Urban waterbody and riparian encroachment mapping
Satellite water-index time series and sub-metre imagery can distinguish legally protected riparian buffers from seasonally shifting shorelines, giving enforcement agencies defensible evidence of permanent encroachment.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at the equator with both satellites. Provides the MNDWI and AWEI time series needed to characterise seasonal water-extent variation over multi-year baselines. Free archive from 2015.
- Landsat 4–9 archive: 30 m multispectral resolution; archive extends to 1982, giving four decades of shoreline history. Coarser than Sentinel-2 but irreplaceable for establishing pre-encroachment baselines in cities where development began in the 1980s and 1990s.
- Airbus Pléiades Neo: 30 cm panchromatic, 70 cm multispectral. Sufficient resolution to delineate individual building footprints and measure setback distances to registered waterbody boundaries. Required for enforcement-grade evidence; tasked commercially on demand.
- Planet SuperDove: 3 m resolution, daily revisit in eight spectral bands including red-edge. Useful for monitoring active construction events within buffer zones between higher-resolution tasking windows. Red-edge band improves vegetation-to-structure discrimination at the riparian fringe.
Why water indices alone cannot define an encroachment
The Modified Normalised Difference Water Index (MNDWI) and the Automated Water Extraction Index (AWEI) are well-established spectral methods for delineating open water from satellite imagery. Both exploit the strong absorption of shortwave infrared energy by liquid water against the relative reflectance of soil and built surfaces. At Sentinel-2's 10 m resolution, they reliably map water extents to within one or two pixels of the true shoreline under clear-sky conditions.
The problem is that a shoreline is not a fixed line. In monsoon climates, lake surfaces can expand by tens of metres between the dry season minimum and the post-monsoon maximum. A structure built on ground that is inundated for six weeks a year may still sit entirely outside the legal buffer for the other ten months. Conversely, a structure built during a drought year may appear to stand clear of the water in every image taken since construction. Neither the water index nor any single image resolves this ambiguity. What resolves it is the registered legal boundary, combined with a multi-year water-extent distribution that characterises the full seasonal envelope.
Building a defensible seasonal envelope from archive data
The practical workflow starts with the Landsat archive. Because Landsat 5 TM, 7 ETM+ and 8–9 OLI share consistent band geometry and calibration, it is possible to compute MNDWI across scenes spanning 1984 to the present and stack them into a per-pixel water-frequency distribution. A pixel that shows water in more than, say, 80 per cent of cloud-free observations is permanently aquatic. A pixel that shows water in 10 to 40 per cent of observations is the seasonally variable fringe. Structures that sit inside the high-frequency water zone are almost certainly encroaching regardless of when they were photographed.
Sentinel-2's denser time series from 2015 refines the same analysis at 10 m. The combination gives an assessor two things: a long historical baseline to establish what the waterbody looked like before recent urban pressure, and a recent high-cadence record to date when encroachment began. Dating matters legally. A structure built before a riparian protection regulation came into force may be grandfathered; one built after is not.
Cloud cover is the persistent constraint. In monsoon regions, the wet season, when the lake is at maximum extent, is also when cloud cover is heaviest. Synthetic aperture radar from Sentinel-1 C-band can penetrate cloud and detect open water, though it introduces its own ambiguities at the urban fringe where radar backscatter from buildings and wet vegetation can be confused. A combined optical-plus-SAR approach improves wet-season coverage but requires careful interpretation.
What a floating roof gives away
Once the seasonal envelope is established, the enforcement question becomes structural: is there a built object inside the legal buffer? At Sentinel-2 or Landsat resolution, a single-storey building of 50 to 100 square metres is detectable as a spectral anomaly but not reliably classifiable as a building rather than a hard-surfaced yard or a jetty. Pléiades Neo at 30 cm panchromatic changes this entirely. Individual roof materials, wall shadows, and the gap between a structure and the water's edge are all measurable.
Pléiades Neo imagery has been used in published urban-mapping research to achieve building footprint delineation at sub-metre accuracy, which is sufficient to compute a setback distance and compare it against a registered buffer width. The output is not an estimate; it is a measurement. That distinction matters when the deliverable is an enforcement notice rather than a planning study.
The registration problem: when the legal boundary is itself uncertain
Satellite analysis can only measure encroachment relative to a reference boundary. In many cities, the registered waterbody boundary is itself contested, outdated or digitised from paper maps at 1:25,000 scale, which introduces positional errors of 10 to 25 metres before any satellite measurement is made. Where the legal boundary is poorly georeferenced, even 30 cm imagery cannot produce a defensible setback measurement.
The honest answer to this is that the satellite workflow and the cadastral workflow must proceed together. Satellite-derived shoreline positions from high-frequency Sentinel-2 time series can actually help re-establish a defensible full-supply-level contour, which some jurisdictions use as the legal reference line. In that case the satellite data is not just evidence of encroachment; it contributes to defining what the protected boundary should be.
From map to monitoring: keeping the record current
A one-time encroachment survey ages quickly in fast-growing cities. Planet SuperDove's daily revisit at 3 m provides a practical change-detection layer between periodic Pléiades Neo tasking. A new rooftop appearing inside a buffer zone will typically be detectable in SuperDove imagery within days of the roof being laid, because the spectral signature of fresh construction materials differs sharply from the surrounding vegetation and soil. That detection triggers a tasking order for Pléiades Neo confirmation, rather than requiring continuous high-resolution coverage of the entire waterbody perimeter.
Satellize structures this kind of tiered monitoring for clients who need ongoing enforcement support rather than a single baseline map. The Tonga crop-estimation programme is a different domain, but the underlying architecture, open-constellation time series for change detection with commercial tasking for confirmation, applies equally well to riparian enforcement.
Honest limits of the method
Minimum detectable encroachment at Sentinel-2 is roughly 100 square metres of hard surface, which means small informal additions such as a concrete platform or a narrow jetty extension are invisible without commercial imagery. Pléiades Neo can detect structures down to around 10 to 20 square metres, but tasking cost means it is not practical to run continuously across an entire city's waterbody network.
Cloud cover in wet-season tropical environments can reduce usable Sentinel-2 observations to fewer than four or five per monsoon season, which is enough to detect major changes but not to date them precisely. Sentinel-1 SAR partially compensates but is not a substitute for optical classification of built materials. Finally, no satellite method can determine whether a structure has planning permission. The imagery shows what is there and when it appeared; the legal status requires cross-referencing with permit records held by the planning authority.
Typical figures
| Spatial resolution (baseline monitoring) | 10 m (Sentinel-2 MSI) |
| Spatial resolution (enforcement evidence) | 30 cm panchromatic / 70 cm multispectral (Pléiades Neo) |
| Spatial resolution (change detection) | 3 m (Planet SuperDove) |
| Revisit cadence | 5 days (Sentinel-2 dual satellite); daily (SuperDove); on-demand tasking (Pléiades Neo) |
| Archive depth | 1982–present (Landsat); 2015–present (Sentinel-2) |
| Key spectral indices | MNDWI (Green–SWIR1 / Green+SWIR1); AWEI (multi-band linear combination) |
| Minimum detectable encroachment | ~100 m² at Sentinel-2; ~10–20 m² at Pléiades Neo |
| Wet-season optical availability (monsoon tropics) | Typically 4–6 usable Sentinel-2 scenes per monsoon season; Sentinel-1 SAR supplements |
| Positional accuracy (Pléiades Neo) | ~1 m CE90 with ground control; ~3 m without |
| Delivery formats | GeoTIFF water-frequency rasters, GeoPackage encroachment polygons, PDF enforcement reports with image evidence |
Analytics Satellize can run
| Multi-year water-frequency map | Per-pixel MNDWI/AWEI frequency distribution across Landsat and Sentinel-2 archive | GeoTIFF raster showing permanent, seasonal and variable water zones; used as baseline reference |
| Seasonal shoreline envelope | Automated water-body extraction at monthly composites across full archive; percentile contour derivation | Vector polygon set of minimum, median and maximum shoreline positions for each waterbody |
| Encroachment detection layer | Overlay of registered legal buffer boundary against current Pléiades Neo or SuperDove building footprints | GeoPackage of flagged structures with setback distance, area and earliest detectable date of construction |
| Construction-event alerts | Change detection on SuperDove 3 m time series using spectral differencing against rolling baseline | Automated alert with coordinates and thumbnail image, triggering commercial tasking order for confirmation |
| Enforcement evidence package | Pléiades Neo orthorectified imagery with annotated overlays; date-stamped archive imagery for temporal sequence | PDF report per structure: coordinates, area, setback measurement, image time series, sensor metadata |
| Riparian vegetation loss assessment | NDVI time series on Sentinel-2 along buffer zone; comparison of pre- and post-encroachment canopy cover | Area statistics of vegetation loss by waterbody segment; GIS layer for ecological impact reporting |
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.