Comparative Non-Darcian Modeling of Subsurface Preferential Flow Experimental Observations in a Riparian Buffer

Abstract

HighlightsHigh ecohydrological activity drives macropore prevalence in riparian buffers.An abundance of macropore flow (MF) was confirmed in a field riparian buffer in Kenya.Source-response (SR) and multilayer kinematic diffusive wave (MKDW) MF models are compared.A novel MKDW modeling framework efficiently identifies and predicts preferential flow in riparian buffers.Abstract. The significant ecohydrological activity typical of riparian buffers makes them potential hotspots of macropores, i.e., structured preferential flow pathways, through the soil vadose zone. The prevalence of these preferential pathways can allow transported contaminants to bypass the soil matrix and quickly reach a seasonal shallow water table and the adjacent surface waterbody. This quick transport can ultimately limit the role of riparian buffers for runoff pollution control. Currently, there are no management tools that incorporate macropore flow (MF) when assessing riparian buffer performance. The objective of this study was to experimentally quantify and mathematically simulate macropore flow and arrival time in a riparian buffer under field conditions. Three infiltration experiments were conducted with a grid of 20 time-domain transmission (TDT) dielectric soil moisture sensors along a field riparian buffer transect in Kenya to quantify the presence of macropore flow and to test two non-Darcian soil MF models, including the source-responsive (SR) model and the modified kinematic-dispersive wave (MKDW) model developed in this study, by adding a user-defined multilayer convection scheme and a new hysteresis function between water flux and content. The abundance of MF in the riparian buffer was corroborated experimentally. Modeling results showed that the MKDW model was an efficient (average NSE of 0.937 and 0.721 for calibration and testing, respectively), flexible, and robust method to identify and represent non-linear and non-sequential MF signals at any soil depth and antecedent conditions. The SR model was computationally inexpensive and provided good calibration results (NSE = 0.867) but required piecemeal recalibration of the travel time and maximum water content at each layer and yielded lower performance in testing. The Akaike (AIC) and Bayesian (BIC) information criteria showed that MKDW outperformed SR when accounting for the trade-off between model complexity and efficiency. The results support further research focused on independent characterization of model parameters at the field scale, and the inclusion of MKDW in holistic riparian buffer management and decision-support tools such as VFSmod. Keywords: Kinematic-dispersive wave, Macropore flow, Numerical modeling, Preferential flow, Riparian vadose zone.