TY - JOUR
T1 - Use of ordinary cokriging with magnetic susceptibility for mapping lead concentrations in soils of an urban contaminated site
AU - Golden, Nessa
AU - Zhang, Chaosheng
AU - Potito, Aaron
AU - Gibson, Paul J.
AU - Bargary, Norma
AU - Morrison, Liam
N1 - Publisher Copyright:
© 2019, Springer-Verlag GmbH Germany, part of Springer Nature.
PY - 2020/3/1
Y1 - 2020/3/1
N2 - Purpose: Lead contamination is a prevalent issue affecting cities worldwide. Traditional fieldwork and laboratory analysis techniques can be time-consuming and costly. The purpose of this study was to evaluate the performance of ordinary cokriging (CK) when volume magnetic susceptibility (κ) is used as a co-variable for spatial interpolation of Pb in contaminated urban soils. Materials and methods: The study was conducted in contaminated urban soils of a former unregulated landfill site. A total of 76 surface samples (0–10 cm) were collected using a systematic sampling grid separated by 20-m intervals. Magnetic susceptibility measurements were taken at a higher density of 10-m intervals with 288 measurements. Thus, it was used as an auxiliary variable to predict Pb concentrations by the CK procedure with an aim to improve spatial interpolation of Pb. To determine the effectiveness of CK over the ordinary kriging (OK) procedure, the spatial density of samples was reduced prior to interpolation. A total of ~ 15%, 25%, 35%, and 50% of the Pb samples were randomly selected and reserved for validation. Omnidirectional semivariograms and covariograms were fitted using log-transformed data prior to interpolation. Results and discussion: Measurements of κ shared a significant relationship with Pb concentrations by the Spearman’s Rho correlation analysis (rs = 0.676, p < 0.01). The effectiveness of the CK procedure over OK was determined using validation datasets. Statistically, the results showed that lnPb when its auxiliary relations with lnκ were used in CK had overall lower “root mean square error” (RMSE) and predicted lnPb values from the CK procedure had a higher r2 value with measured lnPb than OK. A model produced by the CK procedure with a reduced spatial density of 49 Pb points provided the more accurate map with a RMSE of 0.550 and an r2 value of 0.730, p < 0.01 level. Conclusions: This technique can potentially reduce fieldwork and soil analysis costs considerably. Measurements of Pb and κ must share a substantial level of spatial continuity to implement CK effectively. Where applicable, it can be used in the site-specific evaluation of hazard posed by Pb exposure to ecosystems, human health or water bodies in urban green spaces, roadside soils, allotments or brownfield sites.
AB - Purpose: Lead contamination is a prevalent issue affecting cities worldwide. Traditional fieldwork and laboratory analysis techniques can be time-consuming and costly. The purpose of this study was to evaluate the performance of ordinary cokriging (CK) when volume magnetic susceptibility (κ) is used as a co-variable for spatial interpolation of Pb in contaminated urban soils. Materials and methods: The study was conducted in contaminated urban soils of a former unregulated landfill site. A total of 76 surface samples (0–10 cm) were collected using a systematic sampling grid separated by 20-m intervals. Magnetic susceptibility measurements were taken at a higher density of 10-m intervals with 288 measurements. Thus, it was used as an auxiliary variable to predict Pb concentrations by the CK procedure with an aim to improve spatial interpolation of Pb. To determine the effectiveness of CK over the ordinary kriging (OK) procedure, the spatial density of samples was reduced prior to interpolation. A total of ~ 15%, 25%, 35%, and 50% of the Pb samples were randomly selected and reserved for validation. Omnidirectional semivariograms and covariograms were fitted using log-transformed data prior to interpolation. Results and discussion: Measurements of κ shared a significant relationship with Pb concentrations by the Spearman’s Rho correlation analysis (rs = 0.676, p < 0.01). The effectiveness of the CK procedure over OK was determined using validation datasets. Statistically, the results showed that lnPb when its auxiliary relations with lnκ were used in CK had overall lower “root mean square error” (RMSE) and predicted lnPb values from the CK procedure had a higher r2 value with measured lnPb than OK. A model produced by the CK procedure with a reduced spatial density of 49 Pb points provided the more accurate map with a RMSE of 0.550 and an r2 value of 0.730, p < 0.01 level. Conclusions: This technique can potentially reduce fieldwork and soil analysis costs considerably. Measurements of Pb and κ must share a substantial level of spatial continuity to implement CK effectively. Where applicable, it can be used in the site-specific evaluation of hazard posed by Pb exposure to ecosystems, human health or water bodies in urban green spaces, roadside soils, allotments or brownfield sites.
KW - Co-regionalisation modelling
KW - Soil-contamination spatial modelling
KW - Topsoil contamination
KW - Volume magnetic susceptibility
UR - http://www.scopus.com/inward/record.url?scp=85076882756&partnerID=8YFLogxK
U2 - 10.1007/s11368-019-02537-7
DO - 10.1007/s11368-019-02537-7
M3 - Article
AN - SCOPUS:85076882756
SN - 1439-0108
VL - 20
SP - 1357
EP - 1370
JO - Journal of Soils and Sediments
JF - Journal of Soils and Sediments
IS - 3
ER -