5° × 2.5°) and time (6 h), some regional details and cyclones could be missing. Therefore, for the wind-field snapshots during the maximum sea levels at Pärnu we have chosen the regional reanalysis Baltan65+ (Luhamaa et al. 2011), with a spatial resolution of 0.1°. We looked for deep cyclones that see more might cause high sea level events at Pärnu (above + 150 cm) and Tallinn (above + 100 cm): for only one case out of 31 was it not possible to detect the corresponding cyclone (1 November 1983, see Table 1). All the high sea level events listed in Table 1 took place during the storm season, i.e. from September to March. Extreme sea levels

were not always observed at both stations on the same days, however, as this depends on the cyclone’s exact position, lifecycle phase and velocity; but in really extreme cases, sea levels were high over a larger area of the sea along the entire Estonian coast. The cyclones that passed over the Baltic Sea and caused these 31 extreme events in 1948–2010 were not exclusively deep, and there was no obvious correlation between the minimum air pressure of the cyclones and the extreme sea level. Table 2 presents, separately for Tallinn and Pärnu, the average values of the cyclone characteristics for extreme sea level events. The atmospheric pressure at sea level at their centre is lower than the average value in the northern Baltic region – 985 hPa (Link & Post 2007). We counted the number

of cyclones in the research area during 60-day periods to test the hypothesis about the series of cyclones causing these high water events. Here we used two options: either the learn more extreme event was in the middle of the counting period (N 60_c) or we counted the cyclones that preceded the storm surge (N60_b). The number of cyclones was higher if the high sea level event was in the middle of the counting period (see Table 2). The same result is supported by Figure 1, where the secondary maximum sea levels are of the same magnitude before and after the main event. The average values of the real cyclone characteristics compared to the values modelled by Averkiev &

Klevannyy (2010) are presented in Table 2 and Figure 2. The dangerous cyclones for Tallinn and Pärnu sea levels are slightly different: Sitaxentan for Tallinn the position of the deepest phase of the cyclone should be shifted to the north by about two degrees, but the longitudes are considered to be the same. The ideal Pärnu cyclone has a stronger meridional track component (the slope of the trajectory is 0.304 instead of 0.223). On average, the most accurately predicted characteristic of a dangerous cyclone is the latitude of the deepest state; at both sites this coincides with the modelled value within one degree. In fact, the cyclones propagate somewhat more slowly than predicted and therefore their minimum pressure also occurs some 4–5 degrees farther to the west than predicted.