A Bayesian quantile regression framework for modeling trends in air temperature: The case for Greece

This study investigates the trend dynamics of some extreme air temperature bounds estimated via quantile models. In practice, we investigate the dynamics of the 5% lower, the median and the 95% upper quantile measures of the minimum and maximum temperature in various regions of Greece. For this purpose, we propose some semi-parametric quantile trend models. Our estimation is based on a Bayesian early-rejection Markov chain Monte Carlo algorithm. The climatic data used, are the mean monthly homogenized minimum and maximum air temperatures observed at several meteorological stations of the Hellenic Meteorological Service located in Greece for the 1960-2010 period. Results based on several margins of datas' distribution, reveal the strongest and the weakest inter-annual quantile trends for both temperature extremes. The results are very significant since they show the existence of time-trend heterogeneities in temperature extremes under alternative types of quantiles, time periods, geographical zones and seasons.