A lumped-parameter model and least-squares method were used to simulate temporal variations of discharge from Big Spring, Iowa, USA, from 1983 to 1994. The simulated discharge rates poorly match the observed one when precipitation is taken as the sole input. The match is improved significantly when the processes of evapotranspiration and infiltration are considered. The best results are obtained when snowmelt is also included in the model. Potential evapotranspiration was estimated with Thornthwaite's formula, infiltration was calculated through a water-balance approach, and snowmelt was generated by a degree-day model. The results show that groundwater in the limestone aquifer is mainly recharged by snowmelt in early spring and by infiltration from rainfall in later spring and early summer. Simulated discharge was visually calibrated against measured discharge; the similarity between the two supports the validity of this approach. The model can be used to study the effects of climate change on groundwater resources and their quality.