Developing a predictive machine learning model and a kinetic model for the bioremediation of terrestrial diesel spills

Abstract

Terrestrial diesel spills significantly threaten the natural environment and human health, necessitating effective bioremediation strategies for diesel-contaminated soil. This study aims to evaluate the impact of diesel spills on soil water retention capacity and the effectiveness of different bioremediation methods. Four tanks (A-D) were used to compare natural attenuation, bioaugmentation, biostimulation, and a combination of bioaugmentation and biostimulation in enhancing diesel degradation. The findings demonstrated that soil water retention decreases with higher diesel concentrations and increases with more compost. After 21 days, the Diesel Range Organics (DRO) removal efficiencies for Tanks A, B, C, and D were 15.70 %, 23.31 %, 29.65 %, and 49.78 %, respectively. The degradation kinetics primarily followed first-order reaction models, with combined bioaugmentation and biostimulation showing the fastest reaction rate. The projected timelines for complete bioremediation were 44 days for the combined method, 88 days for biostimulation, 116 days for bioaugmentation, and 178 days for natural attenuation. Machine Learning models further supported these findings, with the bilateral Artificial Neural Network outperforming the Linear Regression model (R2 of 0.9990 vs. 0.9713, respectively). This study contributes to understanding the efficiency of different bioremediation techniques and offers practical insights for managing diesel-contaminated soils.

Keywords

• Bioremediation
• Terrestrial diesel spills
• Machine learning
• Reaction kinetics