I've no idea to explain how the witch use their "remote sensing" to find underground water reserve. It's nothing but have worked for centuries. Instead of magic method,
I have noted your query a week ago and hoped that you might have received an expert advice from an expert of many on RG for this commonplace problem. If I have understood you correctly from the caption of the query, because your query may be in good literary taste but is not in a scientific language, I am giving some bits of `my non-expert advice below. You may use them by understanding details how to make them operational by going through standard and expert literature:
• In order to demarcate the groundwater potential zones of a study area different thematic maps are to be prepared from remote sensing data, topographic maps, hydro-geological data and well data.
• The thematic maps on hydro-geomorphology and lineaments are to be prepared.
• Drainage map is to be prepared from large scale topographic maps and satellite data.
• Contour map is to be prepared from large scale topographic maps or from aerial photographs using photogrammetric techniques or crudely from Google Earth that gives elevation as well as location of points. Hydrogeological maps are to be prepared on the basis of borehole logs.
• All primary input layers (hydro-geomorphology, lineament, contour and spot elevation, drainage and hydrogeological) are to be georeferenced and digitized in GIS environment with slope map layer prepared from digital elevation data.
• Using hydrogeological interpretation material, thickness and other aquifer parameters may be determined.
• A land cover layer specially using Normalized Difference Vegetation Index (NDVI) describing the surface vegetation coverage and its density is prepared. Theory is that density and coverage of vegetation indicate soil moisture conditions that are related to groundwater conditions.
• The land surface reflectance, which varies with the surface soil moisture, can be obtained from remote sensing image. Therefore, the relationship among the values of reflectance and NDVI both may predict not only availability but also groundwater depth which could be verified using the well observations of same time period.
• Slope map layer will indicate the relative time available for runoff via land to infiltrate.
• The different polygons in the thematic layers may be labelled separately and then registered. In the final thematic layer initially each of the polygons is to be designated on the basis of interpretation of groundwater occurrence according to that layer and weights in terms of interpreted probability level of occurrence may be assigned to each polygon.
• Finally thematic layers are to be converted into grid with related item weight and then integrated and analysed, using weighted aggregation method. The grids in the integrated layer are grouped into different ground water potential zones by a suitable logical reasoning and conditioning with emphasis on hydrogeological condition of each grid cell. The final groundwater potential zone map thus generated is to be validated through groundtruthing for yield.