Climate Change Data: Use of an Autoregressive (AR) Model in Presence of Change Points under a Bayesian Approach

In this study, we introduce a statistical model applied to climate change data consisting of an autoregressive times series (AR) model which represents a type of random process. A Bayesian approach using MCMC (Markov Chain Monte Carlo) methods is considered to get the inferences of interest. The main goal of the study is to have a model to get good predictions for mean temperature and also good to identify the time of possible change-points that might be present in the time series which could indicate the possible beginning of a change in climate. Applications of the proposed model are considered using annual average temperatures in some locations obtained over a period of time ranging from the end of 1800’s to a popular Bayesian discrimination criterion using MCMC methods.In addition to a good fit of the proposed model for the data, the model also was used to detect the times of climate changes in the different climate stations using CUSUM methodology.