The U.S. Environmental Protection Agency’s (EPA) Superfund—the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) database—has collected and built an open-source database based on nearly 2000 US soil remediation cases since 1980, providing detailed information and references for researchers worldwide to carry out remediation work. However, the cases were relatively independent to each other, so the whole database lacks systematicness and instructiveness to some extent. In this study, the basic features of all 144 soil remediation projects in four major oil-producing states (California, Texas, Oklahoma and Alaska) were extracted from the CERCLA database and the correlations among the pollutant species, pollutant site characteristics and selection of remediation methods were analyzed using traditional and machine learning techniques. The Decision Tree Classifier was selected as the machine learning model. The results showed that the growth of new contaminated sites has slowed down in recent years; physical remediation was the most commonly used method, and the probability of its application is more than 80%. The presence of benzene, toluene, ethylbenzene and xylene (BTEX) substances and the geographical location of the site were the two most influential factors in the choice of remediation method for a specific site; the maximum weights of these two features reaches 0.304 and 0.288.

CO₂-enhanced oil recovery (CO₂-EOR) has gained prominence as an effective oil displacement method with low carbon emissions, yet its microscopic mechanisms remain incompletely understood. This study introduces a novel high-pressure microfluidic visualization system capable of operating at 0.1–10 MPa without confining pressure and featuring stratified porous media with a 63 μm minimum throat size to provide unprecedented insights into CO₂ and CO₂-foam EOR processes at the microscale. Through quantitative image analysis and advanced machine learning modeling, we reveal that increasing the CO₂ injection pressure nonlinearly reduces residual oil saturation, achieving near-complete miscibility at 6 MPa with only 2% residual oil—a finding that challenges conventional thresholds for miscibility in heterogeneous systems. Our work uniquely demonstrates that CO₂-foam flooding not only mobilizes capillary-trapped oil films but also dynamically alters interfacial tension and the pore-scale fluid distribution, a phenomenon previously underexplored. Support Vector Regression (R² = 0.71) further uncovers a nonlinear relationship between the surfactant concentration and residual oil saturation, offering a data-driven framework for parameter optimization. These results advance our fundamental understanding by bridging microscale dynamics with field-applicable insights, while the integration of machine learning with microfluidics represents a methodological leap for EOR research.