An LNG terminal runs three accounts at once: volume, mass, and energy. The LNG tank gauging system measures the first directly, infers the second through density, and supports the third through composition. At the end of the month somebody has to show that the numbers add up — and “add up” does not mean a residual of zero. It means a residual small enough to sit inside a documented uncertainty budget, with a sign and size that can be explained.
Where the missing kilograms actually go
A taxonomy, roughly in order of how often each one turns out to be the answer.
Composition and density. LNG in the tank is not the LNG that was loaded. Methane boils off preferentially, so the liquid left behind gets heavier over time — the weathering effect. If the terminal uses the ship’s composition certificate for weeks after discharge, the inventory mass drifts even though every instrument is working. Density from composition (ISO 6578) is only as good as the sample it starts from; sampling LNG properly is genuinely difficult, and ISO 8943 exists because of it.
Temperature averaging. Density depends on temperature, and LNG in a tank is stratified. A single-point or short-string measurement of a stratified tank gives an average that does not match the actual mass. Multi-point sensing exists precisely for this; a six-point string on a 30 m tank is not a luxury, it is the difference between a balance and a guess.
Boil-off gas that is not metered. BOG sent to the regasification process as fuel, to the flare, back to a ship’s tanks, or vented is part of the balance. Terminals that estimate BOG from tank pressure rise and call it “negligible” are usually carrying their entire residual in that one term. Meter it or model it explicitly, and keep the model visible.
Timing and settling. A tank that has just received a cargo is not in equilibrium: the liquid is still moving, gas is still being displaced, and the level is still settling. Measurements taken during or immediately after transfer are comparably worthless. Define a settling time in the operating procedure and hold to it — commonly several hours, longer after a full discharge.
Level measurement in cryogenic service. Radar, servo, and differential pressure each have failure modes that a calibration certificate will not reveal: foam and two-phase layers at the surface, stillwell effects, tank shell deformation under thermal cycling, and the fact that a floating-roof-style correction does not exist here. A 3 mm level error on a large tank is small in inventory terms, but it is not zero, and it compounds when the same error appears at both ends of the window.
Unmeasured returns and transfers. Vapour return lines, ship-to-shore connections that are purged and vented, line fill and drain-down between transfers, and heel left in a ship’s tanks after discharge. Each is small; together they frequently account for a residual that has been blamed on the gauge for a year.
Real losses. Flange leaks, valve passing, relief valve lifting, and sampling itself — a sample bottle takes mass out of the LNG tank gauging system.
Reading the residual like a diagnostician
A monthly number tells you almost nothing. The shape of the residual over time tells you almost everything.
- Constant sign, constant magnitude. A bias. Something systematic: a stale composition, a level offset at one datum, a BOG route assumed but not measured.
- Random, zero-mean, growing with throughput. Noise from the instruments themselves. This is the acceptable case, and its size should match the uncertainty budget.
- A step change. Something happened: a new cargo supplier, a resampled composition, a changed settling time, a recalibrated gauge, a valve that started passing.
- Seasonal drift. Higher BOG in summer, more frequent cargoes in winter. If the residual tracks ambient temperature, the BOG accounting is the suspect.
- Residual that grows with inventory. Points at level or tank geometry rather than at flow meters.
Two habits make this readable. First, run a daily balance, not a monthly one — a monthly residual hides four weeks of compensating errors. Second, keep a rolling 30-day figure alongside the monthly, and plot both against the tolerance band. A residual that walks slowly out of band is a failing instrument; one that jumps is an event.
A reconciliation workflow that survives an audit
- Snapshot before and after. Fixed settling time, all instruments time-stamped, tank pressure and temperature recorded with the level.
- Receipts per cargo, with the ship figure and the shore figure recorded side by side — never one replacing the other.
- Deliveries per route: send-out, truck loading, bunkering, each with its own meter and its own uncertainty.
- BOG per route, measured where possible, modelled where not, and labelled as modelled.
- Inventory calculated twice — from level/temperature/density, and from a second independent level principle. A disagreement between the two is a finding, not a rounding error.
- Residual computed, signed, and compared to the band. Inside the band: record and move on. Outside: open a case with a hypothesis, not a correction.
- Monthly roll-up with the 30-day trend, the uncertainty budget attached, and a short narrative of any case opened.
- Quarterly review of the budget itself: are the assumed uncertainties still true? Has anything been recalibrated, resampled, or re-routed?
The single most important rule: never adjust a measurement to close the balance. Adjust the model, the assumption, or the procedure — and write down which.
Failure patterns worth recognising
- Using the ship’s figure as truth. The ship and the shore are two independent measurements with two uncertainty budgets. When they disagree, that is information, not an error to be corrected.
- Plugging the gap with density. The easiest number to nudge is the one nobody can independently verify. It is also the number that corrupts every energy calculation downstream.
- Ignoring BOG because it is “small”. At a large terminal, BOG as fuel can be a material fraction of throughput; at a small one it can dominate the residual.
- Measuring during transfer. Produces a residual that looks like a leak and disappears next month.
- One temperature point. Guarantees a density error that varies with stratification, which varies with cargo history — so it looks random and is actually systematic.
- No raw data retention. Turns a solvable discrepancy into a commercial argument you lose.
- Calling the residual “loss” before diagnosing it. Most residuals are measurement artefacts. Some are real losses. They have completely different remedies, and the label determines which one you go looking for.
A terminal that closes well has a boring monthly report: the residual sits in the lower half of its uncertainty band, the rolling trend is flat, BOG is metered on every route, composition is refreshed per cargo, the two independent level measurements agree within a stated tolerance, and the reconciliation file contains a one-page uncertainty budget that has not changed in a year except when something physical changed.
None of that is exotic. All of it depends on the LNG tank gauging system being treated as a measurement chain rather than a display — because the mass balance does not close at the screen. It closes at the sample point, the temperature string, the BOG meter, and the clock.