LNG Carrier Thermal Signature Analysis for Cargo State Estimation
Spaceborne thermal infrared sensors can detect the cold-surface emissivity signature of cryogenic LNG cargo tanks, offering a probabilistic estimate of carrier load state when optical freeboard and AIS data are absent or unreliable.
Sensors
- Landsat 9 TIRS-2: Two thermal bands (Band 10 at 10.6–11.2 µm, Band 11 at 11.5–12.5 µm) at 100 m native resolution, resampled to 30 m in products. Revisit 16 days at the equator. Sufficient to resolve a large QFLEX or Q-Max carrier (~290–345 m LOA) as a thermally distinct object, but sub-pixel mixing with surrounding sea surface limits single-vessel attribution at smaller sizes.
- ECOSTRESS (ISS-mounted): Five TIR bands spanning 8–12.5 µm at approximately 70 m × 38 m pixel size. Non-sun-synchronous orbit from the ISS provides variable overpass times, useful for catching night-time thermal contrast when solar heating of the hull is absent. Revisit is irregular, typically 1–5 days at mid-latitudes, which limits systematic monitoring.
- ASTER TIR: Five thermal bands (8.125–11.65 µm) at 90 m resolution with on-demand tasking. Archive extends to 1999. Useful for retrospective analysis of known anchorage positions. Revisit is 16 days and tasking lead times make it unsuitable for near-real-time operations.
- Sentinel-3 SLSTR: Dual-view thermal channels at 1 km resolution. Far too coarse to isolate individual vessels, but provides sea-surface temperature context for atmospheric correction and can flag anomalous thermal clusters in dense anchorage fields where multiple laden carriers aggregate.
What a cryogenic tank surface actually radiates
LNG is stored at approximately −163 °C and atmospheric pressure. The insulated cargo containment system, whether a Moss spherical tank or a membrane-type design, maintains that temperature throughout a voyage. Even with multi-layer insulation, the outer tank surface or dome is measurably colder than the surrounding steel hull, which equilibrates toward ambient sea-surface temperature over time. The temperature differential between a laden tank dome and the adjacent hull plating can reach several degrees Kelvin under calm, clear conditions, though real-world figures depend heavily on insulation quality, solar loading, and ambient air temperature.
This differential is what thermal infrared sensors are hunting. Emissivity of the steel hull and the insulation cladding differ slightly, adding a secondary spectral signal on top of the temperature contrast. In ballast, the tanks warm toward ambient over a period of days as residual boil-off gas is consumed or vented, so a vessel that has been empty for 48 hours or more will show a progressively reduced cold signature. That decay curve is itself informative: it constrains how recently the vessel discharged cargo.
Boil-off management as an inadvertent broadcast
Modern LNG carriers re-liquefy boil-off gas or burn it in dual-fuel engines rather than venting it. The re-liquefaction plant generates waste heat that exits through conventional cooling systems, producing a localised warm signature at the stern machinery spaces. A laden vessel running re-liquefaction will therefore show a cold forward signature from the tanks and a relatively warm aft signature from the machinery. That spatial pattern, cold bow, warm stern, is a structural indicator of laden status that differs from the more uniform thermal profile of a ballast vessel.
Flaring or venting events, now rare on modern tonnage under MARPOL constraints, would produce a transient hot-spot signature detectable even at coarse resolution. More commonly, gas combustion in the main engine produces exhaust plume temperatures that are detectable in the mid-wave infrared, though that signal is not unique to LNG carriers and is covered separately in the vessel exhaust monitoring use case.
Where the method works and where it breaks down
The honest answer is that current civilian thermal satellites are marginal for single-vessel cargo-state attribution. At 100 m resolution, a standard 270 m LNG carrier occupies roughly 2–3 pixels in the along-track direction. Sub-pixel thermal mixing with the sea surface, which itself varies by several degrees across a scene, degrades the temperature retrieval. Cloud cover is a hard blocker: TIR cannot penetrate cloud, and many high-traffic LNG routes, particularly the North Atlantic and the South China Sea in monsoon season, carry persistent cloud fractions above 60 percent.
Atmospheric water vapour introduces retrieval uncertainty of 1–3 K even after split-window correction using SLSTR or dual-band TIRS data. For a temperature differential that may only be 3–6 K to begin with, that uncertainty is significant. The method is most reliable for large vessels (Q-Flex class and above) at tropical anchorages with clear skies, where the cold tank signature stands out against a warm, stable sea-surface background. For smaller LNG carriers below about 150 m LOA, the signal is likely to be unresolvable from orbit with current open sensors.
Combining thermal data with freeboard measurement from SAR or optical imagery substantially improves confidence. A vessel showing both a cold tank signature and a low freeboard is almost certainly laden. The two methods are complementary rather than redundant, and neither alone is sufficient for high-confidence attribution.
Fusion workflow: from pixel temperature to cargo probability
A practical analytical pipeline begins with atmospheric correction of the TIR scene using concurrent meteorological reanalysis data (ERA5 or NCEP) to estimate column water vapour. The corrected brightness temperature is then compared against the local sea-surface temperature derived from Sentinel-3 SLSTR to compute the vessel-to-background thermal anomaly. Pixels coincident with the AIS-reported vessel position are extracted and the spatial pattern of the anomaly is compared against a simple bow-cold, stern-warm template.
The output is a probability score, not a binary laden or ballast classification. A score above a defined threshold, calibrated against vessels of known status where AIS departure and arrival data are unambiguous, triggers a laden flag. Below the threshold, the result is recorded as indeterminate. Satellize applies this type of multi-source fusion in its analytics work, drawing on open constellations to produce structured outputs rather than raw imagery. The Tonga crop-estimation programme uses a similar probabilistic layer-combination approach, though the physics of crop reflectance differs substantially from cryogenic thermal signatures.
Archive depth matters here. Landsat 9 data is available from 2021 onward; ASTER extends to 1999. For retrospective investigation of a specific vessel's trading history, an analyst can in principle reconstruct cargo-state estimates at each historical overpass where cloud cover was low and the vessel position is known from AIS records.
Practical applications and their honest limitations
The primary buyers for this analysis are commodity trading desks seeking independent verification of LNG cargo flows when AIS data is dark or suspect, and government agencies monitoring compliance with sanctions that restrict LNG exports from specific origins. A vessel that has disabled its transponder near a sanctioned terminal but reappears at a receiving port with a thermal signature consistent with a laden state provides circumstantial evidence of a transfer that AIS alone cannot confirm.
Insurance underwriters and P&I clubs have a secondary interest: knowing whether a vessel claiming to be in ballast is actually carrying cargo affects risk assessment for certain routes. Port state control authorities could use the method to flag vessels for inspection where declared cargo state appears inconsistent with observed thermal signatures.
The limits bear repeating. This is probabilistic evidence, not proof. Courts and regulators will require corroborating data. The method cannot determine cargo origin or volume, only approximate load state. And it is entirely dependent on overpass timing relative to cloud cover, a factor that cannot be controlled with open-access sensors operating on fixed 16-day revisit cycles. Commercial tasking of higher-resolution airborne TIR sensors, or future spaceborne systems with finer ground sampling, would improve attribution confidence considerably.
Typical figures
| Spatial resolution (best available) | ~70 × 38 m (ECOSTRESS); 100 m native / 30 m resampled (Landsat 9 TIRS-2); 90 m (ASTER TIR) |
| Revisit period | 16 days (Landsat 9, ASTER); 1–5 days irregular (ECOSTRESS from ISS); near-daily at coarse resolution (Sentinel-3 SLSTR at 1 km) |
| Spectral bands used | Thermal infrared 8–12.5 µm; split-window pairs at ~10.8 µm and ~12 µm for atmospheric correction |
| Temperature retrieval uncertainty | ±1–3 K after split-window atmospheric correction, depending on water vapour column |
| Minimum detectable vessel size | Approximately 270 m LOA (Q-Flex class) for reliable sub-pixel thermal anomaly detection; smaller vessels marginal |
| Hard blockers | Cloud cover (TIR opaque to cloud); high sea-state spray affecting surface emissivity; solar loading equalising hull-tank differential |
| Archive depth | Landsat 9 from 2021; ASTER from 1999; ECOSTRESS from 2018 |
| Latency (open data) | Landsat 9: typically 12–24 hours post-acquisition via USGS EarthExplorer; ECOSTRESS: 1–3 days via NASA Earthdata |
| Delivery formats | GeoTIFF thermal anomaly layers, vessel-level probability score tables, PDF intelligence reports with scene thumbnails |
Analytics Satellize can run
| Cargo-state probability score per vessel overpass | Split-window brightness temperature retrieval, spatial anomaly pattern matching against bow-cold stern-warm template | Structured CSV or JSON feed: vessel MMSI, overpass timestamp, thermal anomaly delta (K), laden probability score (0–1) |
| Multi-source load-state confirmation | Fusion of TIR thermal score with SAR-derived freeboard estimate; Bayesian combination of independent likelihood scores | Per-vessel intelligence card combining thermal and freeboard evidence, flagged as laden, ballast, or indeterminate |
| Retrospective trading history reconstruction | Archive TIR scene retrieval (Landsat, ASTER) cross-referenced with historical AIS position logs at each clear-sky overpass | Timeline chart of inferred cargo state per voyage leg, exported as GIS layer or PDF report |
| Anchorage thermal cluster alert | Sentinel-3 SLSTR scene differencing to identify anomalous cold-cluster signatures in known LNG anchorage polygons | Email or API alert when a statistically significant cold cluster appears at a monitored anchorage, with scene thumbnail |
| Dark-vessel cargo-state inference | TIR analysis applied to vessels absent from AIS at time of overpass, position estimated from SAR detection; thermal score assigned independently of transponder data | Intelligence report linking SAR-detected vessel position to thermal cargo-state estimate, with confidence interval |
| Boil-off decay curve modelling | Multi-date TIR time series fitted to exponential thermal equilibration model to estimate time since cargo discharge | Estimated hours since discharge with uncertainty range, included in vessel-level report |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.