A series of freshwater subsurface injection, storage, and recovery tests were conducted at an injection-well site near Lake Okeechobee in Okeechobee County, Florida, to assess the recoverability of injected canal water from the Lower Floridan aquifer. At the study site, the Lower Floridan aquifer is characterized as having four local, relatively independent, high-permeability flow zones (389 to 398 meters, 419 to 424 meters, 456 to 462 meters, and 472 to 476 meters below sea level). Four subsurface injection, storage, and recovery cycles were performed at the Lake Okeechobee injection-well site in which volumes of water injected ranged from about 387,275 to 1,343,675 cubic meters for all the cycles, and volumes of water recovered ranged from about 106,200 to 484,400 cubic meters for cycles 1, 2, and 3. The recovery efficiency for successive cycles 2 and 3 increased from 22 to 36 percent and is expected to continue increasing with additional cycles. A comparison of chloride concentration breakthrough curves at the deep monitor well (located about 171 meters from the injection well) for cycles 1, 4, and test no. 4 (from a previous study) revealed unexpected finings. One significant result was that the concentration asymptote, expected to be reached at concentration levels equivalent or close to the injected water concentration, was instead reached at higher concentration levels. The injection to recovery rate ratio might affect the chloride concentration breakthrough curve at the deep monitor well, which could explain this unexpected behavior. Because there are four high-permeability zones, if the rate of injection is smaller than the rate of recovery (natural artesian flow), the head differential might not be transmitted through the entire open wellbore, and injected water would probably flow only through the upper high- permeability zones. Therefore, observed chloride concentration values at the deep monitor well would be higher than the concentration of the injected water and would represent a mix of water from the different high-permeability zones. A generalized digital model was constructed to simulate the subsurface injection, storage, and recovery of freshwater in the Lower Floridan aquifer at the Lake Okeechobee injection-well site. The model was constructed using a modified version of the Saturated-Unsaturated TRAnsport code (SUTRA), which simulates variable-density advective-dispersive solute transport and variable-density ground-water flow. Satisfactory comparisons of simulated to observed dimensionless chloride concentrations for the deep monitor well were obtained when using the model during the injection and recovery phases of cycle 1, but not for the injection well during the recovery phase of cycle 1 even after several attempts. This precluded the determination of the recovery efficiency values by using the model. The unsatisfactory comparisons of simulated to observed dimensionless chloride concentrations for the injection well and failure of the model to represent the field data at this well could be due to the characteristics of the Lower Floridan aquifer (at the local scale), which is cavernous or conduit in nature. To test this possibility, Reynolds numbers were estimated at varying distances from the injection well, taking into consideration two aquifer types or conceptual systems, porous media and cavernous. For the porous media conceptual system, the Reynolds numbers were greater than 10 at distances less than 1.42 meters from the injection well. Thus, application of Darcy's law to ground-water flow might not be valid at this distance. However, at the deep monitor well (171 meters from the injection well), the Reynolds number was 0.08 which is indicative of laminar porous media flow. For the cavernous conceptual system, the Reynolds numbers were greater than 2,000 at distances less than 1,000 meters from the well. This number represents the upper limit of laminar flow, which is the fundamental assumption