Power-plant operating status from thermal discharge plumes
Thermal infrared satellites detect the warm-water discharge plumes that power stations release into rivers, lakes and coastal waters, revealing whether a plant is generating at any given moment. Landsat 8/9 TIRS and ECOSTRESS resolve the 8–15 °C above-ambient signal at 70–100 m, providing a public, verifiable record of operating status.
Sensors
- Landsat 8 TIRS / Landsat 9 TIRS: Band 10 (10.6–11.2 µm) at 100 m native resolution (resampled to 30 m in products) with a 16-day repeat per satellite; the two satellites together achieve an 8-day revisit at mid-latitudes. Radiometric precision of approximately 0.1 K enables detection of discharge plumes well above the 8–15 °C above-ambient signal typical of condenser cooling water. Archive extends to 1984 for trend analysis.
- ECOSTRESS (ISS-mounted): Five thermal infrared bands (8.28–12.13 µm) at 70 m resolution. Non-sun-synchronous orbit from the International Space Station provides variable overpass times, including night acquisitions that eliminate solar-reflection contamination of the thermal signal. Revisit is irregular, roughly 1–5 days depending on latitude, which limits it to opportunistic confirmation rather than systematic monitoring.
- Sentinel-2 MSI: No thermal band, but 10 m optical imagery at 5-day revisit (two satellites combined) resolves cooling-tower vapour plumes and surface turbulence at discharge outfalls on cloud-free days. Useful as a corroborating signal and for plant infrastructure mapping, not for temperature measurement.
- Planet SuperDove: 3–5 m optical imagery with near-daily revisit. Resolves visible surface turbulence and discolouration at discharge outfalls, and clearly shows cooling-tower plume shadows and steam columns. No thermal capability; acts as a high-cadence optical check on Landsat thermal detections.
What a discharge plume actually tells you
Every thermal power station that uses once-through or open-cycle cooling returns water to a river, lake or coastal inlet at temperatures 8 to 15 °C above ambient. That range is not an estimate: it follows from the thermodynamic efficiency limits of steam-cycle plant, where roughly two-thirds of fuel energy is rejected as waste heat. A 1 GW coal unit running at full load may discharge 40 to 60 cubic metres of water per second at elevated temperature. The plume that forms is large, persistent and physically unavoidable.
The operational inference is direct. A plume present and at expected temperature means the condenser is running, which means turbines are turning. A plume absent, or collapsed to ambient, means the plant is offline or on hot standby. Partial plume extent can indicate part-load operation, though translating plume geometry to megawatt output requires calibration against known operating data for each facility. The method does not require any cooperation from the plant operator, any grid-reporting disclosure, or any financial filing.
How Landsat TIRS resolves the signal
Landsat 8 and Landsat 9 each carry the Thermal Infrared Sensor, which images in Band 10 at 10.6–11.2 µm. The native detector pitch yields 100 m ground resolution, resampled to 30 m in distributed products. Radiometric calibration targets a noise-equivalent temperature difference of approximately 0.1 K, far below the several-degree contrast of an active discharge plume against background water. A single Landsat scene covers 185 km × 185 km, so a single pass captures an entire river basin and multiple plants simultaneously.
With Landsat 8 and 9 flying in the same orbit plane offset by 8 days, combined revisit at mid-latitudes is 8 days. That cadence is adequate for weekly operating-status assessments but will miss short outages of a day or two. Cloud cover is the binding constraint: a persistent overcast lasting several weeks, common in monsoon seasons or high-latitude winters, can produce data gaps that no processing can fill. Analysts should plan for 30–50% usable-scene rates in cloudy climates and supplement with ECOSTRESS night passes where available.
Nuclear plants present a different geometry
Nuclear stations are worth treating separately. They operate at higher thermal loads relative to electrical output because their steam cycles run at lower temperatures than coal or gas plant, pushing thermal efficiency down to roughly 33%. Discharge plumes from large nuclear stations are correspondingly larger and warmer per unit of electrical output. Several coastal and riverine nuclear stations in Europe and Asia have well-documented plume signatures in the Landsat archive going back to the 1980s, making them useful calibration benchmarks.
One practical complication: nuclear plant operators sometimes reduce output during summer low-flow periods to comply with river-temperature regulations, not because of any operational or economic problem. A shrinking plume at a nuclear station in July may reflect regulatory compliance rather than a generation shortfall. Contextual knowledge of local environmental rules is therefore part of competent analysis, not an optional extra.
Cooling towers change the observable, not the physics
Plants with closed-cycle or hybrid cooling towers reject much of their waste heat to the atmosphere rather than to a water body. The thermal discharge plume shrinks or disappears. What replaces it is a vapour plume from the tower stack, visible in Sentinel-2 and Planet optical imagery as a white column casting a shadow on clear days. The shadow geometry can indicate whether the tower is operating at load.
Some heat still reaches the water: even closed-cycle towers have a blowdown discharge to manage dissolved solids, and the blowdown is warm. The signal is weaker, perhaps 2–4 °C above ambient rather than 10–15 °C, and requires careful atmospheric correction to isolate. For plants with purely dry cooling, water-body thermal analysis fails entirely. Knowing the cooling-system type for each facility before analysis begins is not optional.
Limits that matter for commodity intelligence
The method is well-suited to answering a binary question: is this plant running or not? It is less suited to precise megawatt estimation. Plume extent depends on river flow rate, wind speed, ambient temperature and mixing, all of which vary independently of plant output. Published academic work has demonstrated statistically significant correlations between plume area and generation at specific well-characterised plants, but those correlations are site-specific and degrade when generalised.
Atmospheric correction is a further source of uncertainty. Landsat TIRS Band 10 retrieves at-sensor brightness temperature, which must be corrected for atmospheric water-vapour absorption to yield surface temperature. The USGS provides a Surface Temperature product using a radiative transfer approach, but residual errors of 1–2 K are common in humid conditions. For a plume that is 10 °C above ambient, that error is tolerable. For a blowdown discharge that is 3 °C above ambient, it is not.
Satellize runs systematic thermal plume monitoring using Landsat and ECOSTRESS archives for clients tracking generation assets across multiple jurisdictions. The Overhead column has discussed the method in the context of energy-market intelligence.
Building an operating-status time series
A useful intelligence product is not a single image but a consistent time series: plant-by-plant, scene-by-scene, with each observation classified as active, inactive, or obscured by cloud. Against that record, analysts can identify unannounced outages, seasonal maintenance patterns, and capacity ramp-ups ahead of public disclosure.
The Landsat archive back to 1984 makes it possible to reconstruct decades of operating history for any plant that has been in service that long. That historical depth is genuinely unusual in commodity intelligence. Cross-referencing the thermal record with publicly filed grid-dispatch data, where available, allows calibration and validation. Where dispatch data are not public, the satellite record stands alone as the primary evidence.
Typical figures
| Thermal spatial resolution | 100 m native (Landsat 8/9 TIRS Band 10); 70 m (ECOSTRESS) |
| Optical corroboration resolution | 10 m (Sentinel-2 MSI); 3–5 m (Planet SuperDove) |
| Revisit (thermal) | 8 days combined Landsat 8+9 at mid-latitudes; irregular 1–5 days for ECOSTRESS |
| Minimum detectable thermal anomaly | Approximately 1–2 °C above ambient after atmospheric correction in favourable conditions; 3 °C practical floor in humid climates |
| Spectral band (primary) | 10.6–11.2 µm (Landsat TIRS Band 10); 8.28–12.13 µm across five ECOSTRESS bands |
| Scene coverage per pass | 185 km × 185 km (Landsat); 400 km × 6 km swath (ECOSTRESS, ISS ground track) |
| Archive depth | Landsat thermal back to 1984 (Landsat 5 TM); ECOSTRESS from 2018 |
| Cloud sensitivity | Thermal and optical both blocked by cloud; expect 30–50% data loss in persistently cloudy climates |
| Latency (open data) | Landsat scenes typically available within 12–24 hours of acquisition via USGS EarthExplorer |
| Delivery formats | GeoTIFF temperature anomaly layers, GIS-ready vector plume extents, CSV time-series tables, PDF monitoring reports |
Analytics Satellize can run
| Binary operating-status classification per plant per overpass | Threshold detection on atmospherically corrected surface temperature anomaly relative to upstream or background water temperature | Time-stamped status table (active / inactive / cloud-obscured) as CSV or GIS attribute layer |
| Plume extent and thermal contrast mapping | Connected-component segmentation of pixels exceeding a temperature threshold above local ambient; plume area and maximum ΔT extracted per scene | GeoTIFF plume mask and vector polygon per acquisition, with ΔT statistics |
| Multi-year operating-history reconstruction | Full Landsat archive stack processed scene-by-scene; cloud-flagged observations excluded; status time series compiled per facility | Annual and monthly operating-rate summaries per plant, with chart-ready CSV |
| Unannounced outage alert | Automated comparison of current-scene status against rolling 90-day baseline; alert triggered when active plant shows no plume on a cloud-free pass | Email or API alert with scene thumbnail and anomaly statistics within 24 hours of Landsat acquisition |
| Optical plume corroboration report | Cooling-tower vapour plume detection in Sentinel-2 or Planet true-colour imagery; shadow-length estimation as qualitative load indicator | Annotated image chips with qualitative load assessment, delivered as PDF or georeferenced PNG |
| Fleet-level generation-capacity monitoring | Aggregation of per-plant status classifications across a defined asset list; portfolio-level active-capacity estimate | Weekly or monthly portfolio dashboard (PDF or interactive web layer) covering all monitored facilities |
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.