Pelagic Sargassum inundation mapping and fisheries impact assessment

Pelagic Sargassum mats can close fishing grounds overnight. Spectral indices applied to MODIS, Sentinel-3 and Sentinel-2 imagery track Atlantic-scale transport and resolve mat density at the scale of individual harbours, giving fleets and fisheries managers days of warning rather than hours.

Sensors

MODIS Aqua/Terra: 250 m bands 1–2 and 500 m ocean-colour bands; daily global revisit (two passes per day combined). The primary synoptic sensor for Atlantic-basin Sargassum transport monitoring. The AFAI (Alternative Floating Algae Index) was originally developed and validated on MODIS data.
Sentinel-3 OLCI: 21 spectral bands at 300 m spatial resolution; approximately daily revisit at mid-latitudes. Full-resolution Level-2 ocean-colour products support FAI and AFAI computation and extend MODIS-era time series under the Copernicus programme.
Sentinel-2 MSI: 10 m (visible and NIR) and 20 m (red-edge and SWIR) spatial resolution; 5-day revisit at the equator with both satellites. Resolves individual mat patches and density gradients at the scale of fishing grounds and harbour approaches. Cloud cover over tropical Atlantic waters limits effective revisit to roughly 10–20 days in practice.
VIIRS SNPP/NOAA-20: 375 m imagery bands and 750 m moderate-resolution bands; daily global coverage. Useful as a gap-filler between MODIS acquisitions and for night-time continuity of ocean-colour retrievals where solar geometry permits.

What the spectral indices actually measure

Pelagic Sargassum floats at the surface, and floating vegetation has a distinctive spectral signature: high near-infrared reflectance from chlorophyll-rich tissue, low reflectance in the red, and near-zero reflectance in the shortwave infrared where water absorbs almost everything. The Floating Algae Index (FAI), developed by Hu (2009) using MODIS bands centred at 645 nm, 859 nm and 1240 nm, exploits this contrast by computing the difference between observed NIR reflectance and a baseline interpolated from the red and SWIR. Positive FAI values indicate floating material; the magnitude loosely tracks surface density.

The Alternative Floating Algae Index (AFAI) is a variant that substitutes the 1640 nm SWIR band, making it applicable to sensors lacking the 1240 nm channel. Both indices are sensitive to sub-pixel coverage: a MODIS 250 m pixel can return a detectable positive signal when Sargassum covers as little as a few percent of the pixel area, though the relationship between index value and actual biomass is non-linear and affected by mat thickness, canopy openness and sun-glint contamination. Published estimates of minimum detectable fractional cover range from roughly 0.1 % to 1 % depending on sensor noise and atmospheric correction quality.

Atlantic transport to harbour entrance: the scale problem

The Great Atlantic Sargassum Belt, documented from satellite observations since the early 2010s, can extend tens of thousands of kilometres across the tropical Atlantic. Monitoring it requires synoptic, daily coverage: MODIS and Sentinel-3 provide that. But a fishing cooperative in Barbados or a trawler working off Senegal needs to know whether the mat will block a specific harbour mouth in 48 hours, not whether the belt is large this year.

Bridging those scales requires a two-stage workflow. MODIS or Sentinel-3 imagery identifies large mat aggregations and their drift direction, using surface current data (from altimetry-derived geostrophic products or operational ocean models) to project arrival timing. Sentinel-2 then provides the high-resolution confirmation pass: at 10 m, individual mat edges are visible, and the density gradient from open water to consolidated mat can be mapped with enough fidelity to assess whether a vessel channel remains navigable. The latency between Sentinel-2 acquisition and a processed product is typically two to four hours through the Copernicus Dataspace pipeline, which is operationally useful but not instant.

False positives are not a minor footnote

Sea foam, floating pumice, turbid plumes and other macroalgae all produce positive FAI signals under some conditions. Cyanobacterial surface scums, relevant on the sibling topic of harmful algal blooms, can also overlap spectrally with Sargassum in broad-band sensors. Sun-glint, even after correction, introduces residual artefacts that mimic low-density mats. In practice, operational Sargassum mapping pipelines use a combination of spatial persistence (a true mat drifts coherently over successive days), texture metrics, and ancillary sea-surface temperature or salinity data to filter implausible detections.

Sentinel-2's red-edge bands at 705 nm and 740 nm add discriminating power that MODIS lacks: Sargassum's red-edge inflection differs measurably from that of seagrass wrack or green macroalgae, reducing but not eliminating confusion. Honest operational accuracy figures from published validation studies sit in the range of 70–90 % precision depending on scene conditions. Users should treat any single-image detection as a hypothesis to be confirmed by the next clear-sky pass.

From mat position to fisheries impact

Mapping where Sargassum is does not automatically answer the question fisheries managers care about: what does it cost? The translation requires overlaying mat extent with fishing ground polygons, vessel track data (from AIS where available), and known gear-deployment areas. A mat covering a productive FAD zone or a seasonal spawning aggregation site has a different economic significance from one drifting through open water.

Nearshore inundation causes at least three categories of harm that satellite data can help characterise. First, physical access: mat density above a threshold makes small-vessel navigation impractical, and harbour-entrance blockage can be flagged when Sentinel-2 shows consolidated coverage across the navigable channel. Second, gear fouling: longlines and gill nets set in mat-dense areas suffer high loss rates; mat position layers can inform gear-deployment advisories. Third, hypoxia: decomposing beached Sargassum consumes dissolved oxygen, creating low-oxygen zones that damage nearshore fish populations. Satellite data cannot directly measure dissolved oxygen, but mat beaching extent and residence time, derivable from multi-day Sentinel-2 composites, serve as a proxy for decomposition load that field teams can prioritise for in-situ verification.

What the archive reveals that a single season cannot

MODIS Aqua has been operational since 2002, and the MODIS Terra record extends to 1999. That archive, combined with Sentinel-3 OLCI data from 2016 onwards, supports multi-decadal analysis of Sargassum seasonality, inter-annual variability and long-term trend. Published research using this record has linked peak belt biomass to Amazon River discharge anomalies and West African upwelling conditions, both of which are observable in satellite data months before the Sargassum arrives at Caribbean or West African coasts.

For a fisheries authority, this means seasonal outlooks are possible, not just reactive alerts. A year with anomalously high river nutrient loading in January and February is a statistical predictor of a severe Caribbean inundation event by May to August. That lead time is long enough to adjust fishing calendars, pre-position clean-up equipment and negotiate temporary access to alternative grounds. Satellize's analytics work, including the crop-estimation programme for the Kingdom of Tonga, is grounded in exactly this kind of multi-year archive analysis rather than single-season snapshots.

Practical limits to set before commissioning analysis

Three constraints shape what is achievable. Cloud cover is the dominant one: the tropical Atlantic and Caribbean, where Sargassum impacts are most severe, have cloud fractions that routinely exceed 70 % in summer months. A 5-day Sentinel-2 revisit becomes an effective revisit of 20 to 40 days in heavily clouded regions, which is too slow for operational early warning on its own. MODIS and Sentinel-3, with daily coverage and larger pixels, are therefore the operational backbone; Sentinel-2 provides periodic high-resolution updates rather than continuous monitoring.

The second constraint is biomass quantification. Spectral indices indicate presence and relative density; converting them to wet or dry biomass tonnes requires empirical calibration against field measurements, and published calibration relationships carry substantial uncertainty. Treat biomass estimates as order-of-magnitude indicators rather than inventory figures.

Third, drift forecasting accuracy degrades beyond 72 to 96 hours. Ocean surface currents at the scales relevant to mat transport are modelled with useful skill over a few days, but week-ahead forecasts have errors large enough to misplace a mat by tens of kilometres, which matters when the target is a specific harbour. Communicate forecast uncertainty explicitly in any operational product.

Typical figures

Synoptic spatial resolution	250 m (MODIS bands 1–2), 300 m (Sentinel-3 OLCI)
High-resolution spatial resolution	10 m visible/NIR, 20 m red-edge/SWIR (Sentinel-2 MSI)
Synoptic revisit	Daily (MODIS Aqua + Terra combined); approximately daily (Sentinel-3 OLCI)
High-resolution revisit (nominal)	5 days at equator (Sentinel-2A + 2B combined)
Effective revisit (cloud-affected tropics)	20–40 days for cloud-free Sentinel-2; MODIS/Sentinel-3 more reliable due to daily cadence
Minimum detectable fractional cover (MODIS AFAI)	Approximately 0.1–1 % of pixel area, depending on atmospheric correction and sun-glint
Key spectral bands	Red (~645 nm), NIR (~859 nm), SWIR (~1240 nm or ~1640 nm) for FAI/AFAI; red-edge 705 nm and 740 nm (Sentinel-2) for discrimination
Archive depth	MODIS Terra from 1999, Aqua from 2002; Sentinel-3 OLCI from 2016; Sentinel-2 from 2015
Product latency (Copernicus Dataspace)	Typically 2–4 hours post-acquisition for Level-1C/2A
Biomass quantification accuracy	Order-of-magnitude only; published calibrations carry high uncertainty without co-located field measurements

Analytics Satellize can run

Daily synoptic Sargassum extent map	AFAI/FAI computed from MODIS or Sentinel-3 OLCI reflectance, with sun-glint masking and multi-day persistence filter	GeoTIFF or vector polygon layer, updated daily, delivered via API or secure file transfer
High-resolution mat density map for specific fishing grounds	FAI and red-edge indices from Sentinel-2 MSI; density classified into navigable, partially obstructed and consolidated-mat categories	10 m raster layer with classification legend, triggered on each clear-sky Sentinel-2 acquisition
72-hour inundation arrival forecast	Mat centroid tracking from MODIS time series combined with Copernicus Marine Service surface current analysis; probabilistic drift envelope	Forecast polygon with uncertainty bounds, issued as alert email and GIS layer when mat is within 300 km of designated ports
Harbour-entrance blockage alert	Sentinel-2 mat coverage fraction computed over predefined harbour-channel polygons; threshold trigger at user-defined coverage percentage	Binary alert with supporting image chip, delivered within 4 hours of Sentinel-2 overpass
Seasonal outlook and anomaly report	MODIS archive analysis of Amazon discharge proxy, West African upwelling SST anomaly and prior-year belt extent; statistical regression against historical inundation severity	Annual pre-season PDF report with probabilistic severity index by sub-region
Beaching extent and residence-time map for hypoxia risk prioritisation	Multi-date Sentinel-2 composites identifying persistent nearshore mat presence; residence time computed as consecutive-day count above density threshold	Raster layer with residence-time values in days, updated weekly, for field-team deployment planning
Fishing-ground impact overlay	Intersection of mat extent polygons with client-supplied fishing ground boundaries and AIS-derived vessel track density	Monthly impact report quantifying mat-covered fishing-ground area by zone, in PDF and GIS format

Sources

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.