The seasonal-to-decadal (s2d) experiments comprise two sets of simulations: the so-called Stream 1 and Stream 2 simulations. For both streams, coordinated forecast experiments over seasonal, interannual and decadal timescales have been performed. Three different approaches are being pursued to represent model uncertainties: the multi-model approach, the perturbed physical parameter approach and the stochastic physics approach.

Description of the s2d experiments

Stream 1:

  • Seasonal and annual ensemble integrations
    • 2 start dates per year: 1st of May and November
    • 7 month hindcasts for May and 14 for November
    • Hindcast production 1991-2001
    • 9-member ensembles
  • Multi-annual ensemble integrations
    • Two hindcasts started on the first of November 1965 and 1994

Stream 2:

  • Seasonal and annual ensemble integrations
    • 4 start dates per year: 1st of February, May, August and November
    • 7 month hindcasts for every start date, except 14 for those starting in November
    • Hindcast production 1960-2005, using ERA-40 up to August 2002 inclusive and ECMWF operational analyses afterwards
    • 9-member ensembles
  • Multi-annual ensemble integrations (minimum set)
    • One hindcast every 5 years
    • 10 year runs starting on 1st of November 1960, with the next one starting in November 1965 and so on
    • Hindcast production 1960-2005
    • 3-member ensembles
  • Additional integrations (case studies)
    • Co-ordinated experiments to assess predictability at several time scales for specific events, e.g., with and without (observed SSTs) ocean coupling, with and without relaxation in the soil, with and without relaxation to atmospheric analysis in the tropics, etc.
    • Set up as similar to the minimum set as possible to take advantage of the minimum set as control experiment
    • Preliminary selection of cases:
      • European cold winters similar to 2005/06
      • European warm summers similar to 2003
      • Strong ENSO events, e.g. 1997/98
      • Warming trend in the Indian Ocean

Perturbation strategies

The perturbation SST and wind stress perturbation files prepared at ECMWF are in ECFS in the directory ec:/ENSEMBLES/pert. The files use the suffix wsx and wsy for zonal and meridional wind stress and sst for SST. They are in NetCDF format, so the way they are organized and the grid information is easily readable. There are monthly files with daily data for the wind stress perturbations (only one perturbation in each file) and one file per month with four instantaneous SST perturbations.

Documents:
  • Description of how the SST and wind stress perturbations are estimated

 PDF icon ENSEMBLES_docu_perturbations.pdf

  • Discussion of the effect of the new versus old wind stress perturbations and assimilating ocean data in the ECMWF system, and the DEMETER versus ENSEMBLES stream1 ocean analyses are compared.

PDF icon ENSEMBLES_docu_oceananalysis.pdf


Boundary forcings

Greenhouse Gas Impact on Seasonal Experiments (cont.) (18/05/2006)

A paper by Doblas-Reyes et al. describing the impact of a variable GHG concentration on seasonal forecasts has been published in GRL.

S2D Boundary Forcings (17/12/2004)

Six different forcings of some relevance to the s2d integrations have been identified:

  • greenhouse gases (CO2, CH4, N2O, CFC12, CFC11)
  • sulphate aerosol (natural and anthropogenic)
  • ozone (tropospheric and stratospheric)
  • black carbon (industrial and biomass burning sources)
  • solar activity
  • volcanoes

We intend to include these forcings (with both seasonal and interannual variability) in the seasonal, annual and multiannual integrations, taking into account that the volcanic forcing should not be used during the hindcast and that the other forcings would have to be linearly extrapolated.

A list of the datasets suggested for some of these forcings follows.

  • Greenhouse gases
    The data and their documentation, prepared by Jean-Francois Royer, can be accessed here. Note that there is a CFC-11 equivalent concentration of all other minor species minor species of CFCS, PFCs and HCFCs included in the SRES table II.2.10 and table II.2.4. The radiative forcing of all these halogenated compounds have been taken from the WMO/UNEP Scientific Assessment of Ozone Depletion: 2002 table 1-6 page 1.32-1.33. The total of the radiative forcing of all these halogenated gases, except CFC-12, has been converted back into the equivalent CFC-11 concentration (CFC-11*) that gives the same radiative forcing. There is a small CFC-11* preindustrial concentration ( 12.48 pptv) that comes from the fact that there are natural sources of CF4 and evidence that the natural concentration of this gas is about 39 pptv (Harnish et al, 1996. Effect of natural tetrafluoromethane. Nature, 384, p 32).
  • Sulphate aerosols
    If partners are able to include this forcing, the dataset prepared by Olivier Boucher for SRES scenarios is the option suggested. The aerosol files and documentation from Olivier Boucher are now available on the RT2A site. Follow the links "Member's site" and "Forcing data" from the menu in the left frame to get to the bullet "Sulfate aerosol forcing concentration: Olivier Boucher". Two links point to the documentation and to the NetCDF with the aerosol concentration, which are part of the CNRM DODS server. The username and password are available upon request from Jean-François Royer.
  • Ozone
    The ozone fields for the stream 2 simulations for years 1850, 1900, 1950, 1980, 2000, and 2100 for scenarios A1, A2 and B1, are available as monthly means (NetCDF format) on a server at the University of Oslo. For more information on the dataset contact Bjørg Rognerud bjorg.rognerud@geo.uio.no. An alternative ozone dataset is offered by the IPCC.
  • Black carbon
    An option to include black carbon is offered by Nozawa's dataset (not available yet).
  • Solar activity
    A monthly solar irradiance dataset has been prepared by Eigil Kaas from the original Krivova and Solanki's series.
  • Volcanoes
    The influence of volcanic aerosols has not been taken into account in the stream 1 hindcast, except for the Met Office integrations with DePreSys. DePreSys adds up the volcanic aerosols (previously split into four latitude bands) at the hindcast start date damping the values as an exponential decay with a time scale of one year. The data used and an example of their application have been made available by Doug Smith. The initial conditions already contain the impact of previous volcanic eruptions, but it is not clear how long the model is able to keep this information for. Modellers are invited to include the volcanic aerosol effect in the stream 2 hindcasts following the strategy already tested by the Met Office. For reference, a dataset of stratospheric aerosol optical thickness including the volcanic effect can be obtained from NASA.
Greenhouse gas impact on seasonal experiments (24/9/2004)

Results of the experiment carried out at ECMWF to assess the impact on seasonal forecasts of a varying greenhouse gas concentration. The figures show the 2-4 month global average T2m for 2 9-member ensemble experiments carried out with the ECMWF coupled model. The difference between the two experiments consists only the treatment of the greenhouse gas concentration, that is constant (fixed to the 1991 value) in the first plot and varies in the second. As usual, the red dots represent the ERA40 values, the blue ones the ensemble mean and the green box and whiskers the ensemble range. The results are for the May start dates and for the period 1958-1969 and 1991-2001.


Table of experiments

Stream 1 simulations

 
multi-model
model version origin other expid method system expver class stream (daily,monthly) years lead (months) start dates ens members
comments
IFS/HOPE
29R2
ECMWF
(ecmf)
eplo
1
1
1004
en,
rd
mmsf
msmm
1991-
2001
7
May 1
9 IC
-
IFS/HOPE
29R2
ECMWF
(ecmf)
epn9
1
1
1004
en,
rd
mmsf
msmm
1991-
2001
14
Nov 1
9 IC
Data for forecast months beyond 12 are affected by a problem with the albedo and should not be used
GloSea
-
MetOffice
(egrr)
-
1
1
1001
en
mmsf
msmm
1991-
2001
14 (12 for the May start date from 1995-2001)
May 1
Nov 1
9 IC
-
GloSea
-
MetOffice
(egrr)
-
1
1
1004
en
mmaf
mmam
1965 &
1994
120
Nov 1
9 IC
-
DePreSys
-
MetOffice
(egrr)
-
10
51
1503
en
mmaf
mmam
1991-
2001
120
May 1
Nov 1
9 IC
-
ARPEGE4/OPA
-
MeteoFrance
(lfpw)
-
1
0
1001
en
mmsf
msmm
1991-
2001
7
May 1
9 IC
-
ARPEGE4/OPA
-
MeteoFrance
(lfpw)
-
1
0
1001
en
mmsf
msmm
1991-
2001
14
Nov 1
9 IC
-
ARPEGE4/OPA
-
MeteoFrance
(lfpw)
-
1
0
1001
en
mmsf
msmm
1965 &
1994
120
Nov 1
9 IC
-
ECHAM5/OM1
-
IfM/Geomar
(ifmk)
-
10
1
1000
en
mmsf
msmm
1991-
2001
7
May 1
9 IC
-
ECHAM5/OM1
-
IfM/Geomar
(ifmk)
-
10
1
1000
en
mmsf
msmm
1991-
2001
14
Nov 1
9 IC
-
perturbed physical parameters
model
version
origin
other expid
method
system
expver
class
stream (daily,monthly)
years
lead
(months)
start dates
ens members
comments
DePreSys
-
MetOffice
(egrr)
-
11-18
51
1504
en
mmaf
mmam
1991-
2001
120
May 1
Nov 1
8 PP
1 ens member for each method (number=0) ;
the 9th member comes from the IC ensemble member 0 (expver=1503, method=10, see above)
stochastic physics
model
version
origin
other expid
method
system
expver
class
stream (daily,monthly)
years
lead
(months)
start dates
ens members
comments
IFS/HOPE
29R2
ECMWF
(ecmf)
epoh
2
1
1001
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
CASBS1.0
IFS/HOPE
29R2
ECMWF
(ecmf)
eq04
2
1
1001
en,
rd
mmsf,
msmm
1991-
2001
14
Nov 1
9 IC
CASBS1.0

Stream 2 simulations

multi-model
model
version
origin
other expid
method
system
expver
class
stream (daily,monthly)
years
lead
(months)
start dates
ens members
comments
IFS/HOPE
31R1
ECMWF
(ecmf)
od
1
1
2001
en
mmsf,
msmm
1960-
2006
7
Feb 1
May 1
Aug 1
9 IC
Same as the operational System 3; note that the MARS variables 201 and 202 appear with the table version 2 numbers 128 and 190; data for the table 190 are the only ones that should be used
IFS/HOPE
31R1
ECMWF
(ecmf)
od
1
1
2001
en
mmsf,
msmm
1960-
2006
14
Nov 1
9 IC
"
IFS/HOPE
33R1
ECMWF
(ecmf)
f5bg/f5de
1
1
2005
en
mmaf,
mmam
1960-
2005 every 5 yrs
120
Nov 1
3 IC
 
ARPEGE4/OPA
ARP4.6
MeteoFrance
(lfpw)
-
1
0
2002
en
mmsf
msmm
1960-
2005
7
Feb 1
May 1
Aug 1
9 IC
31 atm. levels, ocean i.c. from CERFACS (3D-VAR)
ARPEGE4/OPA
ARP4.6
MeteoFrance
(lfpw)
-
1
0
2002
en
mmsf
msmm
1960-
2005
14
Nov 1
9 IC
"
HadGEM2
-
MetOffice
(egrr)
-
1
1
2025
en
mmsf
msmm
1960-2005
7
Feb 1
May 1
Aug 1
9 IC
 
HadGEM2
-
MetOffice
(egrr)
-
1
1
2025
en
mmsf
msmm
1960-2005
14
Nov 1
9 IC
 
HadGEM2
-
MetOffice
(egrr)
-
1
1
2026
en
mmsf
msmm
1960-2005 every 5 years
120
Nov 1
3 IC
 
ECHAM5/OPA8.2
-
INGV
(ingv)
-
0
1
2001
en
mmsf
msmm
1960-
2005
7
Feb 1
May 1
Aug 1
Nov 1
9 IC
 
ECHAM5/OM1
-
Kiel
(ifmk)
-
10
1
2001
en
mmsf
msmm
1960-
2005
7
Feb 1
May 1
Aug 1
9 IC
1) total cloud cover is a 6-hourly average (not instantaneous) - the average of the 6hrs before 00GMT have been archived
2) snow depth on ice and glaciers is set to zero
ECHAM5/OM1
-
Kiel
(ifmk)
-
10
1
2001
en
mmsf
msmm
1960-
2005
14
Nov 1
9 IC
"
perturbed physical parameters
model
version
origin
other expid
method
system
expver
class
stream (daily,monthly)
years
lead
(months)
start dates
ens members
comments
DePreSys
-
MetOffice
(egrr)
-
10-18
51
2501
en
mmaf
mmam
1960-
2005
120
(360 every 10 yr starting in 1965)
Nov 1
9 PP

NoAssim runs;
1 ens member for each method (all number=0);
Please note: all daily fluxes except for precipitation are corrupted

DePreSys
-
MetOffice
(egrr)
-
10-18
51
2502
en
mmaf
mmam
1960-
2007
7
Feb 1
May 1
Aug 1
9 PP

1 ens member for each method (all number=0); initial conditions from long-term climate run;
Please note: all daily fluxes except for precipitation are corrupted

DePreSys
-
MetOffice
(egrr)
-
10-18
51
2502
en
mmaf
mmam
1960-
2005
120
(360 every 10 yr starting in 1965)
Nov 1
9 PP

1 ens member for each method (all number=0); initial conditions from long-term climate run:
Please note: all daily fluxes except for precipitation are corrupted

DePreSys
-
MetOffice
(egrr)
-
10-18
51
2502
en
mmaf
mmam
2006-
2007
14
Nov 1
9 PP

1 ens member for each method (all number=0); initial conditions from long-term climate run;
Please note: all daily fluxes except for precipitation are corrupted

Others

model
version
origin
other expid
method
system
expver
class
stream
years
lead
(months)
start dates
ens members
comments
IFS/HOPE
31R1
ECMWF
(ecmf)
eul4
1
1
1006
en,
rd

mmaf,
mmam,
mmsf

1965
120
Nov 1
9 IC
test run, precip not correct yet
IFS/HOPE
31R1
ECMWF
(ecmf)
eun2
1
1
1006
en,
rd

mmaf,
mmam,
mmsf

1988
120
Nov 1
9 IC
test run, precip correct
IFS/HOPE
31R1
ECMWF
(ecmf)
eujm
1
1
1006
en,
rd

mmaf,
mmam,
mmsf

1994
120
Nov 1
9 IC
test run, precip not correct yet
IFS/HOPE
32R1
ECMWF
(ecmf)
evrf
1
1
1022
en,
rd

mmaf,
mmam

1994
120
Nov 1
3 IC
test run
IFS/HOPE
32R1
ECMWF
(ecmf)
evye
1
1
1021
en,
rd

mmaf,
mmam

1994
120
Nov 1
3 IC
test run, prescribed SSTs
IFS/HOPE
32R1
_for
_32R3
ECMWF
(ecmf)
ewbm
1
1
1020
en,
rd

mmaf,
mmam

1965
120
Nov 1
3 IC
test run, PBL and convection mods
IFS/HOPE
32R1
_for
_32R3
ECMWF
(ecmf)
ewap
1
1
1020
en,
rd

mmaf,
mmam

1994
120
Nov 1
3 IC
test run, PBL and convection mods
GloSea
-
MetOffice
(egrr)
-
1
1
1002
en
mmsf
msmm
1991-
2001
12
May 1
9 IC
lagged ensembles
GloSea
-
MetOffice
(egrr)
-
1
1
1002
en
mmsf
msmm
1991-
2001
14
Nov 1
9 IC
lagged ensembles
GloSea
-
MetOffice
(egrr)
-
1
1
1003
en
mmsf
msmm
1991-
2001
7
mid April
mid Oct
9 IC
started 15 days earlier
IFS/HOPE
31R1
ECMWF
(ecmf)
ex1a
2
1
1002
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
SPBS1.0 test (p=-5/3, r=0.02)
IFS/HOPE
31R1
ECMWF
(ecmf)
exee
2
1
1030
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
SPBS1.0 test (p=-3, r=0.02)
IFS/HOPE
31R1
ECMWF
(ecmf)
exef
2
1
1031
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
SPBS1.0 test p=-3, r=0.05)
IFS/HOPE
31R1
ECMWF
(ecmf)
exhd
2
1
1032
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
CASBS2.0 test
IFS/HOPE
32R3
ECMWF
(ecmf)
ezj3
1
1
1035
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
control
IFS/HOPE
32R3
ECMWF
(ecmf)
f050
1
1
1037
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
stochVC test (VC_MIN=-0.5, VC_MAX=2.5)
IFS/HOPE
32R3
ECMWF
(ecmf)
f0ak
2
1
1037
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
SPBS2.0+stochVC test (p=-1.27, r=0.05)
(VC_MIN=-0.5, VC_MAX=1.5)
IFS/HOPE
32R3
ECMWF
(ecmf)
f0r8
2
1
1038
en,
rd
mmsf,
msmm
1991-
2001
7
May 1
9 IC
SPBS2.0+stochVC test (p=-1.27, r=0.05)
(VC_MIN=-0.25, VC_MAX=1.0)

Legend:

  • IFS/HOPE: IFS as atmospheric model coupled to HOPE as ocean model used for seasonal forecasts at ECMWF; climatological sea ice. This set-up is identical to the operational System 3.
  • GloSea: Global Seasonal prediction system at the UK Met Office
  • DePreSys: Decadal Prediction System at the UK Met Office: The multi-parameter option with flux correction was used.
  • ARPEGE/OPA4: ARPEGE as atmospheric model coupled to OPA as oceanic model used for seasonal forecasts at MeteoFrance
  • ECHAM5/OM1: ECHAM5 as atmospheric model coupled to OM1 as oceanic model used for seasonal forecasts at Institut fuer Meereskunde/Geomar Kiel
  • ECHAM5/OPA8.2: this is the version of the seasonal forecast model used at INGV in Bologna.
  • version: refers to the version or cycle of the model used
  • origin: MARS directive to identify the centre of origin of the simulation
    • ECMWF = ecmf
    • Met Office = egrr
    • Meteo France = lfpw
    • IfM/Geomar = ifmk
    • INGV = ingv
  • expid: MARS directive for experiment identifier at ECMWF
  • method: MARS directive to identify the method used to produce forecast
  • system: MARS directive to identify the prediction system
  • expver: MARS directive to identify the version of prediction system
  • class: MARS directive to identify the class of the archived data
    • rd = research class
    • en = ENSEMBLES class
  • stream: MARS directive to identify the stream under which the data are archived
  • years: start years of the experiments
  • lead: lead time of the forecasts in months
  • start dates: day and month of the start dates of the experiments
  • ens members: number and type of ensemble members
    • IC: initial condition ensemble
    • PP: perturbed physics ensemble