9). In the same way, forbidding see more routes north of Bornholm would produce a curve following the dots to the right

of the gap but with the upper extreme somewhere to the right of the currently depicted curve (green curve in Fig. 9). When both south and north of Bornholm are Permitted, the route will either go south or north of Bornholm (thick curves in Fig. 9). However, for some weighting between the shortest path and the studied measure, the optimal route could proceed equally well south or north of Bornholm. The two optimal points on the curves are defined by the common tangent to the curves. These two points define the gap. In Fig. 10, the mean seasonal cycle averaged over the domain of the average of still-at-sea after 30 days is depicted. The month is determined from the PD98059 in vitro start date of each 30-day period. The result has been tested for significance by dividing the period into two equally long periods and calculating the mean seasonal cycle for each of these periods separately (not shown). We found the same seasonal cycles, including the local minimum in June, for both periods. This result suggests the definition of two seasons, March–September and October–February, henceforth referred to as low- and high-wind seasons. In Fig. 11, the maps of the average of still-at-sea after 30 days are plotted for these two seasons, including optimal

routes. The significance was tested as in the previous paragraph; only differences in details occur, while the overall pattern remains. In Fig. 12, the average of still-at-sea after 30 days for the low-wind and high-wind seasons are depicted. The maximum is clearly located at different

positions for the two seasons. However, a test of significance in which the data are divided into two equally long periods and the results of each are plotted (not shown) demonstrates that the seasonal maxima are not well-defined. The maximum for the low-wind season is still south of the maximum for the high-wind season in both data sets. However, the positions of the maxima in the two data sets differ as much as between the seasons, and the overall shape of the graphs reveals more similarities within each period than within each season. Ovsienko (2002) calculated the risk for a coastal Astemizole hit within 1, 3, 5 and 10 days for releases at 31 different positions during different seasons using the oil spill model OSMS. Of the 31 positions, 21 are located in the Baltic proper, including the entrance area, but one of those is outside the domain investigated in this paper, leaving 20 positions for comparison (see Fig. 13). Comparing the results of our model and the OSMS model demonstrates that twice as much time is required for the first coastal hit in our model than in the results from the OSMS model. One explanation for this difference may be that OSMS is an oil spill model that includes many effects that are missing in our method, e.