E.E. Ellis (1909) described the relation of rock type to source-water areas in Connecticut by carefully examining fracture patterns in outcrops and relating them to hydrologic characteristics observed using water wells. He noted that “the contributing area to a granite well should occupy a space with an approximately uniform radius around the well.” In schist and gneiss, “a single well, instead of drawing water from an area surrounding it on all sides, will draw from long distances through the feeding fractures and vertical fractures connecting with them.” In order to illustrate Ellis’s perceptive conceptual models, groundwater flow was simulated in a typical New England aquifer system consisting of glacial deposits and bedrock. Rock type (granite or schist end-members) was simulated by including fracture orientation in the model. Simulated flow paths that originate in the uplands pass through glacial deposits before reaching bedrock, and water produced by a well can be a mixture of water from various source areas. Source-water areas and contributions are affected mainly by groundwater recharge rates and by rock hydraulic properties. A reduction in groundwater recharge causes a larger percentage of the source-water area to be in the glacial stratified deposits than in the uplands. This result underscores the need to have accurate information on the rate of recharge to bedrock from till under both pumping and non-pumping conditions. Independent estimates of recharge rates could be obtained from streamflow records, tracer concentrations (natural or artificial), and water-level fluctuations.

In a detailed simulation of a valley aquifer system in Connecticut, we found that 50% of the recharge to the uplands flows to bedrock deeper than 18 m. Most water from deep bedrock discharges upward into glacial stratified deposits in the valley. Tritium/helium residence time in a bedrock well (9.9 years) is longer than simulated (2.6 years). This can be explained if the actual flow path is longer than simulated because of tortuous paths through a fracture network (fracture networks are not explicitly included in the simulation); the sample is a mixture of water of a short residence time (through a permeable fracture) and water of a long residence time (from dead-end fractures); or, the simulated bulk rock porosity is too low.