Multi-Source Data Fusion for High-Resolution Characterization of Specific Storage Heterogeneity: A Coupled Synthetic DNA Tracer and Hydraulic Tomography Approach

Characterizing the spatial heterogeneity of specific storage (Ss) at high resolution remains a major challenge in subsurface hydrology. This study developed a precision-weighted fusion framework that combines hydraulic conductivity (K) estimates obtained from steady-state hydraulic tomography (SSHT) and synthetic DNA or dye tracer tomography (TT), along with diffusivity (D) estimates generated from travel-time hydraulic tomography (TTHT), to estimate Ss through the decoupled relationship Ss = K/D. The framework was evaluated through controlled laboratory sandbox experiments with known layered heterogeneity, comparing homogeneous (same-source) and heterogeneous (cross-source) fusion scenarios. Results demonstrate that the heterogeneous fusion of structurally independent data sources (tracer ray-path tomography and pressure-field hydraulic tomography) is critical for accurate Ss characterization. Their spatially offset uncertainty patterns provide complementary constraints that same-source fusion cannot achieve. Synthetic DNA tracers, whose virtually unlimited unique sequences enable simultaneous multi-array tomographic surveys with minimal cross-interference, proved particularly effective in this framework. In independent drawdown validation, the DNA-based heterogeneous fusion achieved superior predictive accuracy (R² = 0.80, MAE = 0.49 cm) compared to its dye-based counterpart (R² = 0.69, MAE = 0.59 cm) at a comparable global RMSE (~0.88 cm), preserving stronger structural fidelity throughout the fusion chain. Uncertainty analysis revealed that while heterogeneous fusion reduced K-field uncertainty (σ(ln K)) by 34–39%, the propagated Ss uncertainty (σ(ln Ss)) decreased by a more modest 12–15%, identifying the shared D-field uncertainty as the dominant binding constraint. Ultimately, the proposed framework is computationally efficient, modular, and readily extensible to three-dimensional field applications.