Crude pipeline pump station activity from ancillary electrical and thermal proxies
Crude pipeline pump stations leave two detectable signatures from orbit: waste heat from motor housings and facility lighting during operations. Combining VIIRS night-time radiance with LWIR thermal anomaly detection can flag abrupt start-stop events, though neither signal alone confirms throughput.
Sensors
- VIIRS Day-Night Band (DNB): Suomi-NPP and NOAA-20 VIIRS DNB provides nightly radiance imagery at approximately 750 m ground sample distance, sensitive to low-level artificial light in the 0.5–0.9 µm panchromatic range. Facility lighting at active pump stations is detectable against dark rural backgrounds; the published noise-equivalent radiance floor is around 2 × 10⁻¹⁰ W cm⁻² sr⁻¹. Daily global coverage with a roughly 12-hour equatorial repeat makes it the primary temporal signal for detecting operational state changes.
- Landsat 8/9 TIRS: The Thermal Infrared Sensor on Landsat 8 and 9 images in Band 10 (10.6–11.2 µm) and Band 11 (11.5–12.5 µm) at 100 m native resolution (resampled to 30 m in products). Pump motor waste heat conducted through building roofs and dissipated by cooling systems produces surface temperature anomalies of several degrees Celsius above ambient. The 16-day exact repeat limits temporal resolution, but archive depth to 2013 (Landsat 8) allows multi-year baseline construction.
- ECOSTRESS (ISS-mounted): The ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station acquires LWIR imagery at approximately 70 m resolution in five thermal bands between 8.3 and 12.5 µm. ISS orbital precession produces irregular revisit (roughly 1–5 days at mid-latitudes, variable), which is a meaningful limitation for systematic monitoring but useful for opportunistic high-resolution thermal snapshots to validate Landsat-derived anomalies.
- Sentinel-1 SAR (C-band): Sentinel-1A/B provide C-band (5.405 GHz) backscatter imagery at 10 m resolution in Interferometric Wide Swath mode with 6–12 day repeat depending on latitude. SAR contributes contextual evidence: parked tanker trucks, pipe-rack shadows and equipment changes within the compound are detectable as backscatter pattern changes. SAR does not measure heat or light, so its role is corroborative, not primary.
What a pump station actually radiates
A crude pipeline pump station is, at its mechanical core, a set of large electric motors driving centrifugal or reciprocating pumps. Motors in the 1–20 MW range are common on major export lines. Even at high efficiency, a meaningful fraction of input power dissipates as heat through motor casings, gearboxes, mechanical seals and cooling circuits. That heat reaches building rooftops and the surrounding hardstand, producing a surface temperature signature that persists for hours after a station comes online.
Facility lighting is the second signal. Pump stations in remote areas run banks of sodium-vapour or LED floodlights whenever personnel are present or operations are under way. At night, these produce a localised radiance spike that VIIRS DNB can detect against the dark rural background typical of pipeline corridors in Central Asia, West Africa or the Middle East. The two signals are physically independent, which is precisely what makes their coincidence informative: a thermal anomaly without corresponding lighting, or lighting without heat, warrants scepticism.
Night-time radiance as an industrial activity proxy: what the published record shows
The use of VIIRS DNB radiance as a proxy for industrial activity has a documented track record in published remote-sensing literature. Researchers have correlated DNB radiance changes with oil-field production levels in the Bakken and Permian basins, with factory output in East Asia, and with conflict-driven economic disruption across the Middle East. The underlying logic is consistent: grid-connected industrial facilities consume more electricity when operating, and a portion of that consumption escapes as light.
Pump stations fit this model well in one respect and poorly in another. They are geographically fixed, isolated and geometrically small, which makes pixel-level attribution straightforward once a station location is known. The complication is that a single VIIRS pixel at 750 m may contain the station compound plus adjacent road infrastructure or small settlements, requiring careful baseline subtraction. Stations on pipelines that cross populated corridors are harder to isolate than those in open desert. Published detection thresholds suggest that facilities consuming more than roughly 1 MW of electrical power are consistently detectable in DNB data when skies are clear, though this figure varies with background radiance and lunar illumination phase.
Thermal inference: what the heat signature can and cannot tell you
LWIR thermal imagery from Landsat TIRS or ECOSTRESS can detect surface temperature anomalies of 1–3 °C above ambient at 70–100 m resolution under clear-sky conditions. A pump station running at full capacity will typically show elevated temperatures on motor building rooftops and around cooling equipment. A station that has been idle for several hours will have cooled toward ambient, particularly in climates with strong nocturnal radiative cooling.
The honest limits matter here. Cloud cover renders LWIR imagery useless, which is a serious constraint in tropical or monsoon-affected pipeline corridors. Landsat's 16-day revisit means that a station restarted and shut down again within that window may leave no thermal trace in the archive. ECOSTRESS improves temporal sampling but its ISS-derived orbit produces irregular coverage gaps. Neither sensor can distinguish between a station running at 30% capacity and one running at 90%; the thermal signal saturates well before it becomes a reliable throughput meter. The method is most defensible for binary state detection: on or off, active or dormant.
Building a monitoring architecture from mismatched revisit rates
The practical challenge is that the two primary sensors operate on very different cadences. VIIRS provides a nightly observation but at coarse spatial resolution. Landsat provides a sharp thermal image but only every 16 days. ECOSTRESS fills some of the gap but unpredictably. A sensible monitoring architecture treats VIIRS DNB as the trigger layer: a sustained change in nightly radiance at a known station location initiates a request for the next available LWIR acquisition to corroborate.
Sentinel-1 SAR adds a third, weather-independent layer. Changes in the backscatter pattern within a pump station compound, such as the arrival of maintenance vehicles or the repositioning of portable equipment, can be detected at 10 m resolution regardless of cloud or darkness. SAR cannot confirm the station is running, but it can confirm that human activity is occurring, which is corroborating context. Combining all three layers into a probabilistic state estimate is more defensible than relying on any single sensor, and it is honest about the fact that the result is an inference, not a measurement.
Satellize applies this multi-sensor inference approach on open constellations, with commercial tasking added where a client requires higher temporal density or tasked ECOSTRESS-class resolution.
Where the method earns its keep, and where it does not
The strongest use case is detecting abrupt operational changes on pipelines where throughput data is not publicly available: a station that has been dark for six months suddenly illuminates in VIIRS, and a Landsat pass three days later shows a clear thermal anomaly on the motor building. That combination is a credible signal of resumption. Commodity traders monitoring export routes, governments assessing sanctions compliance and infrastructure insurers tracking operational risk all have legitimate reasons to want that signal.
The method is far weaker as a continuous throughput estimator. There is no published physical relationship that reliably maps LWIR rooftop temperature or DNB radiance to barrels per day through a given station, because throughput depends on pump configuration, fluid viscosity, back-pressure and operating mode, none of which are visible from orbit. Analysts who present radiance or temperature as a throughput proxy without this caveat are overstating the method. The honest product is an operational state flag, not a flow-rate estimate.
Typical figures
| Primary radiance sensor | VIIRS DNB, ~750 m GSD, nightly global coverage |
| Primary thermal sensor | Landsat 8/9 TIRS Band 10, 100 m native (30 m resampled), 16-day repeat |
| High-resolution thermal (opportunistic) | ECOSTRESS, ~70 m, irregular ISS-derived revisit (1–5 days at mid-latitudes) |
| Contextual SAR | Sentinel-1 IW mode, 10 m, 6–12 day repeat depending on latitude |
| VIIRS DNB noise floor | ~2 × 10⁻¹⁰ W cm⁻² sr⁻¹; facilities >~1 MW electrical load typically detectable in dark rural settings |
| LWIR temperature sensitivity | Anomalies of 1–3 °C above ambient detectable under clear sky; cloud renders LWIR unusable |
| Archive depth | VIIRS: 2012–present; Landsat 8 TIRS: 2013–present; Landsat 9: 2021–present |
| Minimum detectable target | Isolated compound with active facility lighting; dense urban or industrial backgrounds degrade detection |
| Latency (open data) | VIIRS: 1–2 days; Landsat: 1–3 days after acquisition; ECOSTRESS: variable |
| Output state resolution | Binary (active/dormant); continuous throughput estimation is not supported by the physics |
Analytics Satellize can run
| Station operational state flag | VIIRS DNB radiance change detection against multi-month baseline, with lunar-phase and cloud masking applied | Nightly alert feed (JSON or GIS layer) flagging radiance exceedance at monitored station coordinates |
| Thermal anomaly confirmation | Landsat TIRS or ECOSTRESS surface temperature retrieval; anomaly scored against rolling seasonal ambient baseline | Per-acquisition thermal anomaly report with confidence score and cloud-cover flag |
| Multi-sensor state probability estimate | Bayesian combination of DNB radiance flag, LWIR anomaly score and SAR compound-activity indicator into a single operational probability | Weekly station status report with probability band and supporting sensor evidence |
| Historical state reconstruction | Archive VIIRS DNB and Landsat TIRS time-series analysis from 2013 to present, gap-filled with cloud-occurrence statistics | Multi-year operational state time series per station in CSV and GIS formats |
| SAR compound change detection | Sentinel-1 coherence and backscatter intensity change detection within defined station compound boundary | Change event log with acquisition date, change magnitude and annotated image chip |
| Pipeline corridor operational context | Aggregated state flags across all monitored stations on a named pipeline segment to infer segment-level activity pattern | Pipeline segment activity dashboard updated on each VIIRS pass |
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.