Inland water body monitoring for freshwater fisheries management
Freshwater bodies present spectral conditions that defeat open-ocean water-quality models. Sentinel-2 MSI and Landsat 8/9 OLI offer the spatial resolution needed to monitor turbidity, cyanobacterial bloom extent and seasonal inundation in lakes and reservoirs that matter to inland fisheries managers.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible bands, 20 m in red-edge and shortwave infrared. Five-day revisit at the equator with two satellites. The red-edge bands (705 nm, 740 nm) are particularly useful for detecting phycocyanin-related cyanobacterial signals and separating bloom from suspended sediment. Minimum mappable water body roughly 0.5 ha.
- Landsat 8/9 OLI: 30 m multispectral resolution, 16-day revisit per satellite (8 days combined). The coastal/aerosol band (443 nm) and the high signal-to-noise OLI design improve shallow-water and turbidity retrievals compared with earlier Landsat sensors. Archive extends to 1972 for Landsat 1, giving decadal surface-area change context.
- Sentinel-3 OLCI: 300 m resolution, daily revisit. Twenty-one spectral bands purpose-built for ocean colour, including 620 nm and 665 nm bands sensitive to phycocyanin and chlorophyll. Practical only for water bodies wider than roughly 1 km; pixel contamination from surrounding land makes it unreliable on smaller lakes.
- MODIS-Aqua/Terra: 250 m in red and near-infrared bands, 500 m for the full spectral suite. Daily revisit is valuable for tracking rapid bloom development, but the coarse pixel footprint restricts use to large systems such as the African Great Lakes or Lake Titicaca. Archive runs from 2000, useful for long-term phenology studies.
Why inland water confounds the algorithms written for the sea
Open-ocean colour algorithms were designed for Case 1 waters, where phytoplankton and their co-varying products dominate the optical signal. Most freshwater bodies are Case 2: suspended inorganic sediment, coloured dissolved organic matter (CDOM) and cyanobacterial pigments all vary independently, and they absorb and scatter light in ways that overlap spectrally. Apply a standard chlorophyll-a retrieval to a turbid reservoir and you may be reading suspended clay, not algae.
The practical consequence is that every inland water body needs its own optical characterisation before satellite retrievals can be trusted. Empirical algorithms built on field measurements from the target lake outperform globally tuned models in most published validation studies. This is not a minor caveat; it is the central design constraint for any freshwater monitoring programme. Managers who skip the ground-truth step should treat satellite-derived chlorophyll figures as ordinal indicators of relative change, not absolute concentrations.
What a floating roof gives away: surface area and inundation as fisheries proxies
Surface-area change is the most reliable satellite product for inland waters because it requires only a water/land boundary, not a calibrated radiometric retrieval. The Modified Normalised Difference Water Index (MNDWI), computed from Sentinel-2 green and shortwave infrared bands, consistently delineates open water at 10 m resolution with high accuracy across a wide range of turbidity conditions. Landsat's 30 m pixels are coarser but the 50-year archive makes them indispensable for trend analysis.
For fisheries, surface area matters because it tracks habitat availability directly. Seasonal drawdown in reservoirs concentrates fish and can improve catch per unit effort, but extreme drawdown strands spawning habitat and exposes nests. Floodplain inundation extent determines whether migratory species can access lateral feeding and spawning grounds. A time-series of MNDWI composites, produced at monthly intervals from Sentinel-2, can quantify inundation duration at roughly 10 m spatial precision, which is sufficient to map connectivity between a main channel and adjacent oxbow lakes or wetland fringes.
Turbidity, sediment and the light environment fish actually live in
Turbidity governs the photic zone depth, which controls primary productivity and the visual hunting range of predatory fish. High suspended sediment loads, often driven by catchment erosion or reservoir stratification breakdown, can suppress spawning success in species that require visual mate selection or that deposit eggs on light-dependent substrates. Satellite-derived turbidity, expressed as the diffuse attenuation coefficient Kd(490) or as a proxy index from red and near-infrared reflectance ratios, tracks these conditions at the lake-wide scale that field sampling rarely achieves.
Sentinel-2's 10 m red band (665 nm) is sensitive to suspended particulate matter across a wide concentration range and has been used in published studies to map turbidity plumes entering reservoirs from inflowing rivers. The shortwave infrared band (1610 nm) adds discrimination between mineral sediment and organic particles in highly turbid conditions. Honest limits apply: at very high suspended sediment concentrations (above roughly 1000 mg/L), the near-infrared signal saturates and retrievals become unreliable without in-situ correction.
Cyanobacterial blooms: detecting them early enough to matter
Cyanobacterial blooms produce cyanotoxins that kill fish directly and close recreational and commercial fisheries. The spectral signature of cyanobacteria differs from green algae principally because of phycocyanin, a pigment with an absorption peak near 620 nm. Sentinel-3 OLCI's dedicated band at 620 nm was designed for exactly this detection, and on large lakes it performs well. On smaller water bodies, Sentinel-2's red-edge band at 705 nm provides a workable alternative through the Cyanobacterial Index or the Maximum Chlorophyll Index, both of which are published and validated methods.
The practical detection floor matters. Published studies suggest Sentinel-2 can detect surface bloom concentrations above roughly 5,000 to 10,000 cells per millilitre under favourable atmospheric conditions, but sub-surface accumulations and thin surface films are routinely missed. Cloud cover is the operational enemy: a bloom that develops and disperses within a cloudy fortnight may never appear in the imagery. Combining satellite observations with in-situ sensor buoys or citizen-science reports substantially improves detection reliability.
Building a monitoring programme that a fisheries authority can actually use
The gap between a satellite dataset and a decision-support tool is wider than most procurement processes acknowledge. A fisheries authority needs outputs calibrated to local conditions, delivered at a frequency that matches the management cycle, and presented in formats that field officers can interpret without remote-sensing expertise. That means investing in local field campaigns to anchor radiometric retrievals, defining alert thresholds in biological rather than spectral terms, and integrating satellite layers with catch records and stocking data.
Satellize runs this class of analytics on open constellations including Sentinel-2 and Landsat, and adds commercial tasking where archive gaps or cloud persistence require it. The Tonga crop-estimation programme demonstrates the same design principle in a different domain: satellite-derived indices are only as useful as the ground-truth framework they are anchored to. For a freshwater fisheries client, the right starting point is a scoping analysis of the target water bodies, their typical optical properties and the revisit frequency achievable under local cloud climatology, before committing to an operational monitoring specification.
Archive depth and the question of what 'normal' looks like
One of the underused strengths of the Landsat and MODIS records is their length. Landsat's continuous archive from 1984 onward, and MODIS from 2000, allows analysts to reconstruct multi-decadal surface-area time series for almost any lake or reservoir on Earth. This matters for fisheries management because it establishes what seasonal and interannual variability looked like before a dam was built, before a catchment was cleared, or before a species introduction changed the trophic structure.
Interpreting change against a meaningful baseline is the difference between detecting a genuine regime shift and reacting to normal variability. A reservoir that appears to have lost 15 percent of its surface area in the current dry season may simply be at the low end of its historical range. Conversely, a bloom event that looks modest in absolute terms may be unprecedented in a 30-year context. Neither judgement is possible without the archive.
Typical figures
| Spatial resolution (primary sensors) | 10 m (Sentinel-2 visible/NIR), 20 m (Sentinel-2 red-edge/SWIR), 30 m (Landsat 8/9 OLI), 300 m (Sentinel-3 OLCI) |
| Revisit frequency | 5 days (Sentinel-2A+B combined, equatorial); 8 days (Landsat 8+9 combined); daily (Sentinel-3 OLCI, MODIS) |
| Minimum mappable water body | Approximately 0.5 ha at Sentinel-2 10 m; approximately 1 km width for reliable Sentinel-3 OLCI retrievals |
| Key spectral bands | Green (560 nm), red (665 nm), red-edge (705 nm, 740 nm), NIR (842 nm), SWIR (1610 nm) on Sentinel-2; coastal/aerosol (443 nm) on Landsat OLI; 620 nm phycocyanin band on Sentinel-3 OLCI |
| Bloom detection floor (indicative) | Approximately 5,000–10,000 cyanobacterial cells/mL at surface under clear-sky conditions; sub-surface accumulations not reliably detected |
| Turbidity retrieval range | Approximately 1–1,000 mg/L suspended particulate matter; NIR signal saturation limits reliability above roughly 1,000 mg/L |
| Archive depth | Landsat from 1984 (analysis-ready); Sentinel-2 from 2015; MODIS from 2000; Sentinel-3 OLCI from 2016 |
| Cloud limitation | Optical sensors provide no water-quality data under cloud cover; tropical and monsoon-season revisit gaps of 2–4 weeks are common |
| Typical product latency | Sentinel-2 Level-2A available within 3–5 hours of acquisition via Copernicus Data Space; derived analytics typically 24–48 hours additional processing |
| Delivery formats | GeoTIFF raster layers, vector polygons (bloom/inundation extent), CSV time-series, WMS/WMTS feed, PDF summary report |
Analytics Satellize can run
| Monthly surface-area time series | MNDWI water index applied to Sentinel-2 and Landsat composites; cloud-masked and gap-filled using temporal interpolation | GeoTIFF stack and CSV of monthly water extent per water body, with anomaly flags against historical percentile range |
| Cyanobacterial bloom extent map | Maximum Chlorophyll Index and Cyanobacterial Index computed from Sentinel-2 red-edge bands; threshold-classified into bloom intensity categories | Polygon layer of bloom extent per scene, with area statistics and 5-day change summary delivered as GIS layer and alert email |
| Turbidity and suspended sediment proxy | Red-band reflectance ratio and SWIR-based indices calibrated against available field measurements; expressed as relative turbidity units or estimated mg/L | Raster map per acquisition with spatial distribution of turbidity classes; time-series plot at user-defined monitoring stations |
| Seasonal inundation duration map | Pixel-level count of water-positive observations across a defined season from MNDWI time series; expressed as percentage of cloud-free observations classified as water | Single-band GeoTIFF showing inundation frequency; used to delineate reliable spawning and feeding habitat zones |
| Multi-decadal surface-area trend report | Landsat archive analysis using consistent MNDWI methodology across all available scenes from 1984 onward; Mann-Kendall trend test applied to annual minimum and maximum extents | PDF report with trend charts, anomaly years identified and comparison of pre- and post-intervention periods |
| Bloom early-warning alert | Automated scene-by-scene comparison of Cyanobacterial Index against user-defined baseline; triggered when bloom area exceeds threshold in consecutive acquisitions | Near-real-time alert (within 48 hours of satellite overpass) via email or API, with bloom polygon attached |
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.