Search

[Display omitted] • A data-based model is established to describe cross-scale iron bioreduction rates. • Sensitive analysis is used to provide insights to the data-based 'black box' model. • The model is validated using new data from the experiments in this study. Bioreduction rates of iron (Fe) oxides are a key variable for predicting the fate and transport of Fe and contaminants in subsurface environments. Such rates, however, vary significantly at different spatial scales, and change in orders of magnitude even within the same scale. This study first collected and consistently processed the data of Fe oxide bioreduction rates from literature, which were then used to train a machine learning (ML) model that resulted in a well-fitted relationship between the cross-scale bioreduction rates and common influencing factors including electron donor concentrations, electron numbers, Fe concentrations, cell numbers, and reaction Gibbs free energy. Sensitivity analysis was performed to provide insights into the relative role of the influencing factors and how they affect the rates. The result indicated that their effects were generally consistent with saturation type models. New experiments of iron oxide bioreduction were performed to validate the ML-based cross-scale model. The result showed that ML-based model well predicted the experimental results, indicating the effectiveness of the model. The result has a strong implication for developing the models of cross-scale reaction rates of iron oxide bioreduction in environmental systems and for predicting the fate and transport of iron and contaminants. [ABSTRACT FROM AUTHOR]