Wind farm construction monitoring and turbine installation counting
Offshore wind construction moves fast and often under overcast skies. SAR point-target detection and AIS vessel tracking give independent, cloud-proof counts of installed turbines and installation pace from first monopile to final commissioning.
Sensors
- Sentinel-1 SAR (C-band, ESA): 5 x 20 m ground range resolution in Interferometric Wide Swath mode; 6-day revisit at the equator, 1-3 days at mid-latitudes with both satellites. Monopile foundations and erected tower-nacelle assemblies produce bright point-target returns (radar cross-section well above background sea clutter) that persist regardless of cloud or darkness, enabling automated turbine counting.
- Planet SkySat (optical): 0.5 m native resolution, tasked on demand. Resolves nacelle presence, tower erection stage and foundation type. Revisit is tasking-dependent rather than systematic; useful for confirming ambiguous SAR detections and producing high-confidence installation-stage classification.
- Airbus Pléiades Neo (optical): 0.3 m resolution, stereo and tri-stereo collection available. At this resolution, tower height can be estimated from shadow length and stereo parallax, giving a proxy for turbine model class. Latency from tasking to delivery is typically 24-48 hours.
- Spire Global AIS (RF from orbit): Spire operates more than 100 LEO satellites receiving AIS transponder signals. Jack-up vessels, heavy-lift crane barges and cable-lay ships transmit Class A AIS; their MMSI identifiers, positions and speeds are publicly documented. Vessel presence at a turbine coordinate is a leading indicator of imminent or ongoing installation activity.
What a monopile gives away before the turbine arrives
A steel monopile driven into the seabed is, from C-band radar's point of view, a near-perfect corner reflector. Its cylindrical geometry and the flat water surface around it create a double-bounce return that stands 15-25 dB above the surrounding sea clutter in Sentinel-1 imagery. That signature appears the moment the pile is installed, weeks or months before the tower, nacelle and blades follow. Automated point-target detection algorithms, well established in the published literature on offshore wind SAR analysis, can therefore count foundations independently of construction vessel schedules or developer disclosures.
Onshore sites are noisier. Agricultural fields and industrial terrain produce competing bright returns, so foundation detection relies more on change detection: comparing pre-construction baselines against current imagery and flagging new persistent scatterers at planned turbine coordinates. The method works, but requires a clean baseline and tolerates more false positives than the offshore case.
The installation sequence and how each sensor tracks it
Offshore turbine installation follows a repeatable sequence: seabed survey, monopile driving, transition piece fitting, tower section stacking, nacelle lift, blade attachment. Each stage has a distinct radar and optical signature. Monopile alone: bright SAR point, no optical tower shadow. Tower present: elongated shadow in high-resolution optical, slightly larger SAR return. Nacelle fitted: shadow terminus widens. Blades attached: three-armed shadow pattern visible in sub-metre optical imagery on clear days.
Pléiades Neo at 0.3 m can resolve blade attachment on a clear pass. SkySat at 0.5 m can confirm nacelle presence. Neither can see through cloud, which is the chronic problem for North Sea and East Asian offshore sites where cloud cover exceeds 70% of days in winter months. SAR fills that gap. The practical workflow combines SAR for systematic counting and optical for stage classification when conditions allow.
Onshore, the tower shadow method is more reliable than offshore because land backgrounds are stable. A 100 m tower at solar noon at 50° latitude casts a shadow roughly 85 m long. That is detectable in Sentinel-2 at 10 m resolution as a one-pixel feature, but the geometry is marginal. SkySat or Pléiades Neo make stage classification unambiguous.
Vessel activity as a construction clock
Jack-up installation vessels are among the most distinctive ships in the AIS record. Their MMSI numbers are publicly registered, their dimensions are known, and their operating pattern when installing turbines is characteristic: slow transit to site, hours-long stationary hold at a turbine coordinate, then movement to the next position. A vessel stationary for four to eight hours at a foundation coordinate almost certainly indicates an active lift or pile-driving operation.
Spire's orbital AIS collection provides global coverage with message latency of minutes to a few hours depending on satellite pass timing. Cross-referencing vessel position history against the foundation grid of a wind farm produces an installation timeline with date-stamped events at each turbine location. This is independent of developer announcements and can precede official commissioning notices by weeks.
One honest caveat: AIS can be delayed, spoofed or simply absent if a vessel operator turns off the transponder, though that is rare for large, insured installation vessels operating in regulated waters. SAR vessel detection provides a cross-check: Sentinel-1 can detect ships above roughly 20-30 m length in open water, and a jack-up barge is typically 150 m or more.
What this analysis cannot do
SAR point-target counting is reliable for steel monopiles in open water but degrades near port infrastructure, existing platforms or dense arrays where targets merge in range or azimuth. At 5 x 20 m Sentinel-1 resolution, two foundations separated by less than 500 m in the range direction can appear as a single elongated return rather than two discrete targets. Commercial SAR constellations such as ICEYE or Capella, operating at 0.5-1 m spotlight resolution, resolve this, but systematic revisit at that resolution requires tasking budgets that scale with site size.
Installation-stage classification from shadow geometry requires solar elevation above roughly 20 degrees and cloud-free conditions. In high-latitude winter, both conditions fail simultaneously for weeks at a time. The SAR count remains valid, but the stage label becomes uncertain. Any reporting should carry an explicit confidence flag distinguishing 'foundation confirmed by SAR' from 'full turbine confirmed by optical'.
Neither method gives turbine model, rated capacity or hub height with precision. Shadow-derived height estimates from stereo optical carry uncertainties of 5-10 m depending on baseline geometry and image quality. Capacity estimates require combining height proxies with publicly available turbine specification databases, introducing additional model uncertainty.
From raw detections to a construction progress feed
The analytic product that matters to most clients is not a single count but a time series: how many turbines were installed per week, which foundations are still bare, and what the implied completion date is given current pace. Building that requires a persistent detection archive, a foundation grid derived from planning documents or first-pass SAR mapping, and a state machine that assigns each foundation a status label updated on each satellite pass.
Satellize structures analytics of this type as GIS layers with per-turbine status attributes, updated on a defined cadence and delivered via API or file drop. The methodology draws on the same change-detection and point-target extraction principles used in the company's crop-estimation work for the Kingdom of Tonga, adapted for industrial point targets rather than agricultural spectral indices. Clients typically want alerts when a site crosses installation milestones (25%, 50%, 75% complete) rather than raw imagery, which is what the state-machine approach produces.
For investors and commodity desks, the derived output is often simpler still: a single number, updated weekly, representing confirmed installed capacity in megawatts, using the public nameplate rating of the turbine model where known. That figure, aggregated across a portfolio of sites, feeds directly into power-generation forecasting models.
Typical figures
| SAR spatial resolution (Sentinel-1 IW mode) | 5 m (range) x 20 m (azimuth) ground range detected |
| Optical resolution (Pléiades Neo / SkySat) | 0.3 m / 0.5 m respectively, tasked on demand |
| Sentinel-1 revisit (mid-latitudes, both satellites) | 1-3 days |
| AIS message latency (Spire orbital) | Minutes to a few hours depending on satellite pass |
| Minimum detectable target (SAR, offshore monopile) | Steel cylinder >1 m diameter; returns 15-25 dB above sea clutter |
| Minimum vessel length detectable by SAR | Approximately 20-30 m in open water (Sentinel-1) |
| Shadow-derived tower height uncertainty | ±5-10 m depending on stereo baseline and solar geometry |
| Sentinel-1 archive depth | From 2014 (operational); earlier test data from 2014 onwards |
| Delivery formats | GeoJSON / GeoPackage turbine status layers, CSV installation timelines, API feed, PDF milestone reports |
| Coverage | Global; Sentinel-1 systematic acquisition covers all coastal and most land areas |
Analytics Satellize can run
| Turbine installation count by foundation | SAR point-target detection and persistent scatterer identification (published method class: constant false-alarm rate detection on sigma-naught imagery) | GeoJSON layer with per-foundation status (bare / tower / nacelle / blades) updated on each Sentinel-1 pass |
| Installation-stage classification | Shadow geometry analysis and object classification in sub-metre optical imagery | Attribute table appended to foundation GIS layer; updated on cloud-free optical tasking |
| Construction vessel activity log | AIS trajectory clustering and stationary-hold detection at foundation coordinates | CSV event log with vessel MMSI, foundation ID, arrival/departure timestamps and inferred operation type |
| Weekly installation pace metric | Differencing of sequential turbine count time series | Structured data feed (JSON) with turbines-per-week and rolling four-week average; milestone alerts at 25/50/75/100% completion |
| Implied installed capacity estimate | Foundation count multiplied by nameplate rating from public turbine specification databases, with model uncertainty range | Weekly capacity estimate in MW with confidence interval, delivered as CSV or API endpoint |
| Baseline foundation grid mapping | First-pass SAR change detection against pre-construction imagery, cross-referenced with planning consent documents | Foundation grid GeoPackage used as the persistent reference for all subsequent monitoring passes |
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.