Wildlife water-source mapping and drying trends in drylands
In arid landscapes, water points are the hinge on which wildlife survival turns. Sentinel-1 SAR backscatter, Sentinel-2 multispectral imagery and GPM IMERG precipitation data together reveal which pans are drying, how fast, and why.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground range resolution in Interferometric Wide Swath mode, 6-day repeat at mid-latitudes (12-day for a single satellite). C-band backscatter drops sharply over open water and rises over moist bare soil, making it sensitive to inundation extent and soil-moisture halos around water points. Cloud-penetrating, so it works through the wet season when optical imagery is routinely blocked.
- Sentinel-2 MSI (ESA): 10 m visible and near-infrared bands, 20 m shortwave-infrared bands, 5-day revisit with both satellites. NDWI (Green minus NIR) and MNDWI (Green minus SWIR) indices reliably delineate open water at this resolution. Reliable detection floor is roughly 0.1 ha for isolated water bodies; smaller pans are frequently missed or confused with shadow.
- Planet SuperDove: 3 m resolution, eight spectral bands including red-edge, daily revisit over most land areas on a commercial licence. Resolves water points below 0.1 ha that Sentinel-2 cannot, and captures the wet-soil halo that indicates recent inundation even after surface water has evaporated. Tasking requires a client licence; archive depth varies by location.
- GPM IMERG (NASA/JAXA): Global Precipitation Measurement Integrated Multi-satellitE Retrievals, 0.1-degree (~10 km) spatial resolution, half-hourly to monthly composites from 2000 onwards. Used to compute rainfall anomalies at seasonal and annual timescales and to attribute pan-drying trends to precipitation deficits rather than groundwater abstraction or land-use change.
Why water-point geometry matters more than rainfall totals
Rainfall in drylands is notoriously patchy. A seasonal average that looks adequate at 10 km resolution can conceal a 60 km gap between functioning water points, which is lethal for large ungulates with daily drinking requirements. The spatial configuration of water points, not just their aggregate number, determines whether a landscape is passable for wildlife. Satellite mapping makes that geometry legible for the first time across entire protected areas or transboundary corridors.
Natural pans fill from overland flow and lose water through evaporation and infiltration. Their persistence through the dry season depends on pan morphology, catchment area, soil texture and antecedent rainfall. Artificial water points (boreholes, dams, troughs) add a managed layer that can either supplement or distort natural wildlife movement patterns. Distinguishing the two classes from imagery requires combining spectral water indices with ancillary data on infrastructure, since a concrete trough and a small clay pan can look identical at 10 m resolution.
What a wet-soil halo gives away
Open water is the obvious target, but it is often gone within days of the last rain. The more persistent signal is the moist-soil annulus that surrounds a pan or spring after surface water has retreated. C-band SAR backscatter is sensitive to dielectric properties of the soil surface, and liquid water raises the dielectric constant sharply. A pan that appears dry in a Sentinel-2 true-colour image in late May may still show a statistically significant backscatter depression in the co-polarised (VV) channel relative to the surrounding dry matrix, indicating residual shallow moisture.
Combining Sentinel-1 backscatter time series with Sentinel-2 MNDWI composites produces a more complete inundation history than either sensor alone. The standard approach is to classify each Sentinel-1 acquisition against a dry-reference backscatter stack, flag pixels below a threshold (typically around -15 to -18 dB over bare soil in savanna contexts, though this varies by soil type and vegetation cover), and sum the inundated-day counts per pixel across a season. The result is an inundation-frequency map: how many days in a given period each pixel was detectably wet. Pans that score high in wet years but near zero in dry years are the most drought-sensitive nodes in the network.
Linking pan drying to rainfall anomalies: method and honest limits
GPM IMERG provides a continuous precipitation record from 2000 onwards at half-hourly resolution, aggregated to monthly or seasonal totals for trend analysis. Correlating pan inundation frequency (from Sentinel-1, available from 2014) or MNDWI water-area time series (from Sentinel-2, 2015 onwards, or Landsat-8 for earlier years) against IMERG seasonal anomalies allows analysts to separate rainfall-driven drying from drying caused by groundwater abstraction, borehole failure or catchment modification upstream.
The honest limit here is spatial mismatch. IMERG pixels are roughly 11 km across; individual pans are tens of metres. Attributing a specific pan's drying to a rainfall deficit requires assuming that the IMERG pixel is representative of local catchment rainfall, which fails in areas of strong orographic gradients or highly localised convective storms. Where high-density rain-gauge networks exist, bias-corrected IMERG products (the Final Run, rather than the Early or Late runs) reduce this error. Where they do not, the attribution remains probabilistic. That uncertainty should appear in any delivered product, not be quietly buried in a methodology appendix.
Detection floors and what falls below them
Freely available sensors impose a hard lower bound on what can be reliably detected. Sentinel-2 at 10 m means that a water body must cover roughly 0.1 ha (about 32 x 32 m) to be detected with acceptable commission and omission error rates. Many ecologically significant water points in dryland systems, including seeps, small springs and shallow rock pools, are well below this. SAR can sometimes detect the soil-moisture signal from these features even when the open-water extent is sub-pixel, but the relationship is noisy and requires careful validation against ground truth.
Planet SuperDove at 3 m pushes the detection floor down to features of a few hundred square metres under good conditions, but commercial tasking costs and archive gaps mean it is most practical for targeted monitoring of known high-priority sites rather than systematic landscape-wide surveys. No satellite system currently operational can reliably map water points smaller than a few tens of square metres. For those, field survey or drone-based mapping remains necessary, and satellite data plays a supporting role by directing survey effort to areas of highest uncertainty.
From maps to management: what the analytics actually deliver
The immediate output of a water-source mapping programme is a georeferenced inventory of detected water points with inundation-frequency scores, pan-area time series and a drought-sensitivity classification. That inventory is most useful when it feeds directly into ranger patrol planning, borehole maintenance scheduling and wildlife movement modelling. A pan that was reliably wet for 90 days per year in 2016 to 2019 but has averaged fewer than 30 days since 2021 is a specific, actionable finding: it warrants investigation of whether the catchment has been fenced, whether a borehole has been decommissioned or whether the rainfall deficit is structural.
Satellize runs this class of analytics on open constellations and adds commercial tasking on client licence, applying the same pipeline architecture used in its Tonga crop-estimation programme to dryland water-point contexts. Outputs are delivered as GIS layers, seasonal summary reports and, where agreed, alert feeds that flag anomalous drying events within days of a Sentinel-1 acquisition. The alert latency depends on ESA's Copernicus Open Access Hub processing and distribution schedule, which for Sentinel-1 IW products is typically one to three days after acquisition.
Typical figures
| Spatial resolution (SAR, Sentinel-1 IW) | 10 m range x 10 m azimuth (multi-looked product) |
| Spatial resolution (optical, Sentinel-2) | 10 m (visible/NIR), 20 m (SWIR) |
| Spatial resolution (commercial, Planet SuperDove) | 3 m (client licence required) |
| Revisit frequency | Sentinel-1: 6 days (two satellites); Sentinel-2: 5 days; Planet SuperDove: daily |
| Precipitation product resolution | GPM IMERG: ~0.1 degree (~11 km), half-hourly to monthly |
| Minimum reliably detectable water body | ~0.1 ha (Sentinel-2 optical); smaller features detectable via SAR soil-moisture signal but with higher uncertainty |
| Archive depth | Sentinel-1: from 2014; Sentinel-2: from 2015; GPM IMERG: from 2000; Landsat-8 MNDWI usable as proxy back to 2013 |
| Cloud penetration | Sentinel-1 SAR: all-weather; optical sensors (Sentinel-2, Planet): blocked by cloud |
| Inundation-frequency latency | 1 to 3 days after acquisition (Sentinel-1 Copernicus Hub processing) |
| Delivery formats | GeoTIFF, GeoPackage, cloud-optimised GeoTIFF, PDF seasonal report, optional alert feed (JSON/webhook) |
Analytics Satellize can run
| Pan-network inventory with inundation-frequency scores | Sentinel-1 backscatter thresholding against dry-reference stack; Sentinel-2 MNDWI water classification | GeoPackage of water-point polygons with per-season inundation-day counts and area time series |
| Drought-sensitivity classification of individual pans | Correlation of inundation frequency against GPM IMERG seasonal anomalies across the available archive | GIS layer with drought-sensitivity score per pan; PDF summary for management planning |
| Anomalous drying alert | Change detection on rolling Sentinel-1 backscatter composite; flag when inundation frequency drops below site-specific historical percentile | Alert feed (JSON/webhook) triggered within 1 to 3 days of acquisition; includes pan ID, coordinates and deviation magnitude |
| Rainfall-attribution report | Regression of pan water-area change against GPM IMERG Final Run seasonal totals; residual analysis to flag non-rainfall drivers | Seasonal PDF report with attribution tables and uncertainty ranges; flags pans where drying is inconsistent with rainfall deficit |
| Artificial versus natural water-point classification | Spectral and morphological classification of detected water bodies; cross-reference with available infrastructure datasets | Classified GIS layer distinguishing pans, springs, boreholes and constructed impoundments, with confidence scores |
| Multi-year pan-network connectivity index | Graph-theoretic analysis of inter-pan distances weighted by inundation-frequency scores across dry-season months | Annual connectivity index time series per landscape unit; GIS layer of high-risk network gaps for ranger and maintenance prioritisation |
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.