2012 NGWA Ground Water Summit: Innovate and Integrate

Guidance for Efficient Implementation of the Surface-Subsurface Coupling Approach in MODHMS

Monday, May 7, 2012: 2:30 p.m.
Royal Ballroom E (Hyatt Regency Orange County)
Matthew J. Knowling, Flinders University, National Centre for Groundwater Research & Training;
Jessica E. Liggett, National Centre for Groundwater Research and Training, Flinders University;
Adrian D. Werner, National Centre for Groundwater Research and Training, Flinders University;
Craig T. Simmons, Flinders University, National Centre for Groundwater Research and Training;

Over the last two decades, considerable effort has focused on the development of physically based numerical models that are capable of simulating catchment-scale hydrological processes through fully integrating (i.e. coupling) the surface and subsurface flow domains. These domains are often represented using the surface conductance (SC) coupling approach. However, sufficient guidance on SC parameterisation for block-centred finite-difference codes is currently limited. Common practice is to express the SC coefficient as the quotient of the vertical saturated hydraulic conductivity and the half-cell thickness of the uppermost layer. This research aims to provide guidance on the implementation of the SC approach for block-centred codes by evaluating the application of this approach in the popular hydrology catchment model MODHMS by simulating one-dimensional infiltration experiments under Hortonian conditions. Under these conditions, we show that defining the SC coefficient based on the half-cell thickness of the uppermost subsurface cell inhibits accurate simulation of both infiltration rates (q) and the time taken to initiate runoff (tro) where near-surface vertical grid discretisation is very fine (< 1cm). As sufficient vertical grid refinement allows accurate results - albeit greatly increasing model run times - we show that in analysing the trade-off between model accuracy and run times, the addition of a single thin layer at the surface significantly improves simulation accuracy with only minor increase in computational expense. Increasing the SC coefficient independently of the grid allows accurate simulation of q, but not tro, identifying the need for a thin surface layer. This research aims to aid future industry modelling projects through offering guidance on the implementation of the SC approach in block-centred codes and by highlighting the need for systematic testing of SC parameters (that are otherwise calibrated).