Preferential flow is known to be highly relevant for runoff and solute transport in hydrological systems, but difficult to implement in catchment scale models. Recent work in the upper Weierbach catchment (6 ha, western Luxembourg) has shown an improvement in modeling solute transport at plot scale through the use of a Richard-based dual-permeability model compared to a single permeability approach. Yet, the best performing parameter sets could not be successfully transferred to the catchment scale. In this study we use a Bayesian approach for estimating the preferential flow parameters directly at catchment scale.
Material and Methods
We use DREAM, an adaptive Markov Chain Monte Carlo (MCMC) algorithm, to calibrate different model implementations against catchment discharge. The best performing model realizations are then validated against hydrograph, surface saturation and soil moisture data.
We successfully applied an MCMC algorithm for calibration on different model implementations, which is unusual at this scale due to computational costs. We find that dual-permeability models do not improve the simulation of the discharge hydrograph compared to the single-permeability models. Similarly, differences in surface saturation patterns and differences in performance for soil moisture time series cannot be observed. Matrix conductivities and macropore conductivities do not differ significantly in dual-permeability models.
The results suggest a scale dependency of vertical preferential flow, i.e. the vertical preferential flow effects observed and relevant at plot scale are not relevant for runoff generation processes at catchment scale (and maybe even disconnected to total catchment discharge). Yet it is unclear, if a less constrained implementation of the dual-permeability model might nevertheless improve the catchment response. Moreover, our conclusions are only valid for water transport and not for solute transport at catchment scale.