Uncertainty Analysis of a Dual-Continuum Model Used to Simulate Subsurface Hillslope Runoff Involving Oxygen-18 as Natural Tracer
A one-dimensional dual-continuum model (also known as dual-permeability model) was used to simulate the lateral component of subsurface runoff and variations in the natural 18O content in hillslope discharge. Model predictions were analyzed using the GLUE generalized likelihood uncertainty estimation procedure. Model sensitivity was evaluated by varying two separate triplets of parameters. The first triplet consisted of key parameters determining the preferential flow regime, i.e., the volumetric proportion of the preferential flow domain, a first-order transfer coefficient characterizing soil water exchange between the two flow domains of the dual-continuum system, and the saturated hydraulic conductivity of the preferential flow domain. The second triplet involved parameters controlling exclusively the soil hydraulic properties of the preferential flow domain, i.e., its retention curve and hydraulic conductivity function. Results of the analysis suggest high sensitivity to all parameters of the first triplet, and large differences in sensitivity to the parameters of the second triplet. The sensitivity analysis also confirmed a significant improvement in the identifiability of preferential flow parameters when 18O content was added to the objective function.