Environmental Engineering Reference
In-Depth Information
Table 5.1
Root Mean Square Error (in km) for circular and elliptical initial distributions of oil spill
particles for simulations similar to Fig.
5.10
RMSE (km)
0m
2
/s
1m
2
/s
10m
2
/s
100m
2
/s
D
=
D
=
D
=
D
=
a
=
b
7.44
7.57
8.25
12.7
a
=
2
b
6.89
7.03
8.04
12.48
a
=
b
/
2
8.09
8.15
8.71
13.07
Fig. 5.10
Trajectories of virtual particles deployed inside a circle (
left
), ellipse oriented N-S
(
middle
) and ellipse oriented W-E (
right
). Simulations are performed for 72h with
Ęł
=
0
.
5%and
10 m
2
with
D
s.
Black cloud
indicates the initial location of particles and the
red dots
the final
distribution. The drifter trajectory is depicted with the
blue line
=
/
5.9 Integrating Probability Within an Oil Spill Model
A disadvantage of the multi-model approach (Sect.
5.6
) is the need for systemic,
real-time uncertainty modelling through all models of the forecasting system. This
problem might be reduced by making use of a computationally inexpensive oil spill
model to rapidly create real-time simulation ensembles that can be visualized as
probability maps. In the simplest incarnation, such an approach is complementary to
the multi-model system, i.e. simply replacing the multiple oil spill models and initial
condition perturbations in Fig.
5.3
with a single ensemble model. However, there is
also the potential for an alternative approach that entirely replaces the multi-model
system. Our principal concern is in the range of possible future positions of the oil spill
and the probabilities associated with this range. If the uncertainties in wind, wave,
and hydrodynamic models (driving forces) can be quantified in terms of their effects
on stochastic diffusivity (particle response) [
29
], then the multi-model system could
be replaced with single instances of best estimate wind, wave, and hydrodynamic
models accompanied by an ensemble uncertainty approach. This approach requires
all the uncertainties in the wind, wave, and currents to be re-parameterized in terms of
diffusivity. Methodologies and metrics for translating the chain of uncertainties from
driving force models to particle models remain relatively unexplored, but are likely
Search WWH ::
Custom Search