Solar farm specular glint and aviation glare hazard mapping from satellite geometry
Specular glint from photovoltaic arrays is a regulated aviation hazard in several jurisdictions. Satellite geometry, open DEMs and multispectral reflectance data can screen candidate layouts rapidly before detailed ray-tracing is commissioned.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at mid-latitudes with both satellites. Band 3 (green, 560 nm) and Band 4 (red, 665 nm) surface reflectance anomalies reveal panel tilt and orientation by comparing observed reflectance against Lambertian baseline at known sun-sensor geometry.
- Landsat 8/9 OLI: 30 m resolution, 16-day single-satellite revisit, 8-day combined. Longer archive (Landsat 8 from 2013, Landsat 9 from 2021) supports characterisation of existing installations and multi-angle reflectance sampling across seasonal sun-angle variation.
- Copernicus DEM GLO-30: Global 1 arc-second (approximately 30 m) digital surface model derived from TanDEM-X. Provides terrain elevation for horizon angle computation and slope/aspect extraction needed to resolve panel tilt from flat-roof or ground-mounted geometry.
- SRTM (Shuttle Radar Topography Mission): 30 m void-filled DEM covering latitudes 56°S to 60°N. Useful as a cross-check against GLO-30 in areas where TanDEM-X acquisition quality is uncertain; vertical accuracy approximately 16 m absolute at 90th percentile.
Why specular geometry matters more than average reflectance
A solar panel is not a diffuse reflector. Monocrystalline and polycrystalline silicon modules, and especially thin-film panels with glass covers, have a bidirectional reflectance distribution function (BRDF) that is strongly peaked in the specular direction. When the sun, the panel surface, and an observer's eye align within a few degrees of the specular angle, the reflected luminance can exceed 50,000 candelas per square metre, well above the 10,000 cd/m² threshold at which the FAA and the UK CAA define a potential after-image hazard for pilots.
The critical variable is therefore not the panel's average albedo but the precise geometry: solar azimuth and elevation at the moment of overflight, panel tilt and azimuth, and the angular position of approach or departure corridors relative to the site. A farm that is entirely benign at noon in summer can produce a direct specular flash during a winter morning approach. Satellite data addresses this by supplying three of those four variables from the public record, leaving panel orientation as the quantity to be estimated or measured.
What a floating roof gives away: reading panel orientation from multispectral imagery
Every Sentinel-2 and Landsat acquisition is taken from a known sensor position at a known solar illumination angle. That geometry is recorded in the scene metadata to sub-degree accuracy. When a panel row is oriented so that it reflects sunlight toward the sensor, the surface reflectance value in visible bands rises sharply above the Lambertian expectation for the same land cover. Conversely, panels tilted away from the specular condition appear darker than bare ground of similar albedo.
By sampling an existing installation across multiple acquisitions spanning different solar geometries, it is possible to invert a simplified BRDF model and recover the dominant panel tilt angle and azimuth to an accuracy of roughly plus or minus 5 degrees. This is not a substitute for a site survey, but it is sufficient to seed a glint-angle envelope calculation for a planning-stage hazard screen. The method was described in published literature examining the spectral signature of utility-scale PV in Sentinel-2 imagery, and the underlying physics of BRDF inversion from multi-angle spaceborne data is well established from MODIS and MISR research.
One honest limit: panels that are heavily soiled, partially shaded by vegetation, or covered by a thin cirrus layer during acquisition will produce anomalous reflectance that can bias the orientation estimate. Acquisitions should be quality-filtered for cloud fraction below 5 percent and aerosol optical depth below 0.2 before inversion.
Computing the glint envelope across a candidate layout
Given a proposed site boundary, a DEM, and a set of assumed panel tilt and azimuth values, the glint-angle envelope calculation is a geometric problem. For each point in the layout grid, and for each hour of the day across the year, the solar vector is computed from ephemeris data. The specular reflection vector is then derived from the panel normal. The question is whether that reflected vector intersects any published instrument approach procedure (IAP) corridor, defined as the airspace volume within a specified lateral and vertical tolerance of the published approach path centreline.
The Copernicus GLO-30 DEM supplies terrain height for horizon masking: a glint ray that would be blocked by an intervening ridge before reaching the approach corridor can be excluded. This matters considerably in hilly terrain near regional airports. On flat coastal plains, the terrain mask contributes little and the full glint envelope is determined almost entirely by sun angle and panel geometry.
The output is a time-series of glint-hazard windows: calendar periods and times of day during which specular reflection from a given layout could fall within the approach corridor, expressed as an angular offset from the pilot's line of sight. Regulators typically require this offset to be assessed against published ocular hazard thresholds. The satellite-derived screen identifies whether any such windows exist and how many hours per year they accumulate. It does not replace the certified ray-tracing analysis that many regulators require as a final submission.
Archive depth and the seasonal sun-angle problem
The worst glint geometries often occur at low solar elevations in winter, when the sun is in the southern sky (for northern hemisphere sites) and approach corridors from the north or east are most exposed. A single satellite acquisition in July tells you almost nothing about the December morning hazard. This is why archive depth matters.
Landsat's continuous record from 1984 onward, and Sentinel-2's from 2015, provide acquisitions across the full annual solar cycle. For a new site, the relevant question is not whether glint occurred in any historical image but whether the geometry ever aligns. That can be computed analytically from ephemeris without needing an actual image of the proposed site; the imagery of nearby existing farms is used only to calibrate the panel BRDF model. The two tasks are separable, which is useful when the candidate site is currently bare agricultural land with no panels to observe.
Honest limits of the satellite-geometry approach
Several things this method cannot do are worth stating plainly. It cannot characterise the panel BRDF to the precision required for a certified ocular hazard submission. That requires laboratory or field goniometer measurements of the specific module type, combined with ray-tracing software validated against published standards such as the FAA's Aviation Glare Hazard Assessment Tool (AGHAT) methodology. Satellite-derived orientation estimates carry an angular uncertainty of roughly plus or minus 5 degrees, which propagates into an uncertainty in the predicted glint direction of similar magnitude.
The method also assumes a uniform panel tilt across the layout. Tracking systems that adjust panel angle throughout the day require a time-varying BRDF model that cannot be recovered from a small number of satellite acquisitions. Single-axis trackers in particular can produce glint geometries that a fixed-tilt model would miss entirely. Finally, the DEM-based horizon mask is limited by the 30 m resolution of GLO-30 and SRTM; narrow ridgelines or tall structures near the site boundary may be missed.
Satellize runs this geometric screen as part of its renewable energy analytics portfolio, using open Sentinel and Landsat data combined with GLO-30 terrain. The Tonga crop-estimation programme demonstrated that the same open-constellation workflow can be adapted to small-island geographies where commercial tasking is disproportionately expensive; the same principle applies to island or coastal airports where glint geometry is acute.
From screen to submission: what comes next
A satellite-derived glint screen answers the first regulatory question: does this site have any geometry that warrants detailed assessment? If the answer is no, because the approach corridors are oriented such that the specular vector never intersects them across the full annual solar cycle, that finding can be documented and submitted with the planning application at low cost.
If the screen identifies potential hazard windows, the next step is a certified assessment using panel-specific BRDF data and ray-tracing software. The satellite screen narrows the problem: it identifies which approach corridors, which seasons, and which times of day require detailed modelling, reducing the computational scope of the certified analysis. It also provides a spatial map of the layout cells that contribute most to the hazard, which can inform mitigation options such as anti-reflective coatings, layout rotation, or setback distances from the airport boundary.
To commission a glint-angle envelope screen for a specific site, share the proposed layout boundary, the relevant airport ICAO code, and any known approach procedure identifiers.
Typical figures
| Spatial resolution (optical imagery) | 10 m (Sentinel-2 MSI visible/NIR); 30 m (Landsat 8/9 OLI) |
| DEM resolution | ~30 m (Copernicus GLO-30, SRTM); vertical accuracy ~4 m (GLO-30 LE90) |
| Revisit for panel orientation sampling | 5 days (Sentinel-2 combined); 8 days (Landsat 8+9 combined); cloud-free acquisitions typically 4–12 per season at mid-latitudes |
| Archive depth | Sentinel-2 from 2015; Landsat from 1984 (OLI from 2013) |
| Panel orientation accuracy | ±5° tilt and azimuth from BRDF inversion across multiple acquisitions (honest estimate; not certified) |
| Glint envelope temporal resolution | Hourly solar position steps across full annual cycle (computed analytically from ephemeris) |
| Minimum detectable installation size | Approximately 1 hectare for reliable panel signature extraction in Sentinel-2; smaller sites require Landsat with caution |
| Cloud-cover limit for usable acquisitions | Scene cloud fraction <5%; aerosol optical depth <0.2 recommended for BRDF inversion |
| Deliverable formats | GeoTIFF glint-hazard raster, GeoJSON hazard-window polygons, CSV time-series of glint-angle offsets, PDF summary report |
Analytics Satellize can run
| Annual glint-angle envelope map | Solar ephemeris computation combined with panel-normal specular reflection geometry over GLO-30 terrain; approach corridor intersect test | GeoTIFF raster showing cumulative annual hours of potential glint incidence per layout cell, per approach corridor |
| Panel tilt and azimuth estimate for existing installations | Simplified BRDF inversion from multi-date Sentinel-2 and Landsat surface reflectance at known sun-sensor geometries | GeoJSON layer of dominant panel orientation per installation, with uncertainty bounds |
| Seasonal hazard-window calendar | Hourly ephemeris stepping across full annual solar cycle; specular vector vs. IAP corridor angular offset computation | CSV and PDF calendar identifying dates, times, and durations of potential glint windows by approach procedure identifier |
| Terrain horizon mask | Viewshed analysis on Copernicus GLO-30 DEM from each layout cell toward each approach corridor segment | GeoTIFF mask layer indicating which layout cells have unobstructed line-of-sight to defined airspace volumes |
| Layout sensitivity comparison | Parametric variation of panel tilt (0–35°) and azimuth (±30° from south) across proposed layout; glint envelope recomputed for each scenario | Multi-scenario GeoJSON and summary table supporting layout optimisation or setback decisions |
| Regulatory screen report | Synthesis of glint envelope, terrain mask, and hazard-window calendar against published regulatory thresholds (e.g. FAA AGHAT framework criteria for reference) | PDF planning-submission document confirming presence or absence of geometrically feasible hazard windows, with methodology annex |
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.