Technical note: Biases and uncertainties in estimating water table depths using high-resolution SPOT / Pléiades images of traditional livestock watering practices in the Sahel

Abstract. Groundwater resources in Africa remain largely understudied despite their crucial role in sustaining socio-economic activities, rural livelihoods and climate resilience. This knowledge gap is mainly due to the scarcity of long-term, reliable and spatially extensive in situ monitoring data. In this work, we assess the reliability of a groundwater survey method based on photo-interpretation of high-resolution satellite imagery (SPOT-6, Pléiades and Pléiades Neo; ≤ 1.5 m in resolution, years 2014 to 2025) to estimate the water table depth (WTD) of thousands of wells exploited through traditional livestock watering practices in the Sahel region. These widespread practices involve drawing groundwater from hand-dug wells using ropes and large leather or rubber buckets pulled at a distance by animals. Repeated animal passages create elongated surface tracks that radiate from the well and may extend over several tens of metres, forming radial patterns visible on high-resolution optical images. Using a long-term, pluri-decadal in situ dataset (2003–2025, AMMA-CATCH observatory), we identified 34 wells in SW Niger used during the dry season for intensive livestock watering, for which WTD in situ measurements were coincident in time and could be compared with high-resolution satellite image tracks length estimates. Dry-season imagery (December–March) coincides with both intensive water abstraction for livestock and a cloud-free troposphere, making this approach all the more suitable for detecting well features and surroundings. Measured water table depths ranged from 15 to 59 m below ground level (dry season), while satellite-derived track lengths systematically overestimated in situ WTD measurements, with a median bias of 6.4 ± 2.0 m (median absolute error). This plurimetric offset was shown to reflect the cumulative contributions of i) geometrical (well’s structure and position in the landscape), ii) operational (the way animal herders withdraw groundwater near the well) and iii) hydrogeological (well’s productivity and its changes through time) bias terms. With these limitations and biases duly considered, large-scale regional WTD estimates derived from HR images represent a valuable opportunity to map groundwater depth across large parts of the Sahel region (~3 Mkm²), while also updating and densifying water-point inventories for better planning of water development facilities. This method opens up the possibility of an improved parameterization of large-scale hydrological models with updated and denser datasets at regional, transboundary aquifer scales.