Solar farm construction progress from SAR backscatter
Utility-scale solar sites are built in months, often in climates where cloud cover makes optical monitoring unreliable. C-band and X-band SAR backscatter responds to the specular and double-bounce signatures of panel arrays, enabling weekly progress curves through any weather.
Sensors
- Sentinel-1 (C-band, 5.405 GHz): Free, open-access SAR with 10 m ground range resolution in Interferometric Wide Swath mode. Six-day repeat at mid-latitudes (12-day per pass direction). Adequate for sites above roughly 50 ha; individual panel rows at typical 5-10 m spacing are below the resolution floor. Archive from 2014.
- TerraSAR-X / TanDEM-X (X-band, 9.65 GHz): Spotlight mode reaches 1 m resolution, Stripmap 3 m. X-band is more sensitive to the metallic specular reflection of panel glass and aluminium frames than C-band. Revisit is 11 days at the equator, shorter at higher latitudes with off-nadir tasking. Commercial tasking required.
- ICEYE (X-band): Constellation of small SAR satellites offering Spot mode at approximately 1 m resolution and Strip mode at approximately 3 m. Sub-daily revisit is achievable through constellation scheduling. Useful for high-cadence progress snapshots on fast-moving sites.
- Capella Space (X-band): Spotlight imagery at approximately 0.5 m resolution in Sliding Spotlight mode. The finest commercially available SAR resolution for this application; can resolve individual panel rows on sites where TerraSAR-X Stripmap cannot. On-demand tasking with same-day collection windows.
What a panel field looks like to a radar
A photovoltaic panel tilted at a fixed angle toward the sun presents a flat, low-roughness dielectric surface to an imaging radar. The backscatter response depends on the incidence angle of the sensor relative to the panel tilt. When the radar beam strikes the panel face near-perpendicularly, backscatter is high. When the geometry is oblique, the panel acts as a specular reflector and backscatter drops sharply. This geometry-dependence is the detection mechanism: the same panel field will appear bright in one acquisition and dark in another depending on orbit direction and incidence angle, a behaviour that bare soil or vegetation does not replicate.
Aluminium racking steel adds a second signature. The dihedral formed between a row of racking and the flat ground beneath it produces a double-bounce return that is strong and geometrically stable across incidence angles. This double-bounce appears even before panels are mounted, which means SAR can distinguish three construction states: bare graded earth (low, diffuse backscatter), installed racking without panels (moderate double-bounce return), and fully panelled rows (specular response plus double-bounce from the row-to-ground dihedral). That three-state separation is the basis for progress quantification.
C-band versus X-band: a practical trade-off
C-band wavelength is roughly 5.6 cm. X-band is roughly 3.1 cm. The shorter wavelength interacts more strongly with the metallic panel frames and glass surface texture, producing higher contrast between panelled and un-panelled sections. Published studies using TerraSAR-X have demonstrated clear separation of panel rows in Stripmap mode on sites with row spacing above about 8 m. Sentinel-1 at 10 m resolution struggles to resolve individual rows but reliably detects the aggregate backscatter change as a zone transitions from bare earth to installed panels, which is sufficient for whole-site progress tracking on large projects.
X-band has one disadvantage worth stating plainly: it is more affected by rain on the panel surface. Wet glass changes the dielectric constant and alters the specular return. Acquisitions taken within hours of heavy rainfall should be treated with caution. C-band is less sensitive to this effect. For operational monitoring, cross-checking acquisitions from ascending and descending passes reduces the risk of misclassifying a wet-panel signature as an un-installed zone.
Building a progress curve from sequential acquisitions
The workflow starts with a pre-construction baseline: one or more acquisitions over the cleared, graded site before any racking is installed. Each subsequent acquisition is co-registered to the baseline and the backscatter difference is computed pixel by pixel. Pixels that have shifted from the baseline diffuse-earth signature to a panel-consistent signature are classified as installed. Summing those pixels and multiplying by the known panel density for the project gives an estimated installed capacity in megawatts-peak, assuming the project's layout plan is available.
Revisit cadence drives the resolution of the progress curve in time. Sentinel-1's six-day repeat at mid-latitudes produces roughly five data points per month, enough to track weekly installation rates and flag if a contractor falls behind the programme schedule. ICEYE or Capella tasking can compress that to near-daily observations during critical milestones, such as the final push before a grid-connection deadline. The honest limit: backscatter classification is probabilistic. Mixed pixels at zone boundaries, shadows from tracker-mounted panels that change tilt through the day, and soil moisture variation all introduce noise. Progress estimates carry an uncertainty of roughly plus or minus five to ten percent of installed area, depending on resolution and site geometry.
Resolution limits and the 50-hectare threshold
Below roughly 50 hectares, Sentinel-1's 10 m pixels begin to undercount installed area because each pixel spans multiple construction states simultaneously. A site of 20 ha with rows at 8 m spacing has fewer than three pixels per inter-row gap. The aggregate backscatter still shifts, but the three-state classification loses fidelity and progress estimates become less reliable. For sites in this size range, X-band at 1-3 m resolution is the appropriate choice.
Very small demonstration or community solar sites, say below 5 ha, are effectively invisible to Sentinel-1 for progress tracking purposes and require sub-metre commercial SAR or optical tasking. This is not a limitation of the method; it is a physical consequence of the ratio between wavelength, resolution and target size. Buyers should specify site area early so the sensor selection matches the monitoring requirement.
Connecting the data to project-finance obligations
Lenders and equity investors in utility-scale solar projects typically require independent progress verification at drawdown milestones. The conventional approach uses site inspections, which are expensive, infrequent and subject to access constraints. SAR-based progress curves provide an independent, time-stamped record of installation velocity that can be delivered as a GIS layer or structured report aligned to the project's milestone schedule.
The data does not replace a physical inspection for commissioning sign-off, and it cannot verify panel electrical connectivity or quality. What it does well is flag divergence between the contractor's reported progress and the observed backscatter state, which is precisely the signal a lender needs to decide whether to send an inspector. Satellize runs this kind of progress-monitoring analytics on open and commercial SAR constellations; the Tonga crop-estimation programme is a different application domain but illustrates the same principle of turning raw satellite data into a structured, decision-ready output for a specific client obligation.
What the archive makes possible before a project starts
Sentinel-1's archive runs from 2014. TerraSAR-X archive extends to 2007. For a site where construction has already begun, or where a dispute exists about when installation actually started, retrospective analysis of archived acquisitions can reconstruct the installation timeline independently of contractor records. This has obvious utility for insurance claims, tax-credit compliance (which in some jurisdictions depends on demonstrable construction commencement dates) and dispute resolution.
Archive analysis also supports pre-construction due diligence. Examining multi-year backscatter time series over a proposed site reveals whether the ground has been subject to seasonal flooding, whether adjacent industrial activity produces radar interference, and whether the terrain has been previously disturbed. None of that is visible in a single optical image taken on a clear day.
Typical figures
| Spatial resolution (Sentinel-1 IW) | 10 m range × 22 m azimuth (ground projected) |
| Spatial resolution (TerraSAR-X Spotlight) | ~1 m; Stripmap ~3 m |
| Spatial resolution (Capella Sliding Spotlight) | ~0.5 m |
| Revisit cadence | Sentinel-1: 6 days (mid-latitudes, dual-pass); TerraSAR-X: 11 days baseline; ICEYE/Capella: sub-daily with tasking |
| Radar frequency | C-band 5.405 GHz (Sentinel-1); X-band ~9.65 GHz (TerraSAR-X, ICEYE, Capella) |
| Minimum detectable site area (Sentinel-1) | ~50 ha for reliable three-state classification; ~5 ha for aggregate backscatter shift detection only |
| Progress estimate uncertainty | ±5–10% of installed area (resolution and geometry dependent) |
| Archive depth | Sentinel-1 from 2014; TerraSAR-X from 2007 |
| Weather dependency | Cloud-independent; wet-panel effect on X-band acquisitions within hours of heavy rain |
| Delivery latency | Sentinel-1: typically 1–3 days after acquisition; commercial X-band: same-day to 24 hours with priority tasking |
Analytics Satellize can run
| Installation progress curve | Bitemporal backscatter change detection against pre-construction baseline; pixel-wise three-state classification (bare earth, racking, panelled) | Time-series chart of installed area (ha) and estimated capacity (MWp) with per-acquisition confidence intervals, delivered as PDF report and CSV |
| Zone-level progress map | Classified SAR backscatter raster co-registered to project layout plan; zones coloured by installation state | GeoTIFF and vector GIS layer (GeoJSON or Shapefile) updated each acquisition cycle |
| Milestone verification report | Comparison of observed installed-area estimate against contractor-reported milestone; divergence flagged with supporting imagery | Structured PDF report formatted to lender or equity-investor milestone schedule |
| Installation velocity index | First derivative of the progress curve; rate of change in panelled area per week compared to planned installation schedule | Weekly alert if observed velocity falls more than a defined threshold below the programme baseline |
| Retrospective timeline reconstruction | Archive backscatter time series analysis to establish construction commencement date and phase sequence | Annotated timeline with acquisition-date evidence, suitable for insurance or regulatory submission |
| Pre-construction site characterisation | Multi-year archive backscatter and coherence analysis to identify seasonal flooding, prior disturbance or adjacent interference sources | Site risk summary report with annotated time-series plots |
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.