382
J. Murawski and J. Woge Nielsen
model covers large parts of the north-eastern Atlantic with a resolution of 6 nm.
Salinity and temperature boundary conditions were extracted from the monthly climatology. The 4.5-year-long run was initialized with operational archive data in
August 1990. The first 1.5 years constituted a spin-up period, to ensure that the density field was in balance with the velocity field. The following 3 years, 1992–1994,
were used for the oil drift and fate studies.
The oil drift and fate model has been developed for individual spill prediction,
not for systematic studies of metocean forced and topographically steered pollution
propagation. Therefore its advantages in terms of the enhanced physical description
of weathering and transport processes (e.g., the ability to adequately account for the
wind drift, wave-induced vertical mixing and horizontal dispersion) may become a
disadvantage if the goal is the production of a large pool of simulation results as the
total runtime increases almost linearily with the number of simulations. In order to
keep the computational efforts for the identification of characteristic drift patterns
at a reasonable level, we restricted the spatial resolution of the release sites of oil
spills to 528 locations in the Gulf of Finland and the covered timeframe to three
years (1992–1994).
To match the results with the outcome of purely surface-current-driven trajectory
simulations described in Chaps. 9 and 10 (Soomere et al. 2010, 2011c), the length
of each simulation was set to 10 days. The simulations were restarted every 4th day
with the same configuration of oil spills. This sums up to in total 152,064 simulations of single oil spills. The total physical duration of the runs (about 5 minutes
each run at a single node) would have taken 528 days on a serial computer. The success of the presented study was reached owing to the portability of the drift model
code in a supercomputer facility. All in all 22 runs on 2 computational nodes were
executed in parallel to reduce the physical runtime to about 1 month. This was the
first time when an operational drift model was employed to calculate the statistics
for a whole region, the Gulf of Finland.
11.4 Characteristic Drift Patterns
Oil propagation in the marine environment is often non-uniform and follows regional and seasonal patterns that are enforced by the cumulative effects of the currents and winds. The analysis presented in the previous chapters and in Soomere
et al. (2011d) suggests that in the long-term mean, particles often follow representative pathways that determine when and where they, on average, arrive at the coast.
If all sections of the shoreline are considered to be of equal value (Andrejev et al.
2011; Soomere et al. 2011a), the classical question—where and when the oil hits—
might be reverted to where did the earliest (often the fastest) particle came from and
when it started its motion.
J. Murawski and J. Woge Nielsen
model covers large parts of the north-eastern Atlantic with a resolution of 6 nm.
Salinity and temperature boundary conditions were extracted from the monthly climatology. The 4.5-year-long run was initialized with operational archive data in
August 1990. The first 1.5 years constituted a spin-up period, to ensure that the density field was in balance with the velocity field. The following 3 years, 1992–1994,
were used for the oil drift and fate studies.
The oil drift and fate model has been developed for individual spill prediction,
not for systematic studies of metocean forced and topographically steered pollution
propagation. Therefore its advantages in terms of the enhanced physical description
of weathering and transport processes (e.g., the ability to adequately account for the
wind drift, wave-induced vertical mixing and horizontal dispersion) may become a
disadvantage if the goal is the production of a large pool of simulation results as the
total runtime increases almost linearily with the number of simulations. In order to
keep the computational efforts for the identification of characteristic drift patterns
at a reasonable level, we restricted the spatial resolution of the release sites of oil
spills to 528 locations in the Gulf of Finland and the covered timeframe to three
years (1992–1994).
To match the results with the outcome of purely surface-current-driven trajectory
simulations described in Chaps. 9 and 10 (Soomere et al. 2010, 2011c), the length
of each simulation was set to 10 days. The simulations were restarted every 4th day
with the same configuration of oil spills. This sums up to in total 152,064 simulations of single oil spills. The total physical duration of the runs (about 5 minutes
each run at a single node) would have taken 528 days on a serial computer. The success of the presented study was reached owing to the portability of the drift model
code in a supercomputer facility. All in all 22 runs on 2 computational nodes were
executed in parallel to reduce the physical runtime to about 1 month. This was the
first time when an operational drift model was employed to calculate the statistics
for a whole region, the Gulf of Finland.
11.4 Characteristic Drift Patterns
Oil propagation in the marine environment is often non-uniform and follows regional and seasonal patterns that are enforced by the cumulative effects of the currents and winds. The analysis presented in the previous chapters and in Soomere
et al. (2011d) suggests that in the long-term mean, particles often follow representative pathways that determine when and where they, on average, arrive at the coast.
If all sections of the shoreline are considered to be of equal value (Andrejev et al.
2011; Soomere et al. 2011a), the classical question—where and when the oil hits—
might be reverted to where did the earliest (often the fastest) particle came from and
when it started its motion.