KRIGING WITH CUMULATIVE DISTRIBUTION FUNCTION OF ORDER STATISTICS FOR DELINEATION OF HEAVY-METAL CONTAMINATED SOILS

Abstract

Accurate delineation of contaminated soils is essential for risk assessment and remediation. The probability of pollutant concentrations lower than a cutoff value is more important than the best estimate of pollutant concentrations for unsampled locations in delineating contaminated soils. In this study, a new method, kriging with the cumulative distribution function (CDF) of order statistics (CDF kriging), is introduced and compared with indicator kriging. It is used to predict the probability that extractable concentrations of Zn will be less than a cutoff value for soils to be declared hazardous. The 0.1 M HCl-extractable Zn concentrations of topsoil of a paddy field having an area of about 2000 ha located in Taiwan are used. A comparison of the CDF of order statistics and indicator function transformation shows that the variance and the coefficient of variation (CV) of the CDF of order statistics transformed data are smaller than those of the indicator function transformed data. This suggests that the CDF of order statistics transformation possesses less variability than does the indicator function transformation. In addition, based on cross-validation, CDF kriging is found to reduce the mean squared errors of estimations by about 30% and to reduce the mean kriging variances by about 26% compared with indicator kriging. This suggests that kriging with CDF of order statistics, which takes into account the magnitude of the deviation between an observed value z(x) and a cutoff value zk, is a more accurate and reliable method than is indicator kriging for estimating the probability that a pollutant content is less than a cutoff value at an unsampled location