Method for identifying outliers of soil heavy metal data.

Artificial errors in the experimental process may lead to some outliers, which reduce data quality and cause erroneous judgment in soil pollution assessment. Based on this, a method for detecting outliers of soil heavy metal data was proposed in this study. The As, Cd, and Pb concentrations of the soil in Beijing, China, were taken as samples to verify the validity of the method. Results showed that there were 8, 34, and 38 outliers for the As, Cd, and Pb concentrations in the Beijing soil, respectively. The result of re-analyzed revealed that 75.0, 76.5, and 92.1% of the As, Cd, and Pb outliers, respectively, were caused by artificial errors. After correcting, the interpolation accuracy for data was improved significantly. The mean relative error (MRE) of the As, Cd, and Pb outliers decreased by 48.0, 44.6, and 54.7%, while the mean square error of these outliers decreased by 34.2, 33.3, and 46.4%, respectively. The MRE values of the nearest neighboring points which were influenced by the outliers decreased by 5.2, 20.6, and 27.6%, while the mean square error of these points decreased by 5.3, 17.3, and 33.2%, respectively. To our knowledge, this is the first study on detecting outliers of soil heavy metal data. The method considers both spatial and numerical outliers, which avoids the limitation of single method, and can effectively improve the data quality of soil heavy metal concentrations with a finite sample size and analysis time.