|
Abstract:
|
Multi-species monitoring programs are often implemented to better understand how ecological communities change over time and space. Ordination techniques can be used to visualize high dimensional, multi-species data in a low dimensional space. These visualizations represent species composition across sampled locations and can be used to cluster sampled locations based on species composition. This clustering information can then be used to inform sampling effort or understand how communities respond to disasters, such as wildfires. Recent advances in model-based ordination techniques that use finite mixture models have allowed for simultaneous clustering and ordination in a single coherent modeling framework. However, these techniques require prior specification of the number of groups present in the low dimensional species space, which is often unknown. In this talk, we describe an infinite mixture model capable of simultaneous clustering and ordination of multivariate abundance data without prior specification of the number of groups present in the latent species space. We then use this model to analyze ordinal cover class data collected from Grand Teton National Park
|
