PARENT SESSION
Oral Session # 16: Ecological Modeling I.
Presiding: C Ray
Tuesday, August 5. 8:00 AM to 11:30 AM, SITCC Meeting Room 102.

The effects of error in environmental variables on predictive vegetation modelling.

Van Niel, Kimberly^{*,1, 2}, Austin, Mike³, ¹ The Australian National University, Canberra, ACT, Australia² The University of Western Australia, Perth, WA, Australia³ CSIRO Sustainable Ecosystems, Canberra, ACT, Australia

ABSTRACT- The modelling of species responses to the environment and the prediction of species geographical distributions have important implications for vegetation theory and conservation evaluation. However, error and uncertainty can confuse these procedures and their results. The effect of digital elevation model (DEM) error on environmental variables, and subsequently on predictive vegetation models, is often acknowledged as a concern in modelling, but has not been explored. Based on an error analysis of a DEM, multiple error realisations of the DEM are created and then used to develop both direct and indirect environment variables for input to predictive vegetation models. The effects of this error and the resultant uncertainty of results are explored in the context of the typical steps in the modelling procedure for prediction of forest species presence/absence on the south coast of New South Wales, Australia. Results indicate that all of these steps and results, including the statistical significance of environmental variables, shapes of species response curves in generalised additive models (GAMs), stepwise model selection, coefficients and standard errors for generalised linear models (GLMs), prediction accuracy (Cohens kappa and overall accuracy), and spatial extent of predictions, are greatly affected by this type of error. Unconsidered error in the DEM can affect the reliability of interpretations of model results and level of accuracy in predictions, as well as the spatial extent of the predictions.

Key words: modelling, error, GAM, GLM