In short: Only routine flaring is an economic decision, since safety and upset flaring exist to protect the plant, so the first step is splitting site volumes into those three categories. The World Bank's Global Gas Flaring Tracker has recorded global flaring at roughly 140 to 150 billion cubic metres a year for over a decade, with its 2025 edition reporting about 151 billion cubic metres for 2024, and that persistence through several price cycles points at infrastructure and regulation rather than commodity prices. The options at a stranded site rank by capital cost, running from a gathering tie-in through on-site power, compression, small scale liquefaction and gas to liquids, and the ranking usually turns on distance to the nearest connection. Site level flared volume is frequently an estimate rather than a measurement, so the number the economics run on is worth checking before the case is built.
A well pad brings up oil and, with it, gas that has to go somewhere. The nearest gathering tie-in is forty kilometres away over ground that nobody has surveyed for a right of way. The regional processing plant has been running at inlet capacity for two years. The flare has been lit since first oil, and every barrel the pad produces has been profitable enough that no one has had a strong reason to change the arrangement.
Flared gas is product that was cheaper to burn than to move. That has held at thousands of sites for decades, and what has shifted recently is the regulatory clock now attached to the volume, along with the recognition that a corridor of sites can carry an investment no single site can.
The World Bank's Global Gas Flaring Tracker has put global flaring at roughly 140 to 150 billion cubic metres a year for more than a decade, and its 2025 edition reported about 151 billion cubic metres for 2024, the highest annual figure in its series since 2007. Nine countries have accounted for around three quarters of that total for most of the period. The persistence of the number through several complete price cycles tells you that this is an infrastructure and regulation problem rather than a commodity price problem.
Why the flare is still burning
Gathering infrastructure has to be justified against the gas alone. The oil already has a route out, by truck or by an existing line, so the pipe you are contemplating carries no incremental oil revenue. Its cost scales with distance, terrain and right of way acquisition. Its revenue scales with volume. At a pad producing a few hundred thousand standard cubic feet a day, forty kilometres from a tie-in, that ratio has never worked at any gas price that has existed.
Associated gas also arrives on the oil's schedule. The operator is drilling for oil, the gas comes up in a ratio set by the reservoir, and that ratio rises through field life as pressure declines. The volume at any given site is a by-product whose quantity the operator does not choose and whose growth the original facilities plan did not anticipate. A pad designed with a small separator and a flare tip sized for start-up conditions ends up flaring a stream three times larger five years later.
Processing capacity lags production for a structural reason. Gas has to be dehydrated, sweetened where necessary, and have its liquids removed before it meets pipeline specification. Plants that do this are built on multi-year lead times against a forecast of what the basin will produce. Drilling responds to price in months. The two cycles do not line up, and the gap between them appears as flaring at the wellhead when there is no line, and at the plant inlet when there is a line and no capacity behind it.
A fourth cause is organisational. The operator holding the flare is frequently not the party who would build the gathering system, and that party needs commitments from several operators who each want to see the line before committing. The coordination failure keeps corridors unbuilt long after the aggregate volume in them would support a line.
Routine flaring is the only part that is an economic decision
The World Bank's definition, adopted with its Zero Routine Flaring by 2030 initiative in 2015, is specific: routine flaring is flaring during normal oil production in the absence of sufficient facilities or amenable geology to reinject the gas, use it on site, or send it to a market. That definition is doing real work, because it separates the volume that responds to investment from the volume that does not.
Safety flaring is a design function. A pressure relief system that routes to a flare exists so that an overpressure event ends in a controlled burn rather than in a rupture. You do not want that volume reduced, you want the relief system to work when it is called on, and any programme that counts safety flaring in its reduction target has mixed a hazard control into a commercial metric.
Upset flaring is a reliability outcome. A compressor trip, an instrument air failure, an unplanned unit shutdown, a plant that goes down and pushes its inlet gas to the flare. The volume can be large and it is genuinely wasteful, and no gathering investment touches it, because the gas had somewhere to go and the equipment failed. That volume is addressed through equipment reliability and shutdown avoidance, which is a maintenance and integrity problem with its own methods (N11).
The practical consequence is that a flaring number reported as one total is close to useless as a basis for capital allocation. Split last year's volume by cause at the site level before anything else. The split tells you which sites have an infrastructure problem and which have a reliability problem, and those go to different people with different budgets. A site with high upset flaring will not improve because you built it a pipeline.
The options, ranked by what they cost to build
Reinjection for pressure support. Where compression already exists on site and the reservoir will accept the gas, this is the lowest incremental capital of the four. The gas stays in the reservoir and does work there: maintaining pressure raises the recovery factor on oil, which is the higher value product. The value appears as incremental barrels over subsequent years rather than as gas revenue now, and it has to be discounted accordingly. Whether the reservoir is amenable is a subsurface question with a site-specific answer, and where the answer is no, this option does not exist at any price.
On-site power generation. Reciprocating engines or small turbines burning field gas to displace diesel or a grid import. Capital is moderate, lead time is measured in months, and the gas is valued at whatever it displaces, which at a remote site running on trucked diesel is a high number per unit of energy. The binding constraint is that on-site electrical load is usually small relative to the gas available, so this converts a fraction of the volume. Gas quality matters more here than elsewhere, since a rich stream or one carrying H2S needs conditioning ahead of the engine and that conditioning is part of the capital.
Small-scale liquefaction or compression. Converting the gas into something a truck can carry, either compressed into tube trailers or liquefied in a small modular plant. Capital is higher than power generation and the economics turn almost entirely on trucking distance and on having a buyer inside it. This is the option that fits where a market exists and a pipe does not, and it is sensitive to the delivered price at the buyer in a way the other options are not.
Gathering into a processing and sales system. The highest capital, the longest lead time, and the highest value per unit if it closes, because it puts the gas into a real market and captures the liquids barrel alongside it. Cost is dominated by distance and right of way. Revenue is dominated by the aggregate volume the line can pick up along its route, which is why this option is almost never evaluated correctly at a single site.
Ranking by capital intensity is a useful first pass because it also ranks by lead time and reversibility. A skid-mounted generator set installed in nine months at a pad declining over five years is a different proposition from a line that takes three years to permit and build against the same profile.
Valuing what you recover
The common error is to value recovered volume at a hub price. That price applies to processed gas with a route to market, and what you have is an unprocessed stream at a wellhead with no route, so the two are not comparable without adjustment.
Value the volume against the alternative it displaces at that location, which gives a different figure for each option. For gathering, the netback is the hub price less gathering and processing fees, transport, shrinkage and fuel, plus the liquids credit if you retain the barrel. On a rich stream that liquids credit is frequently the larger half of the value, and because it prices off crude rather than gas it changes the risk profile of the investment: you are underwriting a pipeline with a revenue stream correlated to oil.
For power generation, the value is avoided diesel or avoided grid import at the site, which per unit of energy usually sits well above any gas netback. That single fact explains why generation projects close at sites where gathering studies have repeatedly failed.
For liquefaction or compression, it is the delivered price at the buyer less the conversion cost, trucking, and whatever you lose to boil-off and return logistics. For reinjection, it is the incremental recovery times the oil price, discounted over the years those barrels arrive, less compression fuel and whatever gas sale you gave up to inject.
Two adjustments belong in every one of these cases. Decline is the first: volume falls over the payback period, and a case built on flat volume overstates the answer in a way that worsens the longer the payback. The carbon term is the second, and it goes in where a compliance price actually applies to the flared CO2 or a fee applies to unburned methane, and stays out where it does not. Putting a shadow carbon price into a case in a jurisdiction that has none, then presenting the result as economics, is how flaring projects get approved and quietly cancelled.
The volume in your case is probably an estimate
Very few flares are metered. Most reported flare volumes come from a mass balance across the separator, from a design assumption about what the tip is passing, or from a satellite retrieval.
The World Bank tracker infers volume from satellite radiant heat using VIIRS observations processed by the Earth Observation Group, calibrated against reported volumes. That method suits national and basin totals and does not suit underwriting a site investment, because the calibration is population-level and the per-site error is wide.
A second gap sits at the tip itself. Reporting conventions commonly assume around 98 percent destruction of the hydrocarbon passing through a flare. Plant and colleagues, writing in Science in 2022, measured effective destruction across three United States basins at roughly 91 percent, with unlit and poorly operating flares accounting for most of the shortfall. Measuring and reconciling that gap belongs to a methane programme rather than to this decision (N4). What it means here is that the volume you think you are burning and the volume actually leaving the stack are different numbers, and both feed the business case.
A flare meter costs a small fraction of any of the four options above and converts an argument into a measurement. On a site under serious consideration, metering ahead of the study is cheap insurance against a case built on the wrong number.
Where this stops
At a marginal site the arithmetic frequently does not close on gas value alone. The volume is small, the distance is long, the decline is real, and no valuation method turns that into a positive net present value. Presenting it as though it does loses credibility with a capital committee that has seen the same study before.
So the decision usually turns on one of two things. Regulatory obligation is the first: a permit with an expiry, a routine flaring ban with a date attached, or a fee per unit flared. Any of these converts an optional investment into a cost of continuing to produce, and the comparison becomes the project against shutting in the oil, which is a much larger number and usually resolves quickly.
Aggregation is the second. One pad cannot carry a trunk line. Twelve pads inside a corridor with a shared route frequently can, and the unit economics improve with each tie-in added. That turns a site decision into a network design problem: which pads, which route, what compression, what build order, and what the line is worth if the anchor volume declines faster than the forecast said. It is a capital planning exercise that needs credible volumes from every site in the corridor, which is where the measurement problem returns as the binding constraint on the analysis.
A third possibility is that the right builder is somebody else. A midstream party serving several producers can aggregate across ownership boundaries in a way none of them can individually, and the producer's decision becomes a commercial one about commitment terms and volume dedication rather than a capital one.
Start by splitting last year's flared volume into routine, safety and upset at the site level, then rank only the routine sites by volume against distance to the nearest tie-in. That ranking usually collapses a list of two hundred sites into a handful and one corridor, which is the only version of this problem worth modelling in detail.