Model Accuracy in- and out-of-sample. Plots may take a few seconds to load.
Mean posterior estimates: Fitted coefficients having 95% intervals that exclude zero. Use the dropdown menus below (1) to display regression coefficients for an explanatory variable in the model, (2) to color-code each species by a trait, or (3) to switch between common and scientific name. Environmental variables explain variation in species abundance across the map. Positive values ['Mean posterior estimate'] mean that higher values of a variable (e.g. high elevations) are associated with higher abundance, and vice versa.

Variable Importance: Sensitivity coefficients integrate the effects of predictors across all responses (species) in the model. Variables with high sensitivity account for much of the variation in the community of species.
Commonalities in responses across species help to define communities on the basis of response rather than distribution or abundance. The clustering of responses here can help us rethink community relationships across these diverse species groups. Species are clustered using the E matrix, which is the correlation among species in terms of their responses to the environment. The figure to the right combines the effects of individual species responses with covariance in the environment. Species are displayed as both rows and columns. Warmer colors indicate species that have similar responses to environmental variables. Individual species' responses to environmental variables are shown in the far right column section as the fitted coefficient matrix B.
Sensitivity of predictors compared to mean posterior estimates: In this figure to the right the warmer colors along the diagonal of F indicate covariates with higher sensitivities. Sensitivity coefficients integrate the effects of predictors across all responses (species) in the model.

Communities of small mammals, beetles, and vascular plants

Learn more about the CMIP5 Representative Concentration Pathways (RCPs)
Communities are identified using Gaussian Mixture Modeling with the mclust package in R. Clustering is done on relative abundance-weighted habitat suitability through time. The top map shows 2018 community distributions and the bottom map shows projected future shifts. Communities with similar relative abundance-weighted habitat suitability across the two maps share the same color. The buttons on the left correspond to each community mapped on the right. Select a button to learn how the abundance-weighted habitat suitability for each species is predicted to change across time within the selected community. The figure on the bottom left will then show the mean change in abundance-weighted habitat suitability for species predicted to see the largest change in abundance-weighted habitat suitability within the selected community.

Compositional Changes in Communities

To see how the species composition is changing select a community below.
The abundance-weighted habitat suitability of species within each community is also expected to change. For example, certain species might become more abundant, whereas other species may become rarer, even though overall the community composition indicates that this future community is similar enough to the current community to be considered the same. The figure below shows mean change in abundance-weighted habitat suitability from present for species within the selected community. Only species with the highest predicted change in abundance-weighted habitat suitability are shown. Units depend on the taxa: estimates of abundance per plot (counts, basal area, or cover)

Spatial Shifts in Communities

2018
Future