Information on sediment texture and spatial continuity are inherent to sedimentary depositional facies descriptions, which are therefore potentially good predictors of spatially varying hydraulic conductivity (K). Analysis of complex alluvial heterogeneity in Livermore Valley, California, USA, using relatively abundant core descriptions and field pumping-test data, demonstrates a depositional-facies approach to characterization of subsurface heterogeneity. Conventional textural classifications of the core show a poor correlation with K; however, further refinement of the textural classifications into channel, levee, debris-flow and flood-plain depositional facies reveals a systematic framework for spatial modeling of K. This geologic framework shows that most of the system is composed of very low-K flood-plain materials, and that the K measurements predominantly represent the other, higher-K facies. Joint interpretation of both the K and geologic data shows that spatial distribution of K in this system could not be adequately modeled without geologic data and analysis. Furthermore, it appears that K should not be assumed to be log-normally distributed, except perhaps within each facies. Markov chain modeling of transition probability, representing spatial correlation within and among the facies, captures the relevant geologic features while highlighting a new approach for statistical characterization of hydrofacies spatial variability. The presence of fining-upward facies sequences, cross correlation between facies, as well as other geologic attributes captured by the Markov chains provoke questions about the suitability of conventional geostatistical approaches based on variograms or covariances for modeling geologic heterogeneity.
