Geostatistical Procedures for Developing Three-Dimensional Aquifer Models from Drillers' Logs

The Hydrostratigraphic Drilling Record Assessment (HyDRA) project is developing procedures for employing the vast but highly qualitative hydrostratigraphic information contained in drillers' logs in the development of quantitative three-dimensional (3D) depictions of subsurface properties for use in flow and transport models to support groundwater management practices. One of the project's objectives is to develop protocols for 3D interpolation of lithological data from drillers' logs, properly accounting for the categorical nature of these data. This poster describes the geostatistical procedures developed to accomplish this objective. Using a translation table currently containing over 62,000 unique sediment descriptions encountered during the transcription of over 15,000 logs in the Kansas High Plains aquifer, the sediment descriptions are translated into 71 standardized terms, which are then mapped into a small number of categories associated with different representative property (e.g., hydraulic conductivity [K]) values. Each log is partitioned into regular intervals and the proportion of each K category within each interval is computed. To properly account for their compositional nature, a logratio transform is applied to the proportions. The transformed values are then kriged to the 3D model grid and backtransformed to determine the proportion of each category within each model cell. Various summary measures can then be computed from the proportions, including a proportion-weighted average K and an entropy measure representing the degree of mixing of categories within each cell. We also describe a related cross-validation procedure for assessing log quality.