Abstract ID: 3.12345 | Accepted as Talk | Talk/Oral | TBA | TBA

Understanding the distribution of biodiversity has been a longstanding focus of ecologists; yet explorations beyond individual trajectories to communities remain challenging. The community assembly process is complex with biotic and abiotic factors interacting at varying levels of significance depending on the system. In the northern Andes, the super-páramo sits at the highest elevation reaches (> 4200 m) of the broader páramo ecoregion – a montane ecosystem within a tropical biodiversity hotspot. The geographically isolated super-páramo hosts specific and ecologically unique plant communities that are sensitive to environmental change. Emerging techniques in community modelling are promising for predicting future shifts, yet we still can improve current methods of incorporating biotic interactions in a way that more closely reflects reality. Current modelling frameworks infer biotic interactions through co-occurrence patterns derived from presence-absence data. Yet, abundance data remains underutilised in informing the strength and consistency of these patterns. Here, we use predictive species distribution models (SDMs) and community assembly rules to refine co-occurrence patterns by incorporating an abundance criterion. We adopt two techniques that (i) model first and assemble later; or (ii) assemble first and model later. We fitted SDMs with 556 vegetation plot data obtained from VegAndes and abiotic variables extracted from the CHELSA database. The first technique followed the Spatially Explicit Species Assemblage Modelling framework (SESAM); we built individual SDMs and assembled models by applying richness constraints and probability ranking rules. The second technique employed joint-SDMs that consider biotic interactions by modelling correlated residuals among species after accounting for environmental factors, and from these models build entire communities. For both techniques, the abundance filter was incorporated during the assembly step and weighted relevant to other assembly processes. Overall, the models predicted super-páramo communities well with mismatched areas being quite local, most likely resulting from oversampling in heterogenous landscapes. We found community predictions differed when considering abundance compared to the core techniques, offering the opportunity to improve the ecological relevance of models. These novel improvements could be used to predict future redistribution of super-páramo biodiversity under climate change.