Climatic Research Unit

Regional case studies

Changes in Danube riverflow extremes

The atmospheric circulation states critical for these extremes have been identified and incorporated in a non homogeneous hidden Markov model. Results indicate a slight increase in the probability of occurrence of both low and high discharges, particularly towards the end of the century. Mares et al., 2008; 2009.

Related to the region size predictors we used: European area limited by (30N-65N; 0-40E). On this area we developed the atmospheric fields in EOFs, MEOFs and in the interior of this size we calculated: laplacians, gradients, means centered on: (45N, 12.5E; 42.5N, 17.5 E and 40N, 25E) as well as on (47.5 and 20E). Relating the predictors (precipitation) they are taken in the area: 41-50N; 16-30E.

Now we are processing the precipitation from 10 stations situated in the middle and lower Danube basin, taken as predictor SLP under the different indices, simulated by the four models.

Figures S below represent the mean changes estimated by means of t-test for CNRM (Fig. S1) in the period 2051-2092 and changes for laplacian estimated by CDF with the four models (Fig. S2), where one can see that the most significant change in extremes is revealed by the CNRM model. The laplacian centered on 40N, 25E is a good predictor for precipitation in the lower Danube basin.

Fig. D1 - Probabilities of the emission matrix of the HMM states for discharge level, associated with the atmospheric circulation (sea level pressure) classified in three states.

 

Fig. D2. Probability differences of the occurrence states with low (state 1) and high discharge (state 7) at Orsova * in the two periods in the 21-century estimated by means of nonhomogeneous hidden Markov model (NHMM) with seven states, in association with atmospheric circulation (three types). The link between states of HMM and state of atmospheric circulation is given by emission matrix (Fig. D1), obtained from the daily observations in the springtime (1958-1999). The pressure in the 21st-century is simulated by four models: CNRM_CM3, ECHAM5_MPI, EGMAM and IPSL (SRESA1B, stream 1 ENSEMBLES) for two periods of 42 year (in order to be compatible with observations): 2009-2050 (21C_1) and 2051- 2092 (21C_2). * The Orsova discharge represents an integrator of precipitation from the higher and middle basin of Danube.

 

Fig. S1. Change in spring SLP means (unbiased) simulated by CNRM for the period 2051-2092, related to observations (1958-1999). (Values of t-test > 2.37 indicate a confidence interval of 99 %).

 

Fig. S2. Change in cumulative frequency distributions of SLP laplacian for the period 2051-2092, related to observations (1958-1999).

 

Mares C., Ileana Mares, Mihaela Mihailescu, Heike Hübener, U. Cubasch, P. Stanciu, 2008: 21st century discharge estimation in the Danube lower basin with predictors simulated through EGMAM model. Revue Roumaine de Géophysique, tom 52, 2008.

Mares C., Ileana Mares, A. Stanciu, and Mihaela Mihailescu, 2009: Downscaled precipitation projections in the 21-century in the middle and lower Danube basin by means of nonhomogeneous hidden Markov model. Geophysical Research Abstracts, Vol. 11, EGU2009-11072, 2009 (EGU General Assembly 2009, 19-24 April).

Mares, C., Ileana Mares, and A. Stanciu, 2009: Extreme value analysis in the Danube lower basin discharge time series in the 20 th century. Theoretical and Applied Climatology. 95, 223-233. (DOI: 10.1007/s00704-008-0001-0).

 

return to top