Abstract
Niche-based models, such as transfer functions, are commonly used in a wide variety of ecological systems for quantitative reconstruction of past environments. Recent ecological theory challenges some of the major assumptions underpinning these models. Spatial autocorrelation in species and/or environment data in the training set may lead to overestimation of model fit and bias in the selection of environmental variables. Neutral processes may lead to species—environment relationships that are variable and context-sensitive, and community dynamics that are driven primarily by stochastic processes. Localized interactions and non-linear processes may result in complex dynamics, including step-like changes between stable states, multiple stable states for the same external forcing and hysteresis effects. Null models and independent validation may be used to test for spatial autocorrelation and neutrality. Experimentation with spatially explicit, process-based models is required to understand how complex dynamics affect interpretations of the palaeorecord.
Keywords
Get full access to this article
View all access options for this article.
