Modeling invasion risk of Mimosa pigra L. in Northeastern Thailand using Bayesian count models

Mimosa pigra L. is a globally significant invasive species that threatens wetland and agricultural ecosystems across the tropics. This study models its population density (plants per square meter) in northeastern Thailand using Bayesian Poisson and negative binomial regression, incorporating soil physicochemical properties as predictors. Data were collected from 50 plots across three districts in Maha Sarakham Province, with analyses of soil pH, organic matter, phosphorus, potassium, and electrical conductivity. Model performance was assessed via the Leave-One-Out Information Criterion (LOOIC) and posterior predictive checks. The negative binomial model provided a superior fit by capturing overdispersion, identifying potassium concentration, soil texture classes (e.g., clay loam, sandy clay), stem diameter, and soil structure as key determinants of M. pigra density. This work represents the first Bayesian quantification of edaphic drivers of M. pigra in Southeast Asia, demonstrating the utility of Bayesian count models for invasion ecology and offering practical guidance for habitat prioritization, early detection, and targeted management in high-risk floodplain ecosystems.