252
J. Kjellsson et al.
8.1 Background
Many studies of the ocean rely on Lagrangian trajectories. Knowledge of the origin and destination of a water particle, as well as the spreading of several initially
closely located particles, is necessary for a number of purposes, for example estimating the fate of oil spills (Soomere et al. 2010) or living organisms (Corell et al.
2012), as well as for planning rescue operations or finding lost items. Trajectories
can be either observed using drifters or floats, or simulated using a computer model
of the ocean and a trajectory algorithm. Model-simulated trajectories may be used
to track entire water masses (Döös 1995; Blanke and Raynaud 1997; Döös et al.
2004), or to map transport and dispersion in the ocean (Pizzigalli et al. 2007). As
these studies become more frequent, the need for evaluating the modelled results
against observations is continuously growing.
In the World Ocean, several studies have used surface drifters or floats to validate model-simulated trajectories. These studies have covered, e.g., the North Atlantic (Garraffo et al. 2001; McClean et al. 2002; Lumpkin et al. 2002), the Pacific
(Garfield et al. 2001), and the global ocean (Döös et al. 2011). Although they have
employed different models, and different sets of Lagrangian observations, a common conclusion is that the squared displacement from the initial position (the absolute dispersion) shows a fair agreement between models and observations. Discrepancies between them are often found when studying the squared distance between
two initially paired drifters (relative dispersion), and/or the variability of the currents
(eddy kinetic energy). Relative dispersion and/or eddy kinetic energy is found too
low 1 in models of resolution 1/4 ◦ –1 ◦ , thus the simulated drifters do not separate as
much as the observed ones (Lumpkin et al. 2002; McClean et al. 2002; Döös et al.
2011).
Discrepancies between models and observations can partly be attributed to the
coarse model resolution, which does not take turbulence on small scales into account, and implies a need for parameterizing subgrid-scale motions (Döös et al.
2011; Griffa et al. 2004). For example, an extensive analysis of the performance of
six circulation models was performed for the Gulf of Finland (Myrberg et al. 2010).
Errors in the wind forcing can also explain some of the differences in some models,
e.g., Keevallik and Soomere (2010) highlighted systematic bias between modelled
and measured wind directions and air flow properties in the central part of the Gulf
of Finland.
To be able to implement realistic parameterizations of subgrid turbulence it is
crucial to understand the circulation on small scales. For the Baltic Sea, there have
been several studies using model-simulated trajectories (Döös et al. 2004; Soomere
et al. 2010; Corell et al. 2012), but very little observational Lagrangian data. To
the authors’ knowledge, there has been only one experiment using Surface Velocity
Program (SVP) drifters similar to the ones used in this study (Håkansson and Rahm
1 The notion ‘too low’ is used here to denote the situation when a modelled quantity is systematically smaller than its values estimated from measurements. Similarly, the notion ‘too high’ or ‘too
large’ denotes the case when a modelled quantity systematically exceeds its measured values.
J. Kjellsson et al.
8.1 Background
Many studies of the ocean rely on Lagrangian trajectories. Knowledge of the origin and destination of a water particle, as well as the spreading of several initially
closely located particles, is necessary for a number of purposes, for example estimating the fate of oil spills (Soomere et al. 2010) or living organisms (Corell et al.
2012), as well as for planning rescue operations or finding lost items. Trajectories
can be either observed using drifters or floats, or simulated using a computer model
of the ocean and a trajectory algorithm. Model-simulated trajectories may be used
to track entire water masses (Döös 1995; Blanke and Raynaud 1997; Döös et al.
2004), or to map transport and dispersion in the ocean (Pizzigalli et al. 2007). As
these studies become more frequent, the need for evaluating the modelled results
against observations is continuously growing.
In the World Ocean, several studies have used surface drifters or floats to validate model-simulated trajectories. These studies have covered, e.g., the North Atlantic (Garraffo et al. 2001; McClean et al. 2002; Lumpkin et al. 2002), the Pacific
(Garfield et al. 2001), and the global ocean (Döös et al. 2011). Although they have
employed different models, and different sets of Lagrangian observations, a common conclusion is that the squared displacement from the initial position (the absolute dispersion) shows a fair agreement between models and observations. Discrepancies between them are often found when studying the squared distance between
two initially paired drifters (relative dispersion), and/or the variability of the currents
(eddy kinetic energy). Relative dispersion and/or eddy kinetic energy is found too
low 1 in models of resolution 1/4 ◦ –1 ◦ , thus the simulated drifters do not separate as
much as the observed ones (Lumpkin et al. 2002; McClean et al. 2002; Döös et al.
2011).
Discrepancies between models and observations can partly be attributed to the
coarse model resolution, which does not take turbulence on small scales into account, and implies a need for parameterizing subgrid-scale motions (Döös et al.
2011; Griffa et al. 2004). For example, an extensive analysis of the performance of
six circulation models was performed for the Gulf of Finland (Myrberg et al. 2010).
Errors in the wind forcing can also explain some of the differences in some models,
e.g., Keevallik and Soomere (2010) highlighted systematic bias between modelled
and measured wind directions and air flow properties in the central part of the Gulf
of Finland.
To be able to implement realistic parameterizations of subgrid turbulence it is
crucial to understand the circulation on small scales. For the Baltic Sea, there have
been several studies using model-simulated trajectories (Döös et al. 2004; Soomere
et al. 2010; Corell et al. 2012), but very little observational Lagrangian data. To
the authors’ knowledge, there has been only one experiment using Surface Velocity
Program (SVP) drifters similar to the ones used in this study (Håkansson and Rahm
1 The notion ‘too low’ is used here to denote the situation when a modelled quantity is systematically smaller than its values estimated from measurements. Similarly, the notion ‘too high’ or ‘too
large’ denotes the case when a modelled quantity systematically exceeds its measured values.