Offshore oil and gas platform production status from optical and SAR activity signals
Combining Sentinel-1 SAR vessel detection, VIIRS night-band flare radiance, and high-resolution optical imagery lets analysts classify offshore platforms as producing, on standby, or decommissioned, and flag unannounced curtailments before they appear in official statistics.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground range resolution in IW mode; 6-day repeat at mid-latitudes, 12-day at equator. Detects platform structures and vessel attendance regardless of cloud or darkness. Vessel detection uses CFAR (constant false-alarm rate) algorithms on backscatter intensity.
- VIIRS Day/Night Band (Suomi-NPP / NOAA-20): 750 m nadir pixel; nightly global coverage. The DNB detects gas flares and deck lighting at radiances well above background ocean. VIIRS Nightfire (Colorado School of Mines / NOAA) retrieves flare temperature and radiant heat from multi-band fitting, but cannot resolve individual platform structures at this resolution.
- Planet SkySat: 0.5 m resolution optical; tasked on demand. Resolves individual deck equipment, crane positions and mooring lines. Revisit is task-dependent rather than systematic, so it suits confirmation passes rather than routine monitoring.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral. The highest commercially available optical resolution for platform deck characterisation. Useful for distinguishing active equipment from cold or decommissioned structures, but cloud cover and tasking cost limit frequency.
- Landsat 8/9 OLI (USGS/NASA): 30 m multispectral, 15 m panchromatic; 16-day repeat per satellite, 8-day combined. Detects large flares in shortwave infrared (Band 6, 1.57 µm) during daylight, and resolves clusters of platforms in mature fields. Free archive to 1972 provides long baselines for decommissioning analysis.
Three signals, one classification
No single sensor tells the whole story. SAR sees the platform structure and any vessel alongside it, day or night, through cloud. VIIRS DNB sees whether the flare is lit and whether deck lighting is on. High-resolution optical, when cloud permits, sees whether cranes are rigged, whether the helideck is clear, and whether equipment is in a configuration consistent with active operations. Fused together, these three layers produce a classification that is considerably more reliable than any one of them alone.
The classification scheme has three states: producing, standby, and decommissioned. A platform is classed as producing when at least two of the following are true: a supply or crew vessel is detected within 500 m in SAR imagery on multiple recent passes; VIIRS DNB registers radiance above a field-specific threshold at the platform location; and high-resolution optical shows deck activity consistent with operations. Standby means the structure is present but vessel attendance is absent and flare radiance is at or near background. Decommissioned means the SAR backscatter signature has changed materially, consistent with jacket removal or topsides lift.
What a floating roof gives away, and what this page does not cover
This page is specifically about platform operational status, not about production rate. The method can tell you that Platform Alpha shifted from producing to standby between Tuesday and Thursday. It cannot tell you whether Alpha was producing 10,000 barrels per day or 40,000 before it went quiet. Rate estimation from remote sensing requires thermal or flare-radiance calibration against known production figures, and that is covered separately in the gas flaring volume estimation page.
Similarly, this analysis does not address what is stored on the platform or in any associated FPSO. FPSO heading and mooring status, and laden-versus-ballast discrimination, are handled on their own pages in this library.
Supply vessel attendance as a production proxy
A producing offshore platform requires regular crew rotation and consumables. In most operating regimes, supply or platform support vessels call every few days. Sentinel-1 IW mode at 10 m resolution resolves vessels longer than roughly 20 m against the ocean background using CFAR detection. A vessel parked alongside a platform, or within the platform's 500 m safety zone, is a strong positive signal. Absence of vessel attendance over multiple consecutive Sentinel-1 passes, typically spanning one to two weeks at mid-latitude revisit rates, is a meaningful negative signal.
The limit is that some platforms are supplied by helicopter alone, particularly in benign-weather regions, and will show low vessel attendance even when fully operational. This ambiguity is real and should be flagged in any delivery. Cross-checking against VIIRS flare presence reduces the false-standby rate substantially.
VIIRS night-band radiance: honest about what 750 metres can and cannot do
The VIIRS Day/Night Band has a nadir pixel size of 750 m. A single large flare on an isolated platform is detectable; a cluster of platforms within a few kilometres of each other produces a blended radiance signal that cannot be disaggregated to individual structures without additional information. In mature fields such as the North Sea or the Gulf of Mexico, this is a genuine limitation. The VIIRS Nightfire algorithm (Colorado School of Mines, published and freely available via eogdata.mines.edu) fits a Planck function to the multi-band signal to retrieve flare temperature and radiant heat, but the spatial attribution problem in dense fields remains.
Nightly coverage is the compensating advantage. Every platform on Earth is observed every 24 hours in darkness. For monitoring unannounced curtailments, a sudden disappearance of a previously stable radiance signal at a known platform location is a high-confidence event, even if the absolute radiance figure carries uncertainty. Time series of 30 or more nightly observations establish a reliable baseline; deviations of more than two standard deviations from that baseline are worth investigating.
Detecting unannounced curtailments before the statistics arrive
Official production statistics from national regulators or operators are typically published with a one- to three-month lag. A satellite-derived status change is detectable within days of the event. For commodity traders, production-sharing agreement auditors, or government regulators monitoring contractor compliance, that lead time has direct value.
The detection workflow is straightforward in principle. A platform's historical SAR, VIIRS and optical record establishes a baseline activity profile. Automated comparison of new acquisitions against that baseline flags anomalies. Human review confirms whether the anomaly is consistent with a planned maintenance shutdown, an announced curtailment, or an unannounced production stop. Planned shutdowns typically appear in operator announcements and can be filtered; unannounced stops cannot. The method is most powerful when applied across an entire basin simultaneously, because correlated curtailments across multiple platforms in the same field are a stronger signal than any single platform going quiet.
Satellize runs this kind of multi-signal basin monitoring as a configured analytics product, drawing on the same open-constellation infrastructure used in its Tonga crop-estimation programme. The analytic architecture is the same; the physics and the sensors are different.
Archive depth and what history reveals
Sentinel-1A has been operational since April 2014; Sentinel-1B contributed until 2022. The VIIRS DNB archive on Suomi-NPP runs from 2012. Landsat 8 extends the optical record to 2013, with the broader Landsat archive reaching back to 1972 for coarser historical context. Together, these archives allow analysts to reconstruct the production history of a platform or field over a decade, identify seasonal patterns, and establish when a platform first showed signs of declining activity ahead of formal decommissioning notification.
High-resolution commercial imagery archives are shallower and patchier, because tasking was not systematic over most fields before roughly 2018. Planet SkySat and Pléiades Neo archives improve from 2020 onwards for many offshore regions. For any analysis requiring pre-2018 high-resolution optical, expect significant data gaps and plan the methodology accordingly.
Typical figures
| SAR spatial resolution (Sentinel-1 IW) | 10 m range × 10 m azimuth (after multi-look processing) |
| SAR revisit (Sentinel-1, mid-latitudes) | 6 days single-satellite; 12 days at equator |
| VIIRS DNB pixel size | 750 m nadir; degrades to ~1,600 m at swath edge |
| VIIRS temporal coverage | Nightly global; archive from October 2011 (Suomi-NPP) |
| High-resolution optical resolution | 0.5 m (SkySat), 0.3 m panchromatic (Pléiades Neo); cloud-limited |
| Minimum detectable vessel (SAR) | Approximately 20 m length against open-ocean background in Sentinel-1 IW |
| Flare detection threshold (VIIRS DNB) | Radiant heat detectable to roughly 10–15 MW thermal equivalent in published Nightfire studies; isolated flares only in dense fields |
| Classification latency | 12–48 hours after satellite overpass, depending on processing pipeline and data availability |
| SAR archive depth | Sentinel-1: April 2014 to present; ERS/Envisat extends context to 1991 at coarser resolution |
| Delivery formats | GeoJSON platform status layer, CSV time-series feed, PDF basin report, alert webhook |
Analytics Satellize can run
| Platform operational status classification | Multi-source fusion: CFAR vessel detection on Sentinel-1 + VIIRS DNB radiance thresholding + optical deck-state assessment | GeoJSON layer with per-platform status (producing / standby / decommissioned) updated on each new SAR pass |
| Unannounced curtailment alert | Time-series anomaly detection against 30-observation VIIRS and SAR baseline; z-score threshold trigger | Alert webhook or email within 48 hours of anomaly confirmation across two independent sensors |
| Basin-wide activity heatmap | Aggregated vessel attendance counts and VIIRS radiance sums across all platforms in a defined polygon, weekly cadence | Weekly PDF basin report with trend charts and platform-level change flags |
| Historical production-status reconstruction | Retrospective fusion of Sentinel-1, VIIRS DNB and Landsat SWIR archives; change-point detection on activity time series | Per-platform activity timeline CSV with annotated change points, covering 2014 to present |
| Decommissioning progress tracking | SAR backscatter change detection for structural removal; optical confirmation of topsides lift or jacket removal | Monthly GIS layer update with decommissioning stage classification and supporting imagery chips |
| Vessel attendance frequency report | CFAR vessel detection on all available Sentinel-1 passes over a named platform set; attendance rate calculated as vessel-present passes divided by total passes | Tabular report of attendance frequency per platform per month, with raw detection confidence scores |
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.