Road freight corridor congestion index from vehicle density mapping
Satellite SAR and high-resolution optical imagery can count stationary or slow-moving trucks on major freight corridors, producing a congestion index where loop detectors and GPS probes are absent or unreliable.
Sensors
- Sentinel-1 SAR (C-band, ESA): 5 m x 20 m ground range resolution in IW mode; 6-day revisit at mid-latitudes (12-day per orbit). Moving-target indicator (MTI) processing can detect and localise vehicles moving faster than roughly 3-5 m/s on motorways, displacing them from their true position in azimuth by an amount proportional to their radial velocity. Cloud-independent and night-capable. Free to access.
- ICEYE SAR (X-band): Sub-1 m spotlight resolution; tasking latency can be under a few hours. X-band returns from vehicle superstructures are stronger than C-band, improving detection of individual trucks. Commercially tasked, so revisit is demand-driven rather than fixed.
- Capella Space SAR (X-band): Spotlight mode resolves to approximately 0.5 m, sufficient to distinguish cab-plus-trailer units from cars. Night-time and all-weather collection. Commercially tasked; the constellation supports same-day revisit over priority corridors.
- Planet SuperDove (optical, multispectral): 3 m pixel spacing; daily revisit globally. Resolves heavy goods vehicles clearly in daylight. Cloud-limited: a single overcast day produces a gap. Useful for building free-flow baseline counts on cloud-free days and for visual validation of SAR detections.
What stationary trucks look like to a radar
A heavy goods vehicle parked or crawling on a motorway is a strong, stable radar reflector. Its flat metallic surfaces, particularly the rear panel and cab roof, produce bright returns in synthetic aperture radar imagery that stand out clearly against the lower-backscatter tarmac. At Sentinel-1's IW resolution of roughly 5 m in range, individual trucks are not resolved as shapes, but their point-like bright returns cluster in the traffic lanes and can be counted with reasonable confidence when density is high enough to avoid overlap.
The more interesting physics applies to moving vehicles. SAR forms images by correlating returns across the satellite's flight path over time. A vehicle moving radially relative to the sensor introduces a Doppler shift that displaces its image in azimuth, away from its true ground position. This is the basis of moving-target indicator processing: the displacement magnitude encodes the vehicle's radial velocity component. Sentinel-1 MTI work published in the remote-sensing literature has demonstrated detection of vehicles moving at highway speeds on motorways, though slow-moving or stop-start traffic in congestion is harder to separate from stationary clutter without multi-look processing.
Building a congestion index from a point-in-time count
A single image gives vehicle density at one moment. That number is only meaningful against a baseline. The standard approach is to collect a library of passes over the same corridor under known conditions: public holidays, normal weekday mornings, post-harvest peak seasons. From that archive you estimate a free-flow vehicle density per kilometre of lane. When a new image shows density significantly above that baseline, the excess is the congestion signal.
The index is not a speed estimate in the sense that a GPS probe network provides. It is closer to a queue-length proxy. On a 10 km stretch where free-flow might show 8-12 trucks per kilometre and a congested pass shows 40-60, the ratio is informative even if the absolute count carries a detection-rate uncertainty of 15-30 percent. That uncertainty comes from missed detections (trucks in shadow, under tree canopy at road edges, or with low radar cross-section orientations) and false positives (large roadside objects, bridge abutments). Honest calibration against ground counts on a representative segment is necessary before the index is used operationally.
Optical imagery from Planet SuperDove at 3 m is useful precisely because trucks are visually unambiguous in daylight. A human analyst or a trained object-detection model can count them with high confidence. The limitation is cloud: in humid tropical corridors or during monsoon seasons, optical data may be unavailable for days or weeks at a time. SAR fills that gap, but requires the calibration step described above.
Where this method is worth deploying and where it is not
The method earns its keep on corridors where ground infrastructure is thin. A motorway in Western Europe or North America already has inductive loop detectors, roadside radar, and millions of GPS-emitting smartphones feeding real-time traffic services. Satellite adds little there. The value proposition inverts on major freight arteries in sub-Saharan Africa, Central Asia, or remote parts of South-East Asia, where a single corridor may carry the bulk of a country's import-export freight with no systematic ground monitoring at all.
It is also useful for periodic strategic assessment rather than operational routing. A ministry of transport or a development bank financing corridor upgrades wants to know whether congestion is structural or seasonal, and whether it concentrates at specific chokepoints such as weigh stations, border crossings, or single-carriageway bottlenecks. A time series of satellite passes over 12-24 months can answer those questions without installing any ground infrastructure.
What it cannot do: provide the sub-minute refresh that a live traffic management system needs, detect congestion reliably on urban arterials where trucks mix with cars and motorcycles at sub-3 m separations, or distinguish a truck stopped in a queue from one parked at a layby without contextual analysis.
Sentinel-1 MTI in practice: published capability and honest limits
ESA's Sentinel-1 IW mode covers a 250 km swath at 5 m x 20 m resolution. The along-track baseline between sub-apertures in the IW TOPS mode is short, which limits the minimum detectable velocity for MTI to roughly 3-5 m/s radial component. Vehicles moving slower than that, which is most of the traffic in a genuine queue, appear as stationary clutter and cannot be velocity-tagged. This means Sentinel-1 MTI is best for detecting moving traffic on free-flowing sections and inferring congestion from the absence of moving returns rather than from direct queue measurement.
Commercial X-band SAR from ICEYE or Capella, with sub-1 m spotlight resolution, resolves individual vehicle shapes. At that resolution, a stopped HGV is distinguishable from a car, and a queue of 20 trucks on a two-lane road is directly countable. The trade-off is cost and coverage: a spotlight scene covers perhaps 5 km x 5 km, so corridor-scale monitoring requires multiple scenes stitched together, multiplying acquisition cost.
Delivering a corridor index: from raw counts to an actionable number
The output of this analysis is a corridor congestion index: a dimensionless score, or a set of scores at defined waypoints along a route, compared to a historical baseline and expressed as a percentile or a ratio. Delivered as a GIS layer, it shows which segments are above the 75th percentile of historical density. Delivered as a time-series table, it supports trend analysis by season, day of week, or hour of day (where multi-pass SAR allows).
Satellize applies this class of vehicle-density analysis on open constellations and commercial tasking. The workflow is similar in structure to the crop-area estimation pipeline run for the Kingdom of Tonga, where repeated image analysis produces an index calibrated against known ground conditions. For freight corridors, the calibration anchor is a short field-count campaign rather than crop-cut samples, but the statistical logic is the same: a satellite-derived index is only as good as the ground truth used to validate it.
A practical delivery cadence for a corridor with Sentinel-1 coverage is twice weekly (the 6-day repeat gives two passes per 12-day cycle at mid-latitudes, more at higher latitudes where orbits overlap). Commercial SAR can be tasked to fill gaps or increase temporal resolution during critical periods such as harvest seasons or major infrastructure disruptions.
Typical figures
| Spatial resolution (SAR, Sentinel-1 IW) | 5 m (range) x 20 m (azimuth); individual trucks not resolved as shapes, but detectable as point returns |
| Spatial resolution (commercial SAR, spotlight) | 0.5-1 m (ICEYE, Capella); individual HGV cab-trailer units resolvable |
| Spatial resolution (optical, Planet SuperDove) | 3 m; trucks visually countable in daylight, cloud-limited |
| Revisit (Sentinel-1, mid-latitudes) | 6 days per orbit; combined A+B constellation gave 3-day revisit before Sentinel-1B failure in 2022; single satellite currently operational |
| Revisit (commercial SAR) | Demand-driven; same-day tasking possible over priority corridors |
| Minimum detectable vehicle speed (SAR MTI) | Approximately 3-5 m/s radial component for Sentinel-1 IW; stopped vehicles appear as stationary clutter |
| Cloud penetration | Full (SAR); zero (optical) |
| Archive depth (Sentinel-1) | From 2014; free access via Copernicus Data Space |
| Typical count uncertainty | 15-30% vs ground truth before calibration; reducible with validated detection models |
| Delivery formats | GIS vector layer (GeoJSON, Shapefile), time-series CSV, PDF corridor report |
Analytics Satellize can run
| Free-flow baseline vehicle density | Object detection or SAR bright-target extraction on archive imagery under known uncongested conditions; statistical percentile fitting | Per-kilometre baseline density table by corridor segment, delivered as GIS layer and CSV |
| Point-in-time congestion index score | Comparison of current-pass vehicle count to baseline; expressed as ratio or percentile rank | Corridor-segment GIS layer with congestion score, updated per satellite pass |
| Chokepoint identification | Spatial clustering of above-baseline density peaks across multiple passes; cross-referenced with road geometry | Map of persistent high-density nodes with supporting time-series charts |
| Seasonal congestion trend report | Time-series aggregation of per-pass counts over 12-24 months; regression against calendar variables (harvest season, public holidays, border-crossing schedules) | Annual trend report with monthly index values and annotated anomaly periods |
| MTI-derived moving-vehicle density (SAR only) | Sentinel-1 IW TOPS along-track interferometry or sub-aperture Doppler shift extraction; moving-target localisation | GIS layer of detected moving-vehicle positions with estimated radial velocity band |
| Optical-SAR fusion count (cloud-gap filling) | Planet SuperDove counts on cloud-free days used to calibrate SAR detection rates; SAR fills cloud-obscured periods | Continuous weekly index with source-flagged data points (optical vs SAR) |
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.