Spatial modeling in landscape ecology

Abstract

The Netherlands experienced quite a controversy in January 1999 when an employee of the National Institute of Public Health and the Environment (RIVM) accused his employer, in the media, of relying too much upon unvalidated models instead of empirical data. He argued that the model outcomes were unreliable and that politicians are led to believe that they represent reality, when in fact they represent an artificial universe with no link to real data (Fig. 9.1). He made an interesting point, because models are often used without being calibrated, tested, validated, or analyzed for sensitivity and/or uncertainty. Furthermore, it is usually unclear what part of the model is based upon hard data and where expert knowledge fills in the gaps. This essay is about models, expert knowledge and data, calibration, validation, and model analysis, and how we can apply these for evaluation or prediction. We argue that all these combined produce a more powerful tool than models, experts, or data do alone. We will not discuss the importance of space, or the merits of spatially explicit versus non-spatial or nonspatially explicit models. This issue has been thoroughly discussed elsewhere (Durrett and Levin, 1994a, 1994b; Wiens, 1997). This essay is a little biased toward spatial population models and vegetation dynamics models, which are our primary fields of interest. Although we offer several critical remarks, we are enthusiastic about the merits of spatial modeling for applying landscape ecological knowledge.