Inland dry port and intermodal container depot activity monitoring
Satellite optical time series can track container-stack footprints and truck patterns at inland clearance depots, answering whether dry ports are absorbing coastal congestion or sitting idle. AIS data alone cannot answer that question.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands, 5-day revisit at mid-latitudes with both satellites combined. Sufficient to detect changes in the overall footprint of container stacks and to distinguish loaded from cleared yard sections. Free and open; archive from 2015.
- Planet SuperDove: 3 m resolution, daily revisit globally. Captures the day-to-day rhythm of truck arrivals and stack expansion that a 5-day cadence misses. Colour fidelity across 8 bands helps distinguish containers from surrounding hardstanding. Commercial licence required.
- Maxar WorldView-3: 30 cm panchromatic, 1.24 m multispectral. Used for periodic high-confidence ground-truth: individual containers are resolved, allowing stack-row counting and approximate density validation. Tasked on demand; not a continuous monitoring layer.
- Sentinel-1 SAR (context only): C-band SAR at 10 m IW mode, cloud-independent. Covered in the sibling SAR yard-occupancy page; noted here only because optical and SAR are often fused to handle persistent cloud cover over monsoon-season depots.
What the satellite actually sees at a dry port
A loaded container depot looks different from an empty one, even at 10 m resolution. Dense stacks of steel boxes produce a high-reflectance, geometrically regular signature in visible bands. When a yard clears, the exposed hardstanding has a markedly different spectral response. Sentinel-2's red-edge and near-infrared bands help separate concrete aprons from container tops, which matter when you are trying to distinguish 'yard is full' from 'yard is being resurfaced'.
At 3 m with Planet SuperDove, individual container rows become countable in good conditions, though not individual boxes. The practical detection floor is a change of roughly two to four row-widths, equivalent to perhaps 20 to 40 TEU in a typical stacking configuration. That is honest: this method answers throughput trends, not precise unit counts. For the latter, WorldView-3 at 30 cm is the instrument, used sparingly because tasking costs accumulate.
Reading the weekly rhythm from daily imagery
Dry ports have a working week. Gate activity peaks on certain days; rail lifts happen on fixed schedules; customs dwell times create predictable stack-age patterns. A single Sentinel-2 overpass every five days will miss that structure entirely. Daily Planet coverage lets analysts build a time-series profile of yard occupancy that reveals not just whether a depot is busy, but when it is busy and how quickly it turns over.
The practical method is pixel-level change detection on a normalised difference of consecutive-day composites, followed by object-based segmentation to isolate the container-storage polygon from access roads and administration buildings. Truck presence in gate lanes is detectable as a density signal at 3 m, though individual vehicle classification requires WorldView-3 or equivalent. Published work in Remote Sensing (MDPI) has demonstrated vehicle detection at logistics facilities using similar Planet-class imagery with precision figures in the 70 to 85 percent range depending on vehicle size and shadow conditions.
Cloud is the honest constraint. Depots in South and South-East Asia can lose 60 to 80 percent of optical observations during monsoon months. Fusing with Sentinel-1 SAR mitigates this, but SAR stack-height interpretation is a separate analytic problem covered elsewhere in this library.
Why inland nodes matter when ports are congested
Port congestion events, such as those seen at major hubs during 2020 to 2022, push dwell time inland. Importers divert boxes to inland clearance depots to free berth space; exporters pre-position empty equipment away from saturated marine terminals. The knock-on effect at inland dry ports can be substantial, but it is invisible to vessel-tracking systems. AIS shows the ship anchored offshore. It does not show the depot 200 km inland filling up.
Satellite monitoring of inland nodes therefore complements maritime intelligence rather than duplicating it. A shipper or national trade authority watching both layers simultaneously can identify where the bottleneck actually sits: at the berth, at the gate, or in the inland dwell cycle. That distinction changes the policy or commercial response.
Archive depth and baseline construction
Sentinel-2's archive runs from mid-2015, giving nearly a decade of 10 m observations for most depot locations. That is long enough to establish a credible seasonal baseline: how full is this depot in a normal October, versus a congested one. Deviations from that baseline are the signal. Without the baseline, an analyst seeing a busy yard cannot tell whether it is unusually busy or simply typical for the season.
Planet's commercial archive is shallower and coverage density varies by geography and contract period. For depots where Planet data is sparse before a given date, Sentinel-2 carries the historical load and Planet provides the high-cadence current layer. The two are not interchangeable at the pixel level, but they are compatible for trend analysis when the analytic unit is yard-occupancy fraction rather than absolute container count.
Honest limits of the method
Several things this approach cannot do are worth naming plainly. It cannot identify what is inside a container, distinguish laden from empty boxes, or read container numbers. It cannot reliably detect activity inside covered warehouses or under canopy structures. At 10 m, a small depot of fewer than two or three hectares may fall below the reliable change-detection threshold for Sentinel-2 alone.
Revisit is probabilistic, not guaranteed. Cloud, sensor scheduling and orbit geometry mean that even with Planet's nominal daily revisit, a given site may go three to five days without a usable observation in bad weather. Latency between acquisition and delivered analytic is typically 24 to 72 hours for standard processing pipelines, longer if manual quality checks are required. Buyers should plan monitoring programmes around these constraints rather than assuming continuous real-time visibility.
Satellize runs this class of analysis on open constellations augmented by commercial tasking, structured around the same baseline-deviation logic used in the Tonga crop-estimation programme. The methodology transfers directly to logistics contexts where the question is 'has something changed from normal' rather than 'what is the absolute count'.
Turning observations into an operational index
The most useful deliverable for a logistics operator or trade ministry is not a raw image but a normalised occupancy index: a number between zero and one expressing how full a monitored depot is relative to its historical maximum observed footprint. Plotted weekly, that index shows congestion build-up before it becomes a crisis and recovery after an intervention.
Aggregated across a network of depots, the same index becomes a corridor-level throughput signal. If three dry ports along a single rail corridor all show rising occupancy simultaneously, the constraint is upstream of all of them, probably at the port or at a customs processing bottleneck. That kind of network-level inference is what separates satellite analytics from a site visit.
Typical figures
| Spatial resolution (primary monitoring) | 10 m (Sentinel-2); 3 m (Planet SuperDove) |
| Spatial resolution (validation / audit) | 30 cm pan / 1.24 m multispectral (WorldView-3) |
| Revisit cadence | 5 days at mid-latitudes (Sentinel-2, both satellites); nominally daily (Planet SuperDove) |
| Minimum detectable yard-change | Approximately 2–4 container-row widths (~20–40 TEU equivalent) at 3 m; larger footprint changes only at 10 m |
| Spectral bands used | Visible (RGB), red-edge, NIR (Sentinel-2 bands 2–8A); 8-band visible to NIR (SuperDove) |
| Cloud sensitivity | Full optical outage under cloud; monsoon-season sites may lose 60–80% of observations without SAR fusion |
| Archive depth | Sentinel-2 from mid-2015; Planet commercial archive varies by site and contract |
| Analytic latency | 24–72 hours from acquisition to delivered index under standard processing; longer with manual QA |
| Coverage | Global; Sentinel-2 covers land surfaces between approximately 84°N and 56°S |
| Delivery formats | GeoTIFF change layers, GeoJSON polygons, CSV occupancy index time series, PDF periodic reports |
Analytics Satellize can run
| Yard occupancy index | Object-based image analysis (OBIA) on multitemporal optical composites; normalised against historical maximum observed container-stack footprint | Weekly CSV time series and dashboard tile per monitored depot; alert triggered when index exceeds user-defined threshold |
| Stack-footprint change map | Pixel-level normalised difference change detection between consecutive cloud-free composites; segmentation to isolate storage polygon | GeoTIFF and GeoJSON layer showing expanded, contracted and stable stack zones; delivered per overpass |
| Truck-presence density signal | Vehicle detection on Planet SuperDove imagery using object detection trained on logistics-facility imagery; density counted within gate-lane polygons | Daily vehicle-count estimate per gate zone; weekly trend chart |
| Congestion onset alert | Statistical process control on occupancy index time series; alert fires when index deviates beyond two standard deviations from seasonal baseline | Email or API alert with supporting imagery thumbnail and index chart |
| Network-level corridor occupancy | Aggregation of per-depot occupancy indices across a user-defined depot network; correlation analysis to identify upstream constraint point | Monthly PDF report with network heat-map and narrative interpretation |
| Baseline calibration report | Sentinel-2 archive analysis (2015 to present) to establish seasonal occupancy norms per depot; used to contextualise current observations | One-off PDF report with seasonal occupancy curves and anomaly thresholds; updated annually |
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.