Processing, Analysis, and General Evaluation of Well-Driller Logs for Estimating Hydrogeologic Parameters of the Glacial Sediments in a Ground-Water Flow Model of the Lake Michigan Basin

Abstract

In 2005, the U.S. Geological Survey began a pilot study for the National Assessment of Water Availability and Use Program to assess the availability of water and water use in the Great Lakes Basin. Part of the study involves constructing a ground-water flow model for the Lake Michigan part of the Basin. Most ground-water flow occurs in the glacial sediments above the bedrock formations; therefore, adequate representation by the model of the horizontal and vertical hydraulic conductivity of the glacial sediments is important to the accuracy of model simulations. This work processed and analyzed well records to provide the hydrogeologic parameters of horizontal and vertical hydraulic conductivity and ground-water levels for the model layers used to simulated ground-water flow in the glacial sediments. The methods used to convert (1) lithology descriptions into assumed values of horizontal and vertical hydraulic conductivity for entire model layers, (2) aquifer-test data into point values of horizontal hydraulic conductivity, and (3) static water levels into water-level calibration data are presented. A large data set of about 458,000 well driller well logs for monitoring, observation, and water wells was available from three statewide electronic data bases to characterize hydrogeologic parameters. More than 1.8 million records of lithology from the well logs were used to create a lithologic-based representation of horizontal and vertical hydraulic conductivity of the glacial sediments. Specific-capacity data from about 292,000 well logs were converted into horizontal hydraulic conductivity values to determine specific values of horizontal hydraulic conductivity and its aerial variation. About 396,000 well logs contained data on ground-water levels that were assembled into a water-level calibration data set. A lithology-based distribution of hydraulic conductivity was created by use of a computer program to convert well-log lithology descriptions into aquifer or nonaquifer categories and to calculate equivalent horizontal and vertical hydraulic conductivities (K and KZ, respectively) for each of the glacial layers of the model. The K was based on an assumed value of 100 ft/d (feet per day) for aquifer materials and 1 ft/d for nonaquifer materials, whereas the equivalent KZ was based on an assumed value of 10 ft/d for aquifer materials and 0.001 ft/d for nonaquifer materials. These values were assumed for convenience to determine a relative contrast between aquifer and nonaquifer materials. The point values of K and KZ from wells that penetrate at least 50 percent of a model layer were interpolated into a grid of values. The K distribution was based on an inverse distance weighting equation that used an exponent of 2. The KZ distribution used inverse distance weighting with an exponent of 4 to represent the abrupt change in KZ that commonly occurs between aquifer and nonaquifer materials. The values of equivalent hydraulic conductivity for aquifer sediments needed to be adjusted to actual values in the study area for the ground-water flow modeling. The specific-capacity data (discharge, drawdown, and time data) from the well logs were input to a modified version of the Theis equation to calculate specific capacity based horizontal hydraulic conductivity values (KSC). The KSC values were used as a guide for adjusting the assumed value of 100 ft/d for aquifer deposits to actual values used in the model. Water levels from well logs were processed to improve reliability of water levels for comparison to simulated water levels in a model layer during model calibration. Water levels were interpolated by kriging to determine a composite water-level surface. The difference between the kriged surface and individual water levels was used to identify outlier water levels. Examination of the well-log lithology data in map form revealed that the data were not only useful for model input, but also were useful for understanding th