Crude oil storage volume estimation from floating-roof tank shadow geometry
Floating-roof tanks cast a shadow whose length encodes fill level. High-resolution optical satellites measure that shadow, turning sun geometry into inventory intelligence for traders, insurers and sovereign monitors.
Sensors
- Maxar WorldView-2 / WorldView-3: Panchromatic resolution of 0.31 m (WorldView-3) and 0.46 m (WorldView-2) allows shadow width measurement on tanks as small as 30 m diameter. WorldView-3 also carries a SWIR bundle useful for surface characterisation. Revisit at a given point is roughly 1 to 4.5 days depending on latitude and tasking priority.
- Airbus Pleiades 1A / 1B: 0.5 m native panchromatic resolution, daily revisit globally when both satellites are tasked together. Consistent radiometric calibration across the pair makes shadow-width measurements reproducible between acquisition dates.
- Planet SkySat: 0.5 m panchromatic, with a constellation of around 21 satellites enabling same-day tasking in many regions. Lower signal-to-noise than WorldView at this resolution, so shadow detection is more reliable on larger tanks (diameter above roughly 40 m).
- Maxar WorldView Legion: Next-generation constellation designed for sub-30 cm resolution and revisit rates up to 15 times daily at mid-latitudes. Increases the probability of catching a cloud-free pass during daylight hours with a favourable solar elevation angle, which is the binding constraint for this method.
What a floating roof gives away
A floating-roof tank is designed to minimise vapour loss: a steel or aluminium pontoon sits directly on the crude surface and rises or falls with it. That mechanical fact creates an unintended signal. When the roof is depressed below the top of the tank wall, the wall casts a shadow across the roof. The shadow's width is a deterministic function of three knowable quantities: the tank's inner diameter, the solar elevation angle at the moment of imaging, and the vertical distance between the roof surface and the tank rim.
Because solar elevation angle is calculable to arc-second precision for any location and acquisition timestamp, and because tank inner diameter can be measured from the same image or from prior survey, the only unknown is roof depression depth. Solving for that depth gives fill fraction directly. The method requires no ground sensor, no operator report and no manifest. It reads the physics of the scene.
The trigonometry, and where it breaks
The core relationship is straightforward. If the solar elevation angle is α and the measured shadow width across the roof is s, then the rim-to-roof depth d satisfies d = s × tan(α). Fill fraction follows from d and the tank's known working height. Published remote-sensing research has validated this approach against operator-reported inventory data at major storage hubs, finding volume estimation errors typically in the range of 5 to 15 per cent depending on image resolution and tank diameter.
The method has honest limits that any serious user must understand. First, shadow width shrinks as the sun climbs: at solar elevations above roughly 60 degrees the shadow becomes too narrow to measure reliably at 0.5 m resolution. Conversely, at elevations below about 20 degrees the shadow spills beyond the roof onto the tank wall, breaking the simple geometry. The practical acquisition window for most storage latitudes is a solar elevation band of approximately 25 to 55 degrees. Second, any cloud cover over the tank farm voids the measurement entirely. Third, fixed-roof tanks, which are common for lighter refined products, produce no floating-roof signal at all. Fourth, tanks with external floating roofs that have partially sunk or are undergoing maintenance introduce ambiguity that cannot be resolved from imagery alone.
From pixel to barrel: the measurement chain
Processing starts with orthorectification and pan-sharpening of the tasked image. Each tank is identified by its circular footprint; inner diameter is measured at sub-pixel precision using edge-detection on the tank wall. The shadow boundary on the roof is then located, again by edge-detection, and shadow width is averaged around the illuminated arc to reduce noise from surface irregularities or floating debris.
Solar geometry parameters (elevation, azimuth) are extracted from the image metadata or computed from the acquisition timestamp and tank coordinates using standard ephemeris. Depth d is computed per tank. Multiplying by the tank's cross-sectional area gives a volume in cubic metres, which converts to barrels at 6.2898 barrels per cubic metre. Across a tank farm with, say, 20 measurable floating-roof tanks, the aggregate inventory estimate can be compared with prior acquisitions to derive a change figure, which is often more useful to a commodity analyst than the absolute level.
What this tells a trader or insurer that AIS cannot
AIS-based tanker tracking, covered separately in this library, tells you what crude is moving on the water. It says nothing about what is sitting in onshore storage. The two data streams are complementary: a surge in tanker arrivals at a terminal that is not accompanied by rising tank levels implies either rapid onward distribution or a discrepancy worth investigating. That gap is where the shadow method earns its keep.
For insurers, the application is different. A policy written against declared stock volumes at a storage facility can be cross-checked against satellite-derived estimates at renewal or after a loss event. The method does not replace an independent surveyor's dip measurement, but it provides a rapid, low-cost screen that flags facilities where declared volumes appear inconsistent with observed roof positions across multiple acquisition dates. Repeated inconsistency is a material underwriting signal.
Sovereign energy ministries use similar logic for strategic reserve monitoring: knowing whether a counterparty's declared reserve levels are plausible is a different kind of intelligence from knowing the exact barrel count, and satellite geometry delivers the former reliably.
Practical constraints on coverage and frequency
Commercial tasking of WorldView or Pleiades at a single tank farm costs a fraction of a physical inspection, but it is not free, and cloud probability matters enormously. A facility in the Persian Gulf or US Gulf Coast typically has clear-sky probability above 70 per cent in summer months, making weekly revisit feasible. A facility in the North Sea or coastal China in winter may have clear-sky probability below 30 per cent, meaning monthly reliable coverage is a more honest expectation.
Archive depth is a practical asset. WorldView-2 has been operational since 2009 and WorldView-3 since 2014, so historical inventory reconstruction is possible for major storage hubs going back over a decade. That archive enables backtesting of trading signals and supports litigation or regulatory investigations requiring historical stock-level evidence.
Satellize structures tasking schedules around cloud climatology for each facility, combining archive mining with forward tasking, to maximise the number of usable acquisitions per quarter. The approach mirrors what we developed for agricultural monitoring in the Tonga crop-estimation programme, adapted here for industrial geometry rather than spectral indices.
What the method cannot do
Volume estimation from shadow geometry gives fill fraction, not product identity. A tank whose roof is at 60 per cent could hold crude, condensate or fuel oil; the optical image does not distinguish. Combining shadow-derived levels with manifest data or thermal infrared signatures (covered in the refinery throughput page in this library) narrows the ambiguity but does not eliminate it.
Tanks smaller than about 20 m in diameter are difficult to measure reliably at 0.5 m resolution because the shadow width at typical solar angles is only a few pixels. Underground or partially buried tanks produce no signal. And any facility that deliberately schedules maintenance or roof-cleaning during satellite overpass windows can introduce noise, though sustained manipulation across many tanks and many dates is operationally implausible.
Used honestly, the method is a probabilistic screen, not a certified measurement. Its value is in identifying outliers, tracking trends and providing an independent check on reported figures, not in replacing the auditor.
Typical figures
| Spatial resolution (panchromatic) | 0.31 m (WorldView-3), 0.46 m (WorldView-2), 0.5 m (Pleiades, SkySat) |
| Minimum tank diameter for reliable measurement | ~30 m at WorldView-3 resolution; ~40 m at 0.5 m resolution |
| Usable solar elevation window | Approximately 25° to 55°; outside this range shadow geometry becomes unreliable |
| Cloud requirement | Clear sky over facility; thin cirrus may be tolerable but optically thick cloud voids acquisition |
| Typical volume estimation error | 5 to 15% of tank capacity, depending on resolution, tank diameter and image quality |
| Revisit frequency (commercial tasking) | 1 to 7 days depending on constellation, latitude and cloud climatology |
| Archive depth | WorldView-2 from 2009; WorldView-3 from 2014; Pleiades from 2012 |
| Applicable tank types | External floating-roof and internal floating-roof tanks only; fixed-roof tanks yield no signal |
| Delivery latency (tasked acquisition to report) | Typically 24 to 72 hours after cloud-free acquisition |
| Output formats | Structured data table (barrels per tank), GeoJSON facility layer, time-series CSV, PDF summary report |
Analytics Satellize can run
| Per-tank fill fraction and volume estimate | Shadow-width measurement with solar-ephemeris correction; trigonometric depth inversion | Structured data table giving tank ID, diameter, shadow width, solar elevation, estimated depth, fill fraction and volume in barrels, delivered as CSV or JSON |
| Facility-level aggregate inventory | Summation across all measurable floating-roof tanks at a named terminal or tank farm | Single aggregate figure with per-tank breakdown and confidence interval, in a PDF or machine-readable report |
| Inventory change time series | Repeat-acquisition differencing of fill-fraction estimates across dated image archive | Time-series chart and CSV showing estimated stock build or draw per facility, with acquisition dates and cloud-flag annotations |
| Declared-versus-observed discrepancy flag | Comparison of satellite-derived volume estimates against operator-declared or publicly reported figures | Discrepancy report flagging facilities where satellite estimate differs from declared level by more than a configurable threshold, for underwriting or compliance review |
| Historical inventory reconstruction | Archive trawl of WorldView and Pleiades holdings for a named facility, processed retrospectively | Multi-year time series of estimated fill levels, suitable for backtesting trading signals or supporting legal and regulatory investigations |
| Cloud-adjusted tasking schedule | Cloud climatology analysis per facility location to maximise usable acquisition rate | Recommended tasking calendar with expected clear-sky acquisition probability per month, delivered as a planning document ahead of contract commencement |
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.