Bonded warehouse rooftop expansion as trade volume proxy
Bonded warehouse and free-trade-zone rooftop growth, tracked through optical time-series, gives a spatially explicit, lagging signal of import and re-export demand at specific customs nodes. Combining building-footprint change detection with cadastral overlays separates trade-policy-sensitive storage from general logistics expansion.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands, 5-day revisit at the equator with both satellites. Sufficient to detect new rooftop footprints above roughly 1,000 sq m and to track construction phase transitions across quarters. Free archive from 2015.
- Planet Basemaps (PlanetScope): 3 m resolution, near-daily revisit. Resolves individual warehouse bays and can distinguish rooftop material (bare concrete, metal cladding, membrane) useful for construction-stage classification. Commercial licence required.
- Airbus SPOT 6/7: 1.5 m panchromatic, 6 m multispectral. Useful for confirming footprint boundaries and reading rooftop features such as loading-bay count. Revisit 1 to 3 days with off-nadir tasking. Tasked on demand.
- Google Earth Engine archive: Provides access to historical Landsat and Sentinel-2 stacks for retrospective baseline construction. Landsat archive extends to 1972 at 30 m; useful for decade-scale context on zone establishment, not individual building detection.
Why rooftop area is a credible, if slow, trade signal
Bonded warehouses are not built on speculation in the way that general logistics sheds sometimes are. They require a customs authority licence, a defined catchment of importers or re-exporters, and often a sovereign guarantee or free-trade-zone concession. That regulatory overhead means construction correlates tightly with anticipated throughput: operators do not sink capital into bonded space unless demand is already contracted or the policy environment has changed materially.
The catch is timing. From planning approval to operational status, a large bonded facility typically takes twelve to thirty months. Rooftop area therefore lags the trade signal by several quarters. That makes it a poor tool for spotting this month's import surge but a useful one for reading medium-term structural shifts, such as a customs zone positioning itself for a new preferential trade agreement or a port authority responding to a sustained increase in transshipment volumes.
Separating bonded from general logistics: the cadastral overlay problem
Optical imagery alone cannot tell a bonded warehouse from an ordinary distribution centre. The rooftops look identical. The separation requires a second data layer: cadastral or zoning records that delineate customs-controlled parcels, free-trade-zone boundaries, or special economic zone perimeters. In jurisdictions where these boundaries are published (many are, as they carry legal force for duty exemption), a simple spatial join assigns each detected building footprint to a regulatory category.
Where cadastral data is unavailable or out of date, a reasonable proxy is proximity to a designated port or inland customs depot, combined with building morphology. Bonded facilities tend to be larger than average, set back from public roads, and clustered inside a perimeter fence that itself appears in imagery. Neither heuristic is watertight. A customs-zone boundary that has not been publicly gazetted will defeat the method entirely, and that is worth stating plainly before any client commits to this signal.
The practical workflow: ingest the zone boundary as a polygon, clip the building-footprint change-detection output to that polygon, and report gross roofed area by quarter. Changes outside the polygon are filtered out. Changes inside it are attributed to bonded capacity growth.
Building-footprint change detection: what the methods actually do
The core technique is bi-temporal or multi-temporal image differencing on a normalised spectral index. New rooftops appear as high-reflectance anomalies in the shortwave infrared relative to the pre-construction baseline, which is typically bare soil or low vegetation. Sentinel-2 Band 11 (1.6 µm SWIR) is particularly responsive to the transition from soil to sealed surface. A threshold applied to the change image, followed by morphological filtering to remove small artefacts, yields candidate new-building polygons.
Cloud is the principal operational constraint. Tropical free-trade zones, which are some of the most active globally, can see ten or more consecutive cloud-contaminated Sentinel-2 acquisitions during monsoon season. The standard mitigation is to use the full available time-series and composite the clearest observations per quarter rather than relying on any single date. Planet's near-daily cadence helps here: the probability of at least one cloud-free acquisition per week is substantially higher at 3 m daily than at 10 m every five days, though Planet data carries a licence cost that Sentinel-2 does not.
Minimum detectable footprint at 10 m Sentinel-2 resolution is roughly 400 to 1,000 sq m depending on rooftop contrast and surrounding land cover. A standard bonded warehouse bay is typically 5,000 sq m or larger, so the resolution is adequate for the target class. Smaller ancillary structures, guard posts, generator housings, are below the detection floor and should not be counted.
Reading the lag: construction phase to operational throughput
A quarterly rooftop-area time-series can be segmented into three observable phases. First, site clearance and slab preparation: spectral reflectance rises, vegetation indices drop, but no enclosed roofed area yet exists. Second, structural framing: roofed area begins to appear as bays are enclosed, often non-uniformly across a large shed. Third, operational fit-out: the roof is complete but the facility may not be accepting cargo for several more months while racking, fire suppression, and customs fitments are installed.
Translating rooftop completion into a throughput forecast requires an assumption about the lag between phase three and first cargo movement. Published data on free-trade-zone development timelines is thin, but infrastructure analysts typically apply a six-to-twelve-month lag for large bonded facilities. Clients should treat any throughput estimate derived from rooftop data as a range, not a point forecast, and should cross-reference against port statistics or trade declarations where those are published.
Honest limits of the signal
Rooftop expansion is a one-directional indicator in most optical time-series: buildings appear but rarely disappear. A bonded facility that is decommissioned or converted to general warehousing will not vanish from the footprint count unless the roof is physically demolished. This creates a survivorship bias in long-run area totals. Analysts should check for vacancy signals, such as absence of truck activity or faded rooftop markings in high-resolution imagery, before treating stable rooftop area as stable throughput.
The method also cannot distinguish utilisation rate from capacity. A zone may double its bonded rooftop area while actual throughput stagnates if the new space is built ahead of demand. Rooftop area is a capacity proxy, not a flow proxy. For flow, the correct sibling signal is vehicle or container activity at the facility, covered in separate pages in this library.
Satellize can run this type of footprint time-series as part of its broader analytics stack. The Tonga crop-estimation programme uses a similar change-detection pipeline on agricultural parcels; the bonded-warehouse application substitutes industrial land-cover classes and a customs-zone polygon as the spatial filter.
Typical figures
| Spatial resolution (primary sensor) | 10 m (Sentinel-2 MSI, visible and SWIR bands) |
| Spatial resolution (validation) | 1.5 m panchromatic (SPOT 6/7 on-demand tasking) |
| Revisit cadence | 5 days at equator (Sentinel-2 A+B combined); near-daily (Planet PlanetScope) |
| Minimum detectable rooftop footprint | 400 to 1,000 sq m at 10 m resolution, depending on rooftop-to-background contrast |
| Archive depth | Sentinel-2 from 2015; Landsat for decade-scale context from 1972 at 30 m |
| Key spectral bands | Sentinel-2 Band 4 (red, 665 nm), Band 8 (NIR, 842 nm), Band 11 (SWIR, 1610 nm) for surface-change detection |
| Cloud sensitivity | High in tropical zones; quarterly compositing required; Planet daily cadence reduces gap risk |
| Cadastral overlay requirement | Published customs-zone or free-trade-zone boundary polygon needed to separate bonded from general logistics |
| Throughput lag estimate | 6 to 30 months from construction start to operational status (range from infrastructure literature) |
| Delivery formats | GeoTIFF change layers, GeoJSON footprint polygons, quarterly area time-series in CSV or GIS layer |
Analytics Satellize can run
| Quarterly bonded rooftop area index | Bi-temporal SWIR change detection on Sentinel-2 time-series, clipped to customs-zone polygon | GIS layer and CSV time-series showing gross roofed area by quarter per designated zone |
| Construction phase classification | Multi-date spectral segmentation distinguishing slab, framing, and enclosed-roof stages using Planet 3 m imagery | Per-building status report updated monthly, with phase label and estimated completion quarter |
| New footprint alert | Threshold-based change detection on rolling 90-day Sentinel-2 composite, filtered by minimum area and zone boundary | Alert feed (GeoJSON or email) triggered when a new enclosed footprint above 2,000 sq m appears within a monitored zone |
| Retrospective capacity growth curve | Annual Sentinel-2 and Landsat archive stack processed through supervised land-cover classifier to extract impervious industrial area | Decade-scale rooftop area time-series per zone, delivered as chart-ready CSV and annotated PDF |
| Multi-zone comparative ranking | Standardised area-growth rate calculated across a portfolio of user-specified customs zones, normalised by zone perimeter area | Quarterly ranking table of zones by capacity growth rate, with map overlay in GIS format |
| Vacancy indicator flag | Cross-reference of stable rooftop footprint against vehicle-activity proxy (truck presence in high-resolution imagery) to flag potentially underutilised facilities | Flagged facility list with supporting imagery thumbnails, updated quarterly |
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.