A Site-specific Comparison of Permeability Prediction Models in Alluvial Sediments from Physical and Geoelectrical Measurements

An evaluation of five petrophysical models for hydraulic permeability predication was completed for a site-specific set of alluvial sediments. The models considered were the Kozeny-Carmen (1927-1937), Borner (1996), Revil-Cathles (1999), and two Revil-Florsch (2010) models. The river deposited sediments represent a relatively narrow grain-size distribution, and were acquired adjacent to the Kansas River, in Olathe, Kansas. Using measured physical, electrical, and hydraulic data from Slater and Glaser (2003), a comparison of the performance of these five petrophysical models for hydraulic permeability estimation of soils was completed. For models where the key parameter is effective grain-size, three model variations were considered using d(10), d(50), and d(90), resulting in an evaluation of a total of eleven individual models. Parameters included in the models can be classified as three different types: a) physical parameters only, b) electrical parameters only, and c) physical and electrical parameters together. The performance of each model was rated in terms of linear regression R squared, slope, and y-intercept values when plotted against the measured hydraulic conductivities. The top three models were the Kozeny-Carmen, a modified Revil-Cathles, and the Borner model The Kozeny-Carmen model performed with the highest rating, followed by the modified Revil-Cathles model, and the Borner model rounding out the top three. There was a significant disparity between the rating associated with the top three and the fourth best performing model suggested by Revil and Florsch. However, it should be noted that the Borner model and Revil-Florsch model are based entirely on electrical measurements. The Revil-Cathles model was greatly improved when d(10) was substituted for d(50) for this limited sediment grain-size range.

Automated summary

Five petrophysical models were evaluated for hydraulic permeability prediction in alluvial sediments, with the Kozeny-Carmen model emerging as the top performer. The study also showed significant differences in performance between models relying on physical parameters, electrical parameters, and a combination of both.