The relevance of preferential flow in catchment scale simulations: Calibrating a 3D dual‐permeability model using DREAM

Abstract

The occurrence of preferential flow in the subsurface has often been shown in field experiments. However, preferential flow is rarely included in models simulating the hydrological response at the catchment scale. If it is considered, preferential flow parameters are typically determined at the plot scale and then transferred to largerscale simulations. Here, we successfully used the optimization algorithm DiffeRential Evolution Adaptive Metropolis (DREAM) to calibrate a 3D physics-based dualpermeability model directly at the catchment scale. In order to keep computational costs of the optimization routine at a reasonable level, we limited the number of parameters to be calibrated to the ones that had been shown before to be most influential for the simulation of discharge. We also calibrated parameters of the matrix domain and the macropore domain with a fixed parameter ratio between soil layers instead of calibrating every layer separately. These ratios reflected observed depth profiles of soil hydraulic properties at our study site. The dual-permeability parameter sets identified during calibration were able to simulate observed discharge time series satisfactorily but did not outperform a calibrated single-domain reference model scenario.Saturated hydraulic conductivities of the macropore domain were calibrated such that they became very similar to matrix saturated hydraulic conductivities, thereby effectively removing the effect of macropores. This suggests that the incorporation of vertical preferential flow as represented by the dual-permeability approach was not relevant for reproducing the hydrometric response reasonably well in the studied catchment. We also tested the scale-invariance of the calibrated dual-permeability parameter sets by using the parameter sets performing best at catchment scale to simulate plot-scale bromide depth profiles obtained from tracer irrigation experiments.This parameter transfer proved to be not successful, indicating that soil hydraulic parameters are scale-variant, independent of the direction of parameter transfer. K E Y W O R D S catchment scale, discharge simulation, dual-permeability model, HydroGeoSphere, model calibration, optimization algorithm, preferential flow

Critical Zone Storage Controls on the Water Ages of Ecohydrological Outputs

“…In addition, accounting for local preferential flow may not always improve catchment-scale simulations (Glaser et al, 2019;Hopp et al, 2020). The present simulations further suggest that plants access relatively old water pools (from~6 months on the hillslopes to >2 years in the valley bottom; Figure S4) with an hysteretic relationship to storage not found in streamflow ages.…”

“…Yet combining these simplified formulations with spatially distributed flow paths has allowed capturing water flux and tracer concentrations measured at multiple locations in several critical zone compartments of high‐latitude catchments (Kuppel et al, 2018a; Smith et al, 2019). In addition, accounting for local preferential flow may not always improve catchment‐scale simulations (Glaser et al, 2019; Hopp et al, 2020). The present simulations further suggest that plants access relatively old water pools (from ~6 months on the hillslopes to >2 years in the valley bottom; Figure S4) with an hysteretic relationship to storage not found in streamflow ages.…”

Uncertainty and sensitivity analysis of a travel‐time distribution‐based stream water electrical conductivity model

“…DREAM estimates posterior probability distributions of model parameter values ( θ ) and model predictions using Bayes theorem and undertakes an efficient global search of the parameter space. DREAM has been widely applied in water quality (Benettin et al, 2017; Benettin et al, 2020; Benettin & Bertuzzo, 2018) and hydrological (Hopp et al, 2020) model calibration, treatment of input uncertainty in hydrological modelling watersheds (Vrugt et al, 2008) and other model UA (Liu et al, 2017; Zheng & Han, 2016). We used log likelihood as the fit measure in DREAM assuming an independent identically distributed normal error distribution.…”

An agent-based model that simulates the spatio-temporal dynamics of sources and transfer mechanisms contributing faecal indicator organisms to streams. Part 1: Background and model description

“…More generally, there is a need to consider partial mixing in tracer-aided models (Cain et al, 2019;Knighton et al, 2017) to accommodate possible ecohydrological separation of water in the subsurfacethat is its partitioning between water that is tightly-bound to the soil matrix in small pores and mobile water which may predominantly contribute to groundwater recharge and streamflow (Brooks et al, 2010;Goldsmith et al, 2012;Sprenger et al, 2018). To the authors' knowledge, twopore conceptualisations of the subsurface in physically-based models are only just emerging in catchment studies (Hopp et al, 2020), with efforts largely having been limited to the plot scale (Jackish and Zehe, 2018; Sprenger et al, 2018;Stumpp and Maloszewski, 2010;Vogel et al, 2010).…”