacity has already been destroyed.
Schmitt puts the consequence sharply: qualifying one bridge does not relieve an attacker of assessing the next. The twelfth facility stands on its own footing, and the eleven before it establish nothing about it.
The evidence that answers the question is evidence about the facility, and Section III set out what that looks like. What it does not include is the narrative of the campaign. That an adversary is running short of refining capacity is a fact about the adversary. It is not a fact about the refinery in front of you.
The Divergence Nobody Warns About
Whether an object qualifies is a question every targeting adviser knows he has to answer. What follows is a problem he has no particular reason to look for — and one that does not arise in Ukraine at all. Russia and Ukraine are fighting without coalition partners, so no question of divergent national readings arises between them. But the purpose criterion is applied by States that do plan and fly together, and there the divergence has consequences that only appear when two advisers compare notes.
Section III set out where the manuals part company: four manuals and commentaries requiring intended use, and the DoD Law of War Manual admitting possible use. What matters here is a different question — which of those materials a given adviser actually has in front of him.
The German manual is the instructive case. is the instructive case. It restates the treaty term, Zweckbestimmung, and stops. It does not say that potential use is insufficient to qualify an object as a military objective, or that clear indications of planning are required. A German adviser can find both propositions, but he has to reach past his own manual to find it, in Oeter’s commentary or in the ICRC’s.
So two advisers can look at the same facility, apply their materials correctly, and reach different answers. One asks what the enemy evidently plans to do with it. The other applies a manual that admits possible future use, and asks what could be done with it if the need arose. Neither has misread anything.
Neither is likely to be warned, either. The Annotated Supplement does warn commanders in coalition operations that some allies and partners do not accept that war-sustaining objects are military objectives. That warning covers the divergence the literature already knows about. There is none about the purpose criterion — and on the analysis in Sections III and IV, that is the criterion on which the harder cases turn.
The problem of differing interpretations is not new, and it was identified from the American side. W. Hays Parks wrote in 1990 that many provisions of Additional Protocol I had been drafted in language allowing different nations to come away with different ideas of what each was intended to mean, and he gave the coalition case directly: air assets belonging to one nation supporting ground troops of another, with the two disagreeing about how a provision should be read. His example concerned the presumption in Article 52(3) rather than the purpose criterion, so it does not illustrate the divergence discussed here. The structure of the problem is the same.
My earlier post on lawful targets concluded that practitioners advising forces in coalition need to know which reading their partners apply. That was said of the war-sustaining divergence, and it holds equally here.
A divergence that a manual flags can be planned around. One that no manual mentions has to be noticed first.
Conclusion
The purpose criterion is neither vague nor unusable. An attacker relying on it can say which object the enemy will use for military action, which function is moving to that object, and why it is moving there rather than somewhere else. Those are answerable questions, and answering them is what the criterion asks.
Convertibility reasoning does not meet that description. It asks what could be done with a facility if the need arose, and the answer is available for almost any facility, which is why it cannot be the test. The looser formulation is not an invention of the commentators who reach for it. It is in the DoD Law of War Manual, which states that purpose means the intended or possible use in the future — a word its own cited authorities do not contain.
None of this puts much distance between me and Casey Biggerstaff. He asks whether a particular object can be shown to contribute, and so do I. The disagreement is confined to two sentences about purpose, and to what happens when their reasoning is applied across a campaign that has been running for years.
What that reasoning leaves behind is a criterion that expands as a campaign succeeds, and a divergence between allies that no manual warns anyone about. The first is a problem for the law. The second is a problem for whoever is advising the next mission, and he is unlikely to hear about it from his own materials.