, wind- and wave-driven transports. This discrepancy between the methods can be taken into account by comparing risks for coastal hits within 10 days from Ovsienko (2002) with risks for coastal hits within 20 days from our data (Fig. 14). The risk for a coastal hit is calculated as the number of simulations in which more than 0.1% of the tracer hits a coast after 20 days. Because there are 100 simulations for each location, the number is directly interpreted as the percentage risk of a coastal hit. Because Ovsienko (2002) does not define the time periods for the three seasons (without the ice season) used to model oil spills, we simply averaged the risks for coastal C59 wnt mw hits during

the three seasons and compared the results with our data covering the entire year. Because the winter season is only included in our data, we expect a bias toward a higher risk

of coastal hits due to the seasonal cycle in our data compared to the results of Ovsienko (2002). The averages of the risks are 59.9% (Ovsienko, 2002) and 52.6% (our data), confirming that using day 20 instead of day 10 and including the winter Selleckchem Enzalutamide season has not overcompensated the risk. At the same time, the results from the various oil spill locations are highly correlated. The correlation coefficient between the two data sets is 86%, which seems to be rather high, given that the data are from different years (June 1993–July 1994 vs. June 1961–September 1969) and from different seasons and that the applied methods differ considerably (oil spill model vs. passive tracer following surface currents). However, as expected, both data sets correlate highly with the (negated) distance to the nearest coast (79% respectively 90%). For a better comparison, this effect needs to be removed, which can be done by viewing the data as vectors and projecting them onto the hyper plane orthogonal to the vector representing the distance to the nearest coast. The relevant scalar product for the process is the covariance.

The correlation between the projected vectors is not limited by the original correlation (except if the non-projected vectors are parallel, i.e., correlated +100% or −100%) and can be anything between +100% and −100%. The correlation between the projected data is 57%. However, the correlation is fairly high considering Tau-protein kinase the large differences in the methods outlined above. As expected, the measures are strongly related to the mean currents and their directions. More specifically, if the mean currents are strong, then the local bathymetry has less of an impact, and the bathymetry to where the mean currents are going has a larger impact. This relationship is particularly clear west of Gotland, where the mean currents transport the tracer away from the narrow part between Öland and Gotland out into the open area south of Gotland, while north-west of Gotland, the transport is into the narrow area.