300
T. Soomere
flows it is generally necessary to average over time intervals longer than the typical turnover time of eddies in order to get a flavour of the properties of Lagrangian
transport. This scale is about five days in the Gulf of Finland and in the SW Baltic
Sea, and somewhat longer, some 10–15 days, in the Baltic Proper.
9.3.6 Subgrid Processes and Spreading of Trajectories
The impact of subgrid-scale processes inevitably leads to certain deviations of the
simulated trajectories from the measured paths of water particles or tracers. For
example, in the Ligurian Sea, the simulated and measured trajectories are usually
highly correlated only within 1–1.5 days (Vandenbulcke et al. 2009). The situation
is even more complicated for the Gulf of Finland and in the Northern Gotland Basin,
where the High Resolution Oceanographic Model of the Baltic Sea (HIROMB) circulation model (Kõuts et al. 2010) and Seatrack Web (Verjovkina et al. 2010) (basically also relying on the HIROMB engine) demonstrated a reasonable match during
about 10 hours in autumn 2007. A more satisfactory reproduction of pathways of
surface drifters was obtained in the Gulf of Finland (Gästgifvars et al. 2006) using the OAAS model (see Chap. 10 for its description). Trajectories simulated for
longer than about one day will therefore usually not match the motions of real water
particles in the test areas.
Numerical models will probably never be able to replicate the smallest details
of ocean turbulence. Although a certain increase in the accuracy of simulations of
circulation and particular trajectories is viable, e.g., through an improvement of the
spatial resolution or via a decrease in the time step of the circulation model or by
means of an increase in the complexity of the tracking scheme (Gräwe et al. 2012),
exact matches of longer simulated trajectories with real motions of tracers are very
unlikely. The loss of this exact matching is not critical as a large number of different
pathways may lead to equivalent displacements. These displacements and associated
transport patterns can still be extracted using statistical analysis of a large number
of trajectories. For such an analysis to be reliable, it is necessary to adjust the parameters and options listed above so that these features will be adequately captured
and/or highlighted.
A proper representation of statistical features of transport requires the use of
a sufficient number of single trajectories. The number of simultaneously calculated
independent trajectories is implicitly limited by the number of grid cells in the circulation model. The overview in Chap. 7 demonstrates that the version of the TRACMASS trajectory model used in Soomere et al. (2010, 2011a) does not reproduce
the spreading of real water particles or drifters. This does not mean that the TRACMASS model is wrong. It correctly uses all the data from the circulation model but
cannot guess into which direction and how rapidly the real drifter would move under
the impact of factors that are not resolved by the circulation model.
It is natural to assume that the ignoring of subgrid processes does not considerably alter the statistics of net displacement of the tracked water particles if the
T. Soomere
flows it is generally necessary to average over time intervals longer than the typical turnover time of eddies in order to get a flavour of the properties of Lagrangian
transport. This scale is about five days in the Gulf of Finland and in the SW Baltic
Sea, and somewhat longer, some 10–15 days, in the Baltic Proper.
9.3.6 Subgrid Processes and Spreading of Trajectories
The impact of subgrid-scale processes inevitably leads to certain deviations of the
simulated trajectories from the measured paths of water particles or tracers. For
example, in the Ligurian Sea, the simulated and measured trajectories are usually
highly correlated only within 1–1.5 days (Vandenbulcke et al. 2009). The situation
is even more complicated for the Gulf of Finland and in the Northern Gotland Basin,
where the High Resolution Oceanographic Model of the Baltic Sea (HIROMB) circulation model (Kõuts et al. 2010) and Seatrack Web (Verjovkina et al. 2010) (basically also relying on the HIROMB engine) demonstrated a reasonable match during
about 10 hours in autumn 2007. A more satisfactory reproduction of pathways of
surface drifters was obtained in the Gulf of Finland (Gästgifvars et al. 2006) using the OAAS model (see Chap. 10 for its description). Trajectories simulated for
longer than about one day will therefore usually not match the motions of real water
particles in the test areas.
Numerical models will probably never be able to replicate the smallest details
of ocean turbulence. Although a certain increase in the accuracy of simulations of
circulation and particular trajectories is viable, e.g., through an improvement of the
spatial resolution or via a decrease in the time step of the circulation model or by
means of an increase in the complexity of the tracking scheme (Gräwe et al. 2012),
exact matches of longer simulated trajectories with real motions of tracers are very
unlikely. The loss of this exact matching is not critical as a large number of different
pathways may lead to equivalent displacements. These displacements and associated
transport patterns can still be extracted using statistical analysis of a large number
of trajectories. For such an analysis to be reliable, it is necessary to adjust the parameters and options listed above so that these features will be adequately captured
and/or highlighted.
A proper representation of statistical features of transport requires the use of
a sufficient number of single trajectories. The number of simultaneously calculated
independent trajectories is implicitly limited by the number of grid cells in the circulation model. The overview in Chap. 7 demonstrates that the version of the TRACMASS trajectory model used in Soomere et al. (2010, 2011a) does not reproduce
the spreading of real water particles or drifters. This does not mean that the TRACMASS model is wrong. It correctly uses all the data from the circulation model but
cannot guess into which direction and how rapidly the real drifter would move under
the impact of factors that are not resolved by the circulation model.
It is natural to assume that the ignoring of subgrid processes does not considerably alter the statistics of net displacement of the tracked water particles if the