Informal market and street-trading area mapping
Informal markets are invisible to cadastral records yet visible from orbit. Temporal stacking of very-high-resolution imagery reveals the periodic canopy signatures that betray weekly or daily trading cycles, giving municipalities a defensible evidence base for revenue, health and transport decisions.
Sensors
- Planet SkySat: 50 cm native resolution (resampled product at 50 cm), with the ability to task the same site multiple times per day. That intra-day revisit is the key capability here: a market canopy present at 09:00 and gone by 18:00 is only detectable if you have both acquisitions.
- Maxar WorldView Legion: 30 cm panchromatic resolution from a six-satellite constellation designed to achieve up to 15 revisits per day over high-priority sites. At 30 cm, individual stall structures roughly 1 m wide are marginally resolvable, though reliable detection requires stalls to cluster into groups of at least 3 to 5 metres across.
- Airbus Pléiades Neo: 30 cm panchromatic, four spectral bands, two satellites offering revisit of up to several times daily over the same target. The four-band multispectral product (including red-edge) helps separate synthetic tarpaulin colours from vegetation and permanent roofing materials.
- Satellogic EarthView: 70 cm multispectral with a 1 m panchromatic product and a large constellation targeting sub-daily revisit. Slightly coarser than SkySat or Pléiades Neo, but the hyperspectral mode (30 m, 30 bands) can assist material classification of canopy fabrics where spectral discrimination matters.
What a market canopy gives away
A permanent building has a fixed spectral and geometric signature that barely changes between Monday and Saturday. A market stall does not. The analytical logic here is simple: compare imagery of the same location on a trading day against a non-trading day, and the difference image isolates the temporary structures. In practice, the signal is a cluster of small, high-reflectance objects, often with saturated colour values in the visible bands, appearing in streets, open ground or car parks that are otherwise bare.
Canopy fabrics, typically polyethylene tarpaulins in blue, orange or white, have spectral signatures that differ measurably from concrete, asphalt and clay tile roofing. Pléiades Neo's four-band multispectral product, and Satellogic's hyperspectral mode, can push that discrimination further. The honest limit: at 50 cm resolution, a single 1.5 m stall occupies roughly nine pixels. Reliable automated detection requires stalls to aggregate into patches of at least 10 to 15 square metres. Isolated single vendors are below the detection threshold of any current commercial satellite.
The permanent-awning problem
The most persistent source of false positives is the permanent shop awning. A row of fixed canopies over a retail strip looks, in a single image, almost identical to a line of market stalls. Two things resolve the ambiguity. First, temporal persistence: a permanent awning is present in every acquisition, including those taken on days or hours when no trading occurs. Second, geometric regularity: permanent awnings tend to align flush with building frontages, while informal market canopies cluster in the road reserve or open space and show irregular spacing.
Multi-time-of-day acquisitions are therefore not optional. A single overpass at 10:30 on a Tuesday tells you almost nothing about whether a structure is temporary. A pair of acquisitions, one at 08:00 and one at 17:00 on the same day, combined with a baseline image taken on a known non-market day, gives the three-point comparison needed to classify a feature as periodic rather than permanent. Planet SkySat's tasked multi-collect capability and WorldView Legion's high-revisit architecture are the only commercially available systems that routinely support this within a single day over a specific urban block.
Building the temporal stack
The analytical workflow starts with image co-registration to sub-pixel accuracy, which at 30 to 50 cm resolution means alignment errors must be held below roughly 15 cm, or half a pixel. Standard orthorectification using a lidar or stereo-derived DSM achieves this over flat terrain; hilly markets introduce larger residuals that must be accounted for in the change threshold.
Once co-registered, a temporal stack of six to twelve acquisitions covering at least two full weekly cycles gives enough variance to classify each pixel or object as stable, periodic or erratic. Periodic pixels, those that toggle between high-reflectance and background on a consistent cadence, are the market signature. Erratic pixels often indicate construction activity or parked vehicles and should be flagged separately rather than folded into the market footprint. The output is a probability surface: not a binary market or no-market layer, but a confidence score that lets a municipal officer decide where to invest ground-truth survey effort.
What municipalities actually do with this
Revenue collection is the most direct application. Informal market traders in most jurisdictions owe some form of daily or weekly pitch fee, but collection is manual and patchy. A satellite-derived map of trading-day footprints, showing which areas are active on which days, gives a revenue authority a spatial audit trail that is independent of on-the-ground enforcement records. It does not replace those records, but it identifies the gaps.
Public-health zoning is the second major use. High-footfall informal markets generate waste, wastewater and food-safety risks that are proportional to the active trading area, not the cadastral designation of the land. A health authority that can quantify the periodic extent of a market, rather than relying on a fixed boundary drawn years ago, can allocate sanitation infrastructure and inspection resources more accurately.
Transport planning around market days is a third application that is underused. Markets that double the pedestrian density of a street corridor on specific days create predictable congestion patterns. A satellite-derived market calendar, cross-referenced with road network data, gives traffic engineers a spatial and temporal input that survey counts alone rarely capture at city scale.
Honest limits of the method
Cloud cover is the most disruptive constraint. Many informal markets in tropical cities operate year-round, but optical satellites cannot see through cloud. A wet-season analysis may yield only two or three usable acquisitions per month over a given site, which is insufficient for reliable periodicity classification. SAR data from Sentinel-1 can detect large-scale changes in surface roughness associated with market activity, but at 10 m resolution it cannot resolve individual stall structures. SAR is useful as a scheduling guide, not as the primary detection layer.
Shadows are a secondary problem. In dense urban canyons, market stalls in shadow are spectrally indistinguishable from the road surface. Acquisitions taken within two hours of solar noon minimise shadow length and are strongly preferred. Off-nadir collection angles, which commercial operators sometimes use to increase revisit frequency, increase shadow extent and reduce detection reliability.
Finally, the method maps trading area, not trader count or turnover. It is a spatial and temporal footprint, not an economic census. Satellize's analytics for the Kingdom of Tonga crop-estimation programme use a similar temporal-stack logic to distinguish active from fallow fields; the underlying principle transfers, but the calibration and ground-truth requirements differ substantially between agricultural and urban contexts.
From imagery to a usable planning layer
The deliverable a municipality needs is not a stack of satellite images. It is a GIS polygon layer showing, for each identified market node, the trading-day footprint in square metres, the days and approximate hours of activity, a confidence score, and a flag for nodes that have expanded or contracted relative to a baseline period.
That layer should be updated on a quarterly or annual cycle, not continuously. Markets do shift, but the planning decisions they inform, pitch-fee schedules, sanitation contracts, road-closure orders, operate on administrative timescales. A quarterly refresh is usually sufficient. For cities that want to monitor a specific high-priority market more closely, a tasked revisit programme using SkySat or WorldView Legion can deliver monthly or even weekly updates at additional cost and with the cloud-cover caveats noted above.
Typical figures
| Best available spatial resolution | 30 cm panchromatic (Maxar WorldView Legion, Airbus Pléiades Neo) |
| Minimum detectable market cluster | Approximately 10 to 15 m² of aggregated canopy; single isolated stalls below detection threshold |
| Intra-day revisit (tasked) | Up to 15 passes per day over a priority site (WorldView Legion); multiple collects per day (SkySat tasked) |
| Temporal stack requirement | Minimum 6 acquisitions across at least 2 full weekly cycles for reliable periodicity classification |
| Spectral bands used | Panchromatic plus visible RGB and near-infrared; red-edge (Pléiades Neo) for canopy fabric discrimination |
| Cloud-cover constraint | Optical only; tropical wet seasons may yield fewer than 3 usable acquisitions per month |
| Co-registration accuracy required | Sub-pixel, typically below 15 cm error over flat terrain with DSM-based orthorectification |
| Archive depth | Planet SkySat archive from approximately 2017; WorldView archive from 2009 (WorldView-1/2); Pléiades from 2012 |
| Preferred acquisition time | Within 2 hours of solar noon to minimise shadow extent; multi-time-of-day pairs required for periodicity analysis |
| Delivery format | GeoTIFF change layers, GeoPackage or Shapefile polygon outputs, confidence-scored CSV attribute table |
Analytics Satellize can run
| Trading-day footprint polygons | Bi-temporal and multi-temporal change detection using object-based image analysis on co-registered very-high-resolution stacks | GIS polygon layer with area, day-of-week activity flags and confidence score per market node |
| Market periodicity classification | Time-series variance analysis across 6 to 12 acquisitions; pixels classified as stable, periodic or erratic based on reflectance cadence | Raster probability surface and summary table of active days per identified node |
| Permanent-versus-temporary canopy discrimination | Persistence filter: features present in all acquisitions flagged as permanent; features present only on known trading days retained as informal market structures | Classified vector layer distinguishing permanent awnings from periodic market canopies |
| Market expansion and contraction trend report | Year-on-year comparison of trading footprint polygons; area delta calculated per node against a defined baseline quarter | PDF or dashboard report showing footprint change in m² and percentage, with annotated image pairs |
| Revenue-gap spatial audit | Overlay of satellite-derived active trading area against municipal pitch-fee collection records to identify unregistered or under-recorded nodes | Tabular report and map layer flagging nodes with no corresponding revenue record |
| Market-day transport pressure index | Intersection of trading-day footprint with road network buffer zones; footfall proxy derived from canopy density per street segment | GIS layer of road segments with market-day pressure score, formatted for import into transport planning software |
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.