A Comparison of Three Kriging Methods Using Auxiliary Variables in Heavy‐Metal Contaminated Soils

Abstract

Characterization of the spatial distribution of pollutants in contaminated soils is important for risk assessment and soil reclamation. In this study, three kriging methods using auxiliary variables, cokriging, kriging combined with regression, and kriging combined with Q‐mode factor analysis, were used for interpolation of heavy metal concentrations in a contaminated site. The three interpolation methods were evaluated for whether or not they could make better use of an auxiliary variable for estimation of the spatial distribution of heavy metals. A heavy metal contaminated site about 10 ha in area, situated in Taoyuan, Taiwan, was studied. The results demonstrate that better use of the auxiliary variable in interpolations of the target variable was made using kriging combined with regression compared to using cokriging or kriging combined with Q‐mode factor analysis. The results also show that the spatial distribution patterns of the target variables estimated using kriging combined with Q‐mode factor analysis were more similar to those estimated using kriging combined with regression than were those estimated using cokriging. In addition, kriging combined with regression and kriging combined with Q‐mode factor analysis could avoid negative estimates, which occur in cokriging. Moreover, both of them were more robust than cokriging. Simultaneous estimation of spatial distributions of several target variables using an auxiliary variable was demonstrated for the kriging combined with Q‐mode factor analysis procedure. However, for only one target variable, kriging combined with regression was simpler and less cumbersome than cokriging.

Utility of soil linear alkylbenzenes to assess regional anthropogenic influences with special reference to atmospheric transport

“…The Ordinary Kriging interpolation method was used to characterize the spatial distribution of LABs in soils for assessing sewage contamination, which was carried out with ArcGIS 9.3 (ESRI, Redlands, California, USA) (Juang and Lee, 1998;Carlon et al, 2001;Hu et al, 2006Hu et al, , 2009.…”

Assessing spatial distribution and joint uncertainty of TPH-fractions: Indicator kriging and sequential indicator simulation

“…Juang and Lee (1998b) found that Pb, Cd, and Zn had very high correlation coefficients. They found that incorporating Pb and Cd concentrations as auxillary variables improved the interpolation of Zn.…”

“…They found that incorporating Pb and Cd concentrations as auxillary variables improved the interpolation of Zn. Incorporating auxiliary variables into interpolations is the most advantageous when the primary or target variable is undersampled and there is a strong correlation between the variables (Juang and Lee 1998b;Goovaerts 1997;Deutsch and Journel 1998). This option was investigated to determine if utilizing the different fractions as co-variables would improve interpolation.…”

Upslope length improves spatial estimation of soil organic carbon content

“…In addition to semivariograms of individual variables, a cross-semivariogram is also used in traditional ordinary cokriging to quantify the cross-spatial correlation between the primary and secondary variable Z j and V k , respectively, shown as follows (Juang and Lee 1998):…”