UNCERTAINTY ASSESSMENT OF AQUIFER HYDRAULIC PARAMETERS FROM PUMPING TEST DATA

Abstract

ABSTRACT Groundwater management is almost impossible without a detailed and exact understanding of the aquifer parameters. Groundwater flow forecasts are always subject to uncertainty in estimating aquifer parameters, which are critical for establishing and calibrating groundwater models. Because groundwater predictions cannot be given in a firm, deterministic manner, the proper prediction for the range of aquifer parameters provides groundwater resources managers and decision makers with the probabilistic outcome so that risks associated with decisions can be readily evaluated. Practically speaking, pumping test data is noisy which means that even under ideal conditions no two pump tests will result in the same drawdown traces. As a consequence, various pumping experiments lead to different values for aquifer parameter estimates. The data of pumping tests are analyzed using traditional method (aquifer test, and AQtesolv software) which depends on trial and error technique. During these methods non-unique values of hydraulic parameters are selected which usually have a high level of uncertainty. Several groundwater models are critical, but uncertainty must be taken into account in determining aquifer parameters especially when using these models for decision making. The main goal of this study is to build a comprehensive tool for quantifying uncertainty associated with hydraulic parameters estimation from different pumping test conditions for fully penetrating wells in confined and semi-confined aquifers. In order to achieve the previous objective, a FORTRAN code was developed to apply Markov Chain Monte Carlo (MCMC) algorithm using different likelihood functions (exponential, inverse, and log). This developed tool can be used to detect the most probable range of aquifer parameters that are consistent with pumping test data with level of confidence. The tool was applied to different literature and hypothetical cases (four cases for confined aquifer and three cases for leaky aquifer) using different likelihood functions for tool’s calibration/evaluation and to illustrate the uncertainty in the quantification of aquifer parameters and comparing the results with traditional method’s results. By applying the analytical tool developed from the study, a range of aquifer parameters (Transmissivity, storativity and resistance) was produced near to the real values and with greater reliability instead of using a single value to be used as input in groundwater modeling. To verify and prove the concept of the effectiveness of the developed and comprehensive analytical tool, a hypothetical case was created with known values of hydraulic parameters for confined and semi-confined aquifer layers by building 3D numerical models of groundwater (using Modflow program). The aquifer parameters generated using the new comprehensive analytical tool were very close to the true values entered to the model compared to using the traditional Aquifer Test Software. In addition, the effect of uncertainty quantification of aquifer parameters on the expected drawdown in the groundwater level in the long term was determined. Finally, the new developed tool can be used with high level of confidence to predict the most reliable range of aquifer parameters to reduce the uncertainty in using a single value resulting from using the traditional methods.