Geographically weighted principal component analysis for characterising the spatial heterogeneity and connectivity of soil heavy metals in Kumasi, Ghana

• Main Authors: Eric N. Aidoo, Simon K. Appiah, Gaston E. Awashie, Alexander Boateng, Godfred Darko
• Format: Article
• Language: English
• Published: Elsevier 2021-09-01
• Series: Heliyon
• Subjects: Soil pollution; Heavy metals; Principal component analysis; Spatial heterogeneity; Geographically weighted principal component analysis
• Online Access: http://www.sciencedirect.com/science/article/pii/S2405844021021423

The use of principal component analysis (PCA) for soil heavy metals characterization provides useful information for decision making and policies regarding the potential sources of soil contamination. However, the concentration of heavy metal pollutants is spatially heterogeneous. Accounting for such spatial heterogeneity in soil heavy metal pollutants will improve our understanding with respect to the distribution of the most influential soil heavy metal pollutants. In this study, geographically weighted principal component analysis (GWPCA) was used to describe the spatial heterogeneity and connectivity of soil heavy metals in Kumasi, Ghana. The results from the conventional PCA revealed that three principal components cumulatively accounted for 86% of the total variation in the soil heavy metals in the study area. These components were largely dominated by Fe and Zn. The results from the GWPCA showed that the soil heavy metals are spatially heterogeneous and that the use of PCA disregards this considerable variation. This spatial heterogeneity was confirmed by the spatial maps constructed from the geographically weighted correlations among the variables. After accounting for the spatial heterogeneity, the proportion of variance explained by the three geographically weighted principal components ranged between 85% and 89%. The first three identified GWPC were largely dominated by Fe, Zn and As, respectively. The location of the study area where these variables are dominated provides information for remediation.