Sports and leisure facility mapping for residential catchment valuation
High-resolution optical imagery classifies outdoor sports surfaces and recreational green space by spectral signature and texture, then models proximity buffers to quantify amenity uplift across large property portfolios where ground survey is impractical.
Sensors
- Airbus Pléiades Neo: 0.3 m panchromatic, 0.5 m multispectral (blue, green, red, red-edge, near-infrared). At this resolution individual court lines, penalty arcs and bunker lips are resolved, making surface-type classification unambiguous without texture inference. Revisit roughly 1 day at mid-latitudes with the two-satellite constellation.
- Maxar WorldView-3: 1.24 m multispectral across 8 VNIR bands plus 8 SWIR bands. The SWIR bands distinguish synthetic rubber infill from natural grass even when NDVI values overlap, and the archive stretches back to 2014, enabling multi-year amenity-change histories.
- Planet SuperDove: 3 m resolution, 8 spectral bands including red-edge and near-infrared, daily global revisit. Coarser than Pléiades Neo for court-line detection but adequate for golf-course boundary delineation and green-space area measurement. Useful for portfolio-scale monitoring where hundreds of sites need regular refresh.
- Sentinel-2 (ESA): 10 m in visible and near-infrared bands, 5-day revisit at the equator, freely available. Too coarse to resolve individual courts but reliable for computing NDVI baselines across large recreational green-space polygons and for change detection between annual surveys.
Why surface type matters more than acreage
A hectare of artificial turf and a hectare of amenity grassland carry very different valuation implications. The turf supports structured sport year-round; the grassland may flood or close seasonally. Residential catchment models that treat all green space as equivalent systematically misstate amenity supply. Satellite classification by surface type fixes this without a single site visit.
The spectral logic is straightforward. Natural grass returns a strong near-infrared signal and a characteristic NDVI (Normalised Difference Vegetation Index) that varies with season and maintenance intensity. Artificial turf made from polyethylene or nylon fibres absorbs near-infrared differently and often contains rubber crumb infill that has a distinct SWIR signature, particularly visible in WorldView-3's SWIR bands. Clay courts are high-reflectance in the red channel and low in NIR. Hard courts, whether asphalt or acrylic, have a flat, low-NDVI spectral profile with strong texture regularity at sub-metre resolution. These differences are not subtle; they are the basis of published classification workflows in peer-reviewed remote-sensing literature.
What a floating roof gives away, and what a bunker lip confirms
Golf courses are among the easiest large recreational assets to delineate from orbit. The combination of maintained fairway grass, sand bunker spectral contrast, water-hazard blue, and the geometric regularity of greens produces a morphological signature that automated object-based image analysis (OBIA) identifies reliably at 0.5 m resolution. Pléiades Neo imagery resolves flagstick shadows. That is more than enough to distinguish a nine-hole municipal course from a driving range or a country club with practice facilities.
Swimming complexes present a different challenge. Outdoor pool water has a distinctive blue-green spectral signature, but the more useful discriminator at 0.3 m is the surrounding hard-surface apron, lane-marker geometry and, where present, the curved roof of a leisure centre. Indoor pools are invisible spectrally but their building footprint and car-park geometry can flag likely locations for ground-truth follow-up. Honest caveat: a covered pool with no outdoor element cannot be confirmed from optical imagery alone.
Tennis and multi-use game areas (MUGAs) are resolved clearly at Pléiades Neo resolution. Court lines at 2.74 m width are directly measurable. The colour of the court surface (red clay, green or blue acrylic, green artificial grass) adds a classification layer that distinguishes premium facilities from basic provision, which matters when modelling the amenity premium a property agent would actually cite.
From classified polygons to valuation buffers
Classification produces a geodatabase of facility polygons, each tagged with surface type, estimated area, and a confidence score. The next step is proximity modelling. Published hedonic pricing studies consistently find that residential property values respond to walkable access to recreational amenity, with effects typically measured within 400 m to 1,200 m buffers, though the published range varies considerably by facility type and urban context. Satellize does not assert a specific uplift percentage; the buffer distances and weighting factors are inputs the client's valuation team calibrates against their own transaction data.
What satellite analysis supplies is the denominator: a consistent, auditable count of facility area by type within any buffer, updated on a defined schedule. For a REIT assessing 300 residential assets across multiple cities, that consistency matters more than any single precise uplift figure. A ground survey of 300 sites would take months and introduce surveyor-to-surveyor variation. A satellite pass and an automated classification pipeline take days and apply identical rules everywhere.
Honest limits of the method
Cloud cover is the most obvious constraint. Optical sensors cannot see through cloud. In persistently overcast climates, acquiring a usable Pléiades Neo or WorldView-3 image in a specific month may require multiple tasking attempts or a fallback to archive imagery that could be six to eighteen months old. Planet's daily revisit improves the odds of finding a clear acquisition but at lower resolution.
Indoor facilities are largely opaque to this method. A sports hall, an indoor tennis centre, or a covered swimming pool leaves no spectral trace. Their existence can sometimes be inferred from building morphology and car-park geometry, but that inference carries lower confidence and should be flagged as such in any deliverable.
Minimum detectable target size is a practical floor. At 3 m Planet resolution, a single-court MUGA of roughly 600 square metres is detectable but not cleanly classifiable by surface type. At 0.3 m Pléiades Neo, the same court is fully resolved. Clients with portfolios in dense urban areas, where small facilities matter, should specify sub-metre tasking rather than relying on medium-resolution open data.
Finally, classification accuracy degrades when facilities are poorly maintained. Artificial turf that has lost its infill, or a clay court that has been overseeded with grass, can produce ambiguous spectral signatures. A well-designed workflow includes a confidence score on each polygon; low-confidence features should be flagged for targeted review rather than carried silently into valuation models.
Portfolio-scale delivery and the Satellize approach
For a local authority or REIT covering a large urban region, the practical workflow is a one-time baseline classification using the best available archive imagery at the highest affordable resolution, followed by periodic refresh using Planet's daily constellation to catch new facility construction, closures or surface-type changes. Change alerts can be triggered automatically when NDVI or spectral signatures within a known facility polygon shift beyond a defined threshold.
Satellize runs this classification pipeline on open constellations and adds commercial tasking on client licence. The Tonga crop-estimation programme demonstrated the same underlying logic at national scale: spectral classification of surface types, area measurement, and periodic refresh. The methods transfer directly to urban recreational mapping, with the principal difference being that sports surfaces are more spectrally stable than crops and require less frequent revisit to maintain an accurate baseline.
Outputs are delivered as GIS-ready polygon layers (GeoPackage or shapefile), facility-level attribute tables, and, where requested, a proximity-buffer summary table keyed to the client's existing property asset identifiers. The pipeline is documented so that a client's internal GIS team can rerun it independently on future imagery without returning to Satellize for each update cycle.
Typical figures
| Best spatial resolution (commercial) | 0.3 m panchromatic (Pléiades Neo); 0.5 m multispectral |
| Practical resolution for surface-type classification | 0.5 m to 1.24 m multispectral; 3 m adequate for area measurement only |
| Revisit frequency | Daily (Planet SuperDove); ~1 day (Pléiades Neo, 2-satellite); 5 days (Sentinel-2) |
| Spectral bands used | Blue, green, red, red-edge, NIR for NDVI and surface classification; SWIR (WorldView-3) for synthetic-material discrimination |
| Minimum detectable facility | ~600 m² single court detectable at 3 m; cleanly classifiable by surface type at 0.3–0.5 m |
| Cloud limitation | Optical only; persistent overcast may delay usable acquisition by days to weeks depending on location |
| Archive depth | WorldView-3 from 2014; Sentinel-2 from 2015; Planet from ~2016; Pléiades Neo from 2021 |
| Delivery formats | GeoPackage, Shapefile, GeoTIFF classification raster, CSV attribute table keyed to asset IDs |
| Typical processing latency | 2–5 days from clear image acquisition to classified polygon layer, depending on area of interest size |
Analytics Satellize can run
| Sports surface classification layer | Object-based image analysis (OBIA) combined with NDVI, SWIR band ratios and texture metrics (GLCM); supervised classification trained on labelled surface types | Polygon GIS layer with per-feature attributes: surface type, area (m²), confidence score, acquisition date |
| Golf course facility delineation | Morphological pattern recognition of fairway/bunker/green/water-hazard spectral signatures at sub-metre resolution; OBIA segmentation | Polygon layer distinguishing course boundary, practice areas and clubhouse footprint; attribute table with hole count estimate where resolvable |
| Amenity proximity buffer table | Euclidean and network-distance buffers (400 m, 800 m, 1,200 m) from facility centroids or edges to residential asset coordinates; area-weighted facility score per buffer | CSV summary table keyed to client property asset IDs; optional choropleth GIS layer for portfolio visualisation |
| Facility change detection alert | Bi-temporal spectral differencing on Planet daily imagery within known facility polygons; threshold alert when NDVI or surface-class signature shifts beyond defined tolerance | Automated alert feed (JSON or email) flagging facility closures, surface-type changes or new construction within monitored catchments |
| Recreational green-space area inventory | NDVI thresholding on Sentinel-2 or SuperDove imagery to delineate maintained green space; cross-referenced against classified sports surfaces to separate active from passive green space | Area statistics per local authority zone or REIT portfolio cluster; time-series chart of green-space coverage across available archive |
| Indoor facility inference layer | Building footprint morphology and car-park geometry analysis at 0.3–0.5 m resolution to flag probable indoor leisure centres; explicitly low-confidence, for ground-truth prioritisation only | Point layer of candidate indoor facilities with confidence flag and recommended verification status; not suitable for direct valuation input without ground confirmation |
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.