Construction activity monitoring and project-stage tracking
Satellite imagery and SAR coherence time-series can track individual construction sites from ground-break to topping-out, independent of cloud cover. The method underpins permit compliance, infrastructure investment verification, and economic nowcasting in cities where ground inspection is slow or politically fraught.
Sensors
- Planet SkySat: 50 cm pan-sharpened resolution; on-demand tasking with same-day revisit over priority sites. Resolves individual structural elements such as columns, rebar mats, and formwork. The practical minimum detectable site is roughly 500 square metres at this resolution.
- PlanetScope: 3 m resolution, daily global revisit across a constellation of roughly 200 Dove satellites. Detects bare-soil exposure and rooftop appearance at sites larger than approximately 0.5 hectares; smaller plots are below reliable detection. Four-band (RGBNIR) and eight-band SuperDove variants available.
- Sentinel-1 C-band SAR: 5.4 GHz C-band, 10 m ground range resolution in Interferometric Wide swath mode, 6-day repeat at mid-latitudes (12 days at the equator from a single satellite). Coherence loss between repeat passes flags ground disturbance regardless of cloud or darkness, making it the primary tool for tropical city monitoring.
- Capella Space X-band SAR: Spotlight mode delivers down to 50 cm resolution; commercial tasking with sub-24-hour revisit on request. X-band penetrates light vegetation but not heavy canopy. Coherence analysis at this resolution can resolve individual building footprints under construction.
What the ground tells a radar before it tells anyone else
A construction site announces itself to a synthetic aperture radar long before it is visible in optical imagery. When a site is cleared, the soil surface changes from a rough, vegetated scatter to a smoother, disturbed one. More usefully, the coherence between two SAR passes, a measure of how similar the radar return is from the same patch of ground at two different times, collapses. Undisturbed surfaces maintain coherence over days or weeks. Active earthworks destroy it within a single repeat cycle.
For Sentinel-1, that repeat cycle is 6 days at mid-latitudes. A coherence map differencing two passes will show a newly broken site as a dark patch against a lighter background of stable ground. This works at night, through cloud, and through the persistent wet-season overcast that makes optical monitoring of tropical cities unreliable for months at a time. The limit is spatial: at 10 m resolution, a site must be at least a few hundred square metres before the coherence signal is unambiguous. Narrow plots in dense urban fabric can be masked by strong returns from adjacent buildings.
Cloud cover over tropical cities is a structural problem, not a footnote
Jakarta, Lagos, Dhaka, and Kinshasa share a characteristic that makes optical-only monitoring genuinely difficult: cloud cover exceeds 70 percent of days for large parts of the year. A PlanetScope constellation delivering daily passes still produces usable imagery only when the sky cooperates. In practice, a monitoring programme relying solely on optical data in these cities may accumulate gaps of three to six weeks during the wet season.
SAR coherence time-series fill that gap structurally, not as a backup. The operational approach is to run Sentinel-1 coherence differencing continuously and use optical imagery, whether PlanetScope at 3 m or SkySat at 50 cm, to characterise site stage when a clear acquisition arrives. The two data streams are complementary: SAR detects disturbance events; optical imagery classifies what is happening (earthworks, foundation pour, structural frame, roofing). Neither alone is sufficient for a reliable stage-tracking pipeline in the tropics.
What a floating roof gives away, and what it does not
The appearance of a rooftop in repeat imagery is one of the cleaner signals in construction monitoring. A bare-soil patch that acquires a spectrally bright, geometrically regular surface is almost certainly a completed or near-completed structure. In PlanetScope's NIR band, metal and concrete roofing materials are highly reflective, and the transition from dark soil to bright roof is detectable at sites above roughly 0.5 hectares. Below that threshold, the signal is present but unreliable at 3 m resolution.
SkySat changes the calculus substantially. At 50 cm, individual roofing panels, parapet walls, and plant rooms are distinguishable. A site of 500 square metres, a modest commercial unit or a small apartment block, can be tracked through foundation, frame, and enclosure stages. The cost is tasking frequency: SkySat is a commercial constellation with finite capacity, and daily revisit over many sites simultaneously is expensive. The practical model is to use PlanetScope or Sentinel-1 for broad surveillance across a city, then trigger SkySat tasking on sites that show anomalous activity, stalled progress, or permit discrepancies.
One honest limit applies to both sensors: a completed roof conceals everything beneath it. A building that reaches roofing stage is opaque to optical monitoring of internal fit-out, and SAR coherence stabilises once the structure is enclosed. Interior occupancy, fit-out progress, and services installation are not detectable from orbit.
Permit compliance and investment verification: the two hardest problems
Permit compliance monitoring is attractive in principle and difficult in practice. The analytical question, does what is being built match what was approved, requires a reference dataset of approved plans, a spatial register of permit boundaries, and repeat imagery at sufficient resolution to detect footprint exceedances. Where cadastral and planning data are digitised and georeferenced, the comparison is tractable. Where they are not, satellite data can detect that construction is occurring but cannot assess legality without the reference layer.
Infrastructure investment verification is a cleaner problem. A government or development bank that has disbursed funds for a road, school, or housing block can use satellite imagery to confirm that physical progress matches drawdown schedules. The signal required is simpler: is there evidence of construction activity at the funded location, and does the visible stage correspond to the claimed completion percentage? SAR coherence provides the activity signal; optical imagery provides the stage estimate. The method does not require ground truth at every site, which is precisely its value in contexts where site inspection is logistically difficult or politically sensitive.
Economic nowcasting from construction signals
Construction activity is a leading indicator of economic output. National statistics offices measure it through surveys and permits; both are lagged and subject to reporting gaps. Satellite-derived construction activity indices, aggregating the number of active sites, their estimated stage, and the rate of new ground-breaks across a city or region, can provide a higher-frequency signal than official data.
Published research using Sentinel-1 coherence and PlanetScope imagery has demonstrated the feasibility of city-level construction indices, though the method is not yet standardised across national statistical systems. The honest caveat is that a construction index measures physical activity, not economic value. A city building many small informal structures looks different from one building a single large infrastructure project, and the economic implications differ substantially. Disaggregating by site size, type, and location requires the higher-resolution optical layer to be genuinely useful for economic analysis rather than merely correlated with it.
Satellize runs analytics of this kind on open and commercial constellations for government clients. The crop-estimation programme in the Kingdom of Tonga uses a comparable time-series change-detection architecture; the construction monitoring pipeline applies the same underlying approach to built-environment signals rather than agricultural ones.
What the method cannot see, and where the archive starts
Three limits are worth stating plainly. First, underground construction, basement excavation, tunnelling, and sub-surface utilities, is invisible to both optical and C-band SAR. X-band at very high resolution can detect surface subsidence associated with deep excavation, but this is an indirect and uncertain signal. Second, construction inside an existing building envelope, a full internal refurbishment, for example, produces no detectable surface change. Third, very small sites in dense urban fabric are frequently masked by the radar returns of adjacent structures, particularly in cities with narrow street grids.
The Sentinel-1 archive runs from April 2014, giving roughly a decade of coherence data available for retrospective analysis. PlanetScope's archive begins in 2016 at useful global density. SkySat tasking is prospective; historical SkySat coverage of a specific site depends entirely on whether it was previously tasked. For a new monitoring programme, the practical start date for high-resolution optical tracking is the first tasking order; for SAR and medium-resolution optical, the archive provides a baseline immediately.
Typical figures
| Optical resolution (broad surveillance) | 3 m (PlanetScope SuperDove) |
| Optical resolution (site-level detail) | 50 cm pan-sharpened (Planet SkySat) |
| SAR resolution (disturbance detection) | 10 m (Sentinel-1 IW mode); 50 cm spotlight (Capella Space) |
| Revisit cadence | Daily optical (PlanetScope, cloud-permitting); 6-day SAR (Sentinel-1, single satellite, mid-latitudes); 12-day at equator; sub-24 h on commercial tasking (SkySat, Capella) |
| Minimum detectable site area | ~500 m² at 50 cm optical or X-band SAR; ~0.5 ha at 3 m optical; ~few hundred m² via SAR coherence at 10 m |
| Spectral bands (optical) | RGB + NIR (PlanetScope 4-band); 8-band including red-edge and SWIR (SuperDove); panchromatic + RGB (SkySat) |
| SAR frequency | C-band 5.4 GHz (Sentinel-1); X-band ~9.6 GHz (Capella Space) |
| Archive depth | Sentinel-1 from April 2014; PlanetScope from 2016 (global density); SkySat prospective from first tasking |
| Latency (alert to delivery) | Sentinel-1 NRT products within ~3 h of acquisition; PlanetScope typically same-day; SkySat tasked delivery within hours of pass |
| Delivery formats | GeoTIFF change layers, GeoJSON site polygons, tabular stage-progress reports, API feed for dashboard integration |
Analytics Satellize can run
| Ground-disturbance alert | SAR coherence differencing (Sentinel-1 or Capella repeat-pass pairs); coherence loss below threshold triggers flag | GeoJSON alert layer with site centroid, date of disturbance onset, and coherence-loss magnitude; delivered within 24 h of SAR acquisition |
| Construction stage classification | Multi-date optical change detection using spectral indices (bare-soil NDVI suppression, rooftop NIR reflectance increase) on PlanetScope or SkySat imagery | Per-site stage label (earthworks / foundation / structural frame / enclosed / complete) updated on each clear optical acquisition; GIS polygon layer |
| Progress-versus-schedule comparison | Time-series of SAR coherence and optical stage labels compared against a reference construction programme supplied by client; deviation flagged when observed stage lags schedule by more than one reporting period | Monthly tabular report per site with observed stage, expected stage, and variance; suitable for disbursement verification workflows |
| Permit-boundary exceedance detection | Footprint of disturbed or built area (from SAR coherence or optical classification) intersected with digitised permit boundary polygons; spatial exceedance computed in square metres | Flagged site list with exceedance area and imagery thumbnail; requires client-supplied georeferenced permit register |
| City-level construction activity index | Aggregation of active-site counts and new ground-break events across a defined urban boundary, derived from rolling SAR coherence stack; normalised to a base period | Monthly index value with site-count breakdown by district; time-series chart and underlying GeoJSON for integration into economic dashboards |
| Stalled-site identification | Sites showing initial coherence loss (ground-break) followed by coherence recovery (surface stabilisation without rooftop appearance) over a defined period, typically 90 days, flagged as potentially stalled | Quarterly stalled-site register with duration of inactivity and last-known stage; GIS layer with SkySat tasking recommendations for priority sites |
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.