A Latent Discrete Markov Field Approach for Identifying and Classifying Historical Forest Communities Based on Spatial Multivariate Tree Species Counts

• Activity Number: 190 - Contributed Poster Presentations: Section on Statistics and the Environment
• Type: Contributed
• Date/Time: Monday, July 29, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Stephen Berg* and Jun Zhu and Murray Clayton and Monika Shea and David Mladenoff
• Companies: and University of Wisconsin - Madison and University of Wisconsin-Madison and University of Wisconsin-Madison and University of Wisconsin-Madison
• Keywords: Historical ecology; Stochastic approximation; Markov chain Monte Carlo

Abstract

The Wisconsin Public Land Survey database describes historical forest composition at high spatial resolution across a large spatial extent, and is of interest in ecological studies of forest composition in Wisconsin just prior to significant Euro-American settlement. For such studies, it is useful to identify recurring subpopulations of tree species known as communities, but standard clustering approaches for subpopulation identification do not account for dependence between spatially nearby observations. Here, we develop and fit a latent discrete Markov random field model in the context of identifying and classifying historical forest community types based on spatially referenced multivariate tree species counts across a large region. We show empirically for the actual dataset and through simulation that our latent Markov random field modeling approach improves prediction and parameter estimation performance. We also introduce a new stochastic approximation algorithm for fitting this spatially correlated model. Our algorithm enables computationally efficient estimation and classification of historical forest communities for the large Wisconsin Public Land Survey dataset.