Geostatistical prediction of heavy metal concentrations in stream sediments considering the stream networks

Abstract

Heavy metals in mine wastes can considerably influence surrounding surface waters, soils, and human health. To estimate environmental impact, heavy metal concentrations in stream sediments can be utilized because they are indicators of contamination and change negligibly with time. This study proposes a new Kriging method to predict heavy metal concentrations in stream sediments. The proposed methods compensate for the drawbacks of Kriging based on Euclidean distance because they utilize the stream distance for the prediction by analyzing the stream path and networks using digital elevation models. Moreover, the developed method reduces the exaggeration problem in predicting the concentration of an uncontaminated stream segment by considering the catchment basin area in Kriging. Application of these methods to synthetic and real-world datasets proves that they exhibit improvement in terms of overall error reduction, and they provide reasonable predictions at stream junctions, rather than Kriging based on Euclidean distance.