A Multi-Scale Scheme for Simulating Subsurface Dynamics in Land Surface Models using HydroBlocks

Over recent years, considerable advances have been made in Land Surface Models (LSM) to enhance the representation of small-scale heterogeneity while maintaining reasonable computational efficiency. Such is the case of HydroBlocks, which employs fine-scale clustering to define Hydrologic Response Units (HRUs) or tiles as its core modeling element. These innovations have facilitated a better representation of water and energy balances over large-scale domains by capturing local dynamics and their signature over continental processes.

While the benefits of these advances are substantial, there is still a growing need to understand surface and subsurface dynamics under these novel approaches. In our current study, we propose a novel multi-scale scheme designed to capture subsurface interactions within an LSM. This approach builds upon the abstraction introduced in HydroBlocks and addresses the lateral contributions of soil columns for local, intermediate, and regional subsurface flows. Importantly, this is achieved without compromising the computational efficiency of the already efficient HydroBlocks model. Our approach enables us to capture complex fine-scale interactions between surface and subsurface hydrological processes over continental extents, potentially providing insights that traditional models cannot achieve.

To implement this, we decompose the domain into regional units and compute the subsurface flux exchange, efficiently updating the one-dimensional vertical solution of Richard’s equation within the LSM. A convergence analysis is performed by comparing the efficiency of our framework to that of the quasi-fully distributed solution. The methodology has been tested within a 1.0°x1.0° domain in the United States to evaluate its performance. The inclusion of intermediate and regional groundwater representation led to significant shifts in soil moisture redistribution and streamflow patterns. Notably, we uncover regional water flow patterns from ridges to valleys, often underrepresented in the traditional model. Additionally, we explore the impact of spatial scale on water redistribution, offering profound insights into the uncertainties associated with groundwater structure and its influence on surface fluxes.

Our findings reveal that the multi-scale scheme converges towards a quasi-fully distributed solution for the LSM HydroBlocks emphasizing the efficacy of our method in achieving a comprehensive representation of subsurface dynamics while maintaining computational cost.