Sea turtle habitat suitability mapping for fisheries bycatch risk reduction
Satellite-derived sea surface temperature, altimetry and chlorophyll fronts predict where loggerhead and leatherback turtles aggregate, letting fisheries managers issue probabilistic bycatch risk advisories before vessels set gear.
Sensors
- NOAA VIIRS SST (SNPP and NOAA-20): 750 m pixel resolution at nadir for the I-band thermal channel, daily global coverage under clear skies. The primary SST input for TurtleWatch-style thermal habitat models; cloud cover is the binding constraint, typically reducing usable daily coverage by 30–60% in tropical convergence zones.
- MODIS Aqua chlorophyll-a: 1 km resolution ocean colour product, daily revisit per platform. Frontal chlorophyll gradients are computed from the OC3M band-ratio algorithm. Retrieval quality degrades sharply in turbid or optically complex coastal water, which is why the method is most defensible in open-ocean settings.
- AVISO DUACS sea-level anomaly (multi-mission altimetry): Gridded at 0.25 degrees, updated daily, fusing Jason-3, Sentinel-6 Michael Freilich and other altimeters. Sea-level anomaly identifies mesoscale eddies and convergence zones that concentrate prey; it is the least cloud-affected of the three core predictors because altimetry is not blocked by cloud.
- Sentinel-3 SLSTR: Dual-view thermal infrared at 1 km resolution, with two satellites providing roughly one-day combined revisit at mid-latitudes. Useful as a cross-check against VIIRS SST and for generating frontal gradient products where VIIRS swaths have cloud gaps.
Why turtles are predictable from space
Loggerhead and leatherback turtles do not distribute themselves randomly across the ocean. They follow prey, and prey follows physics. Thermal fronts, where water masses of different temperatures meet, concentrate zooplankton and jellyfish. Mesoscale eddies create convergence zones that aggregate floating Sargassum and the small organisms beneath it. Chlorophyll gradients mark the productive edges of these features. All three phenomena are detectable from orbit.
The NOAA TurtleWatch programme, operational in the Hawaii longline fishery since 2008 and documented in peer-reviewed literature, formalised this relationship into a habitat suitability index. The core insight is that loggerheads in the North Pacific concentrate near the 17.5 °C SST isotherm. That is not a universal constant across species or ocean basins, but the general principle holds: specific thermal windows, combined with elevated chlorophyll and positive sea-level anomaly, predict where turtles are more likely to be. The maps do not count turtles. They quantify environmental suitability, which is a meaningful and operationally useful distinction.
Building the habitat suitability index
The standard workflow has three environmental layers. First, SST from VIIRS or MODIS is processed to identify frontal gradients, typically using edge-detection algorithms such as the Belkin-O'Reilly front detector applied to the daily composites. Second, sea-level anomaly from AVISO DUACS flags positive anomalies associated with warm-core eddies and convergence zones. Third, chlorophyll-a frontal gradients from MODIS Aqua mark the biologically productive boundaries between water masses.
Each layer is normalised and combined into a suitability score, often using logistic regression or Maxent models trained on historical satellite-tracked turtle telemetry data. The telemetry training data is critical: without it, the environmental thresholds are arbitrary. Published studies using ARGOS and Fastloc-GPS tagged turtles have established species-specific and region-specific parameter sets for several major fisheries, including the Hawaii-based North Pacific longline fleet and parts of the Mediterranean.
The output is a gridded probability surface, typically at 0.25-degree resolution matching the altimetry grid, updated daily or every few days depending on cloud cover over the SST inputs. High-probability cells overlaid with fishing vessel positions from AIS produce the bycatch risk layer that managers actually use.
What the maps honestly cannot do
Cloud is the principal operational problem. VIIRS and MODIS SST retrievals require clear skies; in persistently cloudy regions such as the eastern tropical Pacific during La Niña conditions, usable daily coverage can fall well below 50%. Multi-day compositing reduces the gap but introduces temporal lag, which matters when fronts move tens of kilometres in a day.
The suitability index predicts habitat quality, not turtle presence. A high-scoring cell on a given day may contain no turtles at all; a low-scoring cell may contain several. The maps are probabilistic statements about where the environment is favourable, not detections. This distinction matters for fisheries managers who might otherwise treat a risk map as a survey result.
Coastal and shelf-sea settings are genuinely harder. Turbid water corrupts chlorophyll-a retrievals; shallow bathymetry changes the physical interpretation of sea-level anomaly; thermal fronts near river plumes have different biological significance than open-ocean fronts. The approach is most reliable, and most published, for open-ocean longline fisheries. Applying it to coastal gillnet or trawl fisheries requires additional validation work that is not yet in the public literature at the same level of confidence.
From risk map to fisheries management action
The operational value of a habitat suitability map depends entirely on how it is connected to fishing effort. The TurtleWatch model delivers weekly advisory maps to Hawaii longline captains showing areas of elevated loggerhead habitat suitability, with a recommended avoidance zone defined by the 17.5 °C isotherm boundary. Participation has been voluntary, but uptake has been significant because the maps also correlate with swordfish catch rates, giving captains an economic incentive alongside the conservation one.
A similar logic applies to any longline, driftnet or purse-seine fleet operating in a basin where turtle telemetry data exists to calibrate the model. The risk map can be issued as a GIS layer, a vessel-specific alert when a planned set location exceeds a suitability threshold, or a fleet-level summary for observer programmes and port-state control. The latency between satellite overpass and map delivery is typically 24 to 48 hours for operational SST products, which is fast enough for trip-level decision-making on most longline vessels.
Species differences and basin-specific calibration
Loggerheads and leatherbacks occupy different thermal niches. Leatherbacks range into colder water, following jellyfish blooms well outside the thermal windows that attract loggerheads. Published work from the Atlantic and Pacific has identified leatherback habitat associations with cooler SST and strong frontal gradients rather than a single isotherm threshold. Any operational system covering a mixed-species bycatch risk needs separate suitability layers per species, which requires separate telemetry training datasets.
Basin matters too. The 17.5 °C loggerhead threshold established for the central North Pacific does not transfer directly to the Mediterranean or the South Atlantic without recalibration. Satellize can configure the environmental predictor stack and model parameters to a client's specific ocean basin and target species, provided calibration telemetry data is available from a published tagging programme or supplied by the client. The Tonga crop-estimation programme is our only named analytics engagement to date, but the satellite data processing pipeline for environmental indices of this type is operationally the same class of problem.
Practical requirements for a working system
Three things are needed before a habitat suitability advisory is operationally credible. First, a calibrated species-specific model, which means access to published or client-supplied telemetry data from tagged turtles in the target basin. Second, a reliable near-real-time satellite data feed covering SST, chlorophyll and altimetry for the fishery's operating area. Third, a distribution channel that puts the risk map in front of vessel operators or managers before gear is set.
The satellite data side is well-solved. VIIRS and MODIS SST products are available within hours of overpass through NOAA CoastWatch and NASA EARTHDATA. AVISO DUACS altimetry is updated daily. Sentinel-3 SLSTR adds redundancy. The harder work is calibration and institutional integration: getting the map into a fishing vessel's navigation system, or into a national observer programme's data workflow, requires engagement with fisheries authorities that goes beyond remote sensing.
Typical figures
| SST spatial resolution | 750 m (VIIRS I-band), 1 km (MODIS), 1 km (Sentinel-3 SLSTR) |
| Habitat suitability grid resolution | 0.25 degrees (~25 km), constrained by altimetry grid |
| SST revisit | Daily per platform under clear skies; VIIRS SNPP and NOAA-20 combined offer two overpasses per day |
| Altimetry revisit | Daily gridded product (AVISO DUACS multi-mission merge) |
| Operational latency | 24–48 hours from satellite overpass to advisory map, for near-real-time SST products |
| Cloud impact on SST retrieval | Typically 30–60% daily coverage loss in tropical convergence zones; multi-day compositing required |
| Spectral bands used | Thermal infrared (10.5–12.5 µm for SST), visible/NIR (OC3M ratio for chlorophyll-a), radar altimetry (Ku-band) |
| Archive depth | MODIS Aqua from 2002, VIIRS from 2012, AVISO altimetry from 1993 |
| Minimum detectable front gradient | Approximately 0.1–0.5 °C per km for SST fronts using Belkin-O'Reilly detector on 1 km data |
| Output format | GeoTIFF, NetCDF, or OGC WMS layer; vessel alert via API or email digest |
Analytics Satellize can run
| Daily habitat suitability index map | Logistic regression or Maxent model combining SST frontal gradient, sea-level anomaly and chlorophyll-a, calibrated to published telemetry data | Gridded GeoTIFF or NetCDF probability surface, updated daily |
| Bycatch risk overlay | Habitat suitability surface intersected with AIS-derived fishing vessel positions and effort density | GIS layer or PDF advisory chart showing high-risk overlap zones for the coming 24–48 hours |
| SST frontal gradient product | Belkin-O'Reilly or Cayula-Cornillon front detection applied to daily VIIRS or MODIS SST composites | Vector front lines and raster gradient magnitude, ingested as a predictor layer |
| Mesoscale eddy and convergence zone flags | Sea-level anomaly thresholding and eddy-tracking algorithm (e.g. py-eddy-tracker) applied to AVISO DUACS | Polygon layer of positive anomaly features updated daily, tagged with estimated centre and radius |
| Species-specific thermal window exceedance alert | Isotherm boundary detection on SST composite, compared against published species thermal preference ranges | Automated alert when a vessel's planned set location falls within the target thermal window |
| Seasonal risk climatology | Monthly and quarterly aggregation of suitability index over multi-year archive to identify persistent high-risk zones | Static report with maps for fisheries management planning and observer programme deployment |
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.