Prediction of optimal bioremediation conditions for petroleum hydrocarbon contaminated soil by automated machine learning-based analysis

Petroleum hydrocarbons (PH) contaminated soil has become a long-standing problem. By employing microorganisms, plants, or microbial enzymes, bioremediation has the potential to detoxify and remove contaminants from soil and water environments. Given the complexity of environmental variables, it is significant to build a robust model for predicting optimal remediation conditions. In this study, H2O automated machine learning (AutoML) successfully predicted the effects of 8 variables on the remediation efficiency of PH-contaminated soil without human intervention. It was suggested that composting bioremediation and biochar immobilization prompted PH degradation, while bioaugmentation and phytoremediation exhibited lower removal efficiency. Moreover, PH concentration and cultivation period, rather than PH type and soil physicochemical properties, significantly influenced PH bioremediation. Specifically, remediation efficiency enhances when the PH concentration is below 5000 mg/kg and then decreases as it keeps ascending at a gradual rate. Furthermore, optimal cultivation period in the range of 20-40 days is conducive to PH biodegradation. This work successfully demonstrated that AutoML could be a valuable tool for predicting optimal remediation conditions of PH-contaminated soil.