Sensors
- Landsat 8/9 TIRS: Two thermal infrared bands (10.6 µm and 12.0 µm) at 100 m native resolution (resampled to 30 m). Single-channel and split-window retrievals yield sea-surface temperature (SST) with typical accuracy of ±0.5–1.0 K over water under clear skies. 16-day revisit per satellite; combined Landsat 8+9 gives 8-day repeat. Useless through cloud.
- ECOSTRESS (ISS-mounted): Five thermal infrared channels (8–12.5 µm) at 70 m resolution. Opportunistic ISS overpass geometry means variable revisit (roughly 1–5 days at mid-latitudes, irregular timing). Finer spatial resolution than TIRS makes it better for small embayments and narrow coastal strips. Cloud-limited like all thermal sensors.
- SMAP L-band radiometer: 1.41 GHz passive microwave, nominally sensitive to sea-surface salinity in the top centimetre. Spatial resolution ~40 km, 2–3 day global revisit. Useful for broad regional salinity gradients driven by large SGD systems; far too coarse to resolve individual seeps or small embayments.
- Sentinel-2 MSI: No thermal bands, but 10 m visible and near-infrared imagery at 5-day revisit (two satellites). Useful as a complementary layer: turbidity patterns, water colour shifts, and floating freshwater vegetation can corroborate thermal anomalies. Does not directly sense temperature or salinity.
Why fresh groundwater leaves a thermal scar
Groundwater in coastal aquifers is buffered by the thermal mass of the subsurface. In temperate and subtropical settings it typically sits within a few degrees of mean annual air temperature, which in summer is cooler than the sun-warmed sea surface and in winter can be warmer. When this water seeps through permeable seabed sediments, it creates a localised temperature contrast at the surface, sometimes just 0.5 K, sometimes several degrees, depending on discharge rate, water depth, tidal mixing and time of day.
That contrast is what thermal infrared satellites detect. The physics is straightforward: emitted radiance in the 10–12 µm window is a near-direct proxy for skin temperature over water, and split-window algorithms applied to Landsat TIRS bands 10 and 11 can resolve anomalies down to roughly 0.5–1.0 K under favourable conditions. The catch is that the anomaly is a skin signal. Tidal currents, wind mixing and wave action can dilute or smear the plume within hours, so image timing relative to tidal phase matters considerably.
What a thermal anomaly map can and cannot tell you
A well-processed Landsat or ECOSTRESS scene over a known SGD coastline, acquired at low tide and low wind, will often show a coherent cold (or warm) patch that correlates with documented seep locations. Published studies on coastlines from the Mediterranean to the Korean peninsula have demonstrated this repeatedly. The anomaly marks a zone of interest.
It does not tell you the flux. Converting a surface temperature depression to a volumetric discharge rate requires knowing the mixing geometry, the groundwater temperature at the point of discharge, and the ambient current field. None of those are retrievable from the satellite alone. Radon-222 tracer surveys, seepage meters, or salinity transects are the accepted in-situ methods for flux quantification, and satellite mapping is most valuable as a guide to where those field campaigns should focus.
A further honest limit: cloud cover is fatal to thermal retrieval. Coastal zones in humid tropical climates, precisely where SGD is often hydrologically significant, can have persistent cloud cover that renders TIRS and ECOSTRESS unusable for weeks at a time. Multi-year archives help by allowing analysts to composite cloud-free acquisitions across seasons.
SMAP salinity: a coarse but complementary signal
Fresh groundwater lowers salinity. SMAP's L-band radiometer detects the dielectric contrast between fresh and saline water and has been used to map large-scale freshwater plumes in coastal seas, for example, river outflows and estuary fronts. For SGD specifically, the 40 km footprint is a serious constraint: only very large, concentrated discharge systems produce salinity anomalies detectable at that scale.
Where SMAP does add value is in regional screening. A persistent low-salinity anomaly in a SMAP time series, coinciding with a thermal anomaly in Landsat, raises the probability that the feature is real SGD rather than a localised SST artefact from sediment heating or upwelling. The two sensors are not substitutes for each other; they are cross-checks.
Building a detection workflow from open data
A practical satellite SGD screening workflow has three steps. First, assemble a multi-year Landsat 8/9 TIRS archive for the target coastline, apply atmospheric correction and split-window SST retrieval, then filter scenes by cloud fraction, wind speed (from ERA5 reanalysis or similar), and tidal phase. Low-tide, low-wind, early-morning acquisitions are the most diagnostic.
Second, compute a long-term median SST baseline per pixel and flag persistent negative (or positive) anomalies in the nearshore zone. Persistence across multiple seasons is the key discriminator: an artefact from a single warm or cold event will not recur systematically at the same location. Third, cross-reference anomaly locations with SMAP salinity time series and Sentinel-2 water-colour imagery. Convergent evidence from two or more independent sensors substantially reduces false positives before any field work is commissioned.
ECOSTRESS adds a higher-resolution thermal layer when ISS overpass geometry is favourable, and its finer 70 m pixels are particularly useful for small embayments or rocky headland seeps that fall below Landsat's effective detection scale. The two thermal datasets are complementary rather than redundant.
Who needs this, and what they do with it
Coastal water-quality agencies care about SGD because nutrient-laden groundwater, particularly nitrate from agricultural catchments, drives localised eutrophication and harmful algal blooms in bays that appear clean by conventional river-monitoring metrics. Identifying the seep locations from orbit gives regulators a spatial hypothesis to test with water-quality sampling, which is far cheaper than a blanket survey of an entire coastline.
Fisheries managers in tropical island nations face a related problem: reef ecosystems are sensitive to freshwater pulses, and SGD can be a cryptic stressor. Satellite screening is a proportionate first step for governments that lack the vessel time for exhaustive in-situ surveys. Satellize has worked in analogous island-scale remote-sensing contexts, including the Kingdom of Tonga crop-estimation programme, and the data pipelines for open-constellation thermal analytics transfer directly to SGD screening.
Hydrogeologists mapping coastal aquifer boundaries also use SGD location data as a boundary condition: seep locations constrain where the freshwater head in the aquifer is sufficient to overcome hydrostatic pressure from the overlying sea. That is useful input for groundwater models even before a single flux measurement is made.
The honest summary of what orbit delivers
Satellite thermal and salinity data can screen hundreds of kilometres of coastline in a single pass and identify the handful of locations worth investigating on the ground. That is a genuine reduction in field-survey cost. What orbit cannot do is replace the radon survey, the seepage meter, or the salinity transect that turns a candidate zone into a quantified discharge estimate.
The minimum detectable thermal anomaly for Landsat TIRS over water is roughly 0.5 K under ideal conditions, which corresponds to detectable SGD only where discharge is concentrated enough to maintain that contrast at the surface against tidal mixing. Diffuse, low-rate seepage across a wide sandy shelf may produce no detectable signal at all. Honest use of these methods means presenting the satellite output as a ranked list of candidate sites, not a discharge inventory.
Typical figures
| Thermal spatial resolution | 100 m native (Landsat TIRS, resampled to 30 m); 70 m (ECOSTRESS) |
| Revisit (thermal) | 8 days combined Landsat 8+9; 1–5 days ECOSTRESS (irregular ISS orbit) |
| SST retrieval accuracy | ±0.5–1.0 K over water under clear sky (split-window, Landsat TIRS) |
| Minimum detectable thermal anomaly | ~0.5 K under low-wind, low-cloud, low-tide conditions; larger anomalies required in turbulent or cloudy settings |
| SMAP salinity footprint | ~40 km; 2–3 day global revisit; top ~1 cm of water column |
| Sentinel-2 MSI resolution (visible/NIR) | 10 m; 5-day revisit (two satellites); no thermal bands |
| Thermal archive depth | Landsat 8 from 2013; Landsat 9 from 2021; ECOSTRESS from 2018 |
| Cloud limitation | All thermal bands fail under cloud cover; multi-year compositing required in humid coastal climates |
| Delivery formats | GeoTIFF SST anomaly maps, NetCDF time series, GIS polygon layers of candidate SGD zones |
Analytics Satellize can run
| Nearshore SST anomaly map | Split-window atmospheric correction applied to Landsat TIRS bands 10 and 11; long-term median baseline subtraction per pixel | GeoTIFF layer showing persistent cold or warm anomaly magnitude and spatial extent for a defined coastal segment |
| Multi-year persistent anomaly ranking | Temporal compositing of cloud-filtered Landsat scenes; frequency analysis of anomaly recurrence by pixel across seasons | Ranked list of candidate SGD zones with anomaly persistence score, delivered as GIS polygon shapefile |
| ECOSTRESS high-resolution thermal overlay | ISS-pass scene selection for target coastline; co-registration with Landsat baseline; pixel-level comparison | 70 m resolution thermal anomaly GeoTIFF for small embayments or rocky seep features below Landsat detection scale |
| SMAP salinity cross-check layer | SMAP Level-3 daily salinity composites extracted for coastal grid cells; anomaly flagging against seasonal climatology | Time-series chart and spatial grid showing salinity departures coinciding with thermal anomaly locations |
| Sentinel-2 water-colour corroboration | Turbidity and CDOM index retrieval from Sentinel-2 visible bands; spatial overlay with thermal anomaly polygons | RGB and false-colour imagery annotated with thermal anomaly boundaries for field-team briefing |
| Field-campaign prioritisation report | Multi-sensor convergence scoring: thermal persistence, salinity signal, water-colour shift, proximity to known aquifer outcrop | Written report with ranked site coordinates, supporting imagery, and recommended in-situ methods (radon survey, seepage meter) per site |
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.