Olsen, A. R. 2004. Spatially Balanced Survey Design for Groundwater Using Existing Wells. ASA Proceedings of the Joint Statistical Meetings: 3082-3086. WED-04-015
Many states use existing wells to monitor groundwater. Typically, a state maintains a database of all well locations, including limited information on type of well and aquifer. The density of wells tends to be higher near urban areas, resulting in a very uneven spatial distribution of wells. If the objective is to estimate the water quality of wells, then this does not present any particular problem. If the objective is to estimate the quality of underlying groundwater aquifers, then the question is how to design a survey of wells that will adequately sample the aquifer. Using an example set of well locations in a subregion of Florida, we develop an approach that results in a spatially balanced sample of the aquifer. The design is based on the Generalized Random Tessellation Stratified (GRTS) survey design. First, we create an estimated two-dimensional density surface of wells using a two-dimensional kernel density estimator. We use the inverse of the well density surface as unequal inclusion probability densities for use in the GRTS survey design. Our approach is compared to a simple random sample and two-stage stratified area sample of wells.