PROBABILITY-DISTRIBUTIONS FOR HYDRAULIC CONDUCTIVITY OF COMPACTED SOIL LINERS

This paper describes an analysis of hydraulic conductivity data collected from 57 landfill liners and covers throughout North America. Statistical characteristics of the data are summarized, and goodness-of-fit analyses are performed on each set to determine a distributional form that can be used to describe spatial variability of hydraulic conductivity. The statistics of hydraulic conductivity are shown to vary widely. The mean varied from 2.9 x 10(-9) to 1 x 10(-5) cm/s, the coefficient of variation varied between 27 and 767, the coefficient of skewness between -0.23 and 7.7, and the coefficient of kurtosis from 1.75 to 58. Goodness-of-fit tests show that the two-parameter log-normal hypothesis is often rejected. The traditional two-parameter log-normal distribution is compared to four alternative, distributions: the three-parameter log-normal, generalized extreme value (GEV), the three-parameter gamma, and inverse Gaussian. The analyses show that the three-parameter log-normal and GEV result in superior fits to the data.