Predicting vascular plant species richness of fragmented forests in agricultural landscapes in central Belgium

Abstract

This paper tests different methods to assess the species richness of vascular plants in fragmented forests embedded in agricultural landscapes. Certain plant species enabled us to rank forest patches in order of actual species richness, but failed to predict the exact number of plant species. However, the exact species richness could be satisfactorily predicted through multiple linear regression models that used a series of environmental variables (69% explained variance). Using this method, species richness was easier to predict in forests on alluvial soil and of young age (77 and 66% explained variance, respectively) than in those on non-alluvial soil and of old age (47 and 50% explained variance, respectively). The latter needed more observations (improvement to 70 and 56% explained variance, respectively) or more variables (improvement to 61 and 73% explained variance, respectively) to develop a reliable model. Predictions based on data obtained using GIS were satisfactory only for forests on alluvial soil (58% explained variance). Important indicators of species-rich forests were habitat diversity, forest age and forest area. Species richness was substantially larger in forests on alluvial soil compared with those on non-alluvial soil. Lower phosphorus availability to plants and proximity to older forests indicated a higher species richness in forests on alluvial soil and of young age. A higher pH and a lower C/N-ratio of soil indicated a higher species richness in forests on non-alluvial soil and of old age. It is thus possible to predict species richness of vascular plants in different types of forest using environmental variables, which is useful when selecting and managing forests for biodiversity conservation and restoration.