8 Trajectories and Spreading of Observed and Simulated Drifters in the Baltic Sea
261
Baltic Proper 3 and the Gulf of Finland have been frozen during some winters in
1962–2004. The presence of ice definitely impacts both the modelled and real velocity fields but in our comparison the related effects only become evident for simulated
drifters (assuming that the RCO model simulated the sea ice correctly). Including
simulated winter drifters in some ice-free years while excluding those from years
of a frozen Baltic Sea would give extra weight to some years in the statistics and
would thereby bias the comparison. For this reason, all segments recorded by the
SVP drifters during the period December 2010–February 2011 were removed from
the data, leaving 76 drifter segments for the comparisons with the modelled data.
The starting longitude and latitude as well as hour, day, and month for each of the
76 drifter segments were used as starting points for simulated drifters in each full
model year (1962–2004). Thus, 36 simulated drifters were released around each
starting point in each model year, resulting in 36 × 76 = 2736 segments for each
model year. Four of the 36 simulated drifters around each starting point originated
in the same grid box as the surface drifter segment, while the others were spread
horizontally in the eight adjacent grid boxes.
As each grid box was about two nautical miles wide, the described process converted each starting point of a SVP drifter segment into a ‘cloud’ (with a radius of
about 3 nautical miles or ∼5.5 km) of model drifters. Doing so made it possible to
take into account the natural subgrid-scale variability of currents around the starting
point of each drifter. The large number of simulated drifters also resulted in clearer
statistics. In an additional experiment, the number of simulated drifters was doubled, resulting in no significant difference in the statistical parameters. It was thus
concluded that the set of 36 simulated drifters per each SVP drifter formed a sufficient pool of samples to calculate the necessary statistics. No turbulence or diffusion
parameterization was used at this point. In order to better replicate the conditions of
the motion of SVP drifters, the advection of the simulated ones was calculated using
the horizontal velocities at 12–18 m depth, with no vertical velocity. Moreover, only
simulated drifters that stayed for 256 hours in waters deeper than 18 m were used in
the comparison. Simulated drifter segments from the year 1962 (Fig. 8.1b) covered
similar sea regions as the SVP drifter segments but their separation rate was smaller.
The positions of simulated drifters were stored every hour to have the same temporal resolution as for the SVP drifters. However, the velocity fields from the RCO
model were only available every 6 hours. Hence, variations on time scales shorter
than 12 hours were not resolved by the model, and time scales slightly longer were
poorly resolved. The average velocity power spectrum of all SVP drifter segments
(Fig. 8.5, left panel), shows a peak near the frequency of 2 cycles per day, corresponding to a period of ∼14 hours. This peak obviously reflects the presence of
inertial oscillations (see Chap. 2, Sect. 2.3.4 for details).
While this peak was very pronounced for the SVP drifters, it was also visible
for the pool of all simulated drifters in each model year but in this case appeared
as a double peak. This distortion is most likely an effect of the inertial oscillations
3 We use here the notion Baltic Proper to denote the Eastern, Northern and Western Gotland Basin,
Bornholm Basin and Gda´ nsk Bay (Chap. 2, Fig. 2.1).
261
Baltic Proper 3 and the Gulf of Finland have been frozen during some winters in
1962–2004. The presence of ice definitely impacts both the modelled and real velocity fields but in our comparison the related effects only become evident for simulated
drifters (assuming that the RCO model simulated the sea ice correctly). Including
simulated winter drifters in some ice-free years while excluding those from years
of a frozen Baltic Sea would give extra weight to some years in the statistics and
would thereby bias the comparison. For this reason, all segments recorded by the
SVP drifters during the period December 2010–February 2011 were removed from
the data, leaving 76 drifter segments for the comparisons with the modelled data.
The starting longitude and latitude as well as hour, day, and month for each of the
76 drifter segments were used as starting points for simulated drifters in each full
model year (1962–2004). Thus, 36 simulated drifters were released around each
starting point in each model year, resulting in 36 × 76 = 2736 segments for each
model year. Four of the 36 simulated drifters around each starting point originated
in the same grid box as the surface drifter segment, while the others were spread
horizontally in the eight adjacent grid boxes.
As each grid box was about two nautical miles wide, the described process converted each starting point of a SVP drifter segment into a ‘cloud’ (with a radius of
about 3 nautical miles or ∼5.5 km) of model drifters. Doing so made it possible to
take into account the natural subgrid-scale variability of currents around the starting
point of each drifter. The large number of simulated drifters also resulted in clearer
statistics. In an additional experiment, the number of simulated drifters was doubled, resulting in no significant difference in the statistical parameters. It was thus
concluded that the set of 36 simulated drifters per each SVP drifter formed a sufficient pool of samples to calculate the necessary statistics. No turbulence or diffusion
parameterization was used at this point. In order to better replicate the conditions of
the motion of SVP drifters, the advection of the simulated ones was calculated using
the horizontal velocities at 12–18 m depth, with no vertical velocity. Moreover, only
simulated drifters that stayed for 256 hours in waters deeper than 18 m were used in
the comparison. Simulated drifter segments from the year 1962 (Fig. 8.1b) covered
similar sea regions as the SVP drifter segments but their separation rate was smaller.
The positions of simulated drifters were stored every hour to have the same temporal resolution as for the SVP drifters. However, the velocity fields from the RCO
model were only available every 6 hours. Hence, variations on time scales shorter
than 12 hours were not resolved by the model, and time scales slightly longer were
poorly resolved. The average velocity power spectrum of all SVP drifter segments
(Fig. 8.5, left panel), shows a peak near the frequency of 2 cycles per day, corresponding to a period of ∼14 hours. This peak obviously reflects the presence of
inertial oscillations (see Chap. 2, Sect. 2.3.4 for details).
While this peak was very pronounced for the SVP drifters, it was also visible
for the pool of all simulated drifters in each model year but in this case appeared
as a double peak. This distortion is most likely an effect of the inertial oscillations
3 We use here the notion Baltic Proper to denote the Eastern, Northern and Western Gotland Basin,
Bornholm Basin and Gda´ nsk Bay (Chap. 2, Fig. 2.1).