PPT - Roms

Report
Downscaling of atmospheric
forcing for ROMS for the period
1958-2008 :
Application to the Peru-Chile
system
Gildas Cambon IRD, LEGOS, France
1
ROMS_AGRIF development team:
Cambon, G. (a), Marchesiello, P. (a) ; Debreu L. (b) ; Penven P. (c)
(a): IRD/LEGOS, France ; (b): INRIA/LJK, France ; (c) : IRD/LPO, France
Application Humbolt : PEPS project (ANR)
Gildas Cambon (a) , Boris Dewitte (a,c,d), Katerina Goubanova (a,d), Vincent Echevin
(a), Christophe Hourdin (b), Ali Belmadani (b), Timothée Brochier (b), Colas Francois
(e)
(a): IRD/LEGOS, France ; (b): IRD/LOCEAN, France ; (c): IGP, Peru ; (d) : IMARPE, Peru ;
(e) : IGPP, UCLA, USA
2
ROMS_AGRIF general context
ROMS_AGRIF is a branch of ROMS developed in France by IRD and
INRIA and based at LEGOS, Toulouse.
Its main particularity is the AGRIF online nesting capability (Debreu et
al, 2005, 2010) : http://roms.mpl.ird.fr
This French branch of ROMS is developed to respond to the objectives
of IRD in terms of support to developing countries.
In this community experiment, the model code is developed in parallel
with a powerful pre- and post- processing set of tools: the
ROMSTOOLS matlab toolbox (Penven et al, 2008).
3
ROMS_AGRIF general context
A world-wide community of users :
Today, the ROMS_AGRIF community well spread developed around the
world, with more than 1100 registered users.
A more restricted community of about 40 beta testers exist also and is
referenced on our SVN development website at INRIA:
http://gforge.inria.fr/projects/romsagrif.
Here we present an application of
ROMS_AGRIF for the Peru-Chile
current system conducted in
collaboration with Peruvian
partners.
4
Outlines
1. Introduction & Issues
2. The statistical atmospheric model
3. The interannual ROMS simulation
4. Impact of the stat. downscaling
5. Perspectives
6. Summary & Conclusions
5
Introduction & Issues
Scientific context & main objectives
Humboldt Current system (HSC) is the most productive
eastern boundary current system with permanent
upwelling.
Its fisheries represent about 20% of world wide marine fish
catch.
There is a need for evaluating the impact of regional
climate variability on the Peru-Chile upwelling system
In this context:
 Use of global model simulations of the recent past (1958-present)
 ROMS is use for oceanic downscaling
 Downscaling methods for atmospheric forcing are needed.
6
Introduction & Issues
Strategy
• Interannual regional configuration of ROMS_AGRIF at 1/6°
from 1958 to 2008 period.
• For the moment, hind casts over a test period 1992-2005 to
evaluate the realism of the solution and 98 El Nino event.
• Set up of atmospheric downscaling methods
Atmopheric downscaling strategy :
AOGCM model needed for these long periods : NCEP, ERA …but large errors
are found in coastal areas.
 Dynamical downscaling using WRF : experienced in our group in New
Caledonia (Lefevre et al., 2010, Jourdain et al., 2001) and over the Central
Chile region (Renault et al 2009, see poster)
 Statistical downscaling approach, (Goubanova et al , 2010, Clim.
Dyn., revised)
In this study, statistical approach have been choose in reason of his low
7
computational cost but others are also evaluating (not presented)
Outlines
1. Introduction & Issues
2. The statistical atmospheric model
3. The interannual ROMS simulation
4. Impact of the stat. downscaling
5. Perspectives
6. Summary & Conclusions
8
Statistical downscaling of NCEP reanalysis
Description
10m wind speed
from NCEP
10m wind speed
from QuikSCAT
1. Find a statistical relationship F between
large-scale predictors X and regional
variables Y for the observed period
QSCAT
NCEP
Y2000

F
(
X
2008
2000 2008 )  
F – Multiple linear regression
Y – 10 m wind/stress from QSCAT, 0.5x0.5°
X – 10 m wind + SLP from NCEP, 2.5°x2.5°
2. In order to derive regional variables Y for a
period of interest apply the relationship F
to corresponding large-scale predictors X
QuikSCAT is a good product to
reproduce upwelling processes (Capet
et al, 2004) but we only have it for
recent years (2000 – 2008)
NCEP
F ( X1991
1999 )  Y19911999
Goubanova et al 2010 9
Statistical downscaling of NCEP reanalysis
Improvements on spatial structure
In coastal area with NCEP_DS :
•Wind curl is more realistic, which leads to better estimates of along shore
currents. (PCC closer to the coast)
•Wind stronger near the coast : upwelling intensity intensified and more
realistic
10
Statistical downscaling of NCEP reanalysis
Improvements on interannual variability
satellite
data
products
outputs
Differences in
wind amplitude
between the
warm phase and
the cold phase of
the 1997/1998 El
Niño-La Niña.
Data were 6-month
low-pass filtered and
the warm (cold)
phase corresponds
to September 1997
(December 1998).
Stat downscaling induces a better representation of wind interannual
variability near the coast.
11
Statistical downscaling of NCEP reanalysis
Improvements on interannual variability
The statistical downscaling was applied to both the NCEP
(top) and ERA40 (bottom)
• ERA40 and NCEP reanalysis have a very
different pattern near the coast.
The downscaling allows correcting the interannual biases between reanalyses
products.
.
For these reasons, use of these downscaled NCEP wind
stress in an interannual Peru-Chile regional ROMS simulation
12
Outlines
1. Introduction & Issues
2. The statistical atmospheric model
3. The interannual ROMS simulation
4. Impact of the stat. downscaling
5. Perspectives
6. Summary & Conclusions
13
The interannual ROMS experiment : Set-up
Numerical Set-Up
 Ocean model : ROMS_AGRIF
http://roms.mpl.ird.fr
 Resolution : 1/6°
 Test Period : 1992 - 2005
 Bathymetry : Etopo 2
 Atmospheric forcing:
 NCEP heat fluxes (2.5° x 2.5°)
 NCEP_Downscalled wind stress
(0.5°x0.5°)
 Bulk formulation (Fairall et al.) for latent
and sensible heat fluxes
 Lateral forcing : SODA model (Carton &
Giese), 0.5°x0.5°, 5-days T, S, u, v, U2d, V2d, 
imposed at the open boundaries
 OBC : active open boundaries conditions
(Marchesiello et al, 2001)
14
The interannual ROMS experiment : Validations
NECC
SEC
POC
Oceanic circulation
scheme for the eastern
Tropical Pacific (figure
from Montes et al., 2010,
JGR)
PCC
PCC
Annual mean SST and
Surface currents ROMS
Surface currents in agreement
with the obs. and literature :
PCC, SEC and POC
Annual mean SST from
Pathfinder satellite
observation
Mean SST well reproduced
15
The interannual ROMS experiment : Validations
Currents vertical structures across Eq. and across 12.S
86° W Zonal velocity in m/s
12°S Meridional velocity in m/s
Coastal jet
SEC
PCC
NECC
PCUC
pSSCC EUC
sSSCC
U
• Surface equatorial currents SEC and
NECC ok
• Equatorwards subsurface currents
EUC, p-SSCC, s-SSCC: good
amplitude and position.
V
• Along coast poleward current PCUC :
main source for coastal upwelling (Peru–
North Chile)
• Quite well reproduced, consistent with
previous studies description.
This simulation compared well to the obs. It reproduce well the mean state
circulation and thermohaline structure
16
Outlines
1. Introduction & Issues
2. The statistical atmospheric model
3. The interannual ROMS simulation
4. Impact of the atmospheric downscaling
5. Perspectives
6. Conclusions
17
The interannual ROMS experiment : Stat. downscaling impact
On EKE from SSH and SSH variability
TOPEX/ERS
ROMS NCEP_DS
ROMS – NCEP
Irrealistic
offshore
shift of
EKE max
NCEP_DS
NCEP
Annual mean EKE from SSH in cm2/s2
ROMS with NCEP_DS wind stress compared very well to the observations
• Mesoscale variability well represented in NCEP_DS run.
• Mean current baroclinic instability much better represented in NCEP_DS run
• Significative improves compared to the run using original NCEP wind stress
18
The interannual ROMS experiment: Stat. downscaling impact
On the upwelling intensity
SST Pathfinder
With NCEP_DS stress :
Near coast wind are more realistic and stronger  upwelling better represented.
The statistical downscaling methodology employed to downscale the
NCEP wind stresses in the region clearly improves the ROMS simulation
near the coast : upwelling and eddy activity !
19
Outlines
1. Introduction & Issues
2. The statistical atmospheric model
3. The interannual ROMS simulation
4. Impact of the stat. downscaling
5. Perspectives
6. Summary & Conclusions
20
Perspectives
With our statistical downscaling approach, we improve the realism of the
simulation: investigation of the vertical along shore currents structure and its
reverse during El Nino events (as complementary works of Colas et al, 2008)
During El Nino event :
 Collapse of the upwelling
 Strong modifications of the subsurface poleward PCUC, intensified, uplifted to
the surface and moved offshore extensions
 Intensifications of currents and quasi-reverse of currents ay coast
21
Perspectives
• Dynamical atmospheric downscaling using WRF :
Experienced successfully in our group with ROMS_AGRIF and WRF in
Nouméa by Marchesiello’s group (Lefevre et al, Jourdain et al).
 Applied for shorter periods by
Renault et al over the Central
Chile region :to reproduce the
specific wind burst event called
‘Coastal Jet’ and his associated
oceanic response. (98-2000)
Atmospheric
model: WRF
two nested grids :
30km and 10km
Oceanic model:
ROMS : spatial
resolution : 10km
Renault et al., 2010, in preparation and
see poster
Recover in a dynamically
consistent way the energy
spectra of atmospheric
circulation
at
the
mesoscale range (absent
in the NCEP data)
Figure from
Marchesiello et
22
al
Perspectives
• Ocean downscaling using online AGRIF 2 -way procedure capability of
ROMS_AGRIF (Debreu et al 2010, submitted) to improve the ocean downscaling
performance : there is an update on 2D and 3D time stepping for the parent
grid.
Example of 4 grid nested simulation with the 2-way AGRIF procedure (coarser
grid not shown): Menkes C. & Marchesiello P., Debreu L., 2010
Left : vertical velocity, right : vorticity
23
Conclusions (1)
 Interannual ROMS_AGRIF simulation of the Peru-Chile region using the
SODA OGCM interannual boundary conditions and outputs and
adapted NCEP atmospheric forcing.
 Statistical downscaling based on the QuikSCAT (2000 -2008) periods
applied on the NCEP wind-stress (soon the heat fluxes) to improve near
shore winds over long periods.
 Important improvements for the interannual simulation near
the coast, on the PCC equatorwards current drive by the wind
stress curl.
 Large improve on the upwelling structure and intensity due to
stronger and more realistic wind intensity .
 Flexible method and an efficient alternative to the costly
dynamical downscaling that improves the results.
24
Observational programs related
Filamentos, VOCALS-Rex.
Filamentos and VOCALS-Rex
coastal cruises (2008)
Tracks Filamentos and
VOCALS Peru cruises
Long-term cross-shelf sections
Coastal stations (SST, SSS, Winds)
25
Thank you
Picture: Paracas, South of Pisco, Peru
26
27
28
Conclusions (2)
Perspectives :
 WRF atmospheric modeling for
long periods.
 2-way AGRIF nesting to have a
high quality ocean embedding.
 Investigate of the alongshore
vertical structure of PCUC and
especially its reverses during
El Nino events with this
simulation.
29
The interannual ROMS experiment: Validations
Eq.
12° S
ROMS 20°S
Eq.
Some validations to WOA
WOA
Annual Mean 92-2004
• Maximum biases of 2°C in
surface:
 Roms colder near the
equator :  equatorial.
upwelling
 Warmer southern in
upper layer
 Colder near the coast 
coastal upwelling.
12.S
This simulation compared
quite well to the obs.
20.S
It reproduce well the mean
state circulation and
thermohaline structure
30
The interannual ROMS experiment : Stat. downscaling impact
On EKE from SSH and SSH variability
AVISO
ROMS
ROMS
ROMS with NCEP_DS wind stress
compared well observation:
• Mean state pattern well represented.
•As well, significative impact of the
statistical downscaling of NCEP
stresses, especially near the equator
NCEP_DS
NCEP
Annual mean EKE from SSH in cm2/s2
SSH variability in cm
AVISO
ROMS
ROMS
ROMS with NCEP_DS wind stress
compared well observation:
• Mesoscale variability well
represented in NCEP_DS run.
• Significative improves compared to
the run the original NCEP wind
stress
NCEP_DS
NCEP
NCEP
31
Scientific perspectives
Scientific perspectives:
 Evaluation of the pathways from the EUC towards the Peru-Chile
undercurrent (PCUC) at interannual time scale: as complementary works of
Montes et al., JGR, 2010
 Impact of equatorial Kelvin waves on the PCUC during the 1997 El Niño as
complementary works of Colas et al, 2008, PIO.
Equatorial Kelvin wave along the
equator in SODA (baroclinic modes 1
and 2) (in cm)
Mature El
Nino,
Dec 97
EK m=1,2
June 97
Starting El Nino, EK m=1
Out El Nino
Mars 97
32
The interannual ROMS experiment : El Nino signature
On the SST
@12S
P2
El Nino 97-98
 The upwelled water ~[1518°C] near shore( upwelling of
PCUC waters mainly)
P1
Relax
 Offshore ~[21-25°C]
 During El Nino, shift of + 5°C
near shore and 2-3°C offshore
 Different phases in El Nino (Colas et al, 2008) : P1, Relax, P2
 Can be linked to the different modes of Kelvin waves arrivals
33
Scientific perspectives
 Evaluation of the pathways from the EUC towards the Peru-Chile
undercurrent (PCUC) at interannual time scale: as complementary works of
Montes et al., JGR, 2010
~60% of EUC alimented by subsurface equatorial
flows (EUC, p-SSCC & s-SSCC): Lagrangian
studies did for climatological runs.
Interest to take in account the interannual
variability of large scale currents, in particular
the ENSO modulation.
 Contribution of the different vertical mode of equatorial Kelvin waves on the
PCUC during the 1997 El Niño as complementary works of Colas et al, 2008,
PIO.
34
Outlines
1. Introduction & Issues
2. The statistical atmospheric model

Description
 Improvements
3. The interannual ROMS simulation
 Numerical set up
 Validations
 Impact of the stat. downscaling
 El Nino signature
4. Scientific perspectives
5. Modeling perspectives
6. Conclusions
35

similar documents