Multivariate Bayesian clustering using covariate-informed components with application to boreal vegetation...

Area of Science:

• Ecology
• Climate Science
• Statistical Modeling

Background:

• Climate change significantly impacts vegetation distribution and abundance, particularly in Arctic regions.
• Understanding vegetation sensitivity and robustness to climate shifts is crucial but limited by a lack of long-term data.
• Plant assemblages exhibit heterogeneous responses to climate change across space and time.

Purpose of the Study:

• To develop a novel statistical model for assessing vegetation sensitivity and robustness to climate change.
• To identify patterns and mechanisms underlying differential vegetation responses.
• To apply the model to vegetation abundance data in Alaska, leveraging spatial extent as a proxy for historical observations.

Main Methods:

• Development of a multivariate statistical model incorporating unknown cluster-specific effects and covariances.
• Utilizing a prototype model for cluster membership that allows flexibility and enforces smoothness across sites.
• Application of the model to vegetation abundance data from Alaska, USA.

Main Results:

• The model successfully identifies site-level cluster labels indicating vegetation sensitivity and robustness.
• Interpretable classifications of vegetation assemblages based on their climate response were achieved.
• The approach did not require strong a priori assumptions about climate sensitivity drivers.

Conclusions:

• The novel statistical model provides a flexible framework for understanding vegetation climate sensitivity.
• Spatial extent can serve as a valuable proxy for unrecorded historical climate and vegetation data.
• The findings enhance our understanding of ecological responses to climate change in high-latitude ecosystems.