A Hybrid Coupled Model for the Pacif i c Ocean–Atmosphere System.Part I:Description and Basic Performance

ZHANG Rong-Hua

Key Laboratory of Ocean Circulation and Waves,Institute of Oceanology,Chinese Academy of Sciences,Qingdao 266071

1.Introduction

The coupled ocean–atmosphere system in the Pacif i c plays a very important role in global climate variability and change.The ENSO is the largest interannual signal originating from the ocean–atmosphere coupling within the tropical Pacif i c(e.g.,Bjerknes,1969;Zebiak and Cane,1987),which affects weather and climate worldwide.In the subtropical and midlatitude regions,the subtropical gyre and the Kuroshio system are the important ocean circulation systems thatdirectlyaffectregionalandbasin-scaleclimatevariability over the Northern Hemisphere.Additionally,the tropics and extratropics exhibit clear interactions over the Pacif i c sector,which give rise to ENSO variability and the existence of lowfrequencyclimate modes,includingthe Pacif i c decadal oscillation(PDO).

While great progress has been made in understanding the coupled ocean–atmosphere system over the Pacif i c,the mechanisms involved in the modulations of ENSO remain poorly understood.For example,as has been demonstrated(Gu and Philander,1997;Kleeman et al.,1999),ENSO can be modulated by extratropical processes through the changes in the subtropical cell of the ocean(e.g.,McCreary and Lu,1994);an interactively coupled loop has been identif i ed that involves interactions between the atmosphere and ocean,and between the tropics and extratropics over the Pacif i c basin(e.g.,Zhang et al.,1998;Fedorov and Philander,2000;Wang and An,2001).In addition,ENSO can also be modulated by a variety of forcing and feedback processes in the region,including stochastic forcing(SF)of wind,freshwater f l ux(FWF),ocean biology-induced heating(OBH),tropical instability waves(TIWs),and so on.As all these effects are intermingled with one another,it is diff i cult to understand how ENSO is affected by these processes individually and collectively.

Numerical models are powerful tools to describe and understand these processes and their complicated interplay involvingdifferentdynamicalregimes.Inthepast,variouscoupled atmosphere–ocean models with varying levels of complexity have been developed and can be used to investigate the interactions between the tropics and subtropics,and betweendifferentforcingandfeedbackprocesses.Forexample,fully coupled general circulation models(CGCMs)are avail-able for use to represent the interplay between processes affectingthe propertiesof ENSO.However,CGCMs are highly complex and complicated tools,with all components being interactively related with each other,making it diff i cult to reveal the effects of a specif i c process on ENSO.Additionally,fully coupled GCMs are computationally very expensive to run.

Here,weintroduceahybridcoupledmodel(HCM)developed for ENSO-related modeling studies,in which an ocean general circulation model(OGCM)is coupled with a simple atmospheric model for interannual wind stress(τ)variability derivedfrom a singular value decomposition(SVD)analysis.TheOGCMusedinthis workis theGent–CaneOGCM(Gent and Cane,1989),a layer model with an explicitly embedded bulk mixed layer model.As an effort to extend our previous tropics-only coupled ocean–atmospheremodel(Zhang et al.,2006),the HCM developed in this work covers almost the entire Pacif i c basin,allowing for representations of ocean processes in the tropics and subtropics,and their interactions.

There are clear advantages of developing such an HCM for ENSO-related modeling studies.As a dominant interannual signal in the region,ENSO produces SST anomalies,which induce atmospheric wind responses that are quick and almost simultaneous in the tropical Pacif i c,leading to coherent relationships between interannual variations in SST and winds during ENSO cycles.Therefore,their statistical relationship from historical data can be used to construct a feedback model to simply capture wind anomalies as a response to SST variability.Therefore,such an HCM can offer an extremely eff i cient modeling tool for the ocean–atmosphere system in the Pacif i c,allowing a large number of experiments to be performed feasibly and affordably.Yet,as will be demonstrated in this paper in this paper,the HCM constructed over the Pacif i c basin can depict reasonably well the coupled ocean–atmosphereinteractions,and in particular can give rise to a realistic simulation of interannualvariability associated with ENSO.As the model covers almost the entire Pacif i c domain,extratropical processes and their interactions with the tropics are naturally represented in the ocean and atmosphere.Hence,the HCM can be adopted to investigate modulations of ENSO that arise from extratropical effects,which are not possible in tropics-only HCMs(e.g.,Zhang et al.,2006).

In additional to extratropical processes that can exert an inf l uence on ENSO,a variety of forcing and feedback processes exist in the Pacif i c.For example,ENSO also acts to produce large and coherent interannual anomalies of FWF and OBH,which can feed back onto ENSO(e.g.,Zhang and Busalacchi,2009a;Zhang et al.,2009,2012).Similar to the treatment of winds,interannual variability of FWF and OBH can be represented as a response to ENSO-related SST anomalies;their corresponding feedback models can also be constructed from historical data and incorporated into the HCM.In particular,various parameterization schemes previously developedin our tropical model can be directly applied to the Pacif i c-domainmodel.For instance,althoughan ocean biology component is not explicitly included in the ocean physical system of the model,its interannual feedback effect on ENSO can still be represented using an empirical parameterization scheme constructed from satellite ocean color data(Zhang et al.,2011).Also,TIW processes have been demonstrated to be important to the heat budget and SST in the eastern tropical Pacif i c,but their effects are still missing in coarse-resolution climate models.At present,highresolutionsatellite dataareavailableandcanbeusedtodepict TIW-induced wind feedback(Zhang and Busalacchi,2008,2009b);an empirical model constructed from daily satellite data of SST and wind can be adopted and incorporated into the HCM to represent TIW-induced wind effects.Another important factor that has been demonstrated to be able to signif i cantly modulate ENSO is SF of atmospheric winds(KirtmanandSchopf,1998),whichis missingintheHCM.Hence,a simple model has also been constructed to represent this wind forcing component(Zhang et al.,2008)and incorporated into the HCM to represent its effect on ENSO.

Taking all these together into an HCM framework,we offer a computationallyeff i cient modeling tool that can be used to represent and understand the coupled ocean–atmosphere system for almost the entire Pacif i c basin,includingthe interactions betweenthe subtropics and tropics.Several processes important to ENSO modulations are taken into account individually and collectively.Additionally,satellite data are used to represent feedback effects for climate modeling.For instance,a novel method is demonstrated for using satellite ocean color data to parameterize OBH in a physical ocean model,and how high-resolution satellite data can be used to representTIW-scalewind feedbackwithin a large-scalemodeling context.Various applications are expected using this modeling framework.In this paper,examples are given to illustrate the ability of the HCM to depict the mean ocean state and interannual variability associated with ENSO.Further applications related to ENSO modulations by various forcings and feedbacks will be presented in Part II of this study.

The remainder of the paper is organized as follows.Section2describesthemodelsandsomeobservationaldataused.The performance of the HCM is examined using its 50-year simulation for mean climatological oceanic f i elds in section 3,and for interannual variability in section 4.Concluding remarks are given in section 5.

2.Data and models

Based on our previous modeling efforts within the tropical Pacif i c,an HCM is constructed over almost the entire Pacif i c domain to represent the coupled ocean–atmosphere system.Figure 1 is a schematic diagram of the HCM,which consists of an OGCM and a simplif i ed representation of the atmosphere(including three forcing f i elds to the ocean:wind stress(τ),freshwater f l ux,and heat f l ux).The total wind stress(τ)can be written as τ = τclim+τinter+τTIW+τSF,whereτclimis the climatological part,τinteris the interannual part,τSFis the stochastic forcing part,and τTIWis the TIW part;τclimis prescribed from observations and other wind componentsare parameterizedin a statistical way(see details below).

The total FWF,represented by precipitation(P)minus evaporation(E),(P–E),is also separated into its climatological part[(P–E)clim]and interannual anomaly part(FWFinter),written as FWF=(P–E)clim+FWFinter=(P–E)clim+(P–E)inter;the climatological f i eld,(P–E)clim,is prescribed in the FWF calculation.The heat f l ux(HF)is interactively determined using an advectiveatmospheric mixed layer(AML)model(Seager et al.,1995);various climatological f i elds are specif i ed in the heat f l ux calculation(Seager et al.,1995).

Furthermore,some related forcing and feedback processes are also included in the HCM to account for their effects on ENSO.For example,ocean biology-induced heating(OBH)is included in the physical ocean model;its effect on the penetrative radiation is simply represented by the attenuation depth of solar radiation in the upper ocean(Hp).Similarly,the total Hpf i eld is separated into its climatological part()and interannual anomaly part(),written as Hp=+.Thepart is derived from remotely sensed ocean color data,andis estimated using its empirical model representing a response to change in the physical system.In this simplif i ed hybrid coupled modeling system,climatological f i elds(τclim,SSTclim,(P–E)clim,and)are all prescribed as seasonally varying from observations;interannual anomaly f i elds(τinter,(P–E)interand H′p)are diagnostically determined from their corresponding empirical submodels relating to interannual SST variability.Additionally,τTIWis estimated using its empirical model constructed from high-resolution satellite data;τSFis estimated empirically as well.All these are brief l y described in this section below.

2.1.Datasets

Various observational and model-based data are used to construct empirical models for perturbation f i elds(τinter,(P–E)interand),as well as to validate model simulations.Long-term climatological f i elds are prescribed in the HCM,including monthly-mean wind stresses(τclim)from the National Centers for Environmental Prediction–National Center for Atmospheric Research(NCEP–NCAR)reanalysis(Kalnay et al.,1996),precipitation from Xie and Arkin(1995),solar radiation from the Earth Radiation Budget Experiment(ERBE),and cloudiness from the International Satellite CloudClimatologyProject(ISCCP).Hpis estimated using multi-year climatological ocean color data averaged over the period of September 1997 to April 2007(McClain et al.,1998).

Interannual anomaly f i elds used to construct empirical models include those of observed SST from Reynolds et al.(2002),and wind stress and P–E obtained from an ensemble mean of a 24-memberECHAM 4.5①The ECHAM4.5 AGCM is a model developed by the Max Planck Institute for Meteorology(MPI)and the European Center for Medium-Range Weather Forecasts;see details in Roeckner et al.(1996).AGCM(Roeckneret al.,1996)simulationovertheperiod1950–99,whichis forcedby observed SST f i elds.Using the ensemble mean data for wind stress and P–E represents an attempt to enhance SST-forced signals by reducing atmospheric noise.

In addition,satellite data are used to represent various specif i c processes that are important to ENSO.For example,chlorophyll(Chl)in the ocean can affect the penetration of solar radiation in the upper ocean,but it is diff i cult to obtain basin-wide in situ measurements.Today,ocean color data are available from satellites and can be used to depict its interannual variability associated with ocean biology.As in our previous modeling studies(Zhang et al.,2009,2011),this feedback is parameterized using remotely sensed ocean color data(McClain et al.,1998).Also,high-resolutionsatellite wind vector data are available and can be used to represent TIW-scale wind feedback(Zhang and Busalacchi,2008,2009b).Daily SST f i elds are from the TropicalRainfall Measuring Mission(TRMM)satellite’s microwave imager(TMI;Wentz et al.,2000),and surface winds are from QuikSCAT during the period 2000–07.

2.2.Hybrid coupled ocean–atmosphere model

As classif i ed to be an HCM(Fig.1),our model’s ocean component is a comprehensive OGCM,while the interannual wind component is a simple statistical model.This is based on the fact that ENSO is a dominant driving force of the coupled ocean–atmosphere system in the Pacif i c,acting to produce large-scale interannual SST anomalies,which in turn induce quick surface wind responses.The coherent relationships between interannual variations in SSTs and winds are used to construct an empirical model.Additionally,some related feedback processes are incorporated in the HCM to represent their possible effects on ENSO.

2.2.1.OGCM

The ocean model used is based on the primitive equation model with the reduced gravity approximation developed by Gent and Cane(1989).In the vertical direction,the ocean model includes a mixed layer and a number of layers below specif i ed on a sigma coordinate(thus referred to as a layer model).The depth of the mixed layer(top layer)and thickness of the last sigma layer(bottom layer)are determined prognostically,while the thickness of the remaining layers are calculated in such a way that the ratio of each sigma layer to the total depth is held to its prescribed value.In addition,several related efforts made in the past have improved this ocean model signif i cantly.For example,a hybrid vertical mixing scheme was developed and embedded into the ocean model(Chen et al.,1994);the OGCM is coupled to an advective AML model to estimate sea surface heat f l uxes(Murtugudde et al.,1996);and salinity effects have been included in the model with freshwater f l ux treated as a natural boundary condition(Murtugudde and Busalacchi,1998).Also,the effects of penetrative radiation in the upper ocean have been taken into account;remotely sensed ocean color data are adopted to prescribe seasonally varying climatological attenuation depths(Murtugudde et al.,2002;Ballabrera-Poy et al.,2007).

TheOGCM coversthe Pacif i c domainfrom40°S to 60°N and from 124°E to 76°W;its zonal resolutionis uniform(1°),and its meridional resolution is stretched in latitude[0.31°within 5°S–5°N,0.5°in the off-equatorial(5°–15°)regions,and 1°poleward of 30°].In the vertical direction,the OGCM has 31 layers,with the f i rst layer treated as a bulk mixed layer.Sponge layers are introduced near the OGCM southern boundaries(poleward of 35°S);that is,a Newtonian term is included in the temperature and salinity equations,acting to relax the model temperature and salinity f i elds to observational data specif i ed using the World Database(Levitus et al.,2005).The OGCM is initiated from the Levitus temperature and salinity f i elds and is integrated for more than 50 years using prescribed atmospheric climatological forcing f i elds(OGCM spinup),including wind stress from the NCEP–NCAR reanalysis products averaged over the period 1950–2000.Note that a 50-year ocean spinup is quite short and may be insuff i cient to achieve a quasiequilibrium state for starting climate simulations.However,the model used is a reduced gravity model designed to describe ocean circulation in the upper ocean(upper 2000 m or so);the shortness of the OGCM spinup period does not matter for the purpose of this specif i c research(i.e.,ENSO simulation).

2.2.2.Empirical model for interannual wind stress variability

The atmosphericτmodel adopted in this work is a statistical one,specif i cally relating its interannual variability to SST anomalies,writtenasτinter=αinterFinter(SSTinter),where SSTinterrepresents interannual SST anomalies,Finterrepresents the relationships between interannual variations inτ and SST,andαinteris a scalar parameter introduced to represent the strength of interannual wind forcing of the ocean.

The Finterfunction is estimated using an SVD analysis of the covariance matrix that is calculated from time series of monthly-mean SSTinterandτinterf i elds(e.g.,Barnett et al.,1993;Chang et al.,2001).In this work,a combined SVD analysis is performed based on the covariance among anomalies of SST,zonal and meridional wind stress components(Zhang et al.,2003,2006).Due to computational limitations,the SVD analysis is performed on a horizontal resolution of 2°zonally,and stretched spacing meridionally from 0.5°within 10°of the equator to 3°poleward(note that the horizontal resolution for SSTinterandτinteranalyses is different from that of the OGCM).In time,the SVD analysis is performed using historical SSTinterandτinterdata over the 1963–96 period(34 years).With these speci fi cations,the dimensions of the matrix for the SVD analyses are 83×128×408(zonal and meridionalgridpoints overthe entire Paci fi c ocean,and 34-year temporal sampling from 1963 to 1996).

Figure 2 illustrates the spatial patterns of the fi rst SVD mode derived for SST and wind stress in the Paci fi c basin.Here,the SVD analysis is performed on all time series data during1963–96(i.e.,irrespectiveof seasons)to obtain singular vectors,singular values,and the correspondingtime coeffi cients.The fi rst fi ve singular values are about 239,55,32,21 and 16,with the squared covariance fraction being about 91%,5%,2%,1%and 0.4%,respectively.

The temporal expansion coef fi cients( fi gures not shown)clearly indicate that the f i rst mode describes interannual variability associated with ENSO events.The correspondingspatial patterns of SST and wind(Figs.2a and b)indicate that their large amplitude is located in the tropical Pacif i c;the primary coupled mode is composed of a wind variability center over the western-central Pacif i c that covaries with anomalous SST in the eastern and central equatorial Pacif i c.For example,during El Ni?no,large warm SST anomalies in the eastern equatorial Pacif i c(Fig.2a)are accompanied by westerly wind anomalies over the central equatorial Pacif i c around the date line(Fig.2b).The second mode(f i gures not shown)also shows a coherent relationship between these f i elds,with the spatial patterns representing transition stages of ENSO evolution.Note that the second SVD mode has its largest amplitude located in the extratropics.

Based on this SVD analysis,an empiricalτintermodel can be constructed to relate interannual variations in wind stress to those in SST,representing a dominant wind–SST coupling in thePaci fi c.As demonstratedby Barnettet al.(1993),interannualwindresponsestoagivenSSTanomalyaresensitively dependentonseasonsduringENSOevolution.Thisseasonality needs to be taken into account when constructingtheτintermodel.To this end,the SVD analyses are performed separately for each calendar month,thus yielding 12 seasonally varying sub-models(e.g.,Zhang and Zebiak,2002;Zhang and Busalacchi,2005).Thus,given an SST anomaly,the interannual wind response can be calculated according to the constructedτintermodel.In the considerationof the sequence of the singular values( fi gure not shown)and the reconstruction testing of theτinterfi elds from SST anomalies,the fi rst fi ve leading SVD modes are retained in the empiricalτintermodel for having its reasonable amplitude.

2.3.Representing forcing and feedback processes in the HCM

While interannual wind forcing is a major factor determining ENSO dynamics,other forcing and feedback processes also exist in the Pacif i c and can modulate ENSO in a substantial way(Fig.1),including FWF,OBH,and TIW-scale wind feedback.Additionally,wind forcing is stochastic in nature;as demonstrated by previous studies,SF of atmospheric winds can affect ENSO in a signif i cant way(Kirtman and Schopf,1998;Zhang et al.,2008).In our previous tropical Pacif i c modeling studies,statistical modeling tools were developed to represent these forcings and feedbacks;in this study,they are directly applied to the Pacif i c HCM,as described below.

2.3.1.Stochastic forcing model for wind stress

Wind stress anomalies that force the ocean model consist of an interannually varying signal part and a noise part.The former is estimated using the SVD-based empiricalτintermodel(see section 2.2.2)as a response to interannual SST variations(i.e.,an SST forcedsignal).The latter(the SF part)is of stochastic nature(i.e.,not explicitly related with an external forcing).This SF wind forcing part has been demonstrated to have signif i cant effects on ENSO(e.g.,Zhang et al.,2008),and thus needs to be represented in the HCM.Details of the construction of the SF wind part can be found in Zhang et al.(2008)and will be presented in Part II of this study.

2.3.2.Freshwater f l ux forcing:precipitation minus evaporation

Observational and modeling studies indicate that the interannual variability of P–E and SST are coherently related witheachotheroverthetropicalPacif i c,withadominantSST control of P–E.In particular,interannual variations in P–E closely follow those in SST during ENSO evolution.This provides a physical basis for constructing a statistical feedback model for interannual FWF responses to interannual SST anomalies.

To depict statistically optimized empirical modes of their co-variability,an SVD technique is also adopted for interannual anomaly f i elds of observed SST and P–E estimated from the ECHAM4.5 AGCM ensemble simulations.First,monthly-mean data are normalized in terms of their spatially averaged standard deviation to form the covariance matrix.Then,an SVD analysis is conducted on all time series data irrespective of season over the 1963–96 period(a total of 34 years of data).The f i rst f i ve singular values are about 3023,712,429,343 and 242,with squared covariance fractions of about 89%,5%,2%,1%and 0.6%,respectively.

Figure2cshowsthespatialpatternsofthef i rst SVDmode for FWF.The corresponding temporal expansion coeff i cients(f i gures not shown)indicate that the f i rst mode describes interannual FWF variability associated with ENSO events.The spatial pattern(Fig.2c)illustrates that the primary coupled mode of the variability is composed of a large FWF anomaly center in the central equatorial Pacif i c that co-varies with anomalous SST in the eastern and central equatorial Pacif i c(Fig.2a).The second mode(f i gures not shown)also shows coherentrelationshipbetweenFWF andSSTbothintime and space.

Based on this SVD analysis,an empirical FWF model is constructed using their derived spatial eigenvectors of the SVD modes(e.g.,Zhang and Busalacchi,2009b),written as FWFinter=αFWFFFWF(SSTinter),where FFWFrepresents the relationshipsbetweeninterannualvariationsin FWF and SST determined from the SVD analyses.The FWFintermodel is constructedwith the same period and horizontalresolution as in theτintermodel;the f i rst f i ve leading SVD modes are retained for having reasonable amplitude in the FWFintersimulation.Thus,a given interannual anomaly of SST can be converted into an anomaly of FWF for use in the HCM.

It turns out that the FWFintermodel can successfully capture large-scale interannual FWF variability associated with ENSO evolution.For instance,as shown in Fig.2c,the spatial patterns display interannual FWF anomalies at the mature phase of ENSO events.During El Ni?no,a warm SST anomaly induces an increase both in P and E,but P increases signif i cantly over a broad region in the central basin.As a result,El Ni?no is accompanied by a positive FWF anomaly over the intertropical convergence zone(ITCZ)in the central and eastern tropical Pacif i c(i.e.,anomalousFWF into the ocean due to the dominance of P over E).During La Ni?na,an opposite pattern is seen,with a cold SST anomaly being accompanied by a negative FWF anomaly(i.e.,a net loss of freshwater from the ocean).

2.3.3.Tropical-instability-wave-inducedwind feedback

TIWs are intraseasonal,small-scale phenomena that are prominentlyseen over the central-eastern tropical Pacif i c.As has long been recognized(e.g.,Bryden and Brady,1989),they are an important component in the tropical Pacif i c climate system,having an inf l uence on heat and momentum transport at the equator in the ocean(e.g.,Kessler et al.,1998;Jochum et al.,2005).Recent high-resolution satellite data indicate that TIWs are accompanied by large wind perturbations,giving rise to a wind feedback onto the ocean and coupled air–sea interactions at TIW scales(e.g.,Chel-ton et al.,2001).Due to TIW roles in the climate system of the tropical Paci fi c,it is important to represent TIW-induced surface wind feedback effects on ENSO.However,this process is obviously missing in the HCM since its atmospheric component is a statistical one with a low horizontal resolution designedto capture large-scalewind stress variability.In our previous studies(Zhang and Busalacchi,2008,2009b),high-resolution satellite data were used to develop an empirical model for TIW-induced wind stress perturbations(τTIW)to capture TIW-induced small-scale processes in the largescale HCM;in this work,this method is directly applied to the Paci fi c-domain HCM.

To extract TIW-scale SST and wind stress signals(SSTTIWandτTIW),a spatial high-pass fi lter is applied to their daily data(removing the slow-varying background mean fi elds by subtracting a 12°zonal moving average from the original data).Then,a standard SVD analysis is applied to the resultant SSTTIWandτTIWfi elds to determine their statistically optimized empirical modes,from which an empirical model is constructed forτTIW,written asτTIW= αTIWFTIW(SSTTIW),where FTIWrepresents the SVD-determined empirical relationships between SSTTIWandτTIW,andαTIWis a scalar parameter introduced to represent TIW wind feedback intensity.Thus,given an SSTTIWfi eld,τTIWcan be determined accordingly.The SVD analysis is performed on data over the 2000–07 period and at the OGCM grid(1°×0.5°)in the central-eastern tropical Paci fi c(from 15°S to 15°N and from 180°to 76°W);see details in Zhang and Busalacchi(2009b).

Note that theτintercomponent is determined at a coarse horizontal resolution using theτintermodel,and thus TIWs are not explicitly resolved in the large-scale coupled ocean–atmosphere context.Nevertheless,theτTIWcomponent can be estimated at a relatively high resolution using theτTIW–SSTTIWrelationship.That is,this regionalτTIWmodel can be embedded into the HCM to capture the TIW wind forcing effect on the ocean,allowing for the representation of TIW-induced wind feedback on ENSO variability within a largescale coupled ocean–atmosphere modeling context.Interactions among processes with varying spatiotemporal scales can be represented,includingocean–atmospherecouplingsat TIW scales and ENSO scales.

2.3.4.Ocean biology-induced heating effects

In addition to physical processes,it has been demonstrated that ocean biology can modulate the heat budget in the upper ocean(Lewis et al.,1990).In particular,recent observational analyses and modeling studies have revealed signif i cant effects of ocean biology on the mean climate and its low-frequency variability in the tropical Pacif i c.However,the HCM does not explicitly include a marine ecosystem component,and the related bio–climate coupling is not taken into account.Today,ocean color data from satellites are available and can be used to characterizebasin-scale variability patterns of ocean biology and quantify its relationships with physical parameters.For example,the effect of ocean biology-induced heating can be simply represented by the penetration depth of solar radiation in the upper ocean(Hp).

As a dominant source for ocean biology variability,ENSO induces large Hpanomalies over the tropical Pacif i c,whose spatiotemporal evolution exhibits a good relationship with SST.So,interannual anomalies of Hpcan be treated as a response to those of SST in association with ENSO.Similar toτand FWF,the total Hpf i eld can be separated into its climatological part(seasonally varying)and interannual anomaly part.The former is prescribed by using multi-year ocean color data over the period of September 1997 to April 2007(McClain et al.,1998);the latter can be determined using an empirical model to represent its interannual response to changes in SST(Zhang et al.,2011).

Similar to τinterand(P–E)inter,an empirical Hpmodel can be constructed,written as H′p=αHpFHp(SSTinter),in which FHprepresents the SVD-based statistical relationships between interannual variations in Hpand SST,andαHpis a rescaling parameter that is introduced to represent the amplitude of interannual Hpvariability.As indicated in the SVD analysis performed by Zhang et al.(2011),the f i rst f i ve SVD modes contain about 65%of the covariance between interannual Hpand SST variations.To capture its amplitude using the SVD-based Hpmodel,αHp=2 needs to be taken when the f i rst f i ve modes are retained(Zhang et al.,2011).Thus,given an SST anomaly,the Hpresponse can be determined from its empirical model.The Hpmodel is also embedded into the HCM to represent its penetration effects on solar radiationintheupperocean.As such,interannualHpanomalies are parameterized,with the related ocean biology-induced climate feedback being captured in the HCM.

2.4.The coupled system and model experiment designs

The coupling among these components(Fig.1)is implemented as follows.At each time step,the OGCM calculates SST f i elds,whose interannual anomalies are obtained relative to its uncoupledclimatology(SSTclim,which is predetermined from the OGCM-only run forced by observed climatological atmospheric f i elds).The resultant interannual SST anomaly f i eld is then used to calculate interannual anomalies ofτ,FWF and Hpusing their corresponding empirical models.These interannual anomalies are then added onto their prescribed climatological f i elds for use in the HCM.Also,TIW-scale wind and SF wind parts can be included in the HCM.

The OGCM is initiated from the World Ocean Atlas(WOA01)temperature and salinity f i elds(Levitus et al.,2005),and is integrated for more than 50 years using climatological atmospheric forcing f i elds.Based on this ocean spinup,the HCM is then initiated with an imposed westerly wind anomaly for eight months.Evolution of anomalous conditions thereafter is determined solely by coupled ocean–atmosphere interactions within the system.As examined previously by Barnett et al.(1993),coupled behaviors sensitively depend on the so-called relative coupling coeff icient(αinter);i.e.,the wind stress anomalies from the τintermodel can be further scaled by this parameter.Several tun-ing experiments are performed with different values ofαinterto examine ways the coupled interannual variability can be sustained in the HCM.It is found that takingαinter=1.3 can producea sustainable interannualvariabilityin the HCM.Similarly,the other scalar parameters(αFWF,αTIW,andαHp)are also examined to represent the related feedbacks with a reasonable intensity;see details in Zhang et al.(2006,2009)and Zhang and Busalacchi(2009b).In this paper,a reference run is performed using the HCM in which only large-scale SSTinter–τintercoupling and the FWF effect are taken into account,while other forcing and feedback processes are not taken into account(i.e.,αinter=1.3,αFWF=1.0,αSF=0.0,αTIW=0.0,andαHp=0.0).The related forcing and feedback effects(i.e.,SF,TIW and OBH)will be analyzed in the Part II of this study.

3.Simulated long-term climatology in the ocean

A 50-yearreference run is performedto illustrate the performance of the HCM.The outputs are used to demonstrate its ability to simulate the annual mean,seasonal variations,and interannual variability.In this section,the resultant mean climatological f i elds are presented below.

3.1.Annual-mean ocean state

Figures 3 and 4 illustrate examples of some selected annual-mean f i elds simulated from the reference run.The horizontal distribution of annual-mean SST f i elds bear a strong resemble to the corresponding observed one(Fig.3),withthewarmpoolinthewestandthecoldtongueintheeast.Detailed comparisons between the HCM simulation(Fig.3a)and observation(Fig.3b)also reveal some discrepancies,including the simulated SST being considerably too cold in the far eastern equatorial Pacif i c and along the western coast of South America.The sea level(SL)exhibits well-def i ned trough and ridge structures in the tropical Pacif i c,and gyre patterns in the subtropical and subpolar regions of the North Pacif i c(Fig.4a).The mixed layer depth(MLD)simulated from the HCM(Fig.4b)is in good agreement with the corresponding observed one(e.g.,Monterey and Levitus,1997).Note that the MLD is treated as a prognostic variable in the model,whichis explicitlycomputedusingabulkmixed-layer model(Chen et al.,1994).

More examples of vertical distributions of the simulated equatorial currents and thermal f i elds are displayed in Figs.5 and 6.The HCM realistically captures the current system and temperature structure in the tropical Pacif i c,including the North Equatorial Current(NEC),the South Equatorial Current(SEC),theNorthEquatorialCountercurrent(NECC),and the Equatorial Undercurrent(EUC).Also,as seen in Fig.6a,the HCM depicts a strong meridional boundary f l ow in the western North Pacif i c,with its maximum location along 129°E at 8°N;this boundary f l ow is referred to as the lowlatitude western boundary current(LLWBC),which plays an important role in the water and property exchanges from the subtropics to the tropics.However,it is clear that its amplitude is underestimated signif i cantly.

3.2.Mean ocean circulation pathways in the western Pacif i c

As has been demonstrated(e.g.,McCreary and Lu,1994;Gu and Philander,1997;Rothstein et al.,1998),the tropical Pacif i c Ocean features complicated circulation pathways connecting the extratropics to the tropics,including the subtropical cells(STCs)and LLWBCs.Figure 7 illustrates the structure of mean circulation pathways in the western tropical Pacif i c as represented by SL and mixed-layer currents simulated from the HCM.There is a clear path that makes a connection for waters from the subtropics to the equatorial regions over the western tropical Pacif i c;some subtropical waters of the North Pacif i c f l ow southward through the LLWBC and eastward along the NECC,making direct routes to the equatorin the interiorregions(e.g.,Rothstein et al.,1998;Zhanget al.,1999,2001).Therefore,the LLWBC and NECC pathways indicatea conveyerbelt of water exchangebetween the subtropicsand tropicsin the western NorthPacif i c.At the equator,waters in the west are transportedeastward along the EUC pathway into the eastern Pacif i c(Figs.5a and 6b).As the thermocline shoals along the equator from the west to the east(Fig.5a),thermal conditions at subsurface depths in the west can directly inf l uence SSTs in the eastern equatorial Pacif i c.It is conceivable that thermal conditions in the subtropical subduction regions can affect SST in the eastern equatorial Pacif i c through the STCs,a mechanism for connections between variations in the subtropics and tropics.

3.3.Seasonal variation

Figures 8 and 9 display examples of the simulated annual cycles of SST,MLD and surface zonal currents along the equator.Clearly,the simulated seasonal variation is in good agreement with corresponding observations.For example,large seasonal SST f l uctuations are apparent in the eastern equatorial Pacif i c:a warming occurs during spring and a cooling takes place during fall(Fig.8).Note,however,that the HCM simulation also exhibits some obvious biases compared with corresponding observations.For instance,the simulated seasonal cycle is underestimated with incorrect phase in the eastern equatorial Pacif i c.While the observed warming occurs in March and cooling in September,the simulated warming and cooling take place in April and August,respectively.These model biases could be related to the fact that TIWs and other forcing/feedbackeffects are not adequately representedin the reference HCM run;the effects involved will be examined in part II of this work.

The seasonal cycle of the MLD(Fig.9a)is similar to the observed(e.g.,Monterey and Levitus,1997),including the shoaling in spring and deepening in fall over the eastern equatorialPacif i c.Onenotablefeaturethatis capturedwellin the HCM is the springtime reverse of the SEC in the centraleastern equatorial Pacif i c(Fig.9b),which is clearly related to the seasonal SST warming(Fig.8b).Figure 10 presents one more example of detailed vertical distributions of zonal currents and temperature f i elds and their seasonal variations.Some well-known features simulated well in the HCM include the seasonal variations in the EUC amplitude and its core depth(Fig.10a),the reversal of the SEC and the surfacing of the EUC in spring(Fig.9a),and the corresponding spring-time surface warming(Fig.10b),respectively.The HCM simulation can be well compared with the corresponding observation as presented in Zhang and Zebiak(2002).

4.Simulated interannual variability

Figure 11 displays examples of interannual variability simulated from the reference run.As has been extensively studied,interannual variability in the tropical Pacif i c is dominated by ENSO events,which are determined by the coupling among SST,winds and the thermocline(e.g.,Bjerknes,1969).Encouragingly,the HCM can realistically capture interannual oscillations associated with El Ni?no and La Ni?na events.The overall time scales of simulated interannual variability,the spatiotemporal evolution and coherent phase relationships among various atmospheric and oceanic anomalies are consistent with observations,which have been described before(e.g.,Zhang and Levitus,1997;Zhang and Rothstein,1998).

For example,the total SST f i elds(Fig.11a)clearly display large zonal displacements of the warm pool in the west and cold tongue in the east during ENSO cycles.During El Ni?no,thecold tongueshrinksin theeast,with warmwaters in the west extendingeastward along the equator(e.g.,the 26°C isotherm of SST is seen to extend to east of 120°W).During La Ni?na,the warm pool retreats to the west,whereas the cold tongue in the east develops strongly and expands westward along the equator,with the 25°C isotherm of SST being locatedwest of150°W.InterannualvariationsinSST andsur-face wind do not exhibit obvious propagation at the equator;they are almost in phase in time,with a zonal shift in space.The largest SST anomalies occur in the central and eastern equatorial Pacif i c(Fig.11b),while the largest wind variability is located near the date line(Fig.11c).

Coherent relationships between these anomaly f i elds can be more clearly seen in their horizontal patterns(f i gures not shown).During El Ni?no,for example,warm SST anomalies are located in the eastern equatorial Pacif i c,accompanied by westerly wind anomalies to the west in the tropical Pacif i c and a cyclonic circulation around the Aleutian Low region in the North Pacif i c(i.e.,the Aleutian Low is intensif i ed in response to warm SST anomalies in the tropics).During La Ni?na,an opposite pattern is seen,with cold SST anomalies being associated with easterly wind anomalies in the tropics and an anticyclonic circulation around the Aleutian Low region(i.e.,the Aleutian Low is weakened in response to cold SST anomalies in the tropics).At the transitional stage from El Ni?no to La Ni?na,no signif i cant anomalies are seen in the surface f i elds of SST and wind,but large thermal anomalies are seen at subsurface depths.The commonly adopted Ni?no3.4index is used to quantifythe dominanttime scales of interannualvariability.A waveletanalysis suggeststhatinterannual oscillations in the HCM indicate two peaks at about 2 years and 4 years;the corresponding observation indicates a dominance of about 4.8 years.Interannual SST variability is further quantif i ed in Fig.12.The amplitude and structure are captured well in the tropical Pacif i c.For example,the standard deviations of the Ni?no1+2,Ni?no3,Ni?no3+4,and Ni?no4 SST anomalies are 0.63,1.04,1.22 and 0.89,respectively.Some model biases are also evident.For example,the maximum SST anomalies tend to occur around 120°W–180°E,which is too far west than observed,and the Ni?no1+2 SST anomalies are much weaker than observed.Also,the reference HCM simulation cannot capture the so-called eastern Pacif i c(EP)and central Pacif i c(CP)El Ni?no events.

5.Concluding remarks

ENSO is the largest interannual signal arising from air–sea interactions in the tropical Pacif i c.It has been examined extensively using tropical ocean–atmosphere models in the Pacif i c.Since ENSO can also be modulated by extratropical processes,it is necessary to take into account the modulating effects coming from the subtropics and midlatitudes.In this work,a hybrid coupled ocean–atmosphere model(HCM)is developed for almost the entire Pacif i c basin;its atmospheric component is taken to be a statistical one.Such a conf i guration can be justif i ed by the fact that interannual variability in the region is dominated by ENSO,acting to generate large SST anomalies,which induce atmospheric responses that are quick and coherent at large scales.Therefore,the related surface wind variability can be treated as a feedback process,with the relationships between interannual variations in SST and surface wind being established from historical data.Ac-cording to its classif i cation as an HCM,such a model is economical in terms of computation,and yet can capture dominant SST–wind coupling over the Pacif i c basin,allowing interactions between processes in the tropics and subtropics to be adequately represented.

In addition,ENSO is affected by a variety of processes in the Pacif i c,including various forcings and feedbacks.For example,ENSO acts to produce perturbations in ocean biology,which can feed back to ENSO.As the simulated ENSO properties are sensitively dependent on the ways these processes are represented,their effects need to be adequately included in models.Based on our previous tropical modeling efforts,some important forcing and feedback processes affecting ENSO in the region can be adequately represented in the Pacif i c HCM,including SF,FWF,TIW,and OBH.

As expected,this simplif i ed model tool is computationally eff i cient to run,and physically realistic enough to represent major components of the climate system over the Pacif i c basin.In particular,within this hybridmodelingcontext,forcing and feedback processes can be turned on or off,allowing their effects(individual or collective)on ENSO to be examined in a clean way.In addition,the related feedback intensities can be represented by parameters that are tunable,allowing their effects to be quantif i ed.A variety of applications are anticipated for ENSO-related modeling studies.In this paper,very preliminary results from a reference run are used to demonstrate the HCM’s performance,showing that it can reproduce quite well the mean ocean state,seasonal cycle and interannual variability associated with ENSO.For example,the Pacif i c-domain model depicts the well-def i ned mean water pathwaysfrom the subtropicsto the tropics in the western Pacif i c through the LLWBC and NECC,and from the west to the east at the equatorthroughthe EUC pathways,respectively.As such,thermal conditions in the subtropical subduction regions can affect SST in the eastern equatorial Pacif i c.

In addition,a novel way of making use of satellite data is demonstrated in this paper.Note that some processes are still diff i cult to be adequately represented in current climate models.For example,ENSO can be modulatedby ocean biologyinduced heating,but this effect cannot be represented in physical models without a comprehensive marine ecosystem model.As satellite ocean color data are now available,and can capture interannual ocean biology variability associated with ENSO,theyare used to parameterizethese processes for climate modeling studies.As detailed in Zhang et al.(2011),the relationships between ocean biology f i elds(e.g.,the penetration depth,Hp)and physical states(e.g.,sea surface temperature)can be derived from satellite data,and be used to construct an empirical parameterization that is incorporated in a climate model to represent ocean biology-induced heating effects.Also,TIWs are small-scale intraseasonal signals that are diff i cult to represent in large-scale climate models.However,high-resolution SST and surface wind data from satellites can be used to parameterize TIW wind feedback,which is included in the HCM.As such,vastly different processes at spatiotemporal scales can be integrated within a hybrid modeling context;multiscale interactions between TIW and ENSO can be examined.As demonstrated in this work,these approaches can be directly applied to any fully coupled model to adequately represent the related forcing and feedback processes.

The coupled ocean–atmosphere model developed for the entire Pacif i c in this work is classif i ed as a hybrid model,which has disadvantages as other HCMs do(e.g.,Zhu et al.,2011).For example,the HCM is designed to couple a basinscale OGCM to a statistical atmospheric model,which provides the ocean with three forcing f i elds(wind stress,freshwater f l ux and heat f l ux).The forcings are constructed to consist of the climatological f i elds and,where appropriate,interannual anomaly f i elds;the former are prescribed from observed data and the latter are represented as functions of their corresponding interannual SST anomalies(which are also determined using statistical analysis methods from observed historical data).Thus,the model is designed,more or less,to “nudge”the ocean climate towards the observed one,andthustheHCMessentiallyrepresentsananomalycoupling between the atmosphere and ocean.As a result,the model simulations can be dependent on the ways the climatological f i elds are prescribed and/or statistical anomaly models are constructed(including time period and datasets used).In addition,interannualanomalies are representedas a feedback model without detailed processes in the atmosphere.In particular,the stochastic natureof atmosphericprocesses has not been taken into account adequately.

Some discrepancies are evident in the reference run.For example,the simulated ENSO events are far too regular,dominated by two-year oscillation.These discrepancies are partially related with the fact that some forcing and feedback processes are lacking in the reference HCM simulation.Further improvements are clearly needed and expected whenthese feedbackprocesses areadequatelyincludedin the HCM.Indeed,preliminary testing indicates that stochastic forcing of wind can exert a signif i cant inf l uence on ENSO;when it is included in the HCM,the simulated ENSO events become irregular,yielding a more comparableresult with observations.In Part II of this study,various factors affecting ENSO properties will be examined individually or collectively.The HCM with adequately represented forcing and feedback effects will be used to decipher relationships between forcing/feedback processes and ENSO modulations,multi-process interactions on various spatiotemporal scales,and interplay between processes involving the tropics and subtropics and midlatitudes,respectively.

Acknowledgements.The author would like to thank Drs.HU Dunxin and MU Mu for their comments.The author wishes to thank the two anonymous reviewers for their numerous comments that helped to improve the original manuscript.This research was supported by the CAS Strategic PriorityProject(the Western Pacif i c Ocean System:Structure,Dynamics and Consequences,WPOS),a China 973 project(Grant No.2012CB956000),the Institute of Oceanology,Chinese Academy of Sciences(IOCAS),and the National Natural Science Foundation of China(No.41206017).

Ballabrera-Poy,J.,R.Murtugudde,R.-H.Zhang,and A.J.Busalacchi,2007:Coupled ocean–atmosphere response to seasonal modulation of ocean color:Impact on interannual climate simulations in the tropical Paci fi c.J.Climate,20,353–374.

Barnett,T.P.,M.Latif,N.Graham,M.Flugel,S.Pazan,and W.White,1993:ENSO and ENSO-related predictability.Part I:Prediction of equatorial Paci fi c sea surface temperature with a hybrid coupled ocean-atmosphere model.J.Climate,6,1545–1566.

Bjerknes,J.,1969:Atmospheric teleconnections from the equatorial Paci fi c.Mon.Wea.Rev.,97,163–172.

Bryden H.L.,and E.C.Brady,1989:Eddy momentum and heat fl uxes and their effects on the circulation of the equatorial Paci fi c Ocean.J.Mar.Res.,47,55–79.

Chang,P.,L.Ji,and R.Saravanan,2001:A hybrid coupled model study of tropical Atlantic variability.J.Climate,14,361–390.

Chelton,D.B.,S.K.Esbensen,M.G.Schlax,N.Thum,M.H.Freilich,F.J.Wentz,C.L.Gentemann,M.J.McPhaden,and P.S.Schopf,2001:Observations of coupling between surface wind stress and sea surface temperature in the eastern tropical Pacif i c.J.Climate,14,1479–1498.

Chen,D.,L.M.Rothstein,and A.J.Busalacchi,1994:A hybrid vertical mixing scheme and its application to tropical ocean models.J.Phys.Oceanogr.,24,2156–2179.

Fedorov,A.V.,and S.G.H.Philander,2000:Is El Ni?no changing?Science,228,1997–2002.

Gent,P.,and M.A.Cane,1989:A reduced gravity,primitive equation model of the upper equatorial ocean.J.Compute.Phys.,81,444–480.

Gu,D.-F.,and S.G.H.Philander,1997:Interdecadal climate f l uctuations that depend on exchanges between the tropical and extratropics.Science,275,805–807.

Jochum,M.,R.Murtugudde,R.Ferrari,and P.Malanotte-Rizzoli,2005:The impact of horizontal resolution on the equatorial mixed layer heat budget in ocean general circulation models.J.Climate,18,841–851.

Kalnay,E.,and Coauthors,1996:The NMC/NCAR reanalysis project.Bull.Amer.Meteor.Soc.,77,437–471.

Kessler,W.S.,L.M.Rothstein,and D.Chen,1998:The annual cycle of SST in the eastern tropical Pacif i c,diagnosed in an ocean GCM.J.Climate,11,777–799.

Kirtman,B.P.,and P.S.Schopf,1998:Decadal variability in ENSO predictability and prediction.J.Climate,11,2804–2822.

Kleeman,R.,J.P.McCreary,and B.A.Klinger,1999:A mechanism for generating ENSO decadal variability.Geophys.Res.Lett.,26,1743–1746.

Lewis,M.R.,M.E.Carr,G.C.Feldman,W.Esias,C.McClain,1990:Inf l uence of penetrating solar radiation on the heat budget of the Equatorial Pacif i c.Nature,347,543–546.

Levitus,S.,J.I.Antonov,T.P.Boyer,2005:Warming of the World Ocean,1955–2003.Geophys.Res.Lett.,32,L02604,doi:10.1029GL021592.

McClain,C.R.,M.L.Cleave,G.C.Feldman,W.W.Gregg,S.B.Hooker,N.Kuring,1998:Science quality SeaWiFS data for global biosphere research.Sea Technol.,39,10–16.

McCreary,J.P.,and P.Lu,1994:Interaction between the subtropical and equatorial ocean circulations:The subtropical gyre.J.Phys.Oceanogr.,24,466–497.

Monterey,G.I.Levitus,S.,1997:Seasonal variability of mixed layer depth for the World Ocean.NOAA NESDIS Atlas 14,Natl.Oceanic and Atmos.Admin.,Silver Spring,Md.,100 pp.

Murtugudde,R.,and A.J.Busalacchi,1998:Salinity effects in a tropical ocean model.J.Geophys.Res.,103,3283–3300.

Murtugudde,R.,R.Seager,and A.J.Busalacchi,1996:Simulation of tropical oceans with an ocean GCM coupled to an atmospheric mixed layer model.J.Climate,9,1795–1815.

Murtugudde,R.,J.Beauchamp,C.R.McClain,M.Lewis,and A.J.Busalacchi,2002:Effects of penetrative radiation on the upper tropical ocean circulation.J.Climate,15,470–486.

Reynolds,R.W.,N.A.Rayner,T.M.Smith,D.C.Stokes,and W.Wang,2002:An improved in-situ and satellite SST analysis for climate.J.Climate,15,1609–1625.

Roeckner,E.,and Coauthors,1996:The atmospheric general circulation model ECHAM4:Model description and simulation of present day climate.Rep.218,Max-Planck-Institut fur Meteorologie,90 pp.

Rothstein,L.M.,R.-H.Zhang,A.J.Busalacchi,and D.Chen,1998:A numerical simulation of the mean water pathways in thesubtropical and tropical Pacif i cOcean.J.Phys.Oceanogr.,28,322–343.

Seager,R.,M.Blumenthal,and Y.Kushinir,1995:An advective atmospheric mixed layer model for ocean modeling purposes:Global simulation of surface heat f l uxes.J.Climate,8,1951–1964.

Wang,B.,and S.-I.An,2001:Why the properties of El Nino changed during the late 1970s.Geophys.Res.Lett.,28,3709–3712.

Wentz,F.J.,C.Gentemann,D.Smith,andD.Chelton,2000:Satellite measurements of sea surface temperature through clouds.Science,288,847–850.

Xie,P.,and P.Arkin,1995:An intercomparison of gaobservations and satellite estimates of monthly precipitation.J.Appl.Meteor.,34,1143–1160.

Zebiak,S.E.,and M.A.Cane,1987:A model El Ni?no/Southern Oscillation.Mon.Wea.Rev.,115,2262–2278.

Zhang,R.-H.,and S.Levitus,1997:Interannual variability of the coupled tropical Pacif i c ocean–atmospheric system associated with the El Ni?no-Southern Oscillation.J.Climate,10,1312–1330.

Zhang,R.-H.,and L.M.Rothstein,1998:On the phase propagation and relationship of interannual variability in the tropical Pacif i c climate system.Climate Dyn.,14,713–723.

Zhang,R.-H.,and S.E.Zebiak,2002:Effect of penetrating momentum f l ux over the surface mixed layer in a z-coordinate OGCM of the tropical Pacif i c.J.Phys.Oceanogr.,32,3616–3637.

Zhang,R.-H.,and A.J.Busalacchi,2005:Interdecadal changes in properties of El Ni?no in an intermediate coupled model,J.Climate,18,1369–1380.

Zhang,R.-H.,and A.J.Busalacchi,2008:Rectif i ed effects of tropical instability wave(TIW)-induced atmospheric wind feedback in the tropical Pacif i c.Geophys.Res.Lett.,35,L05608,doi:10.1029/2007GL033028.

Zhang,R.-H.,and A.J.Busalacchi,2009a:Freshwater f l ux(FWF)-induced oceanic feedback in a hybrid coupled model of the tropical Pacif i c.J.Climate,22,853–879.

Zhang,R.-H.,and A.J.Busalacchi,2009b:An empirical model for surface wind stress response to SST forcing induced by tropical Instability waves(TIWs)in the eastern equatorial Pa-cif i c.Mon.Wea.Rev.,137,2021–2046.

Zhang,R.-H.,L.M.Rothstein,and A.J.Busalacchi,1998:Origin of upper-ocean warming and El Ni?no change on decadal scale in the tropical Pacif i c Ocean.Nature,391,879–883.

Zhang,R.-H.,L.M.Rothstein,and A.J.Busalacchi,1999:Interannual and decadal variability of the subsurface thermal structure in the Pacif i c Ocean:1961–90.Climate Dyn.,15,703–717.

Zhang,R.-H.,T.Kagimoto,and S.E.Zebiak,2001:Subduction of decadal North Pacif i c thermal anomalies in an ocean GCM,Geophys.Res.Lett.,28,2449–2452.

Zhang,R.-H.,S.E.Zebiak,R.Kleeman,and N.Keenlyside,2003:A new intermediate coupled model for El Ni?no simulation and prediction.Geophys.Res.Lett.,30,2012,doi:10.1029/2003GL018010.

Zhang,R.-H.,A.J.Busalacchi,and R.G.Murtugudde,2006:Improving SSTanomaly simulations inalayer ocean model with an embedded entrainment temperature submodel.J.Climate,19,4638–4663.

Zhang,R.-H.,A.J.Busalacchi,and D.G.DeWitt,2008:The roles of atmospheric stochastic forcing(SF)and oceanic entrainment temperature(Te)indecadal modulation of ENSO.J.Climate,21,674–704.

Zhang,R.-H.,A.J.Busalacchi,X.Wang,J.Ballabrera-Poy,R.G.Murtugudde,E.C.Hackert,and D.Chen,2009:Role of ocean biology-induced climate feedback in the modulation of El Ni?no-Southern Oscillation.Geophys.Res.Lett.,36,L03608,doi:10.1029/2008GL036568.

Zhang,R.-H.,D.Chen,and G.Wang,2011:Using satellite ocean color data to derive an empirical model for the penetration depth of solar radiation(Hp)in the tropical Pacif i c ocean.J.Atmos.Oceanic Technol.,28,944–965.

Zhang,R.-H.,F.Zheng,J.Zhu,Y.Pei,Q.Zheng,and Z.Wang,2012:Modulation of El Ni?no-Southern Oscillation by freshwater f l ux and salinity variability in the Tropical Pacif i c.Adv.Atmos.Sci.,29,647–660,doi:10.1007/s00376-012-1235-4.