PARENT SESSION
Oral Session #30: Modeling.
Presiding: T. Day
Tuesday, August 6. 8:00 AM to 11:30 AM. Apache Meeting Room, TCC.
Matching scale of modeling with scales of variability in a forested region of Oregon, USA.
Kennedy, Robert*,1, Turner, David1, Guzy, Michael1, 1 Oregon State University, Corvallis, OR
ABSTRACT- When modeling ecosystem fluxes over large areas, a key issue is how to appropriately match spatial and temporal scales of the model, the system being studied, and the array of input variables used to drive the model. Computational constraints often require that daily time-step ecophysiological models be run at relatively coarse grain sizes, although a finer grain size might be more effective in capturing effects of land use and environmental heterogeneity. Here we report on a modeling strategy that allows more efficient modeling at fine grain sizes by reference to model runs derived from a systematically sampled subset of all possible grid cells. The strategy capitalizes on two phenomena: redundancies in input variables caused by spatial autocorrelation, and continuity in modeled output as a function of change in input variables. We applied the strategy to net ecosystem productivity (NEP) values calculated with the ecosystem model Biome-BGC to the forested areas in a 39,000 km2 east-west swath in western Oregon. Results were compared to a more traditional run of the model with full implementation at the finest grain size. We show that the new strategy adequately captures the patterns of NEP across the landscape, while simultaneously allowing for fine-grain (cell size 25 by 25m) variation in key input variables (landcover, leaf area index, and topography). Depending on the autocorrelation and covariance structures of the input variables, we suggest that the method can facilitate better matching between scales of models and input variables.
KEY WORDS: modeling, net ecosystem productivity, forest, Oregon