8 Trajectories and Spreading of Observed and Simulated Drifters in the Baltic Sea
275
(Fig. 8.3a), and to obtain values of mean displacement and absolute and relative
dispersion (Figs. 8.6 and 8.11). Statistical parameters of drift characterizing a substantial part of the Baltic Proper were derived from the SVP drifter data by means
of splitting the time series into basically uncorrelated segments of ∼11 days each.
The resulting statistics was compared to a similar one extracted from the motion of
drifters simulated by a trajectory model starting at the same position and time as the
SVP drifter segments (Fig. 8.1b).
In order to remove inertial oscillations, which were well observed in the drifter
data but not very well in the model results, all drifter segments and the simulated
drifters were filtered using a 14-hour running mean. As the ocean model data were
not available for 2010 and 2011, the drifter segments had to be compared to simulated drifters in the years available, 1962–2004. As such, the motion of a drifter
segment could not be directly compared to any specific simulated drifter. However,
some indications followed from a statistical comparison.
The mean displacement and absolute dispersion were found to be significantly
lower for the simulated than for the observed drifters (Fig. 8.6). This property was
attributed to the model velocities being lower and less variable, as shown by comparing the probability distributions of Lagrangian velocities from drifters to that of
the model trajectories (Fig. 8.7).
Near-surface currents are, to some extent, wind-driven on time scales comparable
to the duration of the drifter segments (Leppäranta and Myrberg 2009). The quality
of simulated near-surface currents thus partly depends on the quality of the wind
forcing, mixed-layer depth, and parameterization. Meier (2002) compared temperature and salinity profiles modelled using the RCO model to observations and found
good agreement in mixed-layer depth. The wind forcing (ERA-40 winds, dynamically downscaled by the RCA model) was corrected using a parameterization of
wind gusts by Höglund et al. (2009) as the wind speeds were found not to be variable
enough. The correction yielded somewhat more realistic frequency distributions of
the wind speeds. However, this does not imply that the wind at a specific point or
time became more realistic. In particular, the root-mean-square errors may very well
have increased with this correction. Furthermore, the study of Höglund et al. (2009)
was limited to the Swedish coastal regions, as no observations over open water were
available. Thus, there is no information about the quality of the wind forcing over
open water, although it is likely to share some of the problems of the coastal winds.
It is thus conceivable that errors in the subsurface currents, to a large part, are due
to errors in the wind field from the RCA model output.
In principle, the discrepancy may reflect the possibility that the years 2010–2011
could have been ‘extreme’ in terms of absolute dispersion. However, Fig. 8.6, suggests that such ‘extreme’ years are quite uncommon, and the likelihood that they
would occur over a two-year period is thus even smaller. The number of SVP drifters
is also relatively small, and the data may be somewhat biased as two of the drifter
pairs stayed together for nearly 20 days, and thus both drifters in the pairs sampled
the same dynamical region. However, this only affects a few segments out of the
total 76 used.
SVP drifters are seldom fully submerged into the water and may thus be affected
by the winds. Langmuir currents, Stokes drift, and waves may also effect the drifters
275
(Fig. 8.3a), and to obtain values of mean displacement and absolute and relative
dispersion (Figs. 8.6 and 8.11). Statistical parameters of drift characterizing a substantial part of the Baltic Proper were derived from the SVP drifter data by means
of splitting the time series into basically uncorrelated segments of ∼11 days each.
The resulting statistics was compared to a similar one extracted from the motion of
drifters simulated by a trajectory model starting at the same position and time as the
SVP drifter segments (Fig. 8.1b).
In order to remove inertial oscillations, which were well observed in the drifter
data but not very well in the model results, all drifter segments and the simulated
drifters were filtered using a 14-hour running mean. As the ocean model data were
not available for 2010 and 2011, the drifter segments had to be compared to simulated drifters in the years available, 1962–2004. As such, the motion of a drifter
segment could not be directly compared to any specific simulated drifter. However,
some indications followed from a statistical comparison.
The mean displacement and absolute dispersion were found to be significantly
lower for the simulated than for the observed drifters (Fig. 8.6). This property was
attributed to the model velocities being lower and less variable, as shown by comparing the probability distributions of Lagrangian velocities from drifters to that of
the model trajectories (Fig. 8.7).
Near-surface currents are, to some extent, wind-driven on time scales comparable
to the duration of the drifter segments (Leppäranta and Myrberg 2009). The quality
of simulated near-surface currents thus partly depends on the quality of the wind
forcing, mixed-layer depth, and parameterization. Meier (2002) compared temperature and salinity profiles modelled using the RCO model to observations and found
good agreement in mixed-layer depth. The wind forcing (ERA-40 winds, dynamically downscaled by the RCA model) was corrected using a parameterization of
wind gusts by Höglund et al. (2009) as the wind speeds were found not to be variable
enough. The correction yielded somewhat more realistic frequency distributions of
the wind speeds. However, this does not imply that the wind at a specific point or
time became more realistic. In particular, the root-mean-square errors may very well
have increased with this correction. Furthermore, the study of Höglund et al. (2009)
was limited to the Swedish coastal regions, as no observations over open water were
available. Thus, there is no information about the quality of the wind forcing over
open water, although it is likely to share some of the problems of the coastal winds.
It is thus conceivable that errors in the subsurface currents, to a large part, are due
to errors in the wind field from the RCA model output.
In principle, the discrepancy may reflect the possibility that the years 2010–2011
could have been ‘extreme’ in terms of absolute dispersion. However, Fig. 8.6, suggests that such ‘extreme’ years are quite uncommon, and the likelihood that they
would occur over a two-year period is thus even smaller. The number of SVP drifters
is also relatively small, and the data may be somewhat biased as two of the drifter
pairs stayed together for nearly 20 days, and thus both drifters in the pairs sampled
the same dynamical region. However, this only affects a few segments out of the
total 76 used.
SVP drifters are seldom fully submerged into the water and may thus be affected
by the winds. Langmuir currents, Stokes drift, and waves may also effect the drifters