Abstract: Polycyclic aromatic hydrocarbons (PAHs) are among the primary organic contaminants in groundwater, and numerical modeling of PAH transport is a crucial tool for efficient groundwater pollution remediation. Under actual groundwater contamination conditions, the co-transport mechanism of contaminants and colloids is often neglected due to the difficulty in accurately characterizing the types and distribution of colloids within the aquifer matrix. This omission introduces structural errors into the model, resulting in significant biases in predictive outcomes. This study focuses on fluoranthene and phenanthrene, addressing the neglected co-transport mechanisms of PAHs and colloids by employing Gaussian Process Regression (GPR) to correct structural model errors. A PAH transport model coupling data-driven and physical mechanisms was developed. Through saturated sand column experiments on PAH transport, the predictive performance of models using uncoupled and coupled data-driven approaches was compared and analyzed. The results indicate that groundwater PAH transport models neglecting the co-transport mechanisms of PAHs and colloids exhibit significant structural errors. Direct parameter calibration fails to compensate for the omitted co-transport mechanisms, leading to substantial prediction biases. The application of the GPR model effectively compensates for the PAH-colloid co-transport mechanisms and corrects structural errors in the groundwater model. During the validation period, the 95% confidence interval coverage of observed concentrations improved by 56.84% for fluoranthene and 19.04% for phenanthrene. NSE increased by 40.09% and 21.73%, while RMSE dropped by 33.10% and 55.38%, MAE by 32.00% and 46.34%, respectively. The predictive performance of the groundwater PAHs transport model improved significantly. The coupled data-driven and physics-based approach proposed in this study provides a viable framework for accurate simulation of PAHs transport in site groundwater, contributing to precise and efficient groundwater contamination remediation.