9 th LBA-ECO Science Team Meeting Assessing the Influence of Observational Data Error on SiB 2 Model Parameter Uncertainty Luis A. Bastidas 1, E. Rosero 1, S. Pande 1, W.J. Shuttleworth 2 1 Civil and Environmental Engineering and Utah Water Research Laboratory, Logan, Utah 2 Hydrology and Water Resources, SAHRA – NSF Science and Technology Center University of Arizona, Tucson, Arizona

usu.edu 2/16 Modeling Modified from Refsgaard, 2001 Model Construction Construction Reality ConceptualModel Model Model Code Analysis Simulation Programming Model Calibration Model Validation Model Qualification Code Verification

usu.edu 3/16 Components of a Model I Inputs X State Variables Model Structure X t = F ( X t-1, , I t-1 ) O t = G ( X t, , I t ) Model Structure O Outputs  Parameters XoXo Initial States

usu.edu 4/16 Pareto Optimality Parameter  1 Parameter  2 Criterion f 1 Criterion f 2 α β γ γ δ δ α β Parameter SpaceCriterion Space f1f1 f2f2

usu.edu 5/16 Sensitivity Analysis. SiB Santarem Km 83 Pre-logging Post-logging MOGSA Algorithm Bastidas et al., JGR,1999 Multi Objective Generalized Sensitivity Analysis The further away from the center the more sensitive the parameter

usu.edu 6/16 Pareto Set of Parameters

usu.edu 7/16 Observed vs Computed. Pre-logging

usu.edu 8/16 Dry Typical Periods

usu.edu 9/16 Wet Typical Periods

usu.edu 10/16 Daily Average E HCO 2

usu.edu 11/16 Performance Measures E HCO 2 E H R RMSE NSE BIAS

usu.edu 12/16 Objectives for Radiation and Precipitation Errors Heteroscedastic Error Added

usu.edu 13/16 Parameter Distributions

usu.edu 14/16 Parameter Distributions 01

usu.edu 15/16 Daily Average E HCO

usu.edu 16/16 Beware …. “A fool with a tool is still a fool”