Altitudinal changes of summer air temperature trends in the Romanian Carpathians based on serially correlated models

This paper investigates the relationship between altitude and summer temperature trends in the Romanian Carpathian Mountains. We considered 20 weather stations and three types of topography: summit, slope and depression. We used a change-point regression model with serially correlated errors and compared it with a mainstream literature change-point model with independent errors. For both models, we identified decreasing trends before the change-point and increasing trends afterwards for most summer temperature series. The model with serially correlated errors gives change-points and trends that are more variable for lower altitude and more similar with each other at higher altitudes. One of the possible factors explaining this behavior is that temperatures from some weather stations in depression areas may be influenced by human settlings and industrial activities, and this influence would dissipate with altitude. On the other hand, the model with independent errors shows no pattern in the variability of change points or trends. The cluster analysis also reveals that weather stations are best grouped by altitude.