Observed changes of temperature extremes in Serbia over the period 1961 − 2010

• Trends in extreme temperature indices in Serbia during 1961–2010 were investigated.
• Hot indices displayed a cooling trend until 1980 and a warming trend afterwards.
• A general warming tendency of cold indices over the entire period.
• No significant changes of diurnal temperature range.
• Significant lengthening of the growing season since 1980.

The analysis of spatiotemporal changes of temperature extremes in Serbia, based on 18 ETCCDI indices, was performed using daily minimum and maximum temperature observations from 26 meteorological stations over the period 1961–2010. The observation period was divided into two sub-periods (1961–1980 and 1981–2010) according to the results of the sequential Mann–Kendall test. Temporal trends were evaluated by a least-squares linear regression method. The average annual minimum temperature displayed a mixed pattern of increasing, decreasing, and no trends over 1961–1980 and a significant increasing trend over 1981–2010 across the whole country, with a regionally averaged rate of 0.48 °C per decade. The average annual maximum temperature showed a decreasing trend during 1961–1980 and a significant increasing trend at all stations during 1981–2010, with a regionally averaged rate of 0.56 °C per decade. Hot indices exhibited a general cooling tendency until 1980 and a warming tendency afterwards, with the most pronounced trends in the number of summer and tropical days during the first period and in the frequency of warm days and nights in the second. Cold indices displayed a mostly warming tendency over the entire period, with the most remarkable increase in the lowest annual maximum temperature and the number of ice days during the first period and in the frequency of cool nights during the second. At most stations, the diurnal temperature range showed a decrease until 1980 and no change or a slight increase afterwards. The lengthening of the growing season was much more pronounced in the later period. The computed correlation coefficient between the annual temperature indices and large-scale circulation features revealed that the East Atlantic pattern displayed much stronger association with examined indices than the North Atlantic Oscillation and East Atlantic/West Russia pattern.