4 Studying the Baltic Sea Circulation with Eulerian Tracers
121
sonable amount of computer power. At the same time it is crucial to sample the oil
spill with a large enough number of particles to properly represent the complete oil
spill at later stages when it is more spread out. If too few particles are released, there
will be a risk of missing a potential coastal hit.
With an Eulerian approach the complete field is covered without the need of
initial choices. On the other hand, small-scale processes are missed if a fixed grid
size is used. It would be possible to use an unstructured grid with higher resolution
close to the initial release or even an adaptive grid. To our knowledge no model
for oil spills uses such an advanced method today. However, there are examples of
tracking oil spills where Eulerian principles have been used. Two examples using
purely Eulerian methods and one example with a hybrid approach are presented
below to illustrate the potential of the Eulerian approach.
4.4.1 Ensemble Approach
One use of Eulerian tracers to model oil spills was demonstrated by Höglund and
Meier (2012). They simulated a large number of oil spills in different seasons and in
different weather conditions in the Baltic Proper. From the ensemble of simulations,
maps of several statistical measures of the fate of the oil spills were drawn showing
how severe consequences are that one may expect from an oil spill at different locations (Fig. 4.7). Further, Höglund and Meier (2012) calculated maritime routes that
minimize the risks of an accidental oil spill demanding that both the chosen measure
and the length of the ship track are as small as possible. Examples of routes with
different weights for distance relative to the measure are shown in Fig. 4.8.
The oil spill simulations were done with a 2D Eulerian tracer advected with
the currents in the uppermost model layer of the ocean circulation model. This
approach considers both advection and diffusion of the oil slick and is similar to
the studies described in Chaps. 9–11. Explicitly prescribed horizontal diffusion parameterizes those subgrid-scale dispersive processes that are not resolved by the
ocean circulation model and that are not directly wind dependent, e.g., turbulent
motions.
The initial concentration of the tracer was set to zero except for one grid point
representing the location of the original release. Per definition, all grid points with
land next to them were perfect sinks of the tracer (just setting the concentration to
zero) to simulate that the oil sticks to the coast. As there were only sinks and no
sources, the total content of the tracer must be decreasing (not necessarily strictly
decreasing). The output of each simulation was a time series containing the evolution of the total remaining tracer content in time.
The use of ensembles to cover different seasons as well as many different weather
conditions implies that a fairly large number of simulations must be done to get good
statistics. This obviously requires considerable computer resources and forces some
other simplifying approximations like the restriction to the surface layer (see below).
However, the ensemble approach has the advantage to compensate non-systematic
121
sonable amount of computer power. At the same time it is crucial to sample the oil
spill with a large enough number of particles to properly represent the complete oil
spill at later stages when it is more spread out. If too few particles are released, there
will be a risk of missing a potential coastal hit.
With an Eulerian approach the complete field is covered without the need of
initial choices. On the other hand, small-scale processes are missed if a fixed grid
size is used. It would be possible to use an unstructured grid with higher resolution
close to the initial release or even an adaptive grid. To our knowledge no model
for oil spills uses such an advanced method today. However, there are examples of
tracking oil spills where Eulerian principles have been used. Two examples using
purely Eulerian methods and one example with a hybrid approach are presented
below to illustrate the potential of the Eulerian approach.
4.4.1 Ensemble Approach
One use of Eulerian tracers to model oil spills was demonstrated by Höglund and
Meier (2012). They simulated a large number of oil spills in different seasons and in
different weather conditions in the Baltic Proper. From the ensemble of simulations,
maps of several statistical measures of the fate of the oil spills were drawn showing
how severe consequences are that one may expect from an oil spill at different locations (Fig. 4.7). Further, Höglund and Meier (2012) calculated maritime routes that
minimize the risks of an accidental oil spill demanding that both the chosen measure
and the length of the ship track are as small as possible. Examples of routes with
different weights for distance relative to the measure are shown in Fig. 4.8.
The oil spill simulations were done with a 2D Eulerian tracer advected with
the currents in the uppermost model layer of the ocean circulation model. This
approach considers both advection and diffusion of the oil slick and is similar to
the studies described in Chaps. 9–11. Explicitly prescribed horizontal diffusion parameterizes those subgrid-scale dispersive processes that are not resolved by the
ocean circulation model and that are not directly wind dependent, e.g., turbulent
motions.
The initial concentration of the tracer was set to zero except for one grid point
representing the location of the original release. Per definition, all grid points with
land next to them were perfect sinks of the tracer (just setting the concentration to
zero) to simulate that the oil sticks to the coast. As there were only sinks and no
sources, the total content of the tracer must be decreasing (not necessarily strictly
decreasing). The output of each simulation was a time series containing the evolution of the total remaining tracer content in time.
The use of ensembles to cover different seasons as well as many different weather
conditions implies that a fairly large number of simulations must be done to get good
statistics. This obviously requires considerable computer resources and forces some
other simplifying approximations like the restriction to the surface layer (see below).
However, the ensemble approach has the advantage to compensate non-systematic