The Northern Guam Lens Aquifer is an island karst aquifer in young, highly conductive limestone units that transmit groundwater readily through diffuse flow, but exhibit evidence of significant transport through discrete pathways as well. We applied a numerical model as a heuristic tool to estimate aquifer properties and test hypotheses regarding aquifer hydrology. Estimated average regional hydraulic conductivity obtained from history-matching of simulated and observed levels is about 6 km/day. The juxtaposition of this value against local values of meters to hundreds of meters per day from pumping tests at newly installed production wells suggests a dual-porosity model would be appropriate for the aquifer, though Darcian models suffice for simulating certain large-scale behaviors. Discrepancies between model predictions based on specified assumptions and observations from field study and instruments provide insights for refining hypotheses regarding infiltration, recharge, and discharge. For our numerical analysis, we estimated minimum infiltration to be about 67% of mean annual rainfall, based on differences between daily rainfall and pan evaporation rates. Analysis of historical daily rainfall records indicates about 20% of rainfall arrives in daily amounts of 0.6 cm or less from light showers and under conditions that make much of it unlikely to escape evapotranspiration. Rapid rises and recessions in well hydrographs associated with rainfall in daily amounts of 5 cm or more suggest another 20% of rainfall infiltrates through the vadose zone too rapidly for the fresh water lens to capture it in long-term phreatic storage. Maximum exploitable recharge, i.e., the amount of recharge that is retained by the lens long enough to be available for extraction is thus estimated to be about 60%. Calculations of mean monthly water levels in which monthly recharge is assumed to equal monthly infiltration, consistent with assumptions in previous studies, predict significantly higher amplitude in seasonal water level curves than observed at observation wells,. This suggests a significant portion of monthly infiltration during the wet season is retained in vadose storage and released gradually during the dry season. This conclusion is corroborated by the observation that water-level peaks associated with storms arriving early in the dry-to-wet season transition are greatly attenuated and exhibit significant lag compared to responses to similar wet-season rainfall events. Discrepancies between numerical estimates of long-term average discharge and estimates of discharge observed in the field during the dry-to-wet season transition suggest that either modeled estimates are too high for some sectors of the coast, substantial amounts of groundwater discharge escaped detection during our surveys of the beach and near-shore waters, or that there is significant temporal variation in discharge, or that some combination of these possibilities applies. Temporal variations in discharge could include short-term, storm-driven variation and seasonal variation. Discrepancies between calculated and observed depths to the fresh water/salt water interface suggest there might be unaccounted-for vertical velocity in groundwater flow and/or significant stratigraphic inhomogeneity in regional hydraulic conductivity.