Pacote GeoPoisson: Implementação e Aplicações

Abstract

The application of geostatistical models have been growing rapidly in the last decades. Geostatistical models for counting via Poisson process have been a versatile tool for environ- mental sciences in predicting anomalous events (excessive rains, lethal CO concentration). However, the counting via homogeneous Poisson process model is applicable to a very narrow scope of phenomena. For thus a geostatistical model was developed for counting via non homogeneous Poisson process. We propose a new package in R software with a set of functions to estimate the model above. The usage of the package is exemplified with data obtained from 29 measuring stations in Piaui and Maranhao - Brazil, from 1980 to 2010 published by ANA (Agencia Nacional de Aguas - Brazilian National Water Agency). The main function obtains credibility intervals to the parameters of the model in a Bayesian approach through MCMC, specifically the Metropolis-Hastings algorithm within Gibbs sampling. In addition, we made a function which returns an interpolation map of anomalous events in neighboring unobserved regions.