Assessing Sensitivity of Surface Water-Groundwater Exchanges to Model Initialization

Abstract:

Integrated groundwater-land surface models characterize catchment responses by coupling the subsurface flow with land surface processes. One of the uncertainties of these models is the specification of the initial condition and its influence on progressive simulations. The key challenge is that initializations often require spatially distributed information on model states, groundwater levels and soil moisture, which is often unavailable. One approach to reduce uncertainty in model initialization is to run the model recursively using a single or multiple years of forcing data until the system equilibrates with respect to state and diagnostic variables (spin-up process). Here, we performed a multiple criteria analysis using different spin-up measures and thresholds to define equilibrium state in an integrated groundwater-land surface model, ParFlow.CLM, and evaluated model performance as a function of various spin-up criteria. The ParFlow.CLM model was developed over a 208 km² catchment in Denmark with a temperate climate. Results illustrated that equilibrium of subsurface storages required 20 years of spin-up simulations which was computationally intensive (26 days of computation using 32 processors). In the next stage, we assessed sensitivity of surface water-groundwater exchanges to initializations obtained from multiple equilibrium states through recursive simulations. Our results indicated sensitivity in modeled recharge and stream flow to the different equilibrium initializations, but reduced sensitivity in modeled energy fluxes to the same changes. Due to computational cost of spin-up for model initialization, new methodologies are needed for reducing computation time of spin-up approaches.