Post on 27-Feb-2018
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
1/58
Slide 1
OCEAN WAVE FORECASTING
AT E C M W F
Jean-Raymond Bidlot
Marine Prediction Section
Predictability Division of the Research Department
European Centre for Medium-range Weather Forecasts
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
2/58
Slide 2
Ocean waves:
We are dealing with wind generated waves at the surface of the oceans,
from gentle to rough
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
3/58
Slide 3
Ocean Waves
1.010.0 0.03 3x10-3 2x10-5 1x10-5Frequency (Hz)
Forcing:
wind
earthquakemoon/sun
Restoring:gravity
surface tension Coriolis force
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
4/58
Slide 4
A Wave Record
Individual Waves,
Significant Wave Height, Hs,
Maximum Individual Wave Height, Hmax
Hmax
Hs= H1/3
Individual waves
etc.
Surface elevation time series from platform Draupner in the North Sea
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
5/58
Slide 5
Wave Spectrum
The irregular water surface can be decomposedinto (infinite) number of simple sinusoidal
components with different frequencies(f) and
propagation directions(
).
The distribution of
wave energy
among those
components is
called:
wave spectrum,
F(f,).
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
6/58
Slide 6
OceanWave
Modellin
g
Modern ocean wave prediction systems are based on
statistical description of oceans waves (i.e. ensemble
average of individual waves).
The sea state is described by the two-dimensional wavespectrum F(f,).
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
7/58
Slide 7
Ocean Wave Modelling
The 2-D spectrum follows from the energy balance equation (in its
simplest form: deep water case):
Where the group velocity Vgis derived from the dispersion relationship
which relates frequency and wave number.
dissnling SSSFVt
F
Sin: wind input source term (generation).
Snl: non-linear 4-wave interaction (redistribution).
Sdiss: dissipation term due to whitecapping (dissipation).
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
8/58
Slide 8
Once you know the wave spectrum F, any other seastate parameters can be estimated. For example, the
mean variance of the sea surface elevation due towaves is given by:
The statistical measure for wave height, called thesignificant wave height(Hs):
The term significant wave heightis historical as this valueappeared to be well correlated with visual estimates of wave heightfrom experienced observers. It can be shown to correspond to theaverage 1/3rdhighest waves (H1/3).
Ocean Wave Modelling
dfdfF ),(2
2
4Hs
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
9/58
Slide 9
ECMWF Wave Model Configurations
From 5 N to 90 N and 98 W to 56 E.
11 km grid spacing.
36 frequencies and 36 directions
Forced by 10m neutral wind fields
from the global system.
Data assimilation of altimeter wave
heights (Jason 2).
2 daily forecasts (from 0 & 12 UTC)
extending to day 5, output every
hour.
1) Limited area model (LAW)
Ocean wave Forecasting
t+54 forecast wave heightvalid 31 January 2014, 6 UTC
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
10/58
Slide 10
ECMWF Wave Model Configurations
From 5 N to 90 N and 98 W to 56 E.
11 km grid spacing.
36 frequencies and 36 directions
Forced by 10m neutral wind fields
from the global system.
Data assimilation of altimeter wave
heights (Jason 2).
2 daily forecasts (from 0 & 12 UTC)
extending to day 5, output every
hour.
Imposed surface currents from
TOPAZ4 system.
1) Limited area model (LAW)
Ocean wave Forecasting
TOPAZ4
surface currents
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
11/58
Slide 11
ECMWF Wave Model Configurations
Global from 81 S to 90 N
Coupled to the atmospheric model
(IFS) with feedback of the sea
surface roughness change due to
waves. The Stokes drift is also
returned for the parameterisationin the skin layer model for the daily
cycle of the SST.
The interface between WAM and
the IFS has been generalised to
include air density and gustinesseffects on wave growth and neutral
winds.
Data assimilation of Jason-2
altimeter wave heights.
2) Global models
Atmosphe
ric
model
Wavemodel
neutral wind
wind gustiness
air density
roughness
Stokes drift
Ocean wave Forecasting
Global from 81 S to 90 N
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
12/58
Slide 12
ECMWF Wave Model Configurations
Ensemble forecasts
(EPS)
55 km grid spacing.
30 25 frequencies *.
24 12 directions *.
Coupled to TL639 TL319
model *.
(50+1) (10+5) day forecasts from
0 and 12Z (monthly once a
week).
Coupling to ocean model.
High resolution
28 km grid spacing.
36 frequencies.
36 directions.
Coupled to the TL1279 model.
Analysis every 6 hrs and 10 day
forecasts from 0 and 12 UTC.
*Change in resolutions after 10 daysNB: also in seasonal forecast at lower resolutions
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
13/58
Slide 13
IFS-WAM-NEMO coupled system.
atmospherewaves - oceans
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
14/58
Slide 14
Wave Model Products
The complete description of thesea state is given by the 2-Dspectrum, however, it is a fairlylarge amount of data (e.g. 1296values at each grid point in theglobal model (36x36).
It is therefore reduced tointegrated quantities:
1-D spectrum obtained by
integrating the 2-D
spectrum over all directions
and/or over a frequencyrange.
Wave model
2-Dspectrum
1-Dspectrum
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
15/58
Slide 15
Wave Model Products
When simple numbers arerequired, the followingparameters are available:
The significant wave height
(Hs).
The peak period (period of
the peak of the 1-Dspectrum).
Mean period(s) obtained
from weighted integration
of the 2-D spectrum.
Integrated mean direction.
Few others.
Complete list at: http://www.ecmwf.int/services/archive/d/parameters/order=/table=140/
f
E(f) peak
area under spectrum=
< f>(mean frequency)
fp(peak frequency)
T= 1 /f
2
4Hs
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
16/58
Slide 16
Wave Model Products2-D spectrum are used to specify boundary conditions
for limited area wave model.
Spectraused asboundaryconditionsby Met ireann
windseaswell
total sea
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
17/58
Slide 17
Wave Model Products
Plot of 2-D spectrum can become very busy !
windseaswell
total sea
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
18/58
Slide 18
Wave Model Products
Except if you only look at one location
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
19/58
Slide 19
Wave Model Products
Use simple parameters:
total wave height and mean propagation direction
Wave height and mean direction:
Analysis : 14 February 2009, 00 UTC
10m winds and mean sea level pressure:
Analysis : 14 February 2009, 00 UTC
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
20/58
Slide 20
Wave height and mean direction:
Analysis : 14 February 2009, 00 UTC
Wave Model Products
PEAK PERIOD:
Analysis : 14 February 2009, 00 UTC
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
21/58
Slide 21
Wave Model Products
Situation might be more complicated !
Wave height and mean direction:
Analysis : 15 February 2009, 00 UTC
10m winds and mean sea level pressure:
Analysis : 15 February 2009, 00 UTC
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
22/58
Slide 22
Wave Model Products
Situation might be more complicated:
Wave height and mean direction:
Analysis : 15 February 2009, 00 UTC
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
23/58
Slide 23
Wave Model ProductsA scheme is used to split the global wave fields into waves which are
under the direct influence of the forcing wind, the so-called windseaor
wind waves, and those waves that are no longer bound to the forcing
wind, generally referred to as swell. Period and mean direction are alsodetermined for these split fields.
Wave height and swell mean direction:
Analysis : 15 February 2009, 00 UTC
Wave height and windseamean direction:
Analysis : 15 February 2009, 00 UTC
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
24/58
Slide 24
Wave Model Products
Windsea and swell: opposing sea
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
25/58
Slide 25
Wave Model Products
Windsea and swell: cross sea
swell
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
26/58
Slide 26
Wave model deterministic products on the web
Significant wave height and mean direction
Wave products available by default on the centres web pages:
(Home -> Products -> Forecasts -> Ocean Wave Forecasts :
http://www.ecmwf.int/products/forecasts/wavecharts/index.html#forecasts
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
27/58
Slide 27
Ocean wave forecasts
At the end of December 2013 and beginning of January 2014, the UKand western Europe were battered by large waves:
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
28/58
Slide 28
Ocean wave forecasts
Wave height forecast and wind from 4 January 2014, step 12hours
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
29/58
Slide 29
Ocean wave forecasts
Wave height forecast and wind from 4 January 2014, step 18hours
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
30/58
Slide 30
Ocean wave forecasts
Wave height forecast and wind from 4 January 2014, step24 hours
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
31/58
Slide 31
Ocean wave forecasts
Wave height forecast and wind from 4 January 2014, step 30hours
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
32/58
Slide 32
Ocean wave forecasts
Wave height forecast and wind from 4 January 2014, step 36hours
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
33/58
Slide 33
Ocean wave forecasts
Wave height forecast and wind from 4 January 2014, step 48hours
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
34/58
Slide 34
Ocean wave forecasts
Wave height forecast and wind from 4 January 2014, step 54hours
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
35/58
Slide 35
Ocean wave forecasts
Wave height forecast and wind from 4 January 2014, step 60hours
Nazare, Portugal, Jan 6, 2014
Porthcawl, South Wales, 6 Jan 2014
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
36/58
Slide 36
Long swell forecasts
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
37/58
Slide 37
Long swell forecasts
Wave height and long swell forecast from 4 January 2014, step 24
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
38/58
Slide 38
Long swell forecasts
Wave height and long swell forecast from 4 January 2014, step 48
Ocean wave Forecasting
So far everything has been presented as output from the
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
39/58
Slide 39
0 1 2 3 4 5 6 7 8 9 10
Forecast day
2
4
6
8
10
12
14
Significant wave height (m) at Heidrun
8 m
So far, everything has been presented as output from the
deterministic forecast system. BUT, forecast should actually be
more probabilistic. Nowadays, weather centres rely on ensemble
techniques :
From an ensemble of waveforecasts it is possible to derive
probabilities for certain wave
conditions.
ECMWF Newsletter 95Autumn 2002Significant wave height above 8 m
06 Nov. 2001 12 UTC ECMWF EPS probability forecast t+120
DRAUGEN
HEIDRUNMIKE
55N
60N
65N
25W
25W
20W
20W
15W
15W
10W
10W
5W
5W
0
0
5E
5E
10E
10E 15E
15E
20E
20E
25E
25E
Surface: signific ant wave height probability >8
Tuesday 6 November 2001 12UTC ECMWF EPS Probabili ty Forecast t+120 VT: Sunday 11 November 2001 12UTC
0
5
10
15
20
25
30
35
40
45
50
54.47
50%
25%
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
40/58
Slide 40
probability for set thresholds (4m)
Basic EPS Wave Model Products
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
41/58
Slide 41
probability for set thresholds (6m)
Basic EPS Wave Model Products
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
42/58
Slide 42
probability for set thresholds (8m)
Basic EPS Wave Model Products
Ocean wave Forecasting
A bit t W EPS
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
43/58
Slide 43
A bit more compact: Wave EPSgram:
Isles of Scilly, west of Cornwall
Like normal EPSgram but for
wind direction, wind speed,
significant wave height,
mean wave direction and
mean period.
Ocean wave Forecasting
A bit t W EPS
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
44/58
Slide 44
A bit more compact: Wave EPSgram:
Each octant is coloured based on the distribution ofthe significant wave height associated with each
mean direction. The coloured areas correspond to
the fractional number of ensemble members with
wave height in the range specified by the coloured ruler.
Ocean wave Forecasting
Hs
EFI plots
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
45/58
Slide 45
EFI plots
From the new model climate, it is possible to derive indices that
indicate deviations in probabilistic terms from what is expected.
Extreme Forecast Index (EFI): 1 means that all EPS are above climate.
Ocean wave Forecasting
Since June 2012 : new set of EFI plots
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
46/58
Slide 46
Since June 2012 : new set of EFI plots
From the new model climate, it is possible to derive indices that
indicate deviations in probabilistic terms from what is expected.
Extreme Forecast Index (EFI): -1 means that all EPS are below climate.
Ocean wave Forecasting
We are not always dealing with nice
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
47/58
Slide 47
We are not always dealing with nice
predictable waves:
Ocean wave Forecasting
Individual Waves,
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
48/58
Slide 48
Individual Waves,
Significant Wave Height, Hs,
Maximum Individual Wave Height, Hmax, and
Freak Wave
Hmax Hs= H1/3
Individual waves
etc.
If Hmax> 2.2 Hs freak wave event
Ocean wave Forecasting
Wave Model Products: Extreme Waves
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
49/58
Slide 49
Wave Model Products: Extreme Waves
We have recently introduced a new parameter to estimate the height
of the highest individual wave (Hmax) one can expect. Its value can be
derived from the 2d wave spectrum:
March 3, 2010, 15UTC
Forecasts fields from Friday 2 March, 2010, 0 UTC
Hs Expected Hmaxin 3 hour records
See ECMWF Tech Memo 288 for derivation and discussionhttp://www.ecmwf.int/publications/library/do/references/list/14
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
50/58
Slide 50
Verification:
What Do You Compare Against ?
Using in-situ data (from buoys and platforms).
Scarce, only few hundreds worldwide
Using satellite data (altimeters, SAR).
Assimilated in the model
Using the model analysis.
For forecast verification only
Comparison to other centres.
Ocean wave Forecasting
C f f
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
51/58
Slide 51
See also the Wave Forecast Verification Project maintained on behalfof the Expert Team on Waves and Storm Surges of the WMO-IOC
Joint Technical Commission for Oceanography and Marine
Meteorology (JCOMM)
Continued general improvement of model forecasts
http://www.ecmwf.int/products/forecasts/wavecharts/index.html#verifications
For example: ECMWF forecast wave height against buoy measurements:
analysis
1 day FC
3 day FC5 day FC
7 day FC
In-situ wave observationsR.M.S.E.
2001 2013
Note: GTS wave data now also available forKorea, Australia, Italy and Greece.
http://www.jcomm.info/index.php?option=com_content&task=view&id=131&Itemid=37
Ocean wave Forecasting
Future developments: spectral partitioning
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
52/58
Slide 52
New decomposition:
Future developments: spectral partitioning
windsea
Swell 1
Swell 2
Swell
windsea
total
Operational:
Ocean wave Forecasting
Future developments: unstructured grid
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
53/58
Slide 53
Future developments: unstructured grid
Ocean wave Forecasting
Future developments: unstructured grid
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
54/58
Slide 54
Future developments: unstructured grid
Ocean wave Forecasting
Future developments: unstructured grid
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
55/58
Slide 55
Future developments: unstructured grid
Ocean wave Forecasting
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
56/58
Slide 56
Questions/comments ?
Ocean wave Forecasting
IFS-WAM-NEMO coupled system.
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
57/58
Slide 57
IFS WAM NEMO coupled system.
Ocean wave Forecasting
Ocean Wave Modelling: references
7/25/2019 Ocean Wave Forecasting at Ecmwf Version 201402
58/58
Ocean Wave Modelling: references
The ocean wave modelling at ECMWF is based on the wave
mode WAM cycle 4(Komen et al. 1994), albeit with frequent
improvements (Janssen 2007: ECMWF Tech. Memo 529, Bidlot2012, proceeding of the ECMWF Workshop on Ocean Waves,
25-27 June 2012 ).
Products from different configurations of WAM are currently
available at ECMWF.
http://www.ecmwf.int/publications/manuals/d/gribapi/param/filter=grib1/order=par
amId/order_type=asc/p=1/table=140/
Wave model page on the Centres web site:
http://www.ecmwf.int/products/forecasts/wavecharts/index.html#forecasts
General documentation:
http://www.ecmwf.int/research/ifsdocs/CY38r1/index.html