Forecasting the Drift of
Things in the Ocean
Bruce Hackett,
yvind Breivik, Cecilie Wettre, met.no
Presented at GODAE Summer School, Agelonde, 2004-09-21
Outline
The task
Drift models for
Oil
Ships
Search and Rescue (smaller drifting objects)
Operational services
Geophysical forcing data
User interfacing
Examples
Model evaluation (validation)
Concluding remarks
Norwegian Meteorological Institute met.no
The task
Accidents happen at sea people
fall overboard, ships lose power,
oil is spilled, etc.
These are examples of things
drifting in the sea with
potentially serious consequences
loss of life, maritime safety,
environmental damage.
Most nations have services for
emergency situations:
Search-and-rescue (S&R) services
are ubiquitous (Coast Guards)
Oil spill combatment services in
some countries
Drifting ships and other large
objects tied to S&R services and
Vessel Traffic Services (VTS)
Norwegian Meteorological Institute met.no
Operational services
Emergency response services depend on
quick and reliable access to drift prognoses
response time <30 min
24/7 availability (people & computers)
Critical component for drift forecasting is
access to real-time prognostic forcing data.
hence the close link to operational oceanatmosphere forecasting centers
Norwegian Meteorological Institute met.no
The met.no operational forecast model suite
Atmosphere
20km
10
km
4km
5km
Waves
45km
8km
Atmospheric models run 4x daily
to +48h and +60 h
Wave models run 4x daily to +60
h
Ocean models run daily to +60 h
7 day archive of selected
20km
variables
Ocean
tides
atmospheric
forcing
ice
river runoff
4km
300m
Norwegian Meteorological Institute met.no
Surface drifting objects
Forces acting on a drifting object:
Surface current
Wind
Waves (excitation and damping)
Wind and wave effects are calculated relative
to the current:
Vdrift = Vcurr + Vrel
Wave forces only significant for objects ~50 m
or more
Norwegian Meteorological Institute met.no
Oil drift
Oil spill fate models tend to be complicated
due to complex chemical interactions with
the environment weathering.
Chocolate mousse la Prestige
an emulsion
Norwegian Meteorological Institute met.no
Oil spill processes - weathering
Evaporation removes volatile components
For some oil types (crudes) mass loss can be
considerable, for other types insignificant
Density increases
Emulsification forms water-in-oil mixture
Natural dispersion forms oil-in-water
mixture
Removes oil from the slick (under threshold
concentration)
Norwegian Meteorological Institute met.no
Emulsification vs natural dispersion
Oil spill fate
Daling et al. 2003
Norwegian Meteorological Institute met.no
Oil spill processes
- spreading and advection
Spreading refers to the
motion of the oil fluid as it is
spilled onto the more dense
seawater.
Advection is governed by
geophysical forces (currents,
wind).
Spreading is important in the
initial phase of the spill, but
after some hours weathering
tends to inhibit the fluid
behavior and advective
processes take over.
Effects of weathering on a
spreading / advecting spill
Norwegian Meteorological Institute met.no
Algorithms for oil spill processes
Oil spill processes are generally modeled by
a blend of chemical principles and
experimental data.
Laboratory and field experiments have
focused on the individual processes, but
there is growing awareness of the complex
interplay between all processes.
Complexity is also due to the wide range of
properties in different oil types.
Norwegian Meteorological Institute met.no
Oil spill modelling
Common types of model:
Simple trajectory models (center of mass)
Particle cloud models currently most
popular (our approach)
Polygon slick models under development
Example:
Particle type model at met.no, developed
with SINTEF Applied Chemistry
Norwegian Meteorological Institute met.no
Oil spill modelling OD3D
Prestige spill simulation
Particles seeded
continuously according to
specified flow rate and
duration.
Fixed mass per particle
mass loss effected by
removing particles
Transport and weathering
processes applied on
particle-by-particle basis.
Special seeding module for
deep source (bottom
blowout)
surface
subsurface
stranded
Norwegian Meteorological Institute met.no
Ship drift
Wave forces significant for objects ~50 m or
more, hence wind and wave forces must
balance,
Fwind + Fwave + fform + fwave = 0
Norwegian Meteorological Institute met.no
Ship drift
Fwind and fform are commonly formulated as
Fwind = a (Ah+As) Cd |W10| W10
Fform = w Aw Cd |Vrel| Vrel
Box shapes are a fair approximation to a
tanker hull, box-ship simulations tuned to
full scale experimental data yield a lookup
table.
2D wave spectra from a wave model are used
to estimate the wave excitation and damping
Norwegian Meteorological Institute met.no
Leeway
Objects drift at an
angle to the wind
(divergence angle), and
at a fraction of the
wind speed
The divergence angle
depends on the object
shape, draft,
freeboard, etc
Difficult to model,
relatively simple to
parameterize, hence
empirical methods are
used
Norwegian Meteorological Institute met.no
Empirical leeway data
US Coast Guard has compiled data for 63
classes of S&R objects through extensive
field campaigns, generously made available
to our operational service
GPS/ARGOS antennae
light
6 foot mannequin
Aanderaa Current
Meter
Survival Suit
Norwegian Meteorological Institute met.no
Approximations
Wind speed and
object drift are approximately
Approximations
linearly related
Different objects drift differently
Undrogued life raft
Life raft with drogue
Norwegian Meteorological Institute met.no
Leeway bimodality
Experiments produce two
stable drift directions, left
and right of downwind
In practice, we cannot
know which side will be
selected by the object.
Leeway model must
account for both
possibilities
Wind to
Wind to
Leeway Drift
Direction
Leeway Drift
Direction
L = -25
RWD = -135
L = +25
RWD = +135
Wind
Direction
From
Wind
Direction
From
Wind
to
Norwegian Meteorological Institute met.no
Leeway model - assumptions
Monte Carlo ensemble approach:
S&R object represented by O(500) particles, each with the
characteristics of the object
Uncertainties in the model, initial conditions and forcing are
accounted for by seeding strategy and perturbations to the forcing
The changing cloud of particles represents a probability density for
the object location
The leeway of the object is related linearly to the wind speed
and direction
Wave excitaion and damping is ignored as S&R objects are small
Stokes drift is assumed included in the empirical leeway data
from USCG
Norwegian Meteorological Institute met.no
Monte Carlo modelling
Individual
trajectory
Initial particle cloud
represents
uncertainty in Last
Known Position
(LKP)
Search area grows
with time (error
growth) due to
uncertainties in
Wind field
Leeway properties
Split search area
due to bimodal
leeway divergence
Norwegian Meteorological Institute met.no
Operational services user interface
The operational drift model machinery must
be supplemented with a well-designed userinterface:
Input of initial conditions and reception of
results.
Intuitively understood and used response teams
are in a hurry and have many things on their
minds.
tailored graphical rendition of results essential
Robust, rapid communications.
Norwegian Meteorological Institute met.no
Example: Norwegian S&R service
User is JRCC (Joint
Rescue Coordination
Center Norway).
JRCC goal:
to reduce the time and
resources necessary to
find the lost object
Goal of drift forecasting:
to help JRCC reduce the
size of the search area
as much as possible (but
no more).
Norwegian Meteorological Institute met.no
Search maths
POS = POD x POC
POS: Probability of success (do we find
what we are looking for?)
POS: Probability of success (do we find
what we are looking for?)
POC: Probability of containment (are
we searching in the right place?), our
business
Norwegian Meteorological Institute met.no
Distance from shore
Norwegian Meteorological Institute met.no
Example: Norwegian S&R service
Service consists of 2
parts:
Leeway forecast model
run at met.no
SARA graphical analysis
tool run at JRCC
Procedure:
1. JRCC requests Leeway
forecast using web order
form
2. Model is run automatically
at met.no, data file
returned via email
3. Results fed into SARA
Norwegian Meteorological Institute met.no
Example: Norwegian S&R service
1. Leeway request by the
JRCC:
LKP and its uncertainty entered as 2
time/position data
Object class: Uncertainty
of the object type is
tackled by making several
requests for similar
objects.
Norwegian Meteorological Institute met.no
Example: Norwegian S&R service
2. Return data file
3. Feed into SARA
Display on map
layers digital
navigation charts
Animation
Add search areas
manually
Simulation start life raft
Norwegian Meteorological Institute met.no
Example: Norwegian S&R service
2. Return data file
3. Feed into SARA
Display on map
layers digital
navigation charts
Animation
Add search areas
manually
Simulation start life raft
Norwegian Meteorological Institute met.no
Example: Norwegian S&R service
2. Return data file
3. Feed into SARA
Display on map
layers digital
navigation charts
Animation
Add search areas
manually
Simulation stop life raft
Norwegian Meteorological Institute met.no
Example: Norwegian S&R service
2. Return data file
3. Feed into SARA
Display on map
layers digital
navigation charts
Animation
Add search areas
manually
Simulation stop life raft & sailboat
Norwegian Meteorological Institute met.no
Example: A liferaft in Skagerrak during six days
Norwegian Meteorological Institute met.no
Liferaft release position
Manual search areas
Ensemble search area (10nm x 10nm)
Liferaft pickup +16h
Faroe exercise May 2004
Norwegian Meteorological Institute met.no
Icelandic-Norwegian liferaft exercise Feb 2003
Norwegian Meteorological Institute met.no
Alfred-Wegener-Institut, Germany, lost a benthic
lander on 2002-03-14 when it surfaced due to
malfunctioning.
Lost
lander
Its drift was tracked by the
Argos
satellites (white
dots)
The Leeway-model simulated its drift (as a PIW)
over the following four days (red line line is average
position).
Norwegian Meteorological Institute met.no
Concluding remarks
Emergency drift services are very demanding applications for an
operational ocean forecasting system (cf. GMES) challenge is cool
Operational current, wind, and wave prognoses are the backbone of drift
forecasting
Evaulation of the drift models is a constant requirement, although both the
S&R and oil spill services have shown skill in assisting emergency
operations and exercises
Taxonomy field/lab work:
The drift of objects and ships is based on empirical data and the
taxonomy of S&R objects and ship classes must be continuously
updated and expanded
Oil fate models are also dependent on empirical (laboratory) data for
different oil types, long-term weathering effects are not well
understood (cf. Prestige tar balls).
Our three classes of drifting things (people, ship, and oil) are often related
(crew abandons ruptured tanker).
Particle-based drift models are also applicable to other things (e.g.
plankton).
User interface: Good forecast services need good interfacing to the users.
Speed and reliability of delivery are essential.
Norwegian Meteorological Institute met.no
End
Norwegian Meteorological
Meteorologisk Institute
Institutt met.no
