Urban groundwater recharge zone identification from land-cover and soil data
Satellite-derived land cover, terrain, and soil permeability data can pinpoint the shrinking fraction of urban land where precipitation still infiltrates. The outputs feed directly into drainage bylaws and aquifer protection zones.
Sensors
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands; 5-day revisit at the equator with both satellites. NDVI, NDWI, and bare-soil indices derived from its 13 spectral bands distinguish vegetated, bare, and impervious surfaces with enough spatial detail to resolve individual parks and street medians.
- Landsat 9 OLI-2: 30 m multispectral resolution with a 16-day repeat cycle. Its thermal band (TIRS-2, 100 m resampled to 30 m) adds a surface-temperature signal that correlates with evapotranspiration and soil moisture, providing an independent check on infiltration-capacity estimates. Archive depth to 1972 across the Landsat family.
- TanDEM-X: Global DEM at 12 m posting, with absolute vertical accuracy better than 10 m and relative accuracy around 2 m over flat urban terrain. Slope and flow-accumulation layers derived from TanDEM-X identify low-gradient catchment flats where ponding and infiltration are physically plausible.
- SRTM: Freely available 30 m DEM covering latitudes 56°S to 60°N. Sufficient for city-scale catchment delineation where TanDEM-X is not licensed; vertical accuracy is coarser (roughly 16 m absolute at 90% confidence) and it cannot resolve fine urban drainage structures.
What the surface tells a satellite about what lies beneath
Groundwater recharge is an invisible process, but its preconditions are not. Three surface conditions must coincide for infiltration to occur: the ground must be permeable, the slope must be shallow enough that water does not run off before it can soak in, and the land cover must be non-impervious. Satellites cannot measure soil moisture at depth, but they can characterise all three preconditions directly from orbit.
Sentinel-2's 10 m bands resolve the spectral contrast between asphalt, concrete, bare soil, and vegetation with enough granularity to distinguish a gravel car park from a sealed one, or a grass median from a compacted dirt strip. Landsat 9 adds a longer archive and a thermal channel that responds to latent-heat flux, a proxy for actual evapotranspiration and, by inference, available soil moisture. TanDEM-X supplies the terrain model that determines whether water arriving at a permeable surface has time to infiltrate or is simply routed to the nearest drain.
Building the recharge-potential surface, layer by layer
The analytical workflow stacks four independent layers. First, an impervious-surface classification derived from Sentinel-2 multispectral composites using spectral unmixing or random-forest classifiers trained on published urban spectral libraries. Impervious fraction is expressed per pixel, not as a binary, because a 10 m cell containing a tree canopy over a gravel surface behaves very differently from one that is pure tarmac.
Second, a slope layer from TanDEM-X or SRTM. Slopes above roughly 5 degrees are generally excluded from recharge-potential zones in standard hydrological practice, because surface runoff dominates at steeper gradients before infiltration can act. The threshold is adjustable; clay-rich soils may require even flatter terrain.
Third, a soil hydrological group layer. This typically comes from published national or FAO soil databases rather than from the satellite itself. Soils are classified into groups A through D by the USDA system, ranging from high infiltration rates (sands and gravels) to very low rates (clays and compacted fills). The satellite classification is intersected with this layer to weight permeable land cover appropriately: a sandy vacant lot and a clay-underlain park are both green, but they behave very differently.
Fourth, a vegetation index composite. NDVI from Sentinel-2 identifies actively vegetated surfaces, which generally maintain soil structure and macroporosity better than bare or compacted ground. Seasonal compositing over a full year reduces the risk of misclassifying a seasonally dry lawn as bare soil.
What the method cannot see, and where that matters
Several important limits apply. Optical sensors cannot penetrate cloud, which is a practical problem in humid tropical cities where recharge seasons coincide with persistent overcast. Sentinel-2 and Landsat 9 both require cloud-free compositing over weeks to months to produce reliable land-cover maps; single-date classifications in monsoon climates are unreliable.
The satellite layers characterise surface conditions, not subsurface geology. A highly permeable surface over a clay aquitard will not recharge a deep aquifer regardless of what the imagery shows. The recharge-potential map is therefore a necessary but not sufficient input to aquifer protection decisions; it must be combined with hydrogeological survey data to be defensible in a planning context.
Spatial resolution also sets a floor on what is detectable. At 10 m, Sentinel-2 can identify a small park or a wide road median, but a 2 m-wide tree-pit strip or a permeable paving panel within a car park will be subsumed into the surrounding impervious signal. Very-high-resolution commercial imagery can close this gap for targeted site assessments, but at substantially higher cost and narrower coverage.
Finally, soil databases in many cities are outdated or were compiled at agricultural rather than urban scales. Urban soils are frequently disturbed, compacted by construction, or capped with fill material that bears no relation to the underlying natural soil class. Ground-truthing of soil assumptions is the weakest link in most satellite-derived recharge assessments.
From recharge map to planning instrument
A recharge-potential surface is most useful when it is classified into three or four discrete zones: high, moderate, low, and negligible potential. These zones can then be overlaid with existing zoning maps, infrastructure plans, and proposed development footprints to identify where new impervious cover would eliminate the highest-value recharge areas.
In practice, the most actionable output is often a ranked list of parcels or sub-catchments where recharge potential is high but land use is not yet sealed. These are the sites where sustainable urban drainage systems (SUDS), infiltration trenches, or development restrictions would have the greatest effect on aquifer replenishment. Conversely, the map identifies where recharge is already negligible, so that planners are not protecting land that provides no hydrological benefit.
Change detection between two epochs, using the Landsat or Sentinel-2 archive, quantifies how much recharge-capable surface has been lost to development over a given period. This is a politically legible number: it translates satellite data into a concrete statement about what a city has given up, and at what rate.
Practical considerations for a city-scale commission
A city of 500 km² sits comfortably within a single Sentinel-2 granule. Processing a cloud-free annual composite, running the classification, and intersecting it with terrain and soil layers is a matter of weeks, not months, using cloud-based analysis environments. The main time cost is usually validation: collecting field reference points to assess classification accuracy and reconciling the satellite soil assumptions with local borehole or survey records.
Satellize applies this workflow using open Sentinel and Landsat archives, with TanDEM-X added where clients hold or can acquire a licence. The approach is comparable to the methods used in published academic recharge-mapping studies across cities in sub-Saharan Africa, South Asia, and the Middle East, where groundwater is under the most acute pressure. Clients receive a GIS layer set, a methodology report, and a change-detection baseline that can be updated annually.
For governments considering a broader sovereign data programme, a recharge-zone map is a natural companion to impervious-surface and urban heat island analyses. The three share most of their input data and together address the hydrological, thermal, and drainage dimensions of urban climate adaptation in a single analytical pass.
Typical figures
| Spatial resolution (land cover) | 10 m (Sentinel-2), 30 m (Landsat 9) |
| Spatial resolution (terrain) | 12 m posting (TanDEM-X), 30 m (SRTM) |
| Revisit for land-cover compositing | 5 days (Sentinel-2 two-satellite constellation); 16 days (Landsat 9) |
| Spectral bands used | Visible (Blue, Green, Red), NIR, SWIR-1, SWIR-2 from Sentinel-2 MSI; OLI-2 equivalent bands plus TIRS-2 thermal from Landsat 9 |
| Vertical accuracy of terrain input | ~2 m relative accuracy (TanDEM-X urban); ~16 m absolute at 90% confidence (SRTM) |
| Minimum resolvable permeable surface | Approximately 100 m² at Sentinel-2 10 m resolution; sub-pixel unmixing can detect partial permeability at smaller scales with reduced precision |
| Archive depth | 2015-present (Sentinel-2); 1972-present (Landsat family) |
| Cloud sensitivity | Optical sensors require cloud-free compositing; persistent cloud cover in humid climates extends compositing windows to 3-6 months |
| Delivery formats | GeoTIFF raster layers, vector zone polygons (GeoPackage or Shapefile), methodology report (PDF) |
| Typical area coverage per commission | City to metropolitan scale, 100-5,000 km², within a single Sentinel-2 granule or Landsat scene |
Analytics Satellize can run
| Impervious-surface fraction map | Spectral unmixing or random-forest classification of Sentinel-2 multispectral composites using published urban spectral endmembers | GeoTIFF layer (0-100% impervious per pixel), accuracy report with confusion matrix |
| Recharge-potential zone map | Multi-criteria overlay of impervious fraction, slope (TanDEM-X or SRTM), and soil hydrological group (FAO or national database); scored and classified into four potential tiers | Vector polygon layer with zone classification and scoring attributes, suitable for import into planning GIS |
| High-value recharge parcel ranking | Intersection of recharge-potential zones with cadastral or planning parcel boundaries; parcels ranked by recharge score and current land-use designation | Ranked parcel table (CSV and GIS layer) identifying priority sites for SUDS or development restrictions |
| Recharge surface loss quantification | Bi-temporal change detection using Sentinel-2 or Landsat composites from two epochs; area of permeable-to-impervious conversion calculated per sub-catchment | Change statistics table and difference raster; narrative summary of loss rates by zone |
| Annual update and monitoring layer | Repeat classification on new annual Sentinel-2 composite; differenced against baseline to flag new impervious cover within protected recharge zones | Annual GIS layer update and change-alert report |
| Slope and flow-accumulation terrain analysis | Hydrological conditioning of TanDEM-X or SRTM DEM; D8 or D-infinity flow-routing to delineate sub-catchments and identify low-gradient ponding zones | Slope raster, flow-accumulation raster, and sub-catchment boundary polygons |
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.