ESS: A tool for ecological simulation
Kevin Glass, University of Oregon
For the last four years, I have been researching problems in computational ecology. Specifically, I have considered the application of parallel discrete event simulation to problems in demography. The consideration of discrete event simulation stems from two recent trends in ecological modeling, namely individual oriented modeling and spatially realistic modeling. In an individual oriented model, we model a population by modeling the physiological and social interactions of its members. Recent work has shown that spatial distribution of organisms tends to explain some demographic stability. This has led some ecological modelers including those interested in individual oriented models--to focus on the effect of spatial realism.
With few exceptions, modelers implement spatially explicit, individual oriented models as discrete time models. Discrete time models assume we can represent interactions by updating a model at regular intervals, but ecosystems are not generally synchronous when considering system behavior at the individual level. As a consequence, discrete event models may be more realistic and more efficient to implement.
In the initial phase of my research, I reviewed parallel simulation techniques and implemented a spatially explicit model using the Time Warp Operating System (Time Warp is a major parallel simulation algorithm). This experience led me to two conclusions: the performance of parallel simulation algorithms are not characterized well enough to comment on their appropriateness for a given problem and ecological models must be developed as discrete event simulations before we attempt to fit ecological models to a particular parallel algorithm.
The goal of my research is two fold. First I am working, in conjunction with ecologists, to develop discrete event ecological models. The primary product of this research is the Ecological Simulation System (ESS) a framework for population modeling based on Zeigler’s Theory of Modeling and Simulation. The second goal of my research is a performance analysis of realistic ecological simulation, implemented with parallel simulation algorithms. The end of this research is to develop a general simulation tool for ecological modeling.
Abstract Author(s): Kevin Glass