A Kalman filtering algorithm is developed to forecast groundwater levels in the Upper Floridan aquifer throughout the St. Johns River Water Management District (SJRWMD) in Florida. The algorithm processes historic and currently available head measurements to make optimal predictions of future head levels over a grid of 554 wells spanning the SJRWMD. Measurements are obtained monthly from a subset of 20 wells and semiannually from the remaining wells. The Kalman filter incorporates an empirical spatiotemporal model of regional groundwater fluctuations derived from long‐term historical data records at the 20 monthly measured wells. The algorithm (1) extrapolates the measurements provided by the 20 monthly measured wells to estimate monthly head levels at all 554 wells in the grid and (2) predicts future head levels at each well in the absence of measurements. The performance of the Kalman filtering algorithm is assessed by examining its ability to forecast piezometric head behavior at the 534 well locations where historic data were not used to estimate either the system model or the spatiotemporal correlation structure of the model residual
展开▼
