Sensors
- Sentinel-2 MSI: 10 m optical resolution across 13 spectral bands; 5-day revisit at the equator (2-3 days at mid-latitudes with both satellites). Free archive from 2015. Sufficient to detect new building footprints above roughly 200 m² and field boundary changes. Cloud cover remains the primary operational constraint.
- Planet SuperDove: 3 m resolution, 8-band optical, near-daily revisit globally. Resolves individual structures, access tracks and small polytunnel expansions that Sentinel-2 misses. Commercial tasking required; archive depth varies by site.
- Airbus Pléiades: 0.5 m panchromatic, 2 m multispectral. Used for confirmation imagery on flagged parcels: resolves construction materials, vehicle presence and fence breaches that lower-resolution sensors can only suggest. Tasked on demand; not suitable as a routine area-wide monitor.
- Sentinel-1 SAR C-band: 6-day repeat (12-day single satellite), 10 m IW mode, all-weather day-and-night acquisition. Coherence change detection identifies new hard surfaces and earthworks even under persistent cloud, complementing optical layers where cloud cover exceeds acceptable thresholds.
Why proximity to military land is a valuation risk, not just a planning curiosity
Gazetted military buffer zones and exclusion areas carry statutory powers in most jurisdictions that allow compulsory acquisition, development prohibition or use restriction with limited notice and, in some cases, below-market compensation. A parcel that sits 400 metres from an active training range or airfield clear zone may carry no visible encumbrance in a title search today and a compulsory purchase notice within a planning cycle. The risk is asymmetric: upside from development is capped by proximity constraints; downside from acquisition is uncapped relative to purchase price.
The problem is compounded because military land boundaries are often poorly georeferenced in public cadastral systems. Gazetted descriptions may reference physical features that have since changed, and the operational footprint of an installation sometimes differs from its legal boundary. Buyers, lenders and valuers who rely solely on title documents routinely miss the gap between where the fence sits and where the legal boundary runs.
What change detection actually sees, and what it cannot
Multitemporal optical change detection works by comparing co-registered image pairs or stacks across time, isolating pixels whose spectral signature has shifted beyond a statistically defined threshold. At Sentinel-2's 10 m resolution, a new building cluster of around 200 m² is detectable as a spectral change, though individual small structures may fall below the noise floor. Planet SuperDove at 3 m resolves single-dwelling footprints reliably. Pléiades at 0.5 m can confirm construction type and progress.
SAR coherence change detection adds a complementary layer. When a C-band Sentinel-1 coherence map shows a sudden decorrelation patch adjacent to a boundary, that almost always indicates ground disturbance: earthworks, new foundations or vegetation clearance. It does not tell you what was built, but it tells you something changed, regardless of cloud.
The method has clear limits. It detects surface changes only. Subsurface contamination, unexploded ordnance extent, soil toxicity from historical military use, and underground infrastructure are invisible to any passive or active optical sensor. A parcel flagged as encroaching carries elevated planning and acquisition risk; it does not carry a clean bill of environmental health. Those assessments require ground investigation.
The boundary georeferencing problem
Before change detection is meaningful, the legal boundary of the military zone must be accurately placed on the ground. This is harder than it sounds. Many gazetted military areas in Commonwealth jurisdictions, across parts of Southeast Asia, and in older NATO member states were surveyed to standards that produce positional errors of tens to hundreds of metres when converted to modern geodetic datums. A buffer zone drawn from a 1960s survey description may be offset by 50 m or more from its intended position.
The practical approach is to digitise the legal boundary from the gazette description, then cross-reference it against the visible fence line or cleared perimeter in the highest-resolution available imagery. Where those two lines diverge, the conservative assumption for risk purposes is to treat the outer envelope as the operative boundary. Parcels within that outer envelope are flagged for legal review, not automatically condemned. The satellite layer narrows the search; a solicitor with access to the relevant defence estate records makes the determination.
Designing a monitoring programme that does not cry wolf
The failure mode of any automated change-detection system is alert fatigue. A pipeline that flags every spectral change within a 500 m buffer will bury the genuine encroachments under agricultural seasonal variation, shadow shifts and sensor artefacts. Several design choices reduce false positives materially.
First, mask known seasonal land covers: arable fields change spectrally every growing cycle without any built development occurring. A crop mask derived from multi-year NDVI time series removes most of this noise before the change algorithm runs. Second, require spatial persistence: a change must appear in at least two independent acquisitions, separated by a minimum interval, before generating an alert. Single-date anomalies are usually sensor artefacts. Third, size-threshold the detections: changes smaller than a defined minimum area (typically 150 to 300 m² depending on sensor resolution) are logged but not alerted, unless they cluster.
Even with these filters, some false positives survive. The output should be treated as a ranked list of parcels warranting human review, not a definitive encroachment register.
Archive depth and the retrospective risk question
One underused capability is the retrospective audit. Sentinel-2 data is freely available from late 2015; Landsat archive extends to 1972 at 30 m resolution through the USGS Earth Explorer platform. For a parcel whose planning history is disputed, or where a vendor claims no development has occurred since a particular date, a time series analysis can confirm or contradict that claim with reasonable confidence.
This matters in acquisition due diligence because some jurisdictions treat the date of encroachment as relevant to compensation calculations in compulsory purchase proceedings. A parcel that was already built upon before a buffer zone was extended may have different legal standing than one developed after the gazette notice. Satellite archive evidence is not a substitute for legal advice, but it can frame the question precisely enough to make legal advice cheaper to obtain.
Satellize has applied similar multitemporal change-detection methods in agricultural contexts, including its crop-estimation programme for the Kingdom of Tonga, and the same underlying pipeline adapts to built-environment change with adjustments to the spectral indices and minimum mapping units used.
Integrating satellite flags into property due diligence workflows
The output of a boundary encroachment analysis is most useful when it arrives early in the due diligence process, not as a final check. A GIS layer showing flagged parcels within a defined proximity band around a military installation, colour-coded by confidence tier and change recency, can be overlaid against a transaction pipeline in a standard property data platform. Parcels in the highest-confidence tier go to legal review immediately; lower tiers are noted in the risk register and monitored on a quarterly refresh cycle.
Delivery format matters. A PDF report is useful once; a GIS polygon layer with attribute fields for detection date, confidence score, sensor source and change area integrates into the tools valuers and lenders already use. Alert feeds for ongoing monitoring are typically delivered as API calls or email digests when new detections cross the alert threshold.
The satellite layer does not replace a physical site visit or a search of the relevant defence estate register. It identifies which parcels in a large portfolio or acquisition target list are worth the cost of those investigations. That prioritisation is where the analytical value sits.
Typical figures
| Spatial resolution (routine monitoring) | 10 m (Sentinel-2); 3 m (Planet SuperDove) |
| Spatial resolution (confirmation) | 0.5 m panchromatic (Airbus Pléiades) |
| Revisit frequency | 2–5 days optical (Sentinel-2 + Planet combined); 6–12 days SAR (Sentinel-1) |
| All-weather capability | Sentinel-1 SAR C-band; cloud does not degrade SAR coherence detection |
| Minimum detectable built change | Approximately 150–300 m² at 3–10 m resolution (single structure may be below floor at 10 m) |
| Spectral bands used | Visible, NIR, SWIR (optical change and NDVI); C-band microwave (SAR coherence) |
| Archive depth | Sentinel-2 from 2015; Landsat from 1972 at 30 m via USGS |
| Analysis latency | Routine change flags: 24–72 hours after scene acquisition; confirmation tasking: subject to satellite pass scheduling |
| Coverage | Global; no geographic restriction on open-constellation data |
| Delivery formats | GIS polygon layers (GeoJSON, Shapefile, GeoPackage), PDF risk summary, API alert feed |
Analytics Satellize can run
| Boundary proximity risk layer | Legal boundary digitisation cross-referenced against visible perimeter in high-resolution imagery; conservative outer-envelope buffering | GIS polygon layer with confidence-tier attributes, delivered as GeoJSON or Shapefile |
| Built-change detection alerts | Multitemporal spectral change detection on Sentinel-2 and Planet SuperDove stacks; persistence filter requiring detection in two or more independent acquisitions | Ranked alert list with detection date, area, sensor source and confidence score; refreshed on defined cycle |
| SAR coherence change flags | Sentinel-1 interferometric coherence differencing to detect ground disturbance under cloud cover | Supplementary GIS layer of coherence-loss patches, attributed by date and magnitude |
| Retrospective encroachment timeline | Landsat and Sentinel-2 archive time-series analysis; spectral indices and visual interpretation at annual or seasonal intervals | PDF report with annotated image sequence showing development history by parcel |
| Crop and seasonal-change mask | Multi-year NDVI time-series classification to exclude agricultural spectral variation from built-change alerts | Masking layer applied upstream of change-detection pipeline; reduces false-positive rate |
| Confirmation high-resolution assessment | Pléiades 0.5 m tasked imagery on highest-confidence flagged parcels; manual interpretation of construction type, access and fence condition | Single-parcel PDF with annotated imagery and written assessment of observed change |
| Portfolio encroachment screening | Batch processing of multiple parcels or sites against a defined military boundary dataset; tiered output by proximity band | Spreadsheet or GIS layer covering full portfolio, with per-parcel risk tier and recommended action |
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.