PARENT SESSION
Poster Session #12: Modeling.
Tuesday, August 7, 2001. Presentation from 10:30 AM to 12:00 PM. Exhibition Hall
2
Parameter sensitivity of a land surface model using Monte Carlo analysis and eddy-covariance measurements at the WLEF-TV tower site, Wisconsin.
PRIHODKO, LARA1, DENNING, SCOTT1, HANAN, NIALL1, DAVIS, KENNETH2, BAKWIN, PETER3, 1 2 3
ABSTRACT- Land surface models, which predict fluxes of energy and mass between the land surface and the atmosphere, are becoming increasingly important in the debate over the fate CO2 in the biosphere. It is important, however, not to lose sight of their sensitivities and limitations. Most complex models exhibit some degree of `equifinality¿. Equifinality occurs when compensation between parameters results in physically reasonable simulations across a wide range of parameter sets. The purpose of this study was to evaluate the sensitivity of a complex land surface model (SiB2) to its parameters and to establish acceptable error margins. Micrometeorological and eddy covariance measurements have been made since 1995 on a tall tower located in the Chequamegon National Forest, near Park Falls, Wisconsin. The vegetation of the region is heterogeneous, composed of mixed hardwoods and conifers. Ten thousand constrained-randomized parameter sets were generated and used to simulate fluxes of CO2, latent and sensible heat during a six-month period. The generalized likelihood uncertainty estimation (GLUE) methodology was then used to evaluate the results of the model runs compared to measured data and assess the sensitivity of the model to parameter values. We found that there was an irreducible level of error in the predicted versus measured fluxes, that our default parameter set based on global measurements performed very well and that simultaneous predictions of Net Ecosystem Exchange with energy fluxes reduces equifinalty.
KEY WORDS: Land Surface Model, Parameter Sensitivity, Land Surface Fluxes, Wisconsin