Table 1. Onset dates of used MBIs and their relative intensity Q
1961-03-2612.71970.10.2715.01961-12-0216.41973.11.1327.01963-11-1813.71976.11.3013.31964-02-0311.01982.11.1812.71965-10-3029.51983.01.1312.01969-02-0113.21993.01.1834.01969-10-2929.2
Fig. 1
SLP deviations from the long-term daily mean in relation with MBIs. The composites are based on the selected training events (Table 1). Deviations are shown for 20 (a) and 9 (b) days prior to the MBI as well as the onset day (c). The latter are directly comparable to Fig. 8A+B in Matthäus and Schinke (1994).
Fig. 2
The first four leading EOFs for major Baltic inflows based on 13 selected events (see Table 1). The explained variance of the EOF patterns are 38, 23, 15% and 7%, respectively.
Fig. 3
Thin lines show the mean development of the first four PCs. Thick lines represent the idealised PCs as used for the algorithm. All values are standardised. The day is given relative to the onset of the MBI. Negative days indicate the precursory period whereas positive represent the inflow phase.
Fig. 4
Correlations between the u-wind component and the mean evolution of PC1 to PC4 as shown in Fig. 3. The correlation is the mean over all 13 training events. The full 41 day periods are used in case of PC1, PC2 and PC4 whereas correlation with PC3 is based only on days considered for the computed correlation threshold of the algorithm – day −21 to day 8. The correlation is based on the mean development of the PCs and not the idealised since there is no idealised PC4.
Fig. 5
Same as Fig. 4 but for the v-wind component.
Fig. 6
Development of the PC1, PC2 and PC3 for all MBIs chosen as training events.
Table 2. Threshold values used to identify MBIs
TypePC1PC2PC3PC1+2PC1PC2PC3PC4
I>0.6>0.18>0.95>7>2 (5)>1 (any)>−1.5 (>0)II>0.29>0.5>2.3 (>5)>3.1 (>4)>0.1 (>−0.3)>−1III>0.08 (0.5)>0.1 (0.5)>0.64>2.5 (>6)>0.65 (>−0.5)
[i] Three different types of events are distinguished and the criteria are shown in the corresponding columns. The left part of the table shows thresholds based on correlations. We use correlation thresholds based on PC1, PC2, PC3, as well as the sum of correlation with PC1 and PC2. The right part of the table shows threshold values based on differences between different periods of the PCs development in time (see text). Red numbers highlight the strongest threshold which, in turn, define the type of event. Values in brackets refer to special cases. Here, green numbers indicate a lower than usual thresholds which in turn requires higher levels for other thresholds (blue).
Table 3. Dates of identified MBIs together with some notes, for example, about corresponding observed events if any
1961-03-22Training1979-04-291961-12-02Training1979-09-14Impact on BY51962-10-2516 PSU1981-06-071963-11-18Training1982-11-17Training1964-02-03Training1983-01-13Training1964-11-20Listed MBI1983-03-0717.5 PSU1965-10-30Training1986-05-292002)1967-04-2516 PSU, little on BY51991-12-21Impact on BY51967-10-141993-01-18Training1969-02-01Training1993-12-0917.8 PSU1969-10-29Training1994-03-132002)1970-10-27Training1998-02-2816.5 PSU1972-03-29Impact on BY51998-10-2516 PSU1972-11-1117 PSU1999-04-1016 PSU1973-11-12Training2003-05-102006)1975-12-28Listed MBI2003-12-2018 PSU1976-04-0316 PSU, little on BY52004-09-22IOW, little on BY51976-11-30Training2006-05-261977-11-13Listed MBI2008-02-29IOW1978-09-1616 PSU2008-11-22IOW, BY51979-03-10Impact on BY152009-09-30IOW
[i] Here, Training means that this MBI is part of the events used to train the algorithm. Listed MBI refers to MBIs listed by Matthäus et al. (2008) which were not part of the training events. Other references relate to publications describing barotropic inflows for a corresponding period. IOW corresponds to barotropic inflow events mentioned in the annual report of the Institut of Baltic Research Warnemünde (reports start in 1999, see http://www.io-warnemuende.de/zustand-der-ostsee-2012.html). Salinity declarations refer to high salinity concentrations at BY5 which could have hampered the propagation of saline water into the Baltic Sea (cf. Fig. 7).
Fig. 7
Near bottom salinities at BY5 (top) and BY15 (bottom). Vertical lines show identified MBIs; black lines are training events whereas red lines represent additional events identified by the algorithm Note that the RCA4 simulation ends in 2010 and consequently later events cannot be identified.
Table 4. The number of identified MBIs based on the number of criteria
xxxxx42xxxx–62 (+20)xxx–x62 (+20)xxx––85 (+43)xx–xx60 (+18)xx–x–94 (+52)xx––x79 (+37)x–––x100 (+58)xx–––116 (+74)–xxxx152 (+110)x––––158 (+116)––––x234 (+198)–xxx–278 (+236)
Fig. 8
Same as Fig. 7 but for all potential events identified when the BSI criteria is omitted.
Fig. 9
The evolution of the first three PCs between 2002/12/14 and 2003/01/23. Hence, 2003/01/12 is considered as the onset day which corresponds to the last observed strong MBI.
Fig. 10
Identified MBIs in scenario simulations. MPI and EC-Earth continue after the historical period (1961–2005) with the emission scenarios RCP4.5 (top) and RCP8.5 (bottom), respectively. The ECHAM5 simulation is extended with the A1B scenario. Total numbers of identified events are given for the historical and the future (2055–2099) period.
Table 5. The number of MBIs identified by our algorithm as defined in Section 3.2
MPI-OMall parameter192226MPI-OMno BSI253233MPI-OMno PC4233237MPI-OMno PC3333036MPI-OMno PC3/4383946MPI-OMno PC3/4, no BSI596261EC-Earthall parameter213926EC-Earthno BSI284633EC-Earthno PC4275342EC-Earthno PC3294538EC-Earthno PC3/4386052EC-Earthno PC3/4, no BSI557567ECHAM5all parameter24–34 (A1B)ECHAM5no BSI34–42 (A1B)ECHAM5no PC433–50 (A1B)ECHAM5no PC338–44 (A1B)ECHAM5no PC3/448–60 (A1B)ECHAM5no PC3/4, no BSI64–83 (A1B)
[i] Results are shown for RCA4 simulations driven with three different GCMs and three different greenhouse gas emission scenarios, that is, RCP4.5, RCP8.5 and A1B. Total numbers are given for a control period (1961–2005) and the second half of this century (2055–2099). Results are shown using either all or only certain selected criteria.