Fresh-water lenses and practical limitations of their three-dimensional simulation

Fresh-water lenses are the major sources of water supply in many atoll islands in the Pacific and Indian Oceans, particularly in dry seasons. Several two- and three-dimensional models are currently available for the simulation of atoll-island aquifers; however, 2D models cannot include 3D spatial variability of material properties, they must simplify the boundary conditions, and they cannot correctly simulate pumping wells. In an attempt to overcome these difficulties, a 3D model, SALTFLOW, was adopted for the simulation of Home Island in the Indian Ocean. This exercise required a discretisation on the order of a few metres and time steps of a few hours requiring significantly high CPU times. High CPU demand proved to be a difficult challenge but cannot be considered a serious practical limitation with today's advanced computers. The exhaustive data demands of the model (e.g., 3D distributions of hydraulic conductivity, porosity, dispersivities, and spatial and temporal variations of recharge and extraction rates) proved to be more problematical. Although the Home Island data set is unusually comprehensive by any standards, nonetheless the quality and quantity of the available data proved inadequate to meet the calibration needs of a highly karstic aquifer system. The Home Island modeling demonstrates the practical limitations of 3D models. It raises the concern that our ability to develop computer codes capable of simulating complex systems now exceeds our ability to supply the input data necessary for reliable calibration. Finally, the paper demonstrates the importance of the transient calibration in reliable simulation of various management options and emphasises that transient calibration should be considered as an integral part of any similar 2D or 3D modeling.