HR: 0830h
AN: H51B-01    [Abstracts]
TI: Split-Sample Tests for Vegetation-Type Dependence of Parameters of the Land Dynamics (LaD) Model
AU: * Xia, Y
EM: youlong.xia@noaa.gov
AF: Atmospheric and Oceanic Science Program of Princeton University and NOAA GFDL, Forrestal Campus, Princeton, NJ 08542 United States
AU: Milly, P
EM: Chris.Milly@noaa.gov
AF: US Geological Survey, Forrestal Campus, Princeton, NJ 08542 United States
AU: Dunne, K
EM: Krista.A.Dunne@noaa.gov
AF: US Geological Survey, Forrestal Campus, Princeton, NJ 08542 United States
AB: Poorly constrained vegetation parameters (a stomatal resistance parameter and a plant-root parameter) of the Land Dynamics (LaD) model for global land water and energy balance are estimated by model calibration. Very fast simulated annealing is used to minimize selected measures of the deviation between modeled and measured streamflow over a set of streamgauges (the calibration gauges). Estimated parameters are then transferred for use in other gauged basins that were excluded from the optimization (evaluation gauges). This calibration-evaluation process is performed in two ways: with parameters assumed to be globally constant, and with parameters assumed to depend on vegetation type. Performance of this parameter-transfer procedure is evaluated by computing, for the evaluation gauges, the value of the error measure that had been used in the calibration. Does stratification of the optimization by vegetation type yield predictive power? Comparison of the calibration and evaluation error measures for the global and vegetation-stratified approaches may help to answer that question.
DE: 1836 Hydrologic budget (1655)
DE: 1860 Runoff and streamflow
DE: 1869 Stochastic processes
SC: Hydrology [H]
MN: 2005 Joint Assembly