Geostatistics applied to mine waste characterization at Leadville, Colorado, USA

Abstract

ABSTRACT This paper will describe the process of identifying lead contaminated soil areas exceeding a certain threshold at a historical mining district using geostatistical tools. For this purpose, lead concentration samples coming from Leadville County, Colorado were chosen as the data sets to be examined. In order to check integrity of data files, posting of the data values at sample locations is performed on each of the four data sets available. Statistical analysis is used for identifying possible different populations within the study area Based on the statistical results, the data sets were subdivided into four different populations as they relate to residential domains. More detail geostatistical analysis of data set representing the City of Leadville domain (CL) is performed Variogram analysis is used to characterize spatial variability on the data sets within CL limits. Generation of variograms for such data sets and identification of their parameters such as sill, nugget, range, and anisotropy ratios is performed. Once these parameters are identified, variogram modeling is accomplished. A significant spatial correlation is found within CL boundaries and two geostatistical estimation techniques named ordinary kriging and indicator kriging are applied to determine level of contamination at unsampled locations exceeding 3500 PPM threshold. Such techniques are tested by cross validation in order to check possible overestimation bias over 3500 PPM threshold. Indicator Kriging was found to be the best estimator which provided the minimum overestimation bias of areas of contamination exceeding 3500 PPM threshold compared with ordinary kriging.