LOADING MAP
Global groundwater potential screening
GWPilot estimates groundwater potential for any coordinate on Earth and pairs it with a trajectory reading: whether the water at that point is lasting. It is built for the people who decide where to spend survey budget before a rig is hired, which in practice means water, sanitation and hygiene programmes, the contractors who drill for them, and the donors who fund both.
It tells you where to survey, not where to drill
No satellite can tell you whether a specific spot will yield water. What a screening layer can do is cut the area worth surveying by roughly an order of magnitude, so the geophysics budget goes where it has a chance of paying for itself. Every response GWPilot returns carries that statement in the data rather than in a footer, because the statement has to travel with the number.
How the estimate is made
The method is an AHP weighted overlay, which is the approach the published groundwater potential zone literature has converged on. Seven thematic layers are each reclassified onto a common suitability scale, then combined using weights derived from a pairwise comparison matrix on Saaty’s one to nine scale. The weights are the principal eigenvector of that matrix, and the consistency ratio is 0.0182, against Saaty’s threshold of 0.10.
| Layer | Weight |
|---|---|
| lineamentIntersection | 0.276 |
| Lithology | 0.208 |
| Climatic water balance | 0.155 |
| Mean annual rainfall | 0.114 |
| lineamentDensity | 0.083 |
| Slope | 0.061 |
| Subsoil texture | 0.045 |
| Channel influence | 0.033 |
| Local relief | 0.026 |
Every weight is visible and arguable, which matters when the buyer has to defend the spend in a review meeting. A machine learning model cannot be argued with in a review meeting.
What it does not yet include
Lineament density is absent, and it normally carries one of the larger weights in published models. Slope is computed over about 1.1 km, which understates steep terrain and is why mountainous crystalline ground still scores higher than it should. Both are known limits rather than surprises, and both are being worked on.
Sources
- Lithology:
- Macrostrat compiled geologic maps
- Rainfall and evaporative demand:
- ERA5-Land reanalysis, ECMWF, 1991 to 2020
- Terrain:
- Copernicus DEM, 30 m
- Subsoil texture:
- SoilGrids 250 m, ISRIC
- River discharge:
- GloFAS, Copernicus Emergency Management Service
Working in India? The district level INGRES and CGWB tool is here.