A Dirichlet process model for change‐point detection with multivariate bioclimatic data

Motivated by real-world data of monthly values precipitation, minimum, and maximum temperature recorded at 360 monitoring stations covering the Italian territory for 60 years ( 12 × months), in this work we propose a change-point model multiple multivariate time series, inspired hierarchical Dirichlet process. We assume that each station has its structure and, as main novelties, allow unknown subsets parameters likelihood to stay unchanged before after change-point, possibly share same number weather regimes is estimated random quantity. Owing richness formalization, our proposal enables us identify clusters spatial units parameter, evaluate which are more likely change simultaneously, distinguish between abrupt changes smooth ones. The proposed provides useful benchmarks focus programs regarding ecosystem responses. Results shown whole data, detailed description given three stations. Evidence local behaviors includes highlighting differences potential vulnerability climate Mediterranean ecosystems from Temperate ones locating trends distinguishing continental plains mountain ranges.