Logistics park expansion and land-use change monitoring
Satellite time series can date every phase of a logistics park build, from first ground-break to sealed hardstanding, weeks before a planning register updates. Sentinel-2 and high-resolution optical imagery together give investors and regulators a verifiable construction timeline.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at mid-latitudes with both satellites combined. Reliable for detecting shed footprints above roughly 1,000 m², tracking NDVI loss as vegetation is cleared, and monitoring hardstanding reflectance change over time. Free and openly archived from 2015.
- Planet SuperDove: 3 m resolution, daily revisit over most land areas. Provides the temporal density needed to pin construction milestones to within a few days. Eight spectral bands including red-edge. Commercial licence required; coverage and tasking terms vary by geography.
- Maxar WorldView-3: 30 cm panchromatic, 1.24 m multispectral. Used for confirmation imagery: counting bays, verifying building type, reading lorry dock configurations. Revisit is on-demand rather than systematic; cloud risk must be managed through tasking windows.
- Airbus Pléiades Neo: 30 cm resolution, stereo and tri-stereo capable. Stereo pairs allow digital surface model generation, giving approximate roof height and therefore a cross-check on building volume. Tasked commercially; latency from order to delivery is typically 24 to 48 hours for archive, longer for new acquisition.
What the ground actually looks like when a logistics park starts
The sequence is consistent enough to be algorithmically trackable. Vegetation clearance comes first: NDVI drops sharply across the affected parcel, often within a single Sentinel-2 revisit cycle. Topsoil stripping follows, raising bare-earth reflectance in the shortwave infrared. Then comes compacted sub-base, then concrete or tarmac hardstanding, each with a distinct spectral signature in Sentinel-2 bands 11 and 12 (1.6 µm and 2.2 µm SWIR). A large shed roof, typically clad in profiled steel or membrane, appears as a high-reflectance rectangular object that is stable across revisits.
The practical detection floor for Sentinel-2 is a shed footprint of roughly 1,000 m² or larger, which corresponds to a very small unit in logistics terms. Most modern big-box distribution centres run to 50,000 m² or more, making them unambiguous at 10 m resolution. Access roads are detectable once they reach about 15 m width; internal yard markings are not resolvable at this scale and require commercial imagery for confirmation.
Why a time series beats a single snapshot
A single image tells you what exists. A time series tells you when it appeared, how fast it grew, and whether it matches the approved phasing in a planning consent. That distinction matters enormously to two different audiences.
For a real-estate investor, the construction rate is a proxy for developer confidence and supply-side timing. A shed that should take 18 months and is visibly ahead of schedule after nine months changes the leasing calculus. For a planning authority, a building that appears on imagery before the consent date is granted is prima facie evidence of a breach, and a satellite time series is a defensible, date-stamped record. Sentinel-2's open archive runs back to 2015, which means retrospective analysis of sites already built is possible without any new tasking.
Planet SuperDove's daily revisit is what closes the gap between Sentinel-2 passes. Cloud cover is the main constraint: in northern Europe, a usable clear-sky observation every five to seven days is a reasonable working assumption across a full year, not every day. In drier climates the cadence is better. Analysts should plan for cloud-gap interpolation and flag dates where confidence is lower.
Spectral change detection: the method behind the output
The standard analytical approach is bi-temporal or multi-temporal change detection applied to a pre-processed Sentinel-2 time series. Preprocessing involves atmospheric correction (Sen2Cor is the ESA-published processor), cloud masking, and co-registration to a common grid. Change is then quantified using normalised difference indices: NDVI for vegetation loss, NDBI (Normalised Difference Built-up Index, using SWIR and NIR) for impervious surface gain. A pixel that transitions from positive NDVI to strongly positive NDBI across consecutive images has almost certainly been built upon.
The honest limitation is that NDBI cannot distinguish a logistics shed from a data centre, a factory, or a large retail unit. Spectral change detection identifies that something was built; classification of what was built requires either higher-resolution optical confirmation or contextual analysis of site geometry, access-road configuration, and planning-register cross-referencing. That combination of open-data change detection and commercial-imagery confirmation is where most serious monitoring programmes sit.
High-resolution confirmation: what 30 cm adds and what it costs
Once Sentinel-2 or SuperDove flags a new large footprint, a single WorldView-3 or Pléiades Neo image can resolve dock-door counts, lorry court dimensions, and roof equipment such as HVAC arrays. These details distinguish a chilled-storage facility from an ambient distribution centre, which matters for both valuation and planning-use-class compliance.
The cost of commercial tasking is real and should be factored into programme design. A pragmatic workflow tasks high-resolution imagery only when the change-detection layer has already identified a site of interest, rather than blanketing a region. For a portfolio of, say, twenty active development sites, this approach keeps commercial imagery costs proportionate. Stereo acquisition for DSM generation adds further cost but is justified where building height is contested or where volume estimation is needed for insurance or valuation purposes.
Limits the data cannot overcome
Cloud cover is the most obvious constraint and is discussed above. Less obvious is the ambiguity between construction phases: a freshly poured concrete slab looks spectrally similar to an existing hardstanding that has been resurfaced. Ground-truth visits or planning-document cross-referencing are needed to resolve that ambiguity.
Sentinel-2's 10 m pixel does not resolve individual vehicles or construction machinery. It cannot confirm that a site is actively being worked on, only that the surface has changed. A site placed on hold after ground preparation will look, spectrally, like a completed hardstanding. Temporal cadence helps here: a genuine pause in construction shows no further change across many revisits, whereas an active site continues to evolve.
Finally, the method works best on open, flat sites. Dense urban infill, where new logistics units are inserted between existing buildings, is harder to track because the surrounding built environment confuses the change-detection baseline. Greenfield and edge-of-town sites, which account for the majority of modern logistics park development, are where the method is most reliable.
From satellite signal to investment or compliance decision
The analytical outputs from this method fall into two families. The first is a construction-progress timeline: a dated sequence of footprint outlines, hardstanding extent polygons, and access-road centre lines, delivered as GIS layers that can be overlaid on planning consents. The second is a portfolio-level change dashboard: a ranked list of sites by rate of change, flagging those that are ahead of schedule, stalled, or potentially non-compliant.
Satellize runs change-detection analytics on open constellations including Sentinel-2, adding commercial tasking on client licence for confirmation imagery. The workflow is similar in structure to the crop-estimation programme the company runs for the Kingdom of Tonga: open-data time series for the systematic signal, commercial imagery for targeted ground-truth confirmation.
For a planning authority covering a large administrative area, the value is triage: instead of manually reviewing every planning application site, analysts receive an alert only when a site shows significant spectral change. For an investor tracking a competitor's development pipeline, the value is earlier information than any public register provides. Both use cases rest on the same underlying data; the difference is the question being asked of it.
Typical figures
| Spatial resolution (change detection) | 10 m (Sentinel-2 MSI); 3 m (Planet SuperDove) |
| Spatial resolution (confirmation) | 30 cm panchromatic (WorldView-3, Pléiades Neo) |
| Revisit cadence | 5 days at mid-latitudes (Sentinel-2 dual satellite); daily (Planet SuperDove, weather permitting) |
| Minimum detectable footprint (Sentinel-2) | Approximately 1,000 m² for a high-contrast rooftop against cleared ground |
| Spectral bands used | Visible (B2–B4), NIR (B8), red-edge (B5–B7), SWIR (B11, B12) for NDVI and NDBI change detection |
| Archive depth | Sentinel-2 from 2015; Landsat back to 1972 for coarser baseline context |
| Latency (open data) | Sentinel-2 Level-2A products typically available within 3 hours of acquisition |
| Latency (commercial tasking) | 24–48 hours for archive; new acquisition window dependent on satellite scheduling and cloud |
| Deliverable formats | GeoTIFF change layers, GeoJSON footprint polygons, CSV timeline tables, PDF site reports |
| Cloud cover constraint | Usable clear-sky observations in northern Europe: roughly 5–7 day effective cadence averaged across a year |
Analytics Satellize can run
| Construction-phase timeline | Multi-temporal NDVI and NDBI change detection on Sentinel-2 time series, with phase classification (clearance, sub-base, hardstanding, roofed structure) | GIS polygon layer with dated phase boundaries per site; PDF milestone report |
| Footprint growth curve | Sequential building-footprint extraction using thresholded NDBI and morphological filtering; area calculated per revisit | CSV time series of footprint area (m²) per site; chart overlay on planning-consent boundary |
| Access-road detection and dating | Linear feature extraction on SWIR and panchromatic bands; change date assigned from first clear-sky observation showing road surface | GeoJSON road centre-line layer with acquisition-date attribute |
| Planning-compliance flag | Comparison of satellite-derived construction-start date against planning-consent grant date from client-supplied register; flags where construction precedes consent | Compliance alert table; site-level PDF with annotated imagery |
| Building-type confirmation | High-resolution optical classification of dock-door count, lorry-court geometry, roof equipment; manual analyst review supported by object-based image analysis | Site characterisation report with annotated WorldView-3 or Pléiades Neo image |
| Portfolio change dashboard | Change magnitude ranking across a defined site portfolio using SWIR band difference; sites sorted by rate of change per monitoring period | Weekly or monthly ranked dashboard; email alert on threshold exceedance |
| Building volume estimate | Stereo DSM generation from Pléiades Neo tri-stereo acquisition; roof height extracted and multiplied by footprint area | GeoTIFF DSM; tabulated volume estimate per building with stated uncertainty range |
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.