Trends and multi-annual variability of water temperatures in the river Danube, Serbia.

The relationship between air (Ta) and water temperature (Tw) is very important because it shows how the temperature of a water body might respond to future changes in surface Ta. Mean monthly Tw records of three gauging stations (Bezdan, Bogojevo i Veliko Gradište) were analysed alongside mean monthly discharge (Q) for the same stations. Additionally, Ta series from two meteorological stations (Sombor and Veliko Gradište) were correlated with Tw variations over the period 1950-2012. Locally weighted scatter point smoothing (LOWESS) was used to investigate long-term trends in the raw data, alongside the Mann-Kendall (MK) trend test. Trend significance was established using Yue-Pilon's pre-whitening approaches to determine trends in climate data. Also, the rescaled adjusted partial sums (RAPS) method was used to detect dates of possible changes in the time series. Statistically significant warming trends were observed for annual and seasonal minimum and maximum Tw at all investigated sites. The strongest warming was observed at Bogojevo gauging station for seasonal maximum Tw, with +0.05 °C per year on average. RAPS established that the trend began in the 1980s. This behaviour is linked to climate patterns in the North and East Atlantic which determine the amount of heat advected onto mainland Europe. Statistically significant correlations were found for all Tw on an annual basis. Overall, the strongest correlations ( p < 0.01) between Tw residuals and the North Atlantic Oscillation (NAO) were recorded for the winter period. These findings suggest possible predictability of Tw over seasonal time-scales. Copyright © 2016 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]