Word document

advertisement
NACLIM Deliverable D31.9
Setup of coupled model & hindcasts conducted with initial conditions
corr. to ARGO like samplings
Deliverable title
(Full title: Report on the setup of coupled model and hindcasts
conducted with initial conditions corresponding to ARGO-like sampling.
This deliverable is a report on natural variability in long control
integration of couple model and choice of initial conditions)
WP No.
3.1
WP title
Suitability of the ocean observing system components for initialization
Work duration
1)
Lead beneficiary:
GEOMAR
12
15 Sept
Due delivery deadline:
2013
X
R= report
Actual delivery date:
18
October
2013
P= prototype
Nature of the
deliverable
D= demonstrator
O= Other
PU = public
X
Dissemination
level
PP= restricted to other programme participants, including the
Commission services
RE= restricted to a group specified by the consortium, including the
Commission services
CO= confidential, only for members of the consortium, including the
Commission services
1) Work duration = project month
Lead beneficiary:
GEOMAR
Mojib Latif
Other contributing
partners:
GEOMAR
Wonsun Park
GEOMAR
Fritz Krueger
GEOMAR
Vladimir Semenov
Page 1
Index
1. Executive summary ........................................................................................................... 3
2. Project objectives .............................................................................................................. 3
3. Detailed report on the deliverable ...................................................................................... 4
4. References ...................................................................................................................... 11
5. Dissemination and uptake ............................................................................................... 11
5.1 Dissemination ................................................................................................................ 11
5.2 Uptake by the targeted audience .................................................................................... 12
6. The delivery is delayed:  Yes  No ........................................................................... 12
7. Changes made and difficulties encountered, if any.......................................................... 12
8. Efforts for this deliverable ................................................................................................ 12
10. Dissemination activities ................................................................................................. 13
Page 2
1. Executive summary
With the present deliverable, we provide a setup of coupled model and hindcasts conducted
with initial conditions corresponding to ARGO-like samplings.
This deliverable is a report on choice of initial conditions and natural variability contributing in
the North Atlantic Sector.
We have constructed a coupled atmosphere-ocean-sea ice general circulation model (AOGCM),
the Kiel Climate Model (KCM), and performed long control integration. With the newly updated
version, hindcast experiments have been performed with reduced ocean initial conditions in
which temperature and salinity distributions are ideally reduced, and similar to ARGO-like
observations. We have left only certain amount of initial data that are regarded as observation,
and we have filled the missing values with different strategies.
Hindcast results show that the ocean states deviates from the initial conditions rather quickly,
e.g. within a year, in all different strategies for filling the missing values. This indicates that
reducing the initial shock is important and should be considered in the next step for the setup of
the initialization.
Sophisticated object analysis would be helpful to provide better initial condition. Also, probably
more importantly, initial restoring techniques need to be investigated to provide dynamically
consistent initial conditions that may reconcile with reduced temperature and salinity fields.
2. Project objectives
With this deliverable, the project has contributed to the achievement of the following objectives
(see DOW Section B.1.1):
Nr.
1.
2.
3.
4.
5.
6.
7.
8.
Objective
Assessing the predictability and quantifying the uncertainty in
forecasts of the North Atlantic/Arctic Ocean surface state
Assessing the atmospheric predictability related to the North
Atlantic/Arctic Ocean surface state
Monitoring of volume, heat and fresh water transports across key
sections in the North Atlantic
Quantifying the benefit of the different ocean observing system
components for the initialization of decadal climate predictions
Establishing the impact of an Arctic initialization on the forecast
skill in the North Atlantic/European sector
Quantifying the impact of predicted North Atlantic upper ocean
state changes on the oceanic ecosystem
Quantifying the impact of predicted North Atlantic upper ocean
state changes on socioeconomic systems in European urban
societies
Providing recommendations for observational and prediction
Page 3
Yes
No
X
X
X
X
X
X
X
X
systems
Providing recommendations for predictions of the oceanic
ecosystem
10. Disseminating the key results to the climate service community
and relevant endusers/stakeholders
11. Constructing a dataset for sea surface and sea ice surface
temperatures in the Arctic
9.
X
X
X
3. Detailed report on the deliverable
3.1 Objective
The task for this deliverable was 1) to investigate the benefit of the different ocean observing
system components for the initialization of decadal climate prediction systems, 2) to quantify the
impact of the different observing system components in terms of decadal hindcast skill and 3) to
identify the necessary enhancements and potential reductions of the present observing
systems.
3.2 Setup of coupled model
We have constructed a coupled atmosphere-ocean-sea ice general circulation model (AOGCM)
and we have performed long control integration. The model is an updated version of the Kiel
Climate Model (KCM, Park et al., 2009).
Compared to the previous version described in Park et al. (2009), the updated version uses
higher horizontal resolutions of atmosphere. Different cloud parameterization in the atmosphere
is used, and some of ocean and sea ice parameters are different from the previous version, e.g.
advection scheme. The atmosphere component of the KCM is the EHCAM5 (Roeckner et al.,
2006), atmosphere general circulation model (AGCM) that uses on T42 (2.8x2.8) horizontal
resolution with 19 levels in vertical. Ocean and sea ice model is NEMO (Madec, 2008) that runs
on ORCA2-grid configuration with a nominal 2 degrees horizontal resolution with 31 vertical
level. Tropical ocean has finer grids with 0.5 degree in latitude.
The control run has been integrated with the present levels of greenhouse gases (e.g. CO2=348
ppm). It has been initialised with the Levitus climatology of 3-dimensional temperature and
salinity and runs on 1300 years.
3.3 Experiments with reduced initial conditions
Hindcast experiments have been performed with reduced ocean initial conditions. Perfect initial
conditions, i.e. observations are available at all grid points, have been selected after 1000
model years to skip model’s spin-up phase. We performed ten runs starting at certain dates of
Page 4
the control run (1st of January every 20 years) with reduced initial conditions of temperature and
salinity. The model temperature and salinity can be regarded as ARGO-like observations.
Reducing initial condition first left temperature and salinity only at grid points and filled the rest
missing points with different strategies as described in the next section.
Figure 1 shows some examples of distribution of horizontal known and unknown data points.
The missing values are given also in all layers of the ocean model.
Figure 1: Examples of setup of potential availability of the observed data. 'O' stands for the model grids where
observations are available, 'I' for missing point where data should be provided by using adjacent observation with
relevant Interpolation.
3.4 Interpolation
Several interpolation algorithms were considered to fill the values at missing points. Linear
interpolation has been applied when more observed values than missing ones were available as
illustrated in Fig. 1a. For instance, a linear interpolation was performed at the missing points of
in x-direction by using adjacent values, and then in y-direction. If the values v1 and v 3 at grid
points x1 and x3 are known, the value v 2 at point x2 is obtained by the interpolation,



v2 
| x 3  x 2 | v1  | x 2  x1 | v 2
.
 | x 3  x1 |

Page 5


In the case more known values than missing values were available, as in Figure 1b, an inverse
distance weighted interpolation has been used. For example, if value v at grid point x is
unknown, and the 4 neighbouring known values v1,v2,v3,v4 at grid points x1,x2,x3,x4 are used
 |x x| v
v 
 |x x|
4
in

1
i1
4
i1
i


i
1
i
In the following this interpolation strategy is called INTER.

3.5 Climatology data
In this rather simple strategy, we can also take climatology data to fill the missing points instead
of using interpolation. We calculated monthly mean climatology in the previous 10 years from
the control simulation and used this climatology to provide the initial conditions at the points
where observations were not available. This may provide ocean initial state at decadal time
scales. This strategy is called here MIX.
3.6 Averaging climatology and interpolation value
The missing values were calculated by averaging linear interpolation and the climatology, called
here MIXINTER
vav  (vin  vcl )/2 .
3.7 Smoothing using a Gaussian filter
Since the changed data is not very smooth especially for the last two variants, we considered

also an additional smoothing using a Gaussian filter.
v ga (i, j,z)   gk gl v ik, j l,z
k,l
,
where v denotes the already changed variable by one of the above strategies and g denotes
the 1d-Gaussian filter of length
7 with standard deviation 1:


gk  c exp( k 2 /2),
1
k  3,K ,3, c  ( exp( (k 2  l 2 ) /2) 
k,l
.
Note that the coastal regions are not changed. Additionally, the known values of the original
data arealso changed. In the following we indicated this smoothing variant as SMOOTHED.
Page 6
Figure 2 shows examples of the SST difference of the initial conditions compared to the control
initial state (perfect) case. MIX strategy shows quite unsmooth result, while SMOOTHED seems
to be even smoother than the original data
Figure 2: Initial SST differences to control run for strategies (a) MIX, (b) INTER, (c) MIXINTER and (d) SMOOTHED.
We next checked the global conservations of the reduced initial conditions. Figure 3 shows the
global mean values of sea surface temperature and salinity with different interpolation strategy.
Interpolated field always underestimate the global means, which is not fully understood at this
stage. It is clear that our interpolation schemes underestimate the integral over a concave
function, meaning that the second deviation is always negative. Zonal mean of temperature and
salinity have a "concave-like" shape in latitudinal direction, and we speculate that this may
provide the source of the underestimation. Following the estimation of the trapezoidal rule one
may expect
 v v
new

b  a 2 ''
h v ()
12
,

Page 7
where a, b the boundaries of the calculation region, h the grid size (in an equidistant grid) and
 a value in [a, b]. This systematic underestimation becomes higher when more missing values
are applied.


 It is also found that the smoothing leads to an
underestimation of the mean. The MIX strategy is
 between the climatology and the original value, the MIXINTER, as expected, always
always
between MIX and INTER.
Figure 3: Global mean of (a) sea surface temperature and (b) sea surface salinity from initial conditions obtained from
different interpolation strategies. black: control, blue: INTER, green: MIX, red: MIXINTER. For comparison, the
corresponding mean of
Page 8
Hindcast experiments were performed by starting the KCM with reduced initial conditions.
Figure 4 shows the ensemble mean of the mean absolute error compare to the control run with
different interpolation strategies. All figures correspond for the resolution shown in Fig. 1a (half
resolution). There is not a big difference between the different strategies. In particular, SST
shows an initial "shock", meaning that the error is almost saturated after about 3 months. This
shock appears to be independent of the interpolation or filling strategy. We assume that the
other variables in the model are not adapted to the changed temperature and salinity, and thus
dynamically in consistent start occurs in the hindcast experiments.
Figure 4: The ensemble global mean of absolute error reference to control run with several strategies.
Page 9
Figures 5 and 6 show the time in months when the error to the control run is stagnating, i.e. the
error is not increasing strongly for at least after 10 months. For different initial grid resolution
there is not much effect seen. This also supports that reducing initial shock is the next solution
for a better setup of the initialization. We consider further steps. Sophisticated object analysis
can be used to provide better initial condition to fill the missing values. Also initial restoring
techniques can be used to provide dynamically consistent initial conditions that can be
reconciled with reduced temperature and salinity fields.
Figure 5: Time (months) when the difference of SST to one control run is stagnating, i.e. for at least ten months not
increasing anymore for the MIX strategy cases at different resolutions. (a), (b), and (c) are corresponding to the missing
strategy Fig
Page
10
Figure 6: Time (months) when the difference of SSS to one control run is stagnating, i.e. for at least ten months not
increasing anymore for the MIX strategy cases at different resolutions. (a), (b), and (c) are corresponding to the missing
strategy Fig
4. References
Madec, G. (2008), NEMO ocean engine, Note du Pole de modélisation 27, Institut Pierre-Simon
Laplace, France.
Park, W., N. Keenlyside, M. Latif, A. Stroeh, R. Redler, E. Roeckner, and G. Madec (2009),
Tropical Pacific Climate and Its Response to Global Warming in the Kiel Climate Model, J.
Climate, 22(1), 71–92, doi:10.1175/2008JCLI2261.1.
Roeckner, E., and M.-P.-I. F. Meteorologie (2003), The atmospheric general circulation model
ECHAM5: Part 1: Model description, Report Number 349, Max-Planck-Institut für Meteorologie.
5. Dissemination and uptake
5.1 Dissemination
Peer reviewed articles: None yet.
Publications in preparation OR submitted: None yet.
Page
11
5.2 Uptake by the targeted audience
According to the DOW, the audience for this deliverable is:
The general public (PU)
X
The project partners, including the Commission services (PP)
A group specified by the consortium, including the Commission services (RE)
This reports is confidential, only for members of the consortium, including the Commission
services (CO)
How are you going to ensure the uptake of the deliverables by the targeted audience?
The deliverable has been already circulated to the other partners of the NACLIM consortium. It
will be also made available in the intranet.
Results will be published in peer-reviewed journals, which will provide information mainly to the
observational oceanography and climate modelling community.
6. The delivery is delayed:  Yes
 No
7. Changes made and difficulties encountered, if any
None.
8. Efforts for this deliverable
How many person-months have been used up for this deliverable?
Partner
Person-months
GEOMAR
6
6
Period covered
01/11/2012- 18/10/2013
Total
Total estimated effort for this deliverable (DOW) was 6 person-months.
Page
12
10. Dissemination activities
Add the dissemination activities (starting from November 2012) related to this deliverable. Fill in the table below in all its parts.
[3] Indicate here which type of activities from the following list: Publications, conferences, workshops, web, press releases, flyers, articles published
in the popular press, videos, media briefings, presentations, exhibitions, thesis, interviews, films, TV clips, posters, Other.
[4] Indicate here which type of audience: Scientific Community (higher education, Research), Industry, Civil Society, Policy makers, Medias ('multiple
choices' is possible.
Type of
activities[3]
Main leader
Title (+website
reference)
Presentation
GEOMAR
Presentation
GEOMAR
Date
Place
Type of
audience[4]
Mojib Latif
16-19July 2013
(GEOMAR):
Dynamics and
predictability of
the AMOC
US.AMOC/U.K.
RAPID
International
Science
Meeting‘ AMOC
Variability:
Dynamics and
Impacts’
http://naclim.zma
w.de/Disseminati
on.2509.0.html
Baltimore (USA)
Scientific
Community
(higher
education,
Research)
Mojib Latif
(GEOMAR): A
dynamical/statist
Trieste (IT)
Scientific
Community
(higher
Page
13
1-2 October
2013
Size of
audience
60
Countries
addressed
Have you
sent a copy
to Chiara
(project
office) via
mail?
USA,
Europe
Yes
USA,
Europe
Yes
ical approach to
predict
multidecadal
AMOC variability
and related
North Atlantic
SST anomalies
http://naclim.zma
w.de/Trieste2013.2609.0.htm
l
Page
14
education,
Research)
Page
15
Download