358
T. Soomere
that the sensitivity of the results with respect to small changes in the environmental
criteria may largely vary in different parts of the gulf (Soomere et al. 2011c).
10.7.5 Limitations of Modelling
The construction of different maps using the presented technology and specifically
studies of its reliability and associated uncertainties are computationally expensive
and time-consuming. It is thus natural to address potential simplifications of the calculations involving a minimum loss of accuracy but retaining the reliability of the
results. A generic way of reducing the computational efforts is to decrease the resolution of the underlying hydrodynamic model. Doing so is only feasible until the
model still adequately reproduces the mesoscale dynamics (cf. Albretsen and Røed
2010). As computational costs increase rapidly when increasing the 3D model resolution, such a reduction is still highly desired for the practical use of the technique.
A natural limit is that the model should be at least eddy-permitting. This is a severe condition in the test areas of the Baltic Sea where the baroclinic Rossby radius
is usually about 2–4 km, and may be even down to 0.5–1 km at specific locations
(Fennel et al. 1991; Osi´ nski et al. 2010; Nekrasov and Lebedeva 2002).
Experience from numerical physical oceanography suggests that although the instantaneous simulated currents are quite different for different resolutions and even
the statistics of currents may exhibit substantial changes for different model resolutions (Albretsen and Røed 2010), some other fields such as salinity or temperature
can be reasonably replicated using models that poorly resolve the mesoscale dynamics. Moreover, these fields may remain practically the same at different resolutions
(Andrejev et al. 2010; Myrberg et al. 2010b). For example, the appearance of simulated currents in the Gulf of Finland may change abruptly when the resolution is
increased from 0.5 to 0.25 nm but the salinity and temperature fields are almost the
same as for a resolution of 1 nm (Andrejev et al. 2010).
The overall location of the lowest probabilities of coastal hits in the meridional
cross-sections generally corresponds to the location of the equiprobability lines in
the western and central parts of the gulf in 2 nm simulations that ignored the presence of islands. The increase in the resolution from 2 nm to 1 nm and/or the more
exact representation of islands leads to several obvious differences in the results for
the eastern part of the gulf, especially between Gogland and Kotka (cf. Fig. 10.7 and
Fig. 10.14). In this context, it is natural to ask whether the above-mentioned maps of
environmental risks (reflecting, in essence, long-term statistics of the current-driven
transport), or at least certain parts of their integral features, belong to the family of
those characteristics that are largely insensitive to changes in the resolution of the
underlying ocean model. A related task is to identify an optimum spatial resolution
for the ocean model for different applications in a particular basin.
T. Soomere
that the sensitivity of the results with respect to small changes in the environmental
criteria may largely vary in different parts of the gulf (Soomere et al. 2011c).
10.7.5 Limitations of Modelling
The construction of different maps using the presented technology and specifically
studies of its reliability and associated uncertainties are computationally expensive
and time-consuming. It is thus natural to address potential simplifications of the calculations involving a minimum loss of accuracy but retaining the reliability of the
results. A generic way of reducing the computational efforts is to decrease the resolution of the underlying hydrodynamic model. Doing so is only feasible until the
model still adequately reproduces the mesoscale dynamics (cf. Albretsen and Røed
2010). As computational costs increase rapidly when increasing the 3D model resolution, such a reduction is still highly desired for the practical use of the technique.
A natural limit is that the model should be at least eddy-permitting. This is a severe condition in the test areas of the Baltic Sea where the baroclinic Rossby radius
is usually about 2–4 km, and may be even down to 0.5–1 km at specific locations
(Fennel et al. 1991; Osi´ nski et al. 2010; Nekrasov and Lebedeva 2002).
Experience from numerical physical oceanography suggests that although the instantaneous simulated currents are quite different for different resolutions and even
the statistics of currents may exhibit substantial changes for different model resolutions (Albretsen and Røed 2010), some other fields such as salinity or temperature
can be reasonably replicated using models that poorly resolve the mesoscale dynamics. Moreover, these fields may remain practically the same at different resolutions
(Andrejev et al. 2010; Myrberg et al. 2010b). For example, the appearance of simulated currents in the Gulf of Finland may change abruptly when the resolution is
increased from 0.5 to 0.25 nm but the salinity and temperature fields are almost the
same as for a resolution of 1 nm (Andrejev et al. 2010).
The overall location of the lowest probabilities of coastal hits in the meridional
cross-sections generally corresponds to the location of the equiprobability lines in
the western and central parts of the gulf in 2 nm simulations that ignored the presence of islands. The increase in the resolution from 2 nm to 1 nm and/or the more
exact representation of islands leads to several obvious differences in the results for
the eastern part of the gulf, especially between Gogland and Kotka (cf. Fig. 10.7 and
Fig. 10.14). In this context, it is natural to ask whether the above-mentioned maps of
environmental risks (reflecting, in essence, long-term statistics of the current-driven
transport), or at least certain parts of their integral features, belong to the family of
those characteristics that are largely insensitive to changes in the resolution of the
underlying ocean model. A related task is to identify an optimum spatial resolution
for the ocean model for different applications in a particular basin.