Utility-scale solar farm construction progress from optical time series
Photovoltaic and CSP projects leave a measurable spectral signature from first ground-break to energisation. Optical time series from Planet, Sentinel-2 and high-resolution tasking turn that signature into weekly construction curves, area estimates and schedule-risk flags.
Sensors
- Planet Dove (PlanetScope): 3–4 m native resolution, daily global revisit at the constellation level. Delivers the dense temporal stack needed to resolve week-by-week installation progress and detect brief weather windows between cloud events. Blue, green, red and near-infrared bands capture the low-NIR, low-red reflectance signature of installed panels.
- Sentinel-2 MSI: 10 m resolution in visible and NIR bands (20 m for red-edge and SWIR), 5-day revisit at mid-latitudes with twin satellites. Free and openly archived. Sufficient to track construction progress on projects above roughly five hectares; below that threshold individual panel rows are not reliably separable from background soil.
- Maxar WorldView-3: 30 cm panchromatic, 1.24 m multispectral. Used for milestone verification: confirming panel row count, measuring string spacing, and distinguishing installed panels from racking-only structures. Not suited to routine temporal monitoring due to cost and tasking latency, but decisive for contract-milestone disputes.
- Airbus Pléiades Neo: 30 cm panchromatic, 1.2 m multispectral, same-day stereo acquisition possible. Useful for volumetric checks on earthworks and berm construction alongside panel-installation tracking. Tri-stereo mode can produce a DSM for cut-and-fill verification.
What a panel row gives away spectrally
Silicon photovoltaic panels are engineered to absorb as much incoming radiation as possible. In the red band (roughly 630–690 nm) and near-infrared (760–900 nm), installed panels show markedly lower reflectance than bare soil, dry vegetation or concrete. This is the opposite of the high-NIR signature that makes green crops easy to find. A freshly graded solar site actually brightens in NIR as the topsoil is exposed; once panels go in, that signal reverses sharply. The transition is detectable in Sentinel-2 10 m imagery for any installation patch wider than two or three panel rows.
Concentrated solar power (CSP) plants with parabolic troughs or heliostats behave differently. Their reflective surfaces produce high-NIR returns and specular glint artefacts that require glint-angle correction before classification. The detection logic is essentially inverted relative to PV, but the temporal change signal is equally clear: a field of heliostats appears suddenly bright where bare earth was dark.
Reading the construction sequence from a time stack
A utility-scale solar project passes through recognisable phases, each with a distinct spectral and textural signature. Land clearance removes vegetation and homogenises surface colour; NDVI drops and bare-soil reflectance rises. Road and cable-trench grading introduces linear features visible at 3–4 m resolution. Racking installation creates a regular grid of small shadows without yet changing the bulk reflectance. Panel installation then drives the characteristic NIR suppression across each completed block. Finally, inverter stations and substation construction appear as compact high-reflectance structures adjacent to the array.
Planet Dove's daily cadence is what makes the construction curve legible rather than merely episodic. With Sentinel-2 alone, a five-day revisit combined with cloud cover can leave gaps of two to four weeks in tropical or monsoon-affected regions. Planet fills most of those gaps. The practical output is a time series of installed-panel area, updated weekly, that can be compared against a developer's published construction schedule or a lender's drawdown milestones.
Honest limits: cloud, resolution floors and ambiguity
Optical monitoring stops at cloud. In persistently overcast regions, such as parts of Southeast Asia during monsoon season, useful imagery may be available on fewer than 30 days per quarter. SAR coherence methods (covered in the pipeline construction sibling page) can partially compensate, but panel-row detection from SAR alone is less reliable than from optical because the dielectric and geometric properties of panels vary with tilt angle and look direction.
The Sentinel-2 resolution floor matters. At 10 m, a single pixel covers roughly 25 panel rows of a standard 2 m-pitch residential module, so sub-block installation progress is invisible. For projects below five hectares, Sentinel-2 gives only a coarse installed-versus-not signal. Planet Dove resolves individual rows on large-format commercial modules but still cannot count panels within a row. WorldView-3 or Pléiades Neo are required for panel-count verification. Automated classification also struggles to distinguish installed panels from large dark tarpaulins, shade netting or standing water in early site works; a human QA step on ambiguous detections is not optional.
From pixels to a drawdown curve
The analytic pipeline starts with a site boundary, either supplied by the client or digitised from the earliest clear image showing ground-break. Each new image is cloud-masked, atmospherically corrected to surface reflectance, and classified using a panel spectral index derived from NIR and red bands. The classified area is converted to megawatts-peak using a published or client-supplied capacity density figure, typically in the range of 0.04 to 0.08 hectares per megawatt-peak depending on module efficiency and ground-coverage ratio.
The resulting time series supports several downstream products. Lenders can compare cumulative installed capacity against drawdown schedules. Offtake counterparties can estimate first-power dates from the installation rate in the most recent weeks. Competitors or market analysts can aggregate dozens of sites to estimate near-term additions to national generation capacity, a figure that grid operators and commodity traders both value. None of this requires site access or cooperation from the developer.
Archive depth and what it resolves retroactively
Sentinel-2 archive runs from mid-2015 for Sentinel-2A and from 2017 for the combined constellation. Planet's commercial archive extends from 2016 for many regions. For projects that began construction before a monitoring contract was initiated, this archive allows a retrospective construction curve to be reconstructed, useful for post-hoc due diligence, insurance claims or regulatory compliance reviews.
Landsat 8 and 9 (30 m, 16-day revisit) extend the optical archive to 2013 and earlier via Landsat 7. At 30 m the panel signature is detectable only for very large projects, but the long baseline is valuable for confirming that a site was undisturbed agricultural or natural land before development. Satellize applies this multi-archive approach in its analytics work, including the Tonga crop-estimation programme, where distinguishing land-use transitions over multi-year windows is similarly central to the methodology.
What the data cannot tell you, and what to ask instead
Installed panel area is not the same as grid-connected capacity. A site can be fully panelled but awaiting a grid connection agreement, an inverter delivery or a regulatory sign-off. Optical data cannot observe any of those. The energisation event itself is sometimes visible as a change in thermal emission from inverter stations in TIR imagery, but that is a separate analytic layer not covered here.
If the question is 'when will this project generate power?' rather than 'how much has been installed?', the honest answer from optical data alone is a probabilistic range based on installation rate extrapolation, not a date. Combining the optical construction curve with AIS vessel data for equipment shipments, or with permit-filing records, narrows that range considerably. Buyers should be clear about which question they are actually asking before commissioning a monitoring programme.
Typical figures
| Spatial resolution (routine monitoring) | 3–4 m (Planet Dove); 10 m (Sentinel-2 visible/NIR) |
| Spatial resolution (milestone verification) | 0.3 m pan / 1.2–1.24 m multispectral (Pléiades Neo, WorldView-3) |
| Revisit (dense monitoring) | Daily (Planet Dove constellation); 5 days at mid-latitudes (Sentinel-2 A+B) |
| Minimum detectable installed area | ~0.5 ha (Planet Dove); ~5 ha (Sentinel-2 10 m) |
| Key spectral bands | Red (630–690 nm) and NIR (760–900 nm) for PV index; SWIR (1550–1750 nm) for soil/panel separation in wet conditions |
| Cloud sensitivity | Full outage under cloud; tropical monsoon regions may yield fewer than 30 usable scenes per quarter |
| Archive depth | Sentinel-2 from 2015; Planet from ~2016; Landsat 8/9 from 2013 (30 m) |
| Typical update latency | 24–72 hours from image acquisition to delivered analytic layer |
| Delivery formats | GeoTIFF classification raster, GeoJSON polygon layer, time-series CSV, PDF milestone report |
Analytics Satellize can run
| Weekly installed-panel area time series | Spectral index classification (NIR-red panel index) on atmospherically corrected surface reflectance; change detection against site-boundary mask | GeoJSON polygon layer per epoch plus CSV time series; updated weekly or on each clear acquisition |
| Estimated installed capacity (MWp) curve | Area-to-capacity conversion using client-supplied or published ground-coverage ratio and module efficiency; uncertainty bounds propagated from classification accuracy | Time-series chart and tabular data; suitable for lender drawdown reconciliation reports |
| Construction phase classification | Multi-class land-cover change detection distinguishing bare ground, racking-only, installed panels, inverter/substation structures and access roads | Colour-coded GeoTIFF per milestone date; PDF narrative report |
| Schedule-risk flag | Installation-rate regression over the most recent four to eight weeks compared against developer-published milestones; alert triggered when projected completion date slips beyond threshold | Automated alert (email or API push) with supporting imagery strip and rate chart |
| Retrospective construction curve | Archive processing of Sentinel-2, Planet and Landsat imagery from site boundary back to earliest available date; used for post-hoc due diligence | Full historical time series as CSV and PDF; prior land-use classification for baseline period |
| Regional pipeline aggregation | Multi-site panel-index classification across a defined geography; sites identified from permit records or known project lists; aggregated to national or regional MWp-under-construction figure | Monthly briefing table; GIS layer of all monitored sites with status tags |
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.