A hierarchical Bayesian regression framework to analyze climate data from Central Asia region

Barili, Emerson; Achcar, Jorge; De Oliveira, Ricardo Puziol

doi:10.21608/ejec.2024.278127.1036

A hierarchical Bayesian regression framework to analyze climate data from Central Asia region

Articles in Press

Document Type : Peer-reviewed articles

Authors

¹ Ribeirão Preto Medical School University of São Paulo - USP; Ribeirão Preto; SP, Brazil

² Ribeirão Preto Medical School University of São Paulo - USP; Ribeirão Preto, SP, Brazil.

³ Department of Statistics; State University of Maringá, Maringá, Paraná, Brazil.

10.21608/ejec.2024.278127.1036

Abstract

This study introduces a straightforward framework to analyze the climate data related to the minimum and maximum temperatures of countries in Central Asia (Kazakhstan, Kyrgyzstan, Tadjikistan, Turkmenistan and Uzbekistan) considering annual temperature averages over a long period of time ranging from the early 1900's to the beginning of years 2000. Standard existing multiple linear regression models under a hierarchical Bayesian approach were used in the data analysis assuming as covariates latitude and longitude of the climate stations, temporal factors (linear, quadratic, and cubic effects of years) and altitude of the climate station. The findings shown great accurate results to discover important factors affecting climate change (as the time (year), altitude and spatial factors) and to predict average temperatures in future years. Also, the obtained results are in agreement with many others studies in the literature that climate change is already being observed in all regions of the world. In special, we observed that annual average minimum temperatures in Central Asia are increasing in the five countries assumed in the study at the end of the follow-up period (close to the year 2003). Similar results were also observed for the annual average maximum temperatures.

Keywords