Climate Change: Use of Non-Homogeneous Poisson Processes for Climate Data in Presence of a Change-Point

In this study, non-homogeneous Poisson processes (NHPP) are used to analyze climate data. The data were collected over a certain period time and consist of the yearly average precipitation, yearly average temperature and yearly average maximum temperature for some regions of the world. Different existing parametric forms depending on time and on unknown parameters are assumed for the intensity/rate function lambda(t), t >= 0 of the NHPP. In the present context, the Poisson events of interest are the numbers of years that a climate variable measurement has exceeded a given threshold of interest. The threshold corresponds to the overall average measurements of each climate variable taking into account here. Two versions of the NHPP model are considered in the study, one version without including change points and one version including a change point. The parameters included in the model are estimated under a Bayesian approach using standard Markov chain Monte Carlo (MCMC) methods such as the Gibbs sampling and Metropolis-Hastings algorithms. The models are applied to climate data from Kazakhstan and Uzbekistan, in Central Asia and from the USA obtained over several years.

Automated summary

In this study, non-homogeneous Poisson processes (NHPP) were used to analyze climate data by assuming different parametric forms for the intensity function. The models considered the numbers of years that a climate variable measurement exceeded a given threshold, and employed a Bayesian approach to estimate parameters. The models were applied to climate data from Kazakhstan, Uzbekistan, and the USA.