Integration of DDPM and ILUES for Simultaneous Identification of Contaminant Source Parameters and Non‐Gaussian Channelized Hydraulic Conductivity Field

Identifying highly channelized hydraulic conductivity fields and contaminant source parameters remains a challenging task, primarily due to the non‐Gaussian nature and high dimensionality of the parameter space, as well as the computational burden caused by repeatedly running forward numerical models. This study proposes a novel deep learning parameterization method called AEdiffusion, which combines Diffusion Denoising Probabilistic Model (DDPM) with Variational Autoencoder (VAE) for dimensionality reduction. The method employs a generator‐refiner strategy to generate high‐dimensional aquifer properties from low‐dimensional latent representations. The inversion modeling was performed on a synthetic non‐Gaussian hydraulic conductivity field with line‐source contamination using the Iterative Local Updating Ensemble Smoother (ILUES) algorithm. The results demonstrate that the AEdiffusion‐ILUES framework can accurately identify model parameters. To reduce the computational burden, an AR‐Net‐WL (ARNW) surrogate model was introduced, resulting in an efficient inversion framework (AEdiffusion‐ILUES‐ARNW) with similar prediction accuracy and predictive uncertainty estimation as the AEdiffusion‐ILUES but at a lower computational cost. Plain Language Summary Identifying highly channelized hydraulic conductivity fields and contaminant source parameters is crucial for developing groundwater remediation strategies. However, this remains a challenging task due to the non‐Gaussian nature and high dimensionality of the parameter space, as well as the computational burden caused by repeatedly running numerical models. We propose a novel deep learning‐based inversion framework to identify hydraulic conductivity fields and contaminant sources from sparse and error‐prone observations. Key Points A novel and accurate deep learning parameterization method combining DDPM and VAE is proposed to parameterize non‐Gaussian hydraulic conductivity fields A deep autoregressive neural network is integrated into the inversion framework as a surrogate to alleviate the high computational cost of the forward numerical models The integrated approach is assessed with inverse problems for the identification of a non‐Gaussian conductivity and line contaminant source parameters