High-resolution groundwater modelling in complex terrain: Coupling PCR-GLOBWB and MODFLOW at 1 km resolution for the Piave and Tagliamento basins in Northern Italy
Summary
Hydrological models are increasingly advancing towards hyper-resolution to capture fine-scale processes that govern groundwater–surface water interactions. Yet, it remains unclear whether increasing model resolution or incorporating explicit geological detail provides greater benefits, particularly in regions where lithological contrasts and complex topography shape groundwater flow. This thesis addresses this question through the development of a high-resolution coupled hydrological model for the Piave and Tagliamento basins in northern Italy, situated within the geologically complex Alpine foreland.
The study employs a coupled PCR-GLOBWB–MODFLOW modelling framework at 1 km spatial resolution to assess how explicit geological characterisation affects simulated hydrological behaviour. Three configurations are compared: a five-arcminute (10 km) reference model using standardised global data; a thirty-arcsecond (1 km) downscaled model without geological detail; and a thirty-arcsecond model incorporating a detailed subsurface constructed with ArchPy, a stochastic 3D geological modelling library. The geological model integrates more than 16,000 boreholes, stratigraphic maps, and facies simulations, from which hydraulic conductivity, porosity, and specific yield fields were derived for MODFLOW input.
Model validation focused on river discharge at Alpine and foreland stations, complemented by spatial diagnostics of groundwater recharge and head distributions. Refining spatial resolution from five arcminutes to thirty arcseconds markedly improved discharge simulation by reducing peak inflation, enhancing timing, and yielding positive Kling–Gupta Efficiency (KGE) skill scores across all main validation stations. Explicit geology introduced additional but more localised improvements: it refined hydrograph transitions and produced geologically coherent patterns in recharge and groundwater heads, especially along lithological boundaries and spring-line zones.
Recharge maps revealed that geology mainly redistributes recharge locally, while basin-integrated magnitudes remain stable. Groundwater head differences between the geological and non-geological runs reached several hundred metres in carbonate uplands, indicating sharper hydraulic gradients consistent with lithological contrasts. These effects enhanced the physical realism of subsurface processes, even when discharge metrics showed limited sensitivity.
The findings demonstrate that spatial resolution is the dominant driver of improved hydrological performance, while geological detail primarily enhances local realism in subsurface gradients and groundwater–surface water exchange. At the current one-kilometre resolution, the benefits of refinement outweigh those of geological complexity. However, finer grids combined with recalibrated groundwater response times are expected to unlock the full potential of explicit geological representation.
This study highlights both the promise and the challenges of integrating geological realism into large-scale hydrological models. It shows that hyper-resolution modelling can successfully reproduce key hydrological processes in Alpine forelands and that detailed subsurface characterisation contributes to a more process-consistent depiction of groundwater dynamics. The results provide valuable insights for future hyper-resolution frameworks aiming to balance model realism, computational feasibility, and data availability at continental or global scales.