Environmental Engineering Reference
In-Depth Information
drifter analyses using hindcasts. For model-model comparisons, a hydrodynamic
model can be run with the smallest practical grid and time step over a number of
hindcast periods to provide a set of reference cases. These reference cases can be
considered the best possible simulations for the available model. Statistical analyses
can be used to evaluate the difference between the reference cases and simulations
at the coarser grid scales and larger time steps that are practical for an operational
system. There remains an open question as to the best approach to incorporate such
data into an effective parameterization of uncertainty. The most obvious possibility
is to use the local velocity variance to scale a random perturbation in the velocities
used for the oil spill Lagrangian transport.
5.8.2 Wind Drag Coefficient
Modelled surface oil spills are directly affected by wind; that is they will move with
velocity vectors slightly different than the modelled water currents at the surface. In
hydrodynamics, the wind causes shear stress at the surface that creates 3D turbulence
and transfers momentum down from the surface into the wind-mixed layer. Hydrody-
namic models focus on getting the net downward transfer of energy and momentum,
and the modelled “surface” velocity actually represents the spatially-averaged veloc-
ity in the surface grid cell whose thickness is typically
O
m, depending
on the model scale. In contrast, the wind effect on surface oil (or a floating object)
does not have a significant downward transfer of momentum and must be directly
included in the floating oil particle transport computation. The effect of wind drag
on the oil velocity (
U
wind
) is typically represented using a drag coefficient (
(
1
)
−
O
(
10
)
ʳ
) and
the wind velocity (
V
wind
) such that
U
wind
=
ʳ
V
wind
. Because the drag is generally
small,
is often reported as a percentage of the wind speed. Using winds measured
at 10m above the water surface,
ʳ
ʳ
in the range
[
0
,
3
.
5%
]
have been recommended
[
3
,
29
]. More recently, [
22
] argued for
0%.
The sensitivity of an oil spill model to the selection of
ʳ
values up to 6
.
ʳ
can be evaluated by
model-model hindcast comparisons and the “best”
for a particular combination of
hydrodynamic and oil spill models can be selected by comparison to drifter exper-
iments. As an example, a surface drifter was deployed during a cruise around the
Balearic Sea in October, 2012. An operational ROMS model is available for the same
time period. Using the oil spill model of [
39
], an initially circular spill is transported
as shown in Fig.
5.7
.
The sensitivity of the results to the
ʳ
can be evaluated by running hindcast simula-
tion similar to Fig.
5.7
for a range of values. Figure
5.8
shows the Root Mean Square
Error (RMSE) of the cloud of particles relative to the drifter position at the end of
72h for simulations with 0
ʳ
≤
ʳ
≤
4%.
N
p
1
N
p
d
n
,
RMSE
=
n
=
1
Search WWH ::
Custom Search