Link to bioRxiv paper: http://biorxiv.org/cgi/content/short/2020.08.04.235390v1?rss=1 Authors: Martinez Batlle, J. R., van der Hoek, Y. Abstract: Despite being increasingly threatened by human-induced disturbances, dry forests remain the least studied and protected forest types in the Caribbean region. In contrast to many other forest systems in the world, we have little knowledge of the site-specific variation in vegetation communities within these forests, nor understand how plant species distribution is determined by environmental variables, including among them geological attributes. Here, we assessed the associations between plant communities and habitat types in a semi-deciduous forest of the Dominican Republic. We collected vegetation data from 23 sites within the Ocoa river basin, which we classified into six groups with a Random Forest algorithm, lithology, geomorphology, topography, and last decade history of forest loss as predictor variables. We established three main clusters: one group which encompassed sites with forest over a limestone substrate, four groups of sites with forests over a marlstone substrate with varied degrees of steepness and forest loss history, and one group that gathered all sites with forest over an alluvial substrate. In order to measure the associations of plant communities with groups of sites, we used the indicator value index (IndVal), which indicates whether a plant species is found in one or multiple habitat types, and the phi coefficient of association, which measures species preferences for habitats. We found that 16 species of woody plants are significantly associated with groups of sites by means of their indices. Our findings suggest that the detection of plant species associations with our selection of environmental variables is possible using a combination of indices. We show that there is considerable variation in plant community composition within the semi-deciduous forest studied, and suggest that conservation planning should focus on protection of this variation, while considering the significance and variability of geodiversity as well. In addition, we propose that our indicator groups facilitate vegetation mapping in nearby dry forests, where it is difficult to conduct thorough vegetation or environmental surveys. In short, our analyses hold potential for the development of site-specific management and protection measures for threatened semi-deciduous forests in the Caribbean. Copy rights belong to original authors. Visit the link for more info