Aquaculture pond extent and water-quality monitoring
Multispectral and SAR imagery maps aquaculture pond extent, detects illegal encroachment into mangrove zones, and tracks surface water-quality proxies tied to productivity and disease risk.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands, 5-day revisit at the equator with both satellites. Blue (Band 2), green (Band 3), red (Band 4), red-edge (Bands 5–7) and near-infrared (Band 8) support turbidity and chlorophyll-a index derivation. The 20 m red-edge bands are particularly sensitive to algal concentration in shallow pond water.
- Landsat 8/9 OLI: 30 m resolution, 16-day single-satellite revisit (8-day combined). The coastal-aerosol band (Band 1, 443 nm) improves shallow-water discrimination and suspended-sediment retrieval. Longer archive (Landsat 8 from 2013, Landsat 9 from 2021) supports decadal pond-expansion analysis.
- Sentinel-1 SAR (C-band): 10 m IW-mode ground-range detected imagery, 6-day revisit per satellite, cloud-independent. C-band backscatter distinguishes smooth open water (low return) from vegetated or disturbed surfaces (higher return), enabling detection of pond construction within mangrove canopy that optical sensors miss under cloud cover.
- PlanetScope (commercial tasking): 3–4 m resolution, near-daily revisit on client licence. Useful for resolving individual small ponds below Sentinel-2's practical mapping threshold and for confirming encroachment boundaries identified by SAR.
Why a pond's colour is an operational signal, not just aesthetics
Aquaculture pond water changes colour for reasons that matter commercially. Phytoplankton blooms, which correlate with dissolved oxygen stress and elevated ammonia, shift surface reflectance toward green wavelengths. Suspended sediment from pond disturbance or heavy feeding pushes reflectance into the red. Both shifts are detectable with Sentinel-2's 10 m bands and have been studied extensively in published aquaculture-remote-sensing literature as proxies for chlorophyll-a concentration and turbidity.
The red-edge bands on Sentinel-2 (705 nm and 740 nm) are particularly useful here. Chlorophyll absorption creates a steep reflectance gradient between red and near-infrared that is visible even in small ponds. Indices such as the red-edge chlorophyll index (CIre) and the normalised difference chlorophyll index exploit this gradient and have been validated against in-situ measurements in coastal shrimp and fish pond systems in Southeast Asia and South Asia. These are not perfect water-quality sensors: they measure surface optical properties, not dissolved oxygen or pathogen load directly. A turbid pond may or may not indicate disease. The signal narrows the search; it does not replace water testing.
Mapping pond extent: where optical ends and radar begins
Mapping aquaculture ponds from optical imagery is straightforward in clear conditions. Water absorbs strongly in the near-infrared, so a simple NDWI or MNDWI threshold separates open water from land. At Sentinel-2's 10 m resolution, ponds larger than roughly 0.02 hectares are detectable, though ponds narrower than two pixels in any dimension will be under-mapped. Landsat at 30 m sets a practical lower limit closer to 0.1 hectares for reliable delineation.
The harder problem is tropical cloud cover and mangrove encroachment. In major aquaculture regions across Southeast Asia and West Africa, cloud-free optical windows can be weeks apart during monsoon season. Sentinel-1 SAR fills that gap. Open water returns very low C-band backscatter in calm conditions, typically below minus 15 dB in VV polarisation, while mangrove canopy returns minus 5 to minus 8 dB. A newly cleared pond cut into mangrove forest produces a sharp, persistent low-backscatter patch that persists across acquisitions regardless of cloud. Combining SAR change detection with optical classification in a time-series stack is the most reliable approach for detecting illegal expansion, and several published studies have demonstrated this over the Mekong Delta and Sundarbans.
The mangrove encroachment problem is a legal and ecological one simultaneously
Mangrove-to-pond conversion is prohibited under national law in many countries and under international frameworks including Ramsar Convention commitments, yet it continues because enforcement depends on ground inspection of areas that are physically difficult to access. Satellite change detection changes that calculus. A Sentinel-1 time series with 6-day repeat provides near-continuous monitoring; clearing events that take days to weeks leave a detectable signature before the new pond is even operational.
The practical workflow involves baselining SAR backscatter over a known mangrove extent, then flagging persistent low-backscatter anomalies that appear within that boundary across consecutive acquisitions. False positives arise from tidal inundation of low mangrove areas and from calm-water specular reflection under certain wind conditions. Cross-checking against Sentinel-2 optical composites, when cloud-free windows exist, reduces false-alarm rates substantially. The output is a change polygon with a date range, not a confirmed legal finding, but it is sufficient to direct inspection resources.
Honest limits: what the physics cannot give you
Several constraints are worth stating plainly. First, water-quality indices derived from multispectral imagery are surface proxies. Chlorophyll-a and turbidity estimates degrade in very shallow ponds where bottom reflectance contaminates the signal, and in ponds with mixed algal species where the spectral signature differs from calibration assumptions. Published retrieval uncertainties for chlorophyll-a in inland waters using Sentinel-2 are typically 30 to 50 percent relative error without site-specific calibration.
Second, pond identification fails below the resolution floor. Artisanal ponds in West Africa and parts of South Asia are often under 0.05 hectares. Sentinel-2 will miss many of them; Landsat will miss most. Commercial 3–4 m imagery closes the gap but increases cost and processing time proportionally. Third, revisit frequency is not the same as useful revisit frequency. Cloud cover in humid tropical regions can render optical sensors effectively blind for weeks. SAR solves the cloud problem but cannot provide spectral water-quality information. A monitoring system that relies on either sensor alone will have gaps.
From pixels to a pond-management decision
The analytic chain from raw imagery to an actionable output has several steps, each of which introduces uncertainty. Atmospheric correction of Sentinel-2 data over water is more demanding than over land: aerosol scattering and adjacency effects from nearby vegetation can bias reflectance by several percent, which matters when the index values of interest span a narrow range. The Sen2Cor processor (ESA's standard tool) performs adequately over open ocean but is known to underperform in complex coastal and inland water environments. Alternative processors such as ACOLITE, developed for aquatic remote sensing, are better suited and are publicly documented.
Once corrected, pond delineation is typically done with a supervised classifier or a threshold-based water index, followed by object-based segmentation to separate individual ponds from connected water bodies. Water-quality indices are then computed per-pond and tracked through time. The deliverable that matters to an aquaculture operator is not a reflectance map but a ranked list of ponds showing anomalous turbidity or chlorophyll trends over the past two to four weeks, flagged against a historical baseline for that pond and season. Satellize structures analytics at this output level, drawing on the same principles applied in its Tonga crop-estimation programme: the satellite data is the input, but the product is the decision-relevant signal.
Building a monitoring programme that holds up over time
A one-off pond map has limited value. Aquaculture landscapes change fast: new ponds are constructed, others are abandoned, stocking cycles alter water quality on weekly timescales. A useful monitoring programme requires a consistent time-series approach with fixed processing parameters so that changes in index values reflect changes in the pond, not changes in the algorithm.
Archive depth is an asset here. Sentinel-2 data runs from 2015, Landsat from 1972 in lower resolution. A new client wanting to understand how a coastal aquaculture zone has expanded over the past decade, and which expansions occurred inside mangrove boundaries, can get a defensible answer from public archive data alone. For ongoing operational monitoring, the minimum useful cadence is roughly monthly for extent mapping and fortnightly for water-quality tracking, assuming cloud-free acquisitions are available at that frequency. In persistently cloudy regions, SAR-based extent monitoring at 6-day intervals combined with opportunistic optical water-quality assessment is the realistic operating mode.
Typical figures
| Spatial resolution (pond mapping) | 10 m (Sentinel-2), 30 m (Landsat 8/9), 10 m (Sentinel-1 SAR) |
| Minimum mappable pond size | ~0.02 ha at 10 m (Sentinel-2); ~0.1 ha at 30 m (Landsat); smaller ponds require commercial 3–4 m imagery |
| Revisit frequency | 5 days (Sentinel-2, equatorial); 8 days combined (Landsat 8+9); 6 days per SAR satellite (Sentinel-1) |
| Spectral bands used | Blue (443–490 nm), green (560 nm), red (665 nm), red-edge (705, 740 nm), NIR (842 nm) for water quality; C-band (5.4 GHz) SAR for extent and change |
| Water-quality proxy accuracy | Chlorophyll-a and turbidity retrievals: typically 30–50% relative error without site-specific calibration; improved with ACOLITE atmospheric correction |
| Cloud penetration | SAR only; optical sensors (Sentinel-2, Landsat) are cloud-blocked and require compositing strategies in humid tropical regions |
| Archive depth | Sentinel-2 from 2015; Landsat from 1984 (TM) at 30 m; Sentinel-1 from 2014 |
| Delivery formats | GeoTIFF pond-extent polygons, per-pond time-series CSV, GIS-ready vector layers (GeoPackage / Shapefile), anomaly alert reports |
| Latency (operational monitoring) | 2–5 days after satellite acquisition for processed analytic output, depending on cloud cover and processing queue |
Analytics Satellize can run
| Pond-extent map | MNDWI/NDWI thresholding plus object-based image analysis on Sentinel-2 or Landsat composites | Polygon GIS layer of delineated ponds with area and perimeter attributes, updated quarterly or on demand |
| Mangrove encroachment alert | Sentinel-1 SAR backscatter change detection within a reference mangrove boundary layer, cross-validated with optical composites | Change-polygon shapefile with acquisition date range, flagged for field inspection |
| Surface turbidity index | Normalised difference turbidity index (NDTI) or red-band ratio from atmospherically corrected Sentinel-2 reflectance (ACOLITE processor) | Per-pond turbidity time series, ranked anomaly report (PDF and CSV) on fortnightly cadence |
| Chlorophyll-a proxy | Red-edge chlorophyll index (CIre) from Sentinel-2 Bands 7 and 5, with seasonal baseline normalisation | Per-pond CIre trend chart and alert flag when values exceed two standard deviations above historical mean |
| Decadal expansion analysis | Multi-year Landsat and Sentinel-2 archive classification with change-detection overlay on mangrove extent datasets | Expansion timeline report showing pond area growth by year, with mangrove-loss attribution table |
| Pond condition dashboard | Fusion of SAR-derived extent with optical water-quality indices into a per-pond status layer, updated on each cloud-free acquisition | GIS-ready dashboard layer with traffic-light status per pond, exportable to client GIS or web viewer |
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.