Sharper detection of winter temperature changes in the Romanian higher-elevations

This paper investigates winter temperature trends in the Romanian higher-altitude areas, for three types of topographies: depression, slope and summit. The main challenge is that some winter temperature trends, by comparison with the other seasons, are milder and harder to detect. We used a change-point regression model with statistically dependent errors and compared it with a standard change-point model with independent errors. Statistical theory ensures that the former model gives a more accurate trend analysis than the latter. The model with statistically dependent errors detects change-points in the mid 70s and statistically significant increasing trends both before and after the change-point. On the other hand, the model with independent errors does not detect statistically significant increasing trends after the change-points for the winter series. These general results occur for all topography types. A separate multiple regression model reveals that the winter temperature trend changes in the Romanian higher-elevations can be described by a linear additive effect of several global atmospheric circulation patterns.