A method using drawdown derivatives to estimate aquifer properties near active groundwater production well fields

Abstract

This thesis describes the development of a new inverse modeling approach to estimate aquifer properties in the vicinity of continuously-pumped well fields. The specific emphasis is on deep bedrock aquifers where monitoring well installation is often not practicable due to high drilling costs. In these settings, water levels from groundwater production wells offer a transient dataset that can be used to estimate aquifer properties. Well interference effects, if detectable at neighboring production wells, allow for an interrogated aquifer volume that is larger (and therefore more representative at the well field scale) when compared to single well hydraulic tests. The parameter estimation method utilizes drawdown derivatives to estimate the aquifer transmissivity and storativity. The forward model consists of an initial water level (or a recoverable water level drift function), an analytical solution for aquifer drawdown, and a correction term for well loss. The aquifer drawdown component is based on superposition of the Theis solution, although other analytical solutions are also applicable. The observed dataset was judiciously trimmed to reduce computer run-time while retaining enough points to adequately characterize aquifer and well parameters. By limiting observation points to special domains, the calculated drawdown and observed well water level derivatives with respect to time are independent of well loss, and therefore the transmissivity and storativity can be estimated without knowledge of the recoverable water level or loss coefficient for individual pumping wells. Aquifer properties in the forward model were estimated by minimizing the difference between the modeled and observed drawdown derivatives. The parameter estimation method is tested using hourly water level and pumping data from municipal well fields producing groundwater from sandstone aquifers of the Denver Basin. Data collected over a seven-year period from two distinct well fields, one operating in the Denver aquifer and another operating in the Arapahoe aquifer, are considered. The estimated transmissivities are 30.0 m2/d and 46.5 m2/d for the Denver and Arapahoe aquifers, respectively, whereas the storativities are 4.7×10-4 and 2.0×10-4, respectively. These estimates are within the range of previously reported values, indicating that production well data can be used to derive reasonable aquifer properties. A separate synthetic aquifer test case was considered to further test the parameter estimation methodology, as well as to evaluate the appearance of Theis-like response behavior at the wells. Synthetic water levels were generated using a numerical model with geostatistically-simulated heterogeneity that is characteristic of the Denver Basin (sandstone bodies separated by less permeable inter-bedded siltstone and shale). Analysis of the synthetic water levels revealed meaningful hydraulic properties; the effective hydraulic conductivity (best-fit transmissivity divided by the modeled aquifer thickness) was slightly higher than the geometric mean hydraulic conductivity of the heterogeneous field. In addition to aquifer properties, observed water level data were used to estimate the well-loss coefficient and recoverable water level for individual pumping wells. Loss coefficients obtained for wells in the Denver Basin indicate that this mechanism (head losses due to turbulence around the well screen) may contribute between 20 and 150 m of the total drawdown (based on a pumping rate of 1500 m3/d) commonly observed in these wells. The recoverable water level at each well, when fit with a linear drift function, provides a means of investigating the prevailing trend in aquifer heads due to other regional influences outside the modeled well field.