Abbott, M. H.. Oregon State University, mark@oce.orst.edu

Moisan, J. R.. Long Island University, moisan@vanilla.liu.edu

LESSONS LEARNED FROM CONFIGURING ECOSYSTEM MODEL PATHWAYS AND ESTIMATING MODEL PARAMETERS WITH ASSIMILATION OF OBSERVATIONS FROM LONG-TERM TIME SERIES

Ecosystem models, observations from long-term time series (e.g. U.S. JGOFS time series) and process oriented studies (e.g. NABE), and assimilative techniques are increasingly used together to estimate model parameters that are usually poorly known. Several techniques of data assimilation, such as the simulated annealing and adjoint variational methods, have been successful in estimating the model parameters and improving the fit between model results and observations. However, fitting the full set of observations can be problematic. The use of long-term time series with an adjoint method was found to be a powerful tool for modifying the model pathways and fitting the model results to the full set of observations. Although ecosystem modeling presents its own set of challenges, the utilization of observations also has its own problems. Measurements and modeled concentrations are often expressed in different units. Conversion factors are then required but are often poorly known. Measured concentrations and rates can be overestimated or underestimated as a result of the measurement technique (e.g. filtration with Whatman-GF/F and Nuclepore-PC leads to different estimates of chlorophyll and particulate organic matter). Results and lessons learned from using an eleven component ecosystem model, a variational adjoint technique and observations from two long-term time series will be presented.

Day: Monday, Feb. 1

Time: 04:00 - 04:15pm

Location: Eldorado Hotel

Code: SS44MO0400E