Dennis, B. L. University of Idaho, brian@uidaho.edu

Ives, A. R. University of Wisconsin - Madison, arives@facstaff.wisc.edu

Carpenter, S. R. University of Wisconsin - Madison, srcarpen@facstaff.wisc.edu

MULTIVARIATE STOCHASTIC MODEL FOR ESTIMATING ECOLOGICAL INTERACTIONS: OVERVIEW AND APPLICATION TO DATA FROM WHOLE-LAKE EXPERIMENTS

We have developed a multivariate stochastic model that can be used to estimate ecological interactions and stochastic return times from time series data. The model is a straightforward extension of a univariate autoregressive [AR(1)] model to a multivariate and stochastic situation. Because the model explicitly incorporates stochasticity, it is readily testable, more biologically realistic than deterministic approaches, and shows how ecological processes and noise interact to produce observable patterns. We demonstrate the model's usefulness by applying it to data from the plankton communities of 3 Michigan lakes: Paul, a reference system; West Long, deliberately enriched with nutrients; and Peter, subjected to food web manipulation and nutrient enrichment. For this example, we evaluated the interactions of 4 community components (small [<35 um] phytoplankton, large [>35 um] phytoplankton, Daphnia spp., and non-daphnid zooplankton) assuming that all interactions among the 4 components were possible. Although the nutrient and food web manipulations affected each lake similarly, there were differences in the ecological interactions suggestive of different intrinsic dynamics in each lake. In addition, the stochastic return time was 10x greater in Peter Lake than in Paul and West Long lakes, which may indicate that frequently manipulated lakes are less resilient than less disturbed systems.

Day: Monday, Feb. 1

Time: 03:30 - 03:45pm

Location: Eldorado Hotel

Code: SS44MO0330E