New advances and technologies are leading to an unprecedented high resolution of community data, perhaps making it possible for the first time to unravel the mechanisms of metacommunity assembly. Environmental filtering, species interactions, ecological drift, and dispersal determine community composition in local communities, but disentangling their relative importance has proven elusive, likely due to inappropriate tools. Here, we show that joint species distribution models (JSDM) and variance partitioning can provide a solution. First, JSDM reveals the "internal structure" of communities and species that can be correlated in a second step against environmental and spatial distinctiveness, thereby revealing the importance of metacommunity assembly mechanisms. We demonstrate that this approach can detect environmental filtering and dispersal limitation in a pond metacommunity. We conclude that JSDMs are a powerful tool for metacommunity analysis, especially for large community data.

The Global Inventory of Floras and Traits (GIFT) is a global database of regional plant checklists that has proven successful in documenting biogeographical patterns of plants. Since the release of the first version of GIFT, the database kept on expanding. GIFT version 3.0 contains 5169 plant checklists referring to 3400 regions worldwide. These checklists include a total of 371,148 land plant species, mostly vascular plants, of which 354,848 have accepted species names, and species-level data for 109 functional traits.

This presentation will first introduce the GIFT database and present its structure with examples showing how to retrieve distribution data for specific taxonomic groups, functional traits at the species level, phylogenetic diversity, and environmental data at the regional level. Second, a comparison of data between the GIFT database and the GBIF repository will be presented.

Species distribution modeling (SDM) often performs poorly when evaluated against spatially independent test data. One contributing factor to this issue is the neglect of spatial autocorrelation during model training and validation, resulting in inflated performance metrics and the development of overly complex models. Among the SDM softwares used Maxent is one of the most widely utilized methods, largely attributable to its user-friendly graphical-user-interface (GUI). It has been shown that parameter tuning leads to better Maxent models in terms of complexity and performance. However, in nearly all published applications, Maxent is used with the default settings. The lack of model tuning and the ignoring of spatial autocorrelation may be related to the fact that the Maxent GUI does not include such functionalities.

We implemented tuning and validation functionalities that account for spatial autocorrelation in a software extension for Maxent: spatialMaxent (https://doi.org/10.1002/ece3.10635). We compared our results to models based on Maxent's default settings on a dataset with over 200 species. spatialMaxent outperformed the Maxent models in terms of model complexity and performance on spatially independent test data. All tuning functionalities in spatialMaxent are accessible via the user-friendly GUI, ensuring easy access for researchers and conservation practitioners alike.