We have a new paper out in Ecography. The aim was to link functional traits to environmental gradients. There are existing methods that do this, but they generally involve multiple steps. We created a hierarchical model that effectively joins a species distribution model with species trait values in one step. We were quite happy with the model because it worked well-better than we anticipated for rare species-and, importantly, produced sensible and interpretable results. Here is one example.
Specific leaf area (SLA) represents a tissue allocation strategy of either growing quickly or growing slowly with more tissue devoted to protection or conserving resources. SLA modifies species responses to rock cover. So, species with low SLA (thick, tough leaves) tend to increase in occurrence on increasingly rocky areas. Species with higher SLA (flimsy leaves) tend to be found on deeper, less rocky soil (see Figure).
The y-axis label looks complicated, but it’s simply the expected change in probability of species occurrence for a given change in surface rock cover. (technically, this is a partial response)..
Some key aspects of the model are:
1- species trait values actually modify species responses to environmental gradients. This may be useful for improving species distribution models when trait values are known.
2- rare species borrowed strength from common species. Species that are uncommon or have restricted distributions are usually quite difficult or impossible to model. I think this type of multi-species modelling shows real promise in this area.