Post audit of a numerical prediction of wellfield drawdown in a semiconfined aquifer system

A numerical ground water pow model was created in 1978 and revised in 1981 to predict the drawdown effects of a proposed municipal wellfield permitted to withdraw 30 million gallons per day (mgd; 1.1 X 10(5) m(3)/day) of wate r from the semiconfined Floridan Aquifer system. The predictions are based on the assumption that water levels in the semiconfined Floridan Aquifer re ach a longterm, steady-state condition within a few days of initiation of p umping. Using this assumption, a 75 day simulation without water table rech arge, pumping at the maximum permitted rates, was considered to represent a worst-case condition and the greatest drawdowns that could be experienced during wellfield operation. This method of predicting wellfield effects was accepted by the permitting agency. For this post audit, observed drawdowns were derived by taking the differen ce between pre-pumping and post-pumping potentiometric surface levels. Comp arison of predicted and observed drawdowns suggests that actual drawdown ov er a 12 year period exceeds predicted drawdown by a factor of two or more. Analysis of the source of error in the 1981 predictions suggests that the v alues used for transmissivity storativity, specific yield, and leakance are reasonable at the wellfield scale. Simulation using actual 1980-1992 pumpi ng rates improves the agreement between predicted and observed drawdowns. T he principal source of error is the assumption that water levels in a semic onfined aquifer achieve a steady-state condition after a few days or weeks of pumping. Simulations using a version of the 1981 model modified to inclu de recharge and evapotranspiration suggest that it can take hundreds of day s or several years for water levels in the linked Surficial and Floridan Aq uifers to reach an apparent steady-state condition, and that slow declines in levels continue for years after the initiation of pumping. While the 198 1 "impact" model can be used for reasonably predicting short-term, wellfiel d-scale effects of pumping, using a 75 day long simulation without recharge to predict the long-term behavior of the wellfield was an inappropriate ap plication, resulting in significant underprediction of wellfield effects.