FORECASTING DRIFT
517
scales are smaller than those typically covered by the hydrodynamic
forecast models used to drive oil spill models; these sub-gridscale effects
must be parameterized from the available hydrodynamic data. Advection at scales larger than 1 km can be estimated from the hydrodynamic
model data.
Spreading and advection of oil are also strongly influenced by the
release conditions. So far, we have implicitly assumed a spill on the
surface, or near enough that the oil rises to the surface unaltered. Deep,
underwater releases may result in quite different initial surface slick conditions. The oil may be transformed during its rise to the surface and
it will be advected by subsurface currents. In the case of a wellhead
blowout, the oil may be accompanied by gas, which forms gas hydrates
in contact with the water. Some oil spill model systems include special
initialization modules for deep sources (Johansen, 1998, Wettre et al.,
2001).
3.3
Model implementation
Representing an oil slick on the surface of the ocean and accounting for its motion, deformation and tendency to break up into smaller
slicks, not to mention the changes due to weathering described above,
is a formidable task. Over the years, several concepts have been introduced, ranging from simple center-of-mass trajectories to complex
polygon representations. Perhaps the most popular type of model today
treats the oil as a collection of discrete particles, each representing a certain mass of oil. The model currently in use at met.no is of this type,
and we will explore it as an example. The main advantages of the particle representation are that it is inherently Lagrangian, it is amenable to
a probabalistic approach and it reduces to a series of independent particle integrations. The disadvantage is that the results don’t look much
like an oil slick.
The met.no model OD3D has been developed together with SINTEF
Applied Chemistry, who develops the weathering algorithms. An oil spill
is described by a position, time of start, duration, spill rate, oil type,
plus other optional parameters. Given this information, a prescribed
number of particles is added (seeded) every time step (typically 1 hour)
for the spill duration. The mass of each particle is determined by the
flow rate, and remains constant throughout the simulation; mass loss
due to evaporation and dispersion is effectuated by removing particles.
Advective and weathering processes are applied on a particle-by-particle
basis. The model is forced by atmosphere, ocean and wave data from a
selection of sources, ranging from manually entered values to hindcast
517
scales are smaller than those typically covered by the hydrodynamic
forecast models used to drive oil spill models; these sub-gridscale effects
must be parameterized from the available hydrodynamic data. Advection at scales larger than 1 km can be estimated from the hydrodynamic
model data.
Spreading and advection of oil are also strongly influenced by the
release conditions. So far, we have implicitly assumed a spill on the
surface, or near enough that the oil rises to the surface unaltered. Deep,
underwater releases may result in quite different initial surface slick conditions. The oil may be transformed during its rise to the surface and
it will be advected by subsurface currents. In the case of a wellhead
blowout, the oil may be accompanied by gas, which forms gas hydrates
in contact with the water. Some oil spill model systems include special
initialization modules for deep sources (Johansen, 1998, Wettre et al.,
2001).
3.3
Model implementation
Representing an oil slick on the surface of the ocean and accounting for its motion, deformation and tendency to break up into smaller
slicks, not to mention the changes due to weathering described above,
is a formidable task. Over the years, several concepts have been introduced, ranging from simple center-of-mass trajectories to complex
polygon representations. Perhaps the most popular type of model today
treats the oil as a collection of discrete particles, each representing a certain mass of oil. The model currently in use at met.no is of this type,
and we will explore it as an example. The main advantages of the particle representation are that it is inherently Lagrangian, it is amenable to
a probabalistic approach and it reduces to a series of independent particle integrations. The disadvantage is that the results don’t look much
like an oil slick.
The met.no model OD3D has been developed together with SINTEF
Applied Chemistry, who develops the weathering algorithms. An oil spill
is described by a position, time of start, duration, spill rate, oil type,
plus other optional parameters. Given this information, a prescribed
number of particles is added (seeded) every time step (typically 1 hour)
for the spill duration. The mass of each particle is determined by the
flow rate, and remains constant throughout the simulation; mass loss
due to evaporation and dispersion is effectuated by removing particles.
Advective and weathering processes are applied on a particle-by-particle
basis. The model is forced by atmosphere, ocean and wave data from a
selection of sources, ranging from manually entered values to hindcast