Green-hydrogen site screening via combined solar, wind, and water-availability data

Electrolytic hydrogen needs sun or wind, water, and a route to market, all in the same place. Satellite screening combines GHI climatology, SAR wind proxies, GRACE groundwater anomalies, and JRC surface-water layers to cut a continental candidate list to a workable shortlist before a single survey crew is deployed.

Sensors

SEVIRI (Meteosat) / ABI (GOES): Geostationary radiometers used to derive surface Global Horizontal Irradiance (GHI) via the Heliosat or similar cloud-index methods. SEVIRI provides 15-minute, ~3 km native pixel data over Europe, Africa, and the Middle East; ABI covers the Americas at 10-minute cadence. Multi-year stacks yield GHI climatologies with seasonal and interannual variance, essential for P90 yield estimates.
Sentinel-1 SAR-C (ESA): C-band synthetic aperture radar at 5.405 GHz. Ocean-surface wind speed is retrieved from normalised radar cross-section via geophysical model functions (GMF) such as CMOD5.N, giving 500 m to 1 km effective wind-field resolution with 6-12 day repeat at mid-latitudes. Over land, surface roughness limits direct wind retrieval, but coastal and estuarine fetch zones remain usable. Free archive from 2014.
GRACE-FO (NASA/DLR): Twin satellites measuring gravity-field variations to infer terrestrial water storage anomalies, including groundwater, at roughly 300-500 km spatial resolution and monthly cadence. Useful for identifying regions of persistent groundwater depletion or surplus at basin scale, not for locating a specific aquifer. Honest limit: it cannot resolve individual wells or sub-basin heterogeneity.
JRC Global Surface Water (Landsat archive): Derived from the full Landsat 5/7/8/9 archive (1984 to present) at 30 m resolution. Provides monthly water occurrence, seasonality, and recurrence layers across the globe. Identifies perennial rivers, seasonal wetlands, and reservoirs that could supply process water. Does not measure flow rate or water quality; those require ground surveys.
Sentinel-2 MSI (ESA): 13-band optical imager at 10-60 m resolution, 5-day revisit at mid-latitudes. Used here for land-cover classification to flag competing land uses, protected areas, and terrain features near candidate sites. Also supports proximity analysis to roads, ports, and existing grid infrastructure visible in imagery.

Why four inputs, and why they rarely align

Green hydrogen produced by electrolysis is only economically interesting where the electricity input is cheap and plentiful, the water input is available without destroying local hydrology, and the output can reach a buyer. Each constraint eliminates large swathes of the planet independently. The Atacama Desert scores brilliantly on solar irradiance and poorly on fresh water. The Norwegian coast scores well on wind and water and badly on solar. The Persian Gulf scores on solar and proximity to export terminals and very poorly on non-saline water. Desalination can substitute for fresh water, but it adds capital cost and energy penalty that feed back into the hydrogen cost model.

Satellite screening does not solve this tension. It maps it. The value is in seeing all four constraints simultaneously, at continental scale, before committing to ground surveys that cost real money. A site that fails two constraints at the screening stage is not worth a hydrological study.

What the solar layer actually measures, and what it does not

GHI derived from SEVIRI or ABI is a modelled quantity, not a direct surface measurement. The cloud-index method compares top-of-atmosphere radiance against a clear-sky reference; the surface irradiance is then estimated using aerosol optical depth inputs and a radiative transfer model. Published validation studies against pyranometer networks typically report root-mean-square errors of 15-30 W/m² for hourly values, improving substantially when averaged to monthly or annual means. For site screening, annual GHI climatology at 3-5 km resolution is fit for purpose.

What it cannot tell you: the micro-topographic shading from ridgelines or escarpments at sub-kilometre scale, the precise aerosol loading from dust events in any given future year, or grid-connection quality. A site showing 2,200 kWh/m²/year GHI in the satellite climatology still needs a bankable irradiance study with on-site sensors before a lender will finance it.

Wind from radar: what the ocean surface gives away

Sentinel-1's C-band backscatter over water surfaces responds predictably to wind-driven capillary waves. Geophysical model functions translate that backscatter into 10 m equivalent wind speed with accuracy generally quoted at 1-2 m/s against buoy records. The practical result is a wind climatology at 500 m to 1 km resolution covering coastal and offshore zones, built from thousands of passes since 2014.

Over land, the relationship breaks down. Vegetation, soil moisture, and surface heterogeneity contaminate the backscatter signal. For inland hydrogen sites, wind screening therefore relies on reanalysis products (ERA5 at 31 km, for instance) rather than SAR retrieval, and the resolution penalty is significant. Coastal sites, where hydrogen export by ship is already logistically attractive, are where SAR wind data genuinely earns its place in the screening stack.

Water availability: what GRACE sees and what it misses

GRACE-FO detects changes in the total column of water stored in a region, integrating soil moisture, surface water, snow, and groundwater into a single gravity signal. A persistent negative anomaly over a basin, sustained across several years of monthly observations, is a credible indicator that groundwater is being depleted faster than it is recharged. That matters for hydrogen siting because electrolysis at industrial scale consumes roughly 9 litres of deionised water per kilogram of hydrogen produced, and a site drawing on an already-stressed aquifer faces both regulatory and physical risk.

The JRC Global Surface Water layer adds a complementary picture at 30 m resolution: where does water appear on the surface, how often, and in which months? A perennial river within 20 km of a candidate site is a qualitatively different situation from a seasonal wetland that holds water for three months a year. Together, GRACE and JRC give a coarse-to-fine picture of water availability. Neither replaces a hydrogeological survey. GRACE cannot locate an aquifer; it can only suggest that water storage in a large basin is trending the wrong way.

Building the multi-criteria score

The four satellite layers are normalised and combined into a single raster score. A typical approach assigns each candidate pixel a percentile rank within the study region on each criterion, then applies weights reflecting the client's cost model: a project planning to use desalination might down-weight the water criterion; a project targeting pipeline export might up-weight proximity to existing gas infrastructure. The output is not a recommendation. It is a ranked candidate list, with each site's score decomposed by criterion so an analyst can see exactly why a pixel ranked highly or poorly.

Exclusion masks applied before scoring are often more important than the scoring itself. Protected areas, steep slopes (typically above 5-8 degrees for large ground-mounted plant), urban footprints, and inundation-risk zones derived from satellite flood-frequency layers can remove 60-80 percent of a study region before any resource data is considered. Satellize applies this kind of layered screening in its analytics work; the Tonga crop-estimation programme involved similar multi-layer spatial filtering, though at a very different scale.

Honest limits of the whole framework: satellite screening operates at 30 m to 5 km resolution depending on the layer. It cannot assess grid-connection stability, local planning constraints, land tenure, or the actual flow rate of a river. The shortlist it produces is an input to feasibility studies, not a substitute for them.

From shortlist to bankable data: what comes next

Once the satellite screen has reduced a continental or national candidate pool to tens of sites rather than thousands, the next steps are sequential and increasingly expensive. Hydrological surveys quantify actual water availability and quality. Wind masts or LiDAR campaigns validate the reanalysis or SAR wind climatology at hub height. Grid operators assess connection capacity and stability. Environmental impact assessments check species and habitat data that no satellite sensor resolves.

The satellite layer continues to add value after the shortlist stage. Time-series monitoring of surface-water extent tracks seasonal variability at shortlisted sites. Updated GHI climatologies can be re-run annually to capture interannual variability linked to large-scale climate modes. SAR-derived wind fields can be compared against mast data to check for systematic bias in the reanalysis. The screening product is a starting point, not a conclusion, and it is most useful when the client understands exactly what each layer can and cannot resolve.

Typical figures

GHI spatial resolution (SEVIRI/ABI)	3-5 km at nadir; degrades toward sensor limb
GHI temporal resolution	10-15 min instantaneous; monthly/annual climatologies used for screening
SAR wind-field resolution (Sentinel-1)	500 m to 1 km effective, coastal and offshore zones
Sentinel-1 revisit	6 days (two-satellite constellation); 12 days single satellite
GRACE-FO groundwater anomaly resolution	~300-500 km spatial; monthly cadence
JRC Global Surface Water resolution	30 m (Landsat); monthly occurrence layers
JRC archive depth	1984 to present (Landsat 5/7/8/9)
Sentinel-1 archive depth	2014 to present
Composite scoring output format	GeoTIFF raster + vector candidate polygons with per-criterion attribute table
Typical study-area latency	2-4 weeks from data pull to scored candidate map, depending on study extent

Analytics Satellize can run

Multi-criteria site-score raster	Weighted linear combination of normalised GHI, wind speed, water-availability, and proximity layers; exclusion masking applied first	GeoTIFF score surface + ranked candidate polygon GIS layer with per-criterion breakdown
GHI climatology with P50/P90 statistics	Multi-year SEVIRI or ABI time-series aggregation; percentile extraction per pixel	Annual and monthly GHI maps (GeoTIFF) with uncertainty band, delivered as report and GIS layer
Coastal wind climatology from SAR	CMOD5.N GMF applied to Sentinel-1 IW/EW mode NRCS; multi-year stack aggregated to seasonal and annual wind-speed percentiles	Gridded wind-speed climatology (GeoTIFF) covering study coastline, with data-gap flags where land contamination is present
Groundwater stress indicator	GRACE-FO terrestrial water storage anomaly time series extracted for candidate basins; trend analysis to flag persistent depletion	Basin-level indicator table (CSV + PDF summary) with monthly anomaly time series and trend significance
Surface-water availability profile	JRC Global Surface Water occurrence and seasonality layers clipped to candidate zones; buffer analysis for proximity to perennial water bodies	Per-site water-availability summary (PDF report) with seasonal occurrence maps (GeoTIFF)
Exclusion-mask layer	Overlay of protected-area boundaries, slope derived from SRTM/Copernicus DEM, urban extent, and satellite-derived flood-frequency layer	Binary exclusion raster (GeoTIFF) with attributed reason codes per excluded zone
Proximity-to-infrastructure index	Euclidean and least-cost distance from candidate pixels to ports, roads, and pipeline corridors digitised from optical imagery and open datasets	Distance-to-infrastructure raster and ranked candidate table (GIS layer + CSV)

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.