Accurately inferring realistic subsurface structures poses a considerable challenge due to the impact of morphology on flow and transport behaviors. Traditional subsurface characterization relies on two primary types of data: hard data, derived from direct subsurface measurements, and soft data, encompassing remotely sensed geophysical information and its interpretation. Existing deep‐learning‐based methodologies predominantly focus on the transition from multiple observations to subsurface structures. However, implicit non‐linear correlations among diverse data sources often remain underutilized, leading to potential bias and errors. In this study, we introduce a multiple‐condition fusion network (MCF‐Net) to characterize subsurface structures based on both hard and soft data. To harness the full potential of multiple‐source subsurface observations, two distinct neural networks extract implicit features from hard and soft data. The integration of these features is achieved through multiple‐condition fusion blocks, designed to capture representative characteristics. These blocks are also adept at reconstructing heterogeneous structures and facilitating hydrological parameterization. MCF‐Net exhibits accuracy in estimating subsurface structures across various types of subsurface observations. Experimental results underscore the utility and superiority of MCF‐Net in applications of hydrogeological modeling.

Increased availability and quality of surface and subsurface observations provide the chance to improve the perception of hydrological phenomena. An important factor that hinders our understanding of hydrological structures is the characterization of heterogeneous patterns with continuous attributes inside the structures (e.g., porosity, permeability, fluid saturation, etc.). Unlike categorical attributes, continuous attributes convey more realistic characteristics but require more computational resources to characterize such complex earth systems. In this work, we propose a novel deep learning approach for the characterization of complex hydrological realism with continuous attributes based on generative adversarial networks (GANs) and self‐attention mechanism, named SA‐RelayGANs. To address the complexity of heterogeneous hydrological structures, we divide the modeling process into two stages: facies construction stage and property reconstruction stage. In the first stage, we employ an improved GAN with self‐attention mechanism to construct the heterogeneous structures while adhering to hard conditioning constraints. In the second stage, we utilize another GAN with an attribute enhancement term to reconstruct realizations based on the constructed structures and observations. SA‐RelayGANs can successfully predict the statistical distributions of heterogeneous structures with continuous attributes by using limited observations. This study highlights the effectiveness of using GANs to characterize the heterogeneous patterns of hydrological realism and the application over large geoscience fields.