Document: MAR-3-87-10

Distinguishing ecologically significant patterns from model artifacts: A spatially-explicit method to identify and represent model uncertainty.

BRUGNACH, M.* ¹, G.A.BRADSHAW ², J.BOLTE ¹ and R.P.NIELSON ¹

Oregon State University Corvallis OR 97331 USA ¹
National Center for Ecological Analysis and Synthesis, Santa Barbara, CA USA ²

Abstract:
Ecological models not only represent a tool for studying ecosystems, but function as a key source for informing environmental policy. As such, these models are expected to present and convey information and uncertainty as accurately as possible. As the complexity of these models increases, the resultant output also becomes more complex and difficult to interpret. Part of the difficulty in model output interpretation is to distinguish ecological variability from computational or representational artifacts. The complexity of most ecological models and intrinsic uncertainty challenge our abilities to accurately distinguish between these two sets of variability; namely, model error and ecologically significant variation. To address this problem, we present a method to "diagnose" the source, type and effect of model uncertainty for the case of highly complex and spatio-temporally explicit models. Currently it is in use for the global vegetation model MAPSS, which seeks to predict changes in vegetation due to climate perturbations. Specifically, the methodology expands the information provided by the model by including measurements that represent the uncertainty embedded in the results. Spatial maps of varying levels of model uncertainty are provided. We employ a stepwise method using genetic algorithms, fuzzy sets and ecological metrics to identify, represent and classify model uncertainty. A novel characteristic of the approach is that uncertainty is expressed entirely in ecological terms, based on ecological criteria, spatial information and frequency of errors.

Keywords: model uncertainty, model assessment, genetic algorithms, fuzzy sets

This abstract is being presented at: 10:30 AM in session:
STATISTICAL ECOLOGY