HR: 10:55h
AN: H21F-03    [PDF]
TI: Full-Inverse Statistical Calibration: A Monte Carlo approach to determining field-scale relationships between hydrologic and geophysical variables
AU: * Moysey, S
EM: moysey@stanford.edu
AF: Stanford University, Department of Geophyiscs, Stanford, CA 94305 United States
AU: Knight, R J
EM: rknight@pangea.stanford.edu
AF: Stanford University, Department of Geophyiscs, Stanford, CA 94305 United States
AB: Successfully defining field-scale relationships between hydrologic and geophysical variables is a fundamental problem that makes hydrologic estimation based on geophysical data possible. Today, site-specific relationships are commonly calibrated empirically using lab measurements, in-situ well logs, or reconstructed geophysical images. This direct approach to calibration can often fail, however, for reasons such as: 1) failure to account for differences in measurement scales, 2) poor representation of heterogeneity in the calibration data set due to sparse sampling, and 3) the introduction of processing artifacts that can impact the interpretation of geophysical data. To overcome issues such as these we introduce the concept of Full-Inverse Statistical (FISt) calibration, a Monte Carlo approach to rock physics calibration. In FISt calibration, geostatistical simulation is used to generate multiple synthetic analogs of a field-site. Each of these analog models is then numerically subjected to the same hydrologic and geophysical experiments, data processing, and inversion procedures used when working with the true data collected in the field. Through this process, we obtain a comprehensive synthetic calibration data set that inherently accounts for differences in measurement scale and processing artifacts, which can then be applied to the true geophysical measurements. As an example, we use FISt calibration to estimate the spatial distribution of water content from a dielectric constant model obtained using cross-borehole ground penetrating radar. We have found that FISt calibration is able to consistently produce better estimates of properties than other calibration methods. In particular, FISt calibration greatly improves estimates of hydrologic properties when the rock physics relationship depends on spatial location, i.e., is non-stationary, due to intrinsic subsurface heterogeneity, experimental design, or inversion artifacts. Because FISt calibration integrates all available information about the subsurface, we anticipate that it will be a useful tool for obtaining the best site-specific relationships possible between field-scale hydrologic and geophysical data.
DE: 0910 Data processing
DE: 0915 Downhole methods
DE: 1829 Groundwater hydrology
DE: 1869 Stochastic processes
DE: 5100 PHYSICAL PROPERTIES OF ROCKS
SC: Hydrology [H]
MN: 2003 Fall Meeting