Research Papers Seasonal Prediction System at …...to seasonal forecast. The atmosphere,...

Research PapersIssue RP0147September 2012

Climate SERvice Division(SERC)

Seasonal Prediction System atCMCC

By Andrea BorrelliCentro Euro-Mediterraneo sui

Cambiamenti Climatici (CMCC),Climate SERvice, Bologna.

[email protected]

Stefano MateriaCentro Euro-Mediterraneo sui


[email protected]

Alessio BellucciCentro Euro-Mediterraneo sui


[email protected]

Andrea AlessandriCentro Euro-Mediterraneo sui


current affiliation ENEA

Silvio GualdiCentro Euro-Mediterraneo sui

Cambiamenti Climatici (CMCC),Climate SERvice, Bologna;

Istituto Nazionale di Geofisica eVulcanologia (INGV), Bologna.

[email protected]

SUMMARY This document describes the structure of the CMCC SeasonalPrediction System (hereafter CMCC-SPS) developed at CMCC, the variousinitialization methods and the experiments performed with the most recentversion. The CMCC-SPS has been used to perform the CliPAS (ClimatePrediction and its Applications to Society) ISO (Intra-Seasonal Oscillation)hindcasts experiments. This report is focused on the current CMCC-SPSversion that includes an atmospheric off-line initialization tool, other than theocean initial state which had been initiailized in the first version. A set ofretrospective forecasts was performed for the period 1989-2010. Everyyear, the model is restarted four times with re-analysis initial conditions, andthen integrated for 6 months. Here we provide some preliminary resultsfrom this forecasting experiment.

02

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i

CMCC Research Papers

Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Model components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Methodologies of initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Verification Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Conclusions and future works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3

04

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i


INTRODUCTION

The basis for dynamical prediction is the as-sumption that large scale, long-lasting anoma-lies in surface fields will impart predictive skillto seasonal forecast. The atmosphere, inter-acting with the slowly varying component ofthe climate system (land surface and ocean)might enhance predictive skill at a regionalscale. Also, atmospheric teleconnections areable to propagate climate variability on monthlyand longer timescales between distant regionsof the globe, and models may use this infor-mation for the purpose of predictability. Thisreport summarizes the structure and the ini-tialization method of the Seasonal PredictionSystem developed at CMCC. The coupledmodel included in the system is an evolutionof the ocean-atmosphere CGCM Scale Inter-action Experiment-Frontier (SINTEX-F; Gualdiet al., 2003a,b; Guilyardi et al., 2003; Luo etal., 2003) [8] [9] [10] [14]. The components ofthe coupled model (Fig. 1) will be describedin section n. 2; details in Scoccimarro et al.,2007 [19]. The initialization methodologies andexperimental set-up is illustrated in section n.3. Results are provided in section n. 4.

MODEL COMPONENTS

The CMCC coupled global circulation model [7]is composed by three parts :

Land-Atmosphere –> ECHAM5-SILVA;

Ocean/Sea-Ice –> OPA-LIM;

Coupler –> OASIS3.

The atmospheric component is constituted bythe atmospheric model ECHAM5 (Roeckner etal., 1996, 2003, 2006) [16] [17] [18], a globalspectral model truncated at zonal wave num-ber 63. The associated Gaussian grid in whichthe physical fields are calculated has 192 points

in the longitudinal direction and 96 points in thelatitudinal direction. Therefore, the horizontalresolution of the atmospheric model is about200 Km.The vertical structure of the model is repre-sented by 19 levels ,which use sigma as thevertical coordinate.ECHAM5 is coupled to the land-surfacescheme SILVA (Alessandri 2006, Alessandri etal., 2007) [1] [3], a dynamic vegetation modelwith the same horizontal resolution.The ocean component is OPA8.2 (Madec et al.,1998) [15], an Ocean Global Circulation Model(OCGM) run on an ORCA2 grid, quasi-isotropetri-polar grid (2 poles in the northern hemi-sphere, one over Canada and the other overSiberia). The horizontal resolution is variable,with a nominal resolution of 1.5◦ in latitude and2◦ in longitude, with an increase to 0.5◦ latitudenear the equator. The vertical structure of themodel is represented by 31 levels with 10-mresolution in the top 100 m. OPA8.2 is cou-pled to the SEA-ICE model LIM (Timmermannet al., 2005) [21] run at the same horizontal res-olution.Fluxes between the atmosphere and the ocean(Scoccimarro et al., 2007) [19] are exchangedthrough the OASIS3 coupler (Valke et al., 2000)[22]. Coupling frequency - No adjustment.

METHODOLOGIES OF INITIALIZATION

Initial conditions for the various components areimposed to the fully coupled CMCC general cir-culation model. The various CMCC-SPS ver-sions differ for the initialized components. (seealso table 1) specifically:

CMCC-SPS1: only the ocean state is ini-tialized;

CMCC-SPS2: both ocean and atmo-sphere are initialized with observed (anal-yses) 3-D fields;

Seasonal Prediction System at CMCC

05

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i

Figure 1:Seasonal Prediction System scheme

CMCC-SPS3: ocean, atmosphere andland surface are initialized;

Here we illustrate and discuss the results ob-tained from CMCC-SPS3 version of the sea-sonal prediction system, in which the atmo-spheric component ECHAM5 is initialized withdata from the ERA Interim reanalysis (Berris-ford et al., 2009) [5]. A detailed discussion ofthe results pbtained from the CMCC-SPS1 ver-sion can be found in Alessandri et al. (2010) [2].In CMCC-SPS2, the multi-level fields (temper-ature, specific humidity, divergence and vortic-ity) are interpolated from the 60 hybrid levels ofERA-Interim to the 19 hybrid levels of ECHAM5through INTERA (Kirchner, 2001) [11]. Surfacepressure, surface geopotential, surface temper-ature, land-sea mask and snow depth form theset of surface initial condition. In order to fit into

the model, the atmospheric initial state neededto be horizontally interpolated from the ERA In-terim resolution (T159) to ECHAM5 resolution(T63). In the following, we describe the way theinterpolation was carried out:

1. SPECTRAL (HIGH RES) –> GRID(HIGHRES)

(a) wind field spectral fields

(b) transform spectral fields

2. GRID(HIGH RES) –> GRID(LOW RES)

(a) Full field interpolation to grid space

(b) humidity correction

3. GRID(HIGH RES) –> TRUNCATION –>GRID(LOW RES)

06

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i


SPS version Atmosphere Ocean Initializationmodel model Atmosphere Ocean

AMIP-type Ensemble of ocean analysisCMCC-SPS1 Echam5-SILVA; OPA8.2; simulations with forced by ERA40/oper.;

T63/L19 2◦/L31 forced SSTs wind stress and SSTperturbations

Ensemble of ocean analysisCMCC-SPS2 Echam5-SILVA; OPA8.2; ERA-Interim forced by ERA40;

T63/L19 2◦/L31 re-analysis wind stress and SSTperturbations

ERA-Interim Ensemble of ocean analysisCMCC-SPS3 Echam5-SILVA; OPA8.2; re-analysis; soil moisture forced by ERA40;

T63/L19 2◦/L31 from satellite data wind stress and SSTperturbations

Table 1Overview of the CMCC Seasonal Prediction Systems

(a) conversion of humidity fields to spec-tral space

i. spectral transform

ii. truncation

iii. back transform

(b) humidity correction

(c) truncate spectral fields

(d) calculate wind field at low resolution

(e) spectral to grid transformation

4. VERTICAL INTERPOLATION

(a) interpolation at ECHAM resolutiongrid

(b) vertical interpolation ERAI (LOWRES) –> ECHAM (LOW RES)

(c) reset log surface pressure

(d) form new spectral fields

(e) calculate vorticity and divergence

(f) humidity field unfiltered

5. FINAL CALCULATION

(a) correction humidity fields

(b) compose SST and land sea mask

(c) correct surface temperature

(d) second part of statistic energycorrection of surface temperaturet(lowestlevel) truncated back trans-form

(e) store the new lowest temperature

(f) interpolate snow

(g) interpolate sst field

6. WRITE INTERPOLATED FIELDS

(a) surface spectral fields

(b) hybrid level spectral fields

(c) surface Gaussian grid fields

(d) hybrid level Gaussian grid fields

The vegetation model SILVA is initialized withsoil moisture and soil temperature data alsotaken from ERA Interim reanalysis.The ocean initialization of the CMCC-SPSs,uses data assimilation products made availableby CMCC-INGV Global Ocean Data Assimila-tion (CIGODAS, Di Pietro and Masina, 2009,Bellucci et al. 2007) [6] [4]. CIGODAS is basedon the assimilation of temperature and salinityprofile and forced by ECMWF operational anal-ysis using an optimal interpolation scheme. In


07

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i

the ocean analyses no observational informa-tion is included about the sea-ice. Therefore,sea-ice initial state is set to a monthly climatol-ogy obtained through a long-lasting simulationof CMCC-SPS.The coupled model was run from 1960 to 2008at a "present climate" state (i.e. the radiativeforcing changes every year according to green-house gases concentrations), and the result-ing sea ice cover climatology (calculated froma 7-day running mean) of the last 20 years wasused. Hence, each start-date contains the sea-ice information as a modeled monthly climatol-ogy.

EXPERIMENTAL SETUP

A set of retrospective forecasts (hindcasts) wasperformed for 22 years (1989 through 2010).Every year, the model is initialized at 4 differ-ent start-dates (Fig. 2) (February 1st, May 1st,August 1st, November 1st) and then integratedfor 6 months. In order to account for the uncer-tainty that characterizes the initial state of thesystem, an ensemble of nine perturbed atmo-spheric initial conditions (ICs) was prepared foreach start date. Specifically, the atmosphericconditions are perturbed, by imposing restartfiles saved every 12 hours during the 4 dayspreceding the start date (see also Fig. 3). Inthis way, we obtain 9 different initial states fromwhich the SPS evolves, producing a probabilitydistribution of the forecast.

VERIFICATION DATA

The predictive skill of the model is assessedcomparing the forecasts with analyses prod-ucts. The ERA-Interim reanalysis (Berrisfordet al. 2009 [5]) is used for verification of theforecasts. The model and the observed (re-analysis) anomalies are defined as deviationsfrom the respective climatology for the period1989-2009.

RESULTS

In this section we briefly present and discussthe performance of the CMCC-SPS2. Whennot differently specified, the maps we show rep-resent the ensemble mean of the nine membersof the forecast. Figure 4 shows the systematicerror for sea surface temperature (SST) andprecipitation. Each plot refers to an average ofthe 3 months following the start date (i.e. for thestart date 1st of February, the March-April-Maymean is shown). Clearly, large biases show upin the polar regions, due to the lack of informa-tion about sea ice. Cold temperatures sistem-atically affect the central equatorial Pacific, aswell as the north eastern Pacific and the northAtlantic. Prediction for equatorial Atlantic aremost of the times warmer than verification.With regard to precipitation, the largest drybiases are shown in equatorial and southernAfrica, and in southeastern Asia. Forecast fornorthern Sahel, Sahara region and Amazonbasin, on the other hand, appear to be sys-tematically wetter than verifications with the ex-ception of the austral summer.To evaluate the skill of the model, we madeuse of the anomaly correlation (ACC) for eachgrid point between the predicted and observedanomalies, calculated as:

ACC =∑M

m=1 x′mo′m√∑Mm=1 x′2m

∑Mm=1 o′2m

(1)

where x′ and o′ are the anomalies of themodel and the observations respectively. Theanomalies are computed with respect to thereference period 1989-2009.Figure (5) shows surface temperature anomalycorrelation at lead time 1 (from 1 to 3 months).The highest skill is found in the Tropics, andin particular over central and eastern tropicalPacific (ENSO region). The ocean-atmospherecoupling in this area makes the climate system

08

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i


predictable up to lead 3 months, while themodel performances degrade almost every-where else as leadtime increases. Figure(6) shows the ACC between predicted andobserved SST in the Nino3.4 region (areabetween 190◦W and 240◦W in longitude andbetween 5◦S and 5◦N in latitude). The ensem-ble mean shows the higher skill compared toany ensemble member.Only a few ocean regions in the northernAtlantic and Pacific shows some predictabilityat lead time 2 and 3 months. In particular,CMCC-SPS shows considerable skill in thecentral sector of northern Atlantic during winter,and along the coasts of Alaska and westernCanada throughout the year.On land, ACC is rather weak, and predictabilityskill degrades consistently as the run pro-ceeds, moving away from the initial condition.Reasonable predictability is detected at highlatitudes during northern winter, thanks tothe snow depth initialization that regulatestemperature anomalies through albedo effect.Also, tropical regions characterized by strongcoupling with the ocean, like the Amazonbasin and part of Australia, associated with theENSO signal, show a moderate predictability.Figure (7) shows the surface temperature RootMean Square Error (RMSE) of the system (2)at lead time 1, that is the season initiating onemonth after the start date (e.g. June-July-August for the start date of May). Discrepancefrom the observations (ERA interim) is verymuch pronounced at high latitudes, where thepresence of a modeled sea ice climatologyaffects sea surface temperatures and landtemperature accordingly.

RMSE =

√√√√ 1M

M∑m=1

(x′m − o′m)2 (2)

The Tropics generally show higher predictabil-

ity with respect to mid and high latitudes, al-though North Atlantic and part of the south-ern Pacific have small RMSE in their respectivecold semesters. Differences to observationsare larger over the continents, since RMSEincreases where the variance is higher. Thisis noticeable in the tropical Pacific, where theerror in the area mainly interested by ENSOis always more pronounced than immediatelynorth and south. The evolution of sea surfacetemperature RMSE in the El Nino 3.4 region(120◦W-170◦W and 5◦S-5◦N) is shown in Fig-ure (8) for the four start dates. The thick blueline represents the ensemble mean of the ninemembers, shown as thinner blue lines. Clearly,the error quickly propagates as the systemmoves away from the initial condition, show-ing the decay of predictability with time. Asexpected, the ensemble mean forecasts havebetter skill than any ensemble member, sincethe average reduces the internal dynamic noisepresent in the individual forecasts (Kirtman andShukla, 2002 [12]). The ACCs (Figure 6) andthe RMSEs (Figure 8) have been computed rel-ative to the ERA Interim SST and persistenceforecasts are made by continuing the ERA In-terim monthly anomaly from the month prior tothe start date of the model forecasts. For ex-ample, SST persistence forecasts for the pe-riod May to September 1991 have been madeby continuing the SSTA found for the observedApril 1991.Another analysis to test the skill of the CMCC-SPS was performed on the indexes Nino3.4 andWAMI. As mentioned above, Nino3.4 is the av-erage sea surface temperature anomaly in theregion bounded by 5◦N to 5◦S, from 170◦Wto 120◦W. WAMI (West African Monsoon In-dex) is computed as the standardized differ-ence between standardized wind modulus at925 hPa and standardized zonal wind at 200hPa in the Sahel (3◦N-13◦N, 20◦W-20◦E). Fig-ure (9) shows the spaghetti plot of Nino34 index


09

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i

for all the start-dates. On figure 9(a-b) the on-set of El Nino 97-98 was predicted quite well.The CMCC-SPS shows systematic high uncer-tainty for neutral episodes, while is confidentfor the strong-one. The correlation of each sin-gle event with observation is high confirmingthe results in figure 6. Figure (10) representsthe WAMI for the start date of May, coveringthe period June-July-August-September. Whilethe monsoon simulated by the CMCC-SPS isgenerally weaker comparing to observations,the correlation is rather high (about 66%) andthe interannual variability is well predicted. Thelarge distance between the ensemble members(shown as blue crosses in figure 10) points outthat predictability of winds in Africa is lower thanpredictability of SSTs in the ENSO region, andthat skill in Sahel is not as high as it is in thetropical Pacific.

CONCLUSIONS AND FUTURE WORKS

The CMCC seasonal prediction system showsgood skill over Tropical ocean, in particular inthe ENSO region (Nino3.4) and the northeast-ern Pacific in general. Other areas character-ized by rather strong predictability are the north-eastern Pacific and, in the winter, the north At-lantic. Over continents, the skill is generallylow except in regions with a strong couplingland surface-oceans, and at high latitudes dur-ing winter, most likely due to the snow coverinitialization from ERA interim reanalysis.The next step will be the full system initializa-tion. In particular, with respect to the versionin current use, land surface and vegetation willbe initialized with an observational data set: inparticular, initial condition of soil moisture andtemperature and leaf area index will be im-posed to the CMCC-SPS. The "soil moisturememory" (Koster et al, 2006 [13]) can be aslong as 6 months/1 year, affecting climate ata seasonal time scale. We expect major en-hancement of the system’s skill, since many

authors have pointed out that small perturba-tion in the land surface condition can propagatevertically and impact regional atmospheric cir-culation (e.g. Steiner et al., 2009 [20]).

10

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i


Figure 2:Experimental set-up of Seasonal Prediction System

Figure 3:The perturbed initial conditions of Seasonal Prediction System for each start-date


11

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i

BIAS Lead time month 1

Figure 4:Systematic error (BIAS) of seasonal mean predicted SST [K] (panels a, c, e, g) and precipitation [mm/day] (panels b, d, f, h) for leadtime month 1 (target months from 2 to 4). SST forecasts with starting dates in (a) 1st February, (c) 1st May, (e) 1st August and (g)

1st November. Precipitation forecasts with starting dates in (b) 1st February, (d) 1st May, (f) 1st August, (h) 1st November. The errorid defined as the differnce between the 1989-2009 climatologies obtained from the forecast ensemble means and from the

ERA-Interim. For SST blue (red) shading indicates values below -0.5 (above 0.5) K. For precipitation blue (red) shading indicatesvalue above 0.5 (below -0.5) mm/day.

12

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i


Figure 5:Ensemble mean forecasts vs ERA-Interim surface air temperature anomalies: point-by-point correlation of lead time month 1 (targetmonths 2-to-4). Forecasts with starting dates (a) 1st February, (b) 1st May, (c) 1st August and (d) 1st November. The grid points inwhich correlations are positve (yellow, orange and red shading) have a good skill. While if correlations are negative (cyan and blue

shading) have a bad skill.

a) b)

c) d)

Figure 6:ACCs between predicted and ERA-Interim SST anomalies over the Nino3.4 region (5◦S-5◦N, 190◦E-240◦E). The 1989-2009

averaged ACCs for the forecasts with starting dates (a) 1st February, (b) 1st May, (c) 1st August, (d) 1st November are plotted as afunction of the leadtime month. Solid thick blue line is for ensmble means, while thin lines show the results for each ensemblemember forecast. The dashed lines stand for the persistence forecasts, obtained by continuing the monthly anomaly observed

during the month prior to the start date of the model forecasts.


13

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i

Figure 7:Ensemble mean forecasts vs ERA-Interim surface air temperature anomalies: point-by-point root mean square error [K] of lead time

month 1 (target months 2-to-4). Forecasts with starting dates (a) 1st February, (b) 1st May, (c) 1st August and (d) 1st November.The grid points in which root mean square errors are near 0-0.5 K (green and light yellow shading) have a good skill. While if root

mean square errors are greater than 1 K (yellow to red shading) have a bad skill.

a) b)

d) c)

Figure 8:RMSEs between predicted and ERA-Interim SST anomalies over the Nino3.4 region (5◦S-5◦N, 190◦E-240◦E). The 1989-2009

averaged RMSEs for the forecasts with starting dates (a) 1st February, (b) 1st May, (c) 1st August, (d) 1st November are plotted asa function of the leadtime month. Solid thick blue line is for ensmble means, while thin lines show the results for each ensemble

member forecast. The dashed black line stand for the persistence forecasts, obtained by continuing the monthly anomaly observedduring the month prior to the start date of the model forecasts, while dashed blue line indicated the ensemble spread. Both

persistence and spread are the upper and lower limits of the forecast error.

14

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i


Figure 9:Spaghetti plot of Nino3.4 index for 1st February start-date (a), 1st May start-date (b), 1st August start-date (c) and for 1st Novemberstart-date (d). Black line stand for HadISST data, while red thin lines represent forecasts. Each forecast is for 6 months integration.

1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 20104

3

2

1

0

1

2

3

YEARS

WAM

I ano

mal

y (m

/s)

JJAS WAMI Anomalies Lon( 20,20) Lat(3,13)

r= 0.66155

OBSCMCC SPS

Figure 10:West African Monsson index (WAMI): the red line is the ERA-Interim index, while blue line is the ensemble mean forecast. Blue

crosses represent forecast ensemble members. WAMI was evaluated for JJAS season on May start date.


15

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i

Bibliography

[1] Alessandri A. Effects of Land Surface andVegetation Processes on the Climate Simu-lated by an Atmospheric General CirculationModel. PhD thesis, Bologna UniversityAlma Mater Studiorum, 2006.

[2] A. Alessandri, A. Borrelli, S. Masina,P. Di Pietro, A. F. Carril, A. Cherchi,S. Gualdi, and A. Navarra. The CMCC-INGV Seasonal Prediction System: Im-proved Ocean Initial Conditions. Mon. Wea.Rev., 138:2930–2952, 2010.

[3] A. Alessandri, S. Gualdi, J. Polcher, andA. Navarra. Effects of land surface-vegetation on the boreal summer surfaceclimate of a GCM. J. Clim., 20:255–277,2007.

[4] A. Bellucci, S. Masina, P. Di Pietro, andA. Navarra. Using temperature–salinity re-lations in a global ocean implementation of

a multivariate data assimilation scheme.Mon. Wea. Rev., 135:3785–3807, 2007.

[5] P. Berrisford, D. Dee, K. Fielding,M. Fuentes, P. Kallberg, S. Kobayashi, andS. Uppala. The ERA-Interim archive. Tech.Rep. ERA Rep. Series 1, ECMWF, 2009.

[6] P. Di Pietro and S. Masina. The CMCC-INGV global ocean data assimilation sys-tem (CIGODAS). Tech. Rep. RP0071,Centro Euro-Mediterraneo per i Cambia-menti Climatici, 2009.

[7] P. G. Fogli, E. Manzini, M. Vichi,A. Alessandri, L. Patara, S. Gualdi,E. Scoccimarro, S. Masina, andA. Navarra. INGV-CMCC Carbon (ICC): Acarbon cycle earth system model. Tech.Rep. 61, Centro Euro-Mediterraneo per iCambiamenti Climatici, 2009.

[8] S. Gualdi, E. Guilyardi, A. Navarra,S. Masina, and P. Delecluse. The in-terannual variability in the tropical indian

16

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i


ocean as simulated by a cgcm. Clim. Dyn.,20:567–582, 2003.

[9] S. Gualdi, A. Navarra, E. Guilyardi, andP. Delecluse. Assessment of the tropi-cal indo-pacific climate in the sintex cgcm.Ann. Geophys., 46:1–26, 2003.

[10] E. Guilyardi, P. Delecluse, S. Gualdi, andA. Navarra. Mechanisms for enso phasechange in a coupled gcm. J. Clim.,16:1141–1158, 2003.

[11] I. Kirchner. The INTERA handbook. Tech-nical report, Max-Planck-Institut fur Mete-orologie, Hamburg, Germany, 2001.

[12] B. P. Kirtman and J. Shukla. Interac-tive coupled ensemble: A new couplingstrategy for CGCMs. Geophys. Res. Lett.,29:1367, 2002.

[13] R. D. Koster and Coauthors. Glace: Theglobal land–atmosphere coupling experi-ment. part i: Overview. J. Hydrometeorol-ogy, 7:590–610, 2006.

[14] J.-J. Luo, S. Masson, S. Becera,P. Delecluse, S. Gualdi, A. Navarra, andT. Yamagata. South pacific origin of thedecadal enso-like variation as simulatedby a coupled gcm. Goephys. Res. Lett.,30(24):2250, 2003.

[15] G. Madec, P. Delecluse, M. Imbard, andC. Levy. OPA version 8.1 ocean generalcirculation model reference manual. Tech.Note 11, LODYC/IPSL, 1998.

[16] E. Roeckner, K. Arpe, L. Bengtsson,M. Christoph, M. Claussen, L. Dumenil,E. Esch, M. Giorgetta, U. Schlese, andU. Schulzweida. The atmospheric generalcirculation model ECHAM-4: Model de-scription and simulation of present-day cli-mate. Tech. Rep. 218, Max-Planck-Institutfur Meteorologie, 1996.

[17] E. Roeckner, G. Bauml, L. Bonaven-tura, R. Brokopf, M. Esch, M. Gior-getta, S. Hagemann, I. Kirchner, L. Korn-blueh, E. Manzini, A. Rhodin, U. Schlese,Schulzweida, and A. Tompkins. Theatmospheric general circulation modelECHAM5. Part I: Model description. Rep.No. 349, Max-Planck-Institut fur Meteo-rologie, Hamburg, Germany, 2003.

[18] E. Roeckner, R. Brokopf, M. Esch,M. Giorgetta, S. Hagemann, L. Korn-blueh, E. Manzini, U. Schlese, andU. Schulzweida. Sensitivity of simulatedclimate to horizontal and vertical resolu-tion in the ECHAM5 atmosphere model. J.Clim., 19:3771–3791, 2006.

[19] E. Scoccimarro, S. Gualdi, A. Bellucci,A. Carril, P. G. Fogli, and A. Navarra.CMCC-SXF025: A high-resolution cou-pled atmosphere ocean general circula-tion climate model. Tech. Rep. 18, Cen-tro Euro-Mediterraneo per i CambiamentiClimatici, 2007.

[20] A. L. Steiner, J. S. Pal, S. A. Rauscher, J. L.Bell, N. S. Diffenbaugh, A. Boone, L. C.Sloan, and F. Giorgi. Land surface cou-pling in regional climate simulations of thewest african monsoon. Clim. Dyn., 98:293–314, 2009.

[21] R. Timmermann, H. Goosse, G. Madec,T. Fichefet, C. Etheb, and V. Duliere. Onthe representation of high latitude pro-cesses in the ORCA-LIM global coupledsea ice ocean model. Technical Report 8,Ocean Modell, 2005.

[22] S. Valke, L. Terray, and A. Piacentini. TheOASIS coupled user guide version 2.4.Technical Report TR 00-10, CERFACS,2000.


17

Cen

tro

Eur

o-M

edite

rran

eosu

iCam

biam

enti

Clim

atic

i

c© Centro Euro-Mediterraneo sui Cambiamenti Climatici 2013

Visit www.cmcc.it for information on our activities and publications.

The Euro-Mediteranean Centre on Climate Change is a Ltd Company with its registered office andadministration in Lecce and local units in Bologna, Venice, Capua, Sassari, Viterbo, Benevento and Milan.The society doesn’t pursue profitable ends and aims to realize and manage the Centre, its promotion, andresearch coordination and different scientific and applied activities in the field of climate change study.

Research Papers Seasonal Prediction System at …...to seasonal forecast. The atmosphere,...

Documents

Transcript of Research Papers Seasonal Prediction System at …...to seasonal forecast. The atmosphere,...