PARENT SESSION
Organized Oral Session 47: Insights, challenges, and future directions in modeling forest dynamics at multiple scales
Organizer(s): C Tripler, C Canham, C Messier, and M Papaik
Thursday, August 11, 1:30 PM - 5:00 PM, Meeting Room 510a, Level 5, Palais des congrès de Montréal
Towards a global individual-based model of forest dynamics.
Pacala, Stephen*,1, Purves, Drew1, Ogle, Kiona1, Lichstein, Jeremy1, Wirth, Christian1, Chen, Anping1, Shevliakova, Elena1, Malyshev, Sergey1, Strigul, Nikolay1, 1 Princeton University, Princeton, NJ, USA
ABSTRACT- We are building a new individual-based forest model intended to be implemented across the range of scales from individual trees to whole continents and biomes. The effort consists of four parts: (1) a physiologically grounded model of an individual tree that competes for light and water, (2) a hierarchical Bayesian statistical package that "inverts" the growth, recruitment, and survivorship records in forest inventories and other data to estimate the species-specific constants in the model, and (3) a set of mathematical methods (moment closure schemes) to scale up individual-based models of forested ecosystems so that they may be implemented at continental or global scales, and (4) an interface that allows the scaled-up system to function as an interactive land surface inside the Geophysical Fluid Dynamics Laboratory's (GFDL) new climate models. Although we review recent results and progress on all dimensions of this work, we focus here primarily on the scaling methods (3). An individual-based forest simulator is a complicated spatial stochastic process that predicts properties of an ensemble of individuals by simulating the fate of every plant throughout its life cycle. We present four alternative approximations that can be used to derive an equation for the first moment of a forest simulator. An equation for the first moment contains only the functional forms and parameter values in the original simulator (no fitted constants), but predicts the size-structured abundances of each species at large scales. It is precisely analogous to the Navier-Stokes equations that predict the behavior of a gas at greater-than-molecular scales, as the first moment of the stochastic process governing a large collection of individual molecules. The first of the four methods was previously published (Moorcroft, Hurtt and Pacala 2001), but the other three are new. Unlike our earlier method, the new methods all are designed handle plastic growth of tree crowns. The new methods range from those with high accuracy of approximation but low numerical speed, to those with lower accuracy but high numerical speed.
Key words: forest model, moment closure, forest inventory