|Kim, Keeewok - Us Environmental Protection Agency (EPA)|
|Whelan, Gene - Us Environmental Protection Agency (EPA)|
|Purucker, Thomas - Us Environmental Protection Agency (EPA)|
|Cyterski, Michael - Us Environmental Protection Agency (EPA)|
|Marirosa, Molina - Us Environmental Protection Agency (EPA)|
|Gu, Yin - Us Environmental Protection Agency (EPA)|
|Guber, Andrey - Michigan State University|
|Franklin, Dorcas - Dory|
Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 12/5/2011
Publication Date: 12/9/2011
Citation: Kim, K., Whelan, G., Purucker, T., Cyterski, M., Marirosa, M., Gu, Y., Pachepsky, Y.A., Guber, A., Franklin, D.H. 2011. Parameter estimation and uncertainty in scaling KINEROS2 runoff parameters. [abstract]. Abstract book 1741B-1032.
Technical Abstract: This study investigates the uncertainty associated with simulating runoff at a field scale based on experimental observations on a plot scale. Thirty-six 0.75-m by 2-m plots were delineated within a farm field, and rainfall simulators were used to control precipitation rates and durations on each plot, resulting in monitored runoff. Rainfall-runoff simulations were performed on all plots in four different seasons, resulting in 144 cases (36 x 4). The parameters of the KINEROS2 watershed model were calibrated for each case using an extended application of the inverse code PEST and minimizing the errors between monitored and simulated runoff time series. Parameters associated with the same plot (e.g., same slope) were consistent across the seasons, where appropriate. The approach resulted in distributions of plot-scale calibrated maximum likelihood estimates and relative parameter sensitivity estimates. At the plot scale, the most sensitive parameters were saturated hydraulic conductivity, micro-topographic spacing, canopy cover, and interception depth, while this sensitivity was not exhibited to the same degree in other parameters (e.g., Manning’s roughness coefficient and pore size distribution index). Statistical bootstrapping techniques were also used to analyze KINEROS2 input and output characteristics to estimate input parameters at the field scale. The field-scale values were then compared to the plot-scale calibrated distribution of values to determine the uncertainty associated with upscaling from plot scale to field scale. Therefore, this study provides relative model sensitivity estimates for KINEROS2/PEST model implementation and recommends initial range distributions for hard-to-define runoff parameters.