Modelling rapid flow response of a tile-drained field site using a 2D physically based model: assessment of 'equifinal' model setups

• Published in: Hydrological processes Vol. 24; no. 12; pp. 1595 - 1609
• Main Authors: Klaus, Julian, Zehe, Erwin
• Format: Journal Article
• Language: English
• Published: Chichester, UK John Wiley & Sons, Ltd 15.06.2010
• Subjects: Assessments; Density; equifinality; Freshwater; Hydrology; Mathematical models; modelling; preferential flow; Soils; tile drain; Two dimensional; Uncertainty; vadose zone; Worms
• ISSN: 0885-6087; 1099-1085; 1099-1085
• DOI: 10.1002/hyp.7687

Rapid flow in connected preferential flow paths is crucial for fast transport of water and solutes through soils, especially at tile-drained field sites. In the present study, we propose a spatially explicit approach to represent worm burrows as connected structures of realistic geometry, high conductivity and low retention capacity in a two-dimensional physically based model. We show that this approach allows successful prediction of a tile-drain discharge and preferential flow patterns in soil observed during the irrigation of a tile-drained hillslope in the Weiherbach catchment. However, we found a considerable equifinality in the spatial setup of the model when key parameters such as the area density of worm burrows, the maximum volumetric water flows inside these macropores and the conductivity of the tile drain were varied within the ranges of either our measurements or measurements reported in the literature. In total, we found that 67 out of 432 model runs were acceptable [Nash-Sutcliffe (NS) ≥ 0·75]. Among these, the 13 best yielded a NS coefficient of more than 0·9, which means that more than 90% of the flow variability is explained by the model. Also, the flow volumes were in good accordance and timing errors were less than or equal to 20 min. It is suggested that this uncertainty/equifinality could be reduced when more precise data on initial states of the subsurface and on the width of the control volume draining into a single drainage tube could be made available. However, such data are currently most difficult to assess even at very well investigated sites such as those studied here. We thus suggest that non-uniqueness of the spatial setup of process-based model seems to be an important factor causing predictive uncertainty at many sites where preferential flow dominates system response. Copyright © 2010 John Wiley & Sons, Ltd.