Ground water levels measured in the vicinity of pumping wells are kriged using a regional-linear and point-logarithmic drift, the latter derived from the approximation to the Theis equation for drawdown in response to a pumping well. Kriging is widely used throughout the hydrogeologic discipline, most commonly as the preferred method for constructing gridded hydrogeologic datasets suitable for contouring. Residuals arising from using the most common (linear) drift to krige water levels in the vicinity of extraction wells often indicate large local departures from the linear drift, which correlate with areas of drawdown. The combined regional-linear and point-logarithmic drift accounts for these drawdowns using a logarithmic approximation for the curvature of the poten-tiometric surface. The drift model approximates the principal physical processes that govern ground water flow and ultimately govern the autocorrelation of ground water elevation data. This approach produces maps of contoured water levels that more realistically represent physical conditions and allow for improved interpretation of measured water-level data by including features and information known to be present. Additional benefits include an improved estimate of the regional (background) hydraulic gradient and generation of an approximately flow-conserved grid suitable for two-dimensional particle tracking.