Internal Stochastic Structure of Annual Discharge Time Series of Serbia’s Large Rivers - page 6

Conclusion

Time series of annual discharges of the Danube River and its tributaries (the Sava, Tisa and Velika Morava) were assessed to reveal their internal structure. The time series of annual discharges were found to be non-stationary in terms of average values, and to exhibit trends and a distinct multiple-year periodicity. Contrary to the standard analysis by means of the Mann-Kendall test, which considers the entire time series, multi-temporal analysis was used in this research. The analysis of the entire time series showed no significant trend at the confidence level of 95%, but distinct downward trends were identified at the stations of Bogojevo and Sremska Mitrovica, as well as an upward trend at Senta. This suggests that the changes were a result of multiple-year periodicity of the time series, which affects trend direction and intensity along the time series, but that there was a substantial decrease in discharge during the last decades. It was also demonstrated that not all the time series exhibited identical trend directions and intensities, so the assumption was that the trends of the analyzed time series were affected by large-scale geophysical processes and physical and geographical characteristics of the river basin. The effect of geophysical processes is reflected in reduced annual precipitation totals in South East Europe, as well as increased totals in Central Europe, resulting in an upward trend of the Tisa River.

 

Given that climate models through the year 2010 anticipate an even more uneven precipitation distribution between northern and southern Europe, as well as increased air temperatures (Djurdjevic et al. 2008, 2010, Danube Study, 2012), there is the question of discharge trends in the coming years, particularly of the Sava and Velika Morava whose catchments will likely experience a considerable decrease in precipitation. The time series of annual discharges were found to have a long memory as the Hurst coefficient was in the range h=0.622-0.738. The periods of the time series were classified into four groups, whose average values were: 9.5, 13.5, 22.5 and 31.5 years. The periods exhibited alternating multiple-year dry and wet episodes in the studied catchments. Knowledge of these periods is of key importance from a water management perspective. Apart from the primary long-term periodicity, secondary microperiods of 3.6-4.9 years were also noted. The analyses described in this paper revealed the internal structure of the hydrologic time series, which is a basis for the development of stochastic hydrologic models that will provide long-term projections of hydrologic parameters in the river basin.

Acknowledgement

The research presented in this paper is funded by Serbia's Ministry of Education and Science as part of an ongoing scientific project titled "Assessment of Climate Change Impact on Water Resources in Serbia" (TR-37005, 2011-2014).