Changes in predictor influence with time and with vegetation type identity in a post-abandonment situation

Abstract

Abstract Variation in the explanatory potential of separate variable groups describing past vegetation patterns and/or abiotic environment were investigated using data from two points in time for three vegetation types, which are variants of the steppe grassland component of the forest steppe biome. First, we compared the predictive power of environmental variables assessed by the performance of generalised linear models (GLMs) explaining the distribution of three vegetation types in 1988 and 2002. Second, we fitted models based on conceptual hypotheses to explain vegetation distribution in 2002. Specifically, we wanted to examine, whether (i) current abiotic topo-environment, (ii) past neighbourhood configuration, or (iii) historical vegetation patterns or a combination of these determine best the current distribution of vegetation types. We developed basic predictor sets for each hypothesis, and using GLMs we tested to what degree these predictor sets were capable of explaining the currently observed patterns of individual vegetation types. We compared model accuracy by AUC and TSS values. Predictive performance of models changed both with time and with vegetation type. The analyses of changes over time showed that two of the three vegetation types had come closer to an equilibrium with abiotic conditions, while the third had moved farther away from equilibrium. Knowledge of past conditions was sufficient to predict the distribution of one of the three investigated vegetation types alone, thus no topo-environmental predictors were needed to successfully predict this type. The other two vegetation types were best explained by current topo-environmental predictors. We conclude that historical conditions clearly improve predictive models, though there may be variation in their contribution to models of different vegetation types and this may depend on how far vegetation types are from an equilibrium state.