Acidophiles enable pollution forensics in soil environments

Abstract

The bacterial community structure of polluted soil differentiates according to toxic pollutants. In this study, the acid pollution source was predicted by using characteristic bacterial community structures which were exposed to HCl, HF, HNO3, and H2SO4. In a soil column, after a simulated acid leak, Bacillus, Citrobacter, Rhodococcus, and Ralstonia sp. were found as acid-resistant bacteria and their relative abundance varied depending on the acid. The complex bacterial community was analyzed by using terminal restriction fragment length polymorphism (T-RFLP) of 16S rRNA gene. Using machine learning models including support vector machine (SVM), K-nearest neighbors (KNN), random forest (RF), and artificial neural network (ANN), the prediction accuracy for acidic pollutants was 72%, 72%, 76%, and 88%, respectively. With data augmentation based on T-RFLP, the accuracy of the ANN model for predicting acidic pollutants improved to 98%. This research provides valuable insights into the potential use of bacterial community structures and machine learning models for the rapid and accurate identification of acid pollution sources in soil.