Modeling runoff and microbial overland transport with KINEROS2/STWIR model: Accuracy and uncertainty as affected by source of infiltration parameters Academic Article uri icon


  • Summary Infiltration is important to modeling the overland transport of microorganisms in environmental waters. In watershed- and hillslope scale-models, infiltration is commonly described by simple equations relating infiltration rate to soil saturated conductivity and by empirical parameters defining changes in infiltration rate with soil water content. For the microbial transport model KINEROS2/STWIR used in this study, infiltration in unsaturated soil is accounted for by a net capillary drive parameter, G , in the Parlange equation. Scarce experimental data and multiple approaches for estimating parameter G introduce uncertainty, reducing reliability of overland water flow and microbial transport models. Our objectives were to evaluate reliability and robustness of three methods to estimate parameter G and associated accuracy and uncertainty in predicting runoff and fecal coliform (FC) transport. These methods include (i) KINEROS2 fitting to the experimental cumulative runoff data; (ii) estimating solely on soil texture; and (iii) estimating by individual pedotransfer functions (PTFs) and an ensemble of PTFs from basic soil properties. Results show that the most accurate prediction was obtained when the G parameter was fitted to the cumulative runoff. The KINEROS2-recommended parameter slightly overestimated the calibrated value of parameter G and yielded less accurate predictions of runoff, FC concentrations and total FC. The PTFs-estimated parameters systematically deviated from calibrated G values that caused high uncertainty in the KINEROS2/STWIR predictions. Averaging PTF estimates considerably improved model accuracy, reducing the uncertainty of runoff and FC concentration predictions. Overall, ensemble-based PTF estimation of the capillary drive can be efficient for simulations of runoff and bacteria overland transport when a single effective value is used across the study area.

publication date

  • January 1, 2014