1. Introduction
The ocean is a key player in climate variability and predictability on various time–space scales. Largely driven by atmospheric forcing, the induced physical changes in the ocean can feedback to the atmosphere by which the principal oceanic quantity felt is sea surface temperature (SST). Numerous studies have identified roles of various forcings and feedbacks in the climate system, including the Bjerknes feedback (e.g., Bjerknes 1969), the wind–evaporation–SST (WES) feedback (e.g., Xie and Philander 1994), the SST–solar radiation feedback (e.g., Waliser et al. 1994), and others. In the past, most studies have emphasized the forcing and feedback effects of atmospheric wind and heat flux on the coupled ocean–atmosphere system. Another less focused atmospheric forcing component to the ocean is freshwater flux (FWF), which has direct effects on ocean salinity, an important variable in climate and the water cycle. While sea surface salinity (SSS) has no direct and immediate influence on the atmosphere, its variations can be forced by atmospheric FWF perturbations, which can further modify the oceanic density fields, the mixed layer depth (MLD), and the mixing and entrainment, all of which can affect SST. For example, FWF forcing and its related salinity effect have been shown to play an important role in climate variability in the North Atlantic, being recognized as a driving force for the thermohaline circulation and its fluctuations (e.g., Schmitt et al. 1989).
In the tropical Pacific, a predominant role of wind forcing has been demonstrated in interannual climate variability associated with ENSO, involving a feedback loop among the SST, winds, and the thermocline (i.e., the Bjerknes feedback). The associated interannual changes in SST induce coherent fluctuations in the atmospheric circulation, including precipitation (P) and evaporation (E), whose interannual variabilities have been well documented in association with ENSO (e.g., Xie and Arkin 1995; Yu and Weller 2007). These large variations of P and E are reflected in those of freshwater flux.
Over the tropical Pacific region, the major contribution to interannual variations in FWF comes from a net difference between P and E, with a dominance of the former over the latter. Indeed, associated with ENSO, interannual FWF variability shows a close relationship with SST in the tropical Pacific. During El Niño, SSTs are warm in the central and eastern equatorial Pacific, accompanied by an increase both in P and E in the central basin. Because of the dominance of P over E, a warming is associated with a positive FWF anomaly (an anomalous flux into the ocean). During La Niña, cold SST anomalies are accompanied by a reduction both in P and E in the central basin. The resultant FWF anomaly is negative (an anomalous loss of freshwater from the ocean). Thus, interannual variations in FWF present a nonlocal positive correlation with SST during ENSO cycles. This is contrasted to those in heat flux, which have been demonstrated to have a negative correlation with SST (e.g., Barnett et al. 1991; Wang and McPhaden 2001).
Recent studies indicate that FWF forcing and its directly related changes in salinity can play an active role in maintaining the Pacific climate and its low-frequency variability through their effects on the horizontal pressure gradients, stratification, and the equatorial thermocline (e.g., Miller 1976; Cooper 1988; Carton 1991; Reason 1992; Weller and Anderson 1996; Murtugudde and Busalacchi 1998; Yang et al. 1999; Huang and Mehta 2004, 2005). One specific example is the identification of the so-called barrier layer that is formed by FWF forcing and the salinity effect in the western tropical Pacific (e.g., Lukas and Lindstrom 1991; Sprintall and Tomczak 1992; Vialard and Delecluse 1998a,b; Maes et al. 2002). More recently, observations and physical understanding of low-frequency salinity variability and its forcing field, FWF, have been significantly advanced (e.g., Schneider and Barnett 1995; Cronin and McPhaden 1999; Kessler 1999; Maes 2000; Lukas 2001; Maes et al. 2002; Lagerloef 2002; Fedorov et al. 2004; Boyer et al. 2005; Delcroix et al. 2007). Some unique salinity-related processes have been identified, including the subduction of thermal anomalies in the ocean (e.g., Zhang et al. 2001; Luo et al. 2005) and the compensated effects of temperature and salinity leading to spiciness phenomenon (e.g., Schneider 2000; Huang and Mehta 2004, 2005; Yeager and Large 2007). Clearly, FWF forcing and its related feedback need to be taken into account in modeling studies because of its large interannual anomalies induced by ENSO.
At present, FWF forcing has not been adequately represented in simplified models. In most previous modeling studies, the effects of FWF forcing have been demonstrated mostly in forced ocean-alone experiments. For example, idealized anomalous FWF forcing fields are perpetually prescribed to examine the response of the ocean (e.g., Reason 1992; Yang et al. 1999; Huang and Mehta 2004, 2005). Since the ocean–atmosphere is not coupled, there is no feedback from the changes in the ocean induced by FWF forcing to the atmosphere. Various coupled ocean–atmosphere models for the tropical Pacific have been developed for use in ENSO-related modeling studies, including intermediate coupled models (ICMs), hybrid coupled models (HCMs), and coupled general circulation models (CGCMs). However, FWF forcing has not been adequately represented in most state-of-the-art coupled models. For example, FWF has not been even included in most ICMs and HCMs used for simulation and prediction of ENSO (e.g., Zebiak and Cane 1987; Barnett et al. 1993; Syu et al. 1995; Zhang et al. 2003, 2005, 2006; Zhang and Zebiak 2004). In CGCMs, the FWF forcing is included but has not been realistically simulated. In particular, the so-called double intertropical convergence zone (ITCZ) problem is still a big challenge to CGCM simulations in the tropical Pacific; most models tend to have excessive precipitation over the ITCZ in the tropical Pacific. This deficiency in precipitation simulation is reflected in the FWF field, resulting in large and systematic biases that affect the ocean. In addition, large uncertainties exist in observational estimates of P and E from different sources and products. Thus, FWF forcing remains a challenge to be represented realistically in diagnostic analyses and coupled modeling studies.
Indeed, previous modeling studies have mostly focused on the roles of atmospheric forcing components of winds and heat flux in the coupled ocean–atmosphere system of the tropical Pacific; FWF forcing and its related salinity effect on climate variability have not been getting much attention. In addition, its effects have been examined mostly in ocean-only modeling studies. In a coupled ocean–atmosphere system, changes in SST induced by FWF forcing can feedback to the atmosphere. But these have not been clearly illustrated in a coupled ocean–atmosphere context. Although CGCMs include the FWF forcing, its impact on interannual variability has rarely been diagnosed explicitly. Furthermore, FWF-induced feedback can also influence the strength of other forcings and feedbacks in the coupled system. For example, the changed SSTs induced by FWF forcing can modulate heat flux forcing, which has been demonstrated to provide a negative feedback to interannual SST variability in the tropical Pacific. Then what are the net effects of these related feedbacks on interannual variability? Moreover, ENSO has been observed to change significantly from one event to another. Many factors have been identified that can modulate ENSO amplitude (e.g., Zhang and Busalacchi 2005; Zhang et al. 2008). As demonstrated in previous forced ocean-only simulations, FWF forcing can induce large changes in SST, indicating the potential for modulation of ENSO. However, the extent to which FWF forcing can play a role is not known.
In this work, a hybrid coupled modeling approach is taken to isolate the influences of anomalous FWF forcing on salinity and interannual variability in the tropical Pacific. The HCM developed at the Earth System Science Interdisciplinary Center (ESSIC; Zhang et al. 2006) consists of a layer ocean general circulation model (OGCM) and an empirical atmospheric model for interannual wind stress variability. This type of coupled model has been used widely and successfully in many tropical coupled ocean–atmosphere modeling studies (e.g., Neelin 1990; Barnett et al. 1993; Syu et al. 1995; Eckert and Latif 1997; Chang et al. 2001; Harrison et al. 2002; Zhang et al. 2003, 2005, 2006; Zhang and Zebiak 2004; for a comprehensive review, see McCreary and Anderson 1991). One obvious assumption behind HCMs is that atmospheric fields (such as wind stress and FWF) in the tropics are determined nonlocally by SST fields throughout the model domain. For example, SST perturbations can be generated by El Niño in the tropical Pacific. As illustrated from observations and modeling studies (e.g., Zebiak and Cane 1987; Barnett et al. 1991), the wind response in the atmosphere is quick and almost simultaneous. As a result, variations in wind and SST are well correlated over the equatorial Pacific, with the dominant influence of SST on winds. The other assumption is that, since the atmospheric fields in the tropics are highly coupled with SST fields, most of their variance can be explained by that associated with SST variations (i.e., portion of variance not captured by coupled modes is relatively small). These arguments provide a physical basis for using simplified HCMs for climate studies, in which the atmosphere can be reasonably well represented by a feedback model in response to SSTs.
As with wind forcing component, an additional empirical model has been developed to take into account interannual FWF variability that is explicitly related to SST anomalies. The FWF model is constructed from a singular value decomposition (SVD) of the covariance matrix that is calculated from time series of monthly mean SST and precipitation minus evaporation (P − E) fields. This statistical approach has been used widely and successfully to construct wind stress anomaly models in many tropical coupled ocean–atmosphere modeling studies (e.g., Barnett et al. 1993; Syu et al. 1995; Chang et al. 2001; Zhang et al. 2003, 2006; Zhang and Zebiak 2004). Then, using this empirical P − E model, an FWF anomaly can be estimated from a given SST forcing, which can be included in the HCM to account for its related possible feedback. In addition, heat flux in the HCM is calculated using an advective atmospheric mixed layer (AML) model developed by Seager et al. (1995). Thus, the HCM has three atmospheric forcings to the ocean: wind stress, heat flux, and freshwater flux. In this study, our focus is on the roles of anomalous FWF forcing in modulating interannual variability.
The paper is organized as follows: section 2 describes the model and some data used. Section 3 presents simulations of interannual variability from the HCM with anomalous FWF forcing (a standard run), followed by two sensitivity experiments in section 4. A positive feedback induced by FWF forcing is highlighted in section 5. The paper is concluded in section 6.
2. Model descriptions
Figure 1 shows a schematic of the various components of an HCM, recently developed at ESSIC (Zhang et al. 2006). The HCM consists of a layer OGCM and a simplified atmospheric representation of three forcing fields to the ocean, including the two empirical submodels for interannual wind stress and FWF variability. These are briefly described in this section, as well as observational and model-based datasets used to construct the empirical models.
a. An ocean general circulation model
The OGCM used is based on a reduced gravity, primitive equation, sigma coordinate model of Gent and Cane (1989), which is developed specifically for studying the coupling between the dynamics and the thermodynamics of the upper ocean. The vertical structure of the model ocean consists of a mixed layer and a number of layers below specified according to a sigma coordinate. The mixed layer depth and the thickness of the last sigma layer are computed prognostically, and the remaining layers are computed diagnostically such that the ratio of each sigma layer to the total depth below the mixed layer is held to its prescribed value.
Several related efforts have improved this ocean model significantly. For example, Chen et al. (1994) developed and embedded a hybrid mixed layer model and studied the effects of vertical mixing, solar radiation, and wind stresses on the seasonal cycle of SSTs in the tropical Pacific. Murtugudde et al. (1996) coupled the OGCM to an advective AML model developed by Seager et al. (1995) to estimate sea surface heat fluxes and showed the nonlocal effects of the atmospheric boundary layer on SST. Complete hydrology has also been added to the model, with freshwater flux treated as a natural boundary condition (e.g., Huang 1993; Murtugudde and Busalacchi 1998). Also, the effect of penetrative radiation on the upper tropical ocean circulation has been taken into account, with attenuation depths derived from remotely sensed ocean color data (Murtugudde et al. 2002). The OGCM has a variety of applications for simulations of the mean ocean state and its variability, including subduction pathways in the Pacific (e.g., Rothstein et al. 1998; Luo et al. 2005) and the coupled response of the tropical Pacific climate system to the seasonal cycle of the ocean color (Ballabrera-Poy et al. 2007).
The details of the OGCM used in this work can be found in Murtugudde and Busalacchi (1998). The OGCM domain covers the tropical Pacific basin from 25°S to 25°N and from 124°E to 76°W, with horizontal resolution of 1° in longitude and 0.5° in latitude, and 31 layers in the vertical. Near the model southern and northern boundaries (poleward of 20°S and 20°N), sponge layers are introduced; that is, a Newtonian term is added to the temperature and salinity equations, relaxing the model solution back to observational temperature and salinity data from the World Ocean Database 2001 (additional information is available online at http://www.nodc.noaa.gov/OC5/WOA01/pr_woa01.html; Levitus et al. 2001). The OGCM, initiated from the Levitus temperature and salinity datasets, is integrated for 20 yr for the OGCM spinup using climatological forcing fields. These data include wind stress from the European Centre for Medium-Range Weather Forecasts (ECMWF) averaged from 1985 to 1998 (e.g., Hackert et al. 2001), precipitation from Xie and Arkin (1995), solar radiation from the Earth Radiation Budget Experiment (ERBE; Harrison et al. 1993), and cloudiness from the International Satellite Cloud Climatology Project (ISCCP; Schiffer and Rossow 1985).
b. Atmospheric empirical models for interannual variability
The atmospheric forcing fields to the ocean consist of three components (Fig. 1): wind stress, heat flux, and FWF. As mentioned above, heat flux is calculated using an advective AML model (Seager et al. 1995). This heat flux parameterization allows for a realistic representation of the feedbacks between mixed layer depths, SSTs, and the heat fluxes (e.g., Murtugudde et al. 1996). Since interannual variations in atmospheric wind stress and FWF fields are determined nonlocally by SSTs in the tropical Pacific and the former can adjust quickly to the change in the latter, these two components are empirically constructed using statistical methods from their historical data, which are described in this subsection.
Various observational and model-based data are used for constructing the empirical models. The observed SST anomaly fields are from Smith and Reynolds (2004). Data of wind stress, E, and P used to construct empirical models are from the ensemble mean of a 24-member ECHAM4.5 atmospheric general circulation model (AGCM) simulation during the period 1950–99, forced by observed SSTs [the ECHAM4.5 AGCM was developed by the Max Planck Institute for Meteorology (MPI) and ECMWF; see Roeckner et al. (1996) for details]. Using the ensemble mean data is an attempt to enhance the SST-forced signal by reducing atmospheric noise.
1) A wind stress (τ) anomaly model
The atmospheric wind stress anomaly model adopted in this work is statistical, specifically relating interannual variability of wind stress (τinter) to large-scale SST anomalies (SSTinter). The τinter model is constructed from an SVD of the covariance matrix that is calculated from time series of monthly mean SST and wind stress fields (e.g., Syu et al. 1995; Chang et al. 2001). In this work, we perform a combined SVD analysis of the covariance among anomalies of SST and zonal and meridional wind stress components. Then an empirical τinter model can be constructed with the base periods from 1963 to 1996. The construction periods chosen include the pre- and postclimate shift periods that took place in the late 1970s to be more representative (e.g., Zhang et al. 2008). As demonstrated by Barnett et al. (1993) and Syu et al. (1995), wind responses to a given SST anomaly are very different from one season to another, which can have an important effect on ENSO evolution. The seasonality is therefore taken into account in the wind stress model by constructing seasonally dependent models for τinter: the SVD analyses are performed separately for each calendar month, and thus the τinter model consists of 12 different submodels, one for each calendar month (e.g., Zhang and Zebiak 2004; Zhang and Busalacchi 2005). To achieve reasonable amplitudes, the first five leading SVD modes are retained in estimating τinter fields from SST anomalies. This wind stress anomaly model has been used for coupled ocean–atmosphere modeling studies in the tropical Pacific by Zhang et al. (2003, 2005, 2006).
2) An FWF anomaly model
Observations and modeling studies indicate that interannual variations in SST and FWF are well correlated over the tropical Pacific, with a dominant SST control on FWF. Figure 2b shows interannual P − E anomaly fields obtained from the 24-member ensemble mean of the ECHAM4.5 AGCM simulations forced by observed SST. In the tropical Pacific, changes in P and E both contribute to variations in FWF. However, interannual variability of FWF is dominated by that of P. So, the resultant interannual variations in P − E look very like those in P, with only a small offset because of anomalous E contribution. For example, during the winter of the 1982/83 El Niño, P increases significantly in the central-eastern tropical Pacific over a very broad region with a maximum enhancement of about 10 mm day−1 in the central basin. Comparisons of Fig. 2b with Fig. 2a clearly indicate that variations in FWF closely follow those in SST during ENSO evolution. During El Niño, a warm SST anomaly causes an increase in P and E both. An anomalous increase in FWF into the ocean (a positive anomaly) is seen because of the dominance of P over E. During La Niña, a cold SST anomaly is accompanied by a negative FWF anomaly (a net loss of freshwater from the ocean), which is attributed to a deficit in P with reduced E. Thus, a clear positive correlation exists between interannual variations in SST and FWF, which provides a physical basis for constructing a feedback model for FWF response to SST anomalies.
An anomaly model for (P − E)inter can thus be derived statistically to relate FWF variability to large-scale SST anomalies. To determine statistically optimized empirical modes of their covariability, an SVD technique is adopted. Monthly data are first normalized by their spatially averaged standard deviation to form the covariance matrix. The SVD analysis is then performed on all time series data irrespective of season to get singular vectors, singular values, and the corresponding time coefficients. The SVD analysis was performed for the period 1963–96 (a total of 34 yr of data) using the SST data (Fig. 2a) and FWF anomaly fields estimated from the ECHAM4.5 ensemble simulations (Fig. 2b).
Figure 3 illustrates the singular values of modes 1–10 for the invariant SVD analysis (the P and E fields are analyzed in unit of mm day−1). The singular values represent the squared covariance accounted for by each pair of singular vectors. The first five singular values have values of about 1932, 412, 184, 173, and 129 with the squared covariance fraction of about 57%, 12%, 5%, 5%, and 4%, respectively.
Figure 4 shows the spatial patterns of the first pair of singular vectors and its associated time series. The temporal expansion coefficients (Fig. 4c) clearly indicate that the first mode describes interannual variability associated with ENSO events. El Niño is accompanied by a positive FWF anomaly (anomalous FWF into the ocean). The FWF changes sign during ENSO evolution. A negative anomaly (anomalous loss of freshwater out of the ocean) is seen during La Niña. The corresponding spatial patterns (Figs. 4a,b) indicate that the primary coupled mode of the variability is composed of a large FWF anomaly in the central Pacific that covaries with anomalous SST in the eastern and central equatorial Pacific. Striking differences are evident in the spatial structure of interannual anomalies between SST and FWF. The largest variability center of SST is located in the eastern equatorial Pacific; that of the FWF is centered over the central equatorial Pacific around 170 °W off the equator at 10 °N. This clearly represents a nonlocal nature of FWF response to SST anomalies. The second mode also shows a coherent relationship between the two fields both in time and space, with the spatial patterns representing different stages of ENSO evolution (figures not shown). The higher modes are typically smaller in amplitude, with less coherent structure in space and higher-frequency variability in time.
Then, we can develop an empirical FWF model using the derived spatial eigenvectors of the SVD modes (e.g., Zhang et al. 2006). The (P − E)inter model is constructed with the same base period as the τinter model (i.e., from 1963 to 1996). While the seasonality of wind response to SST anomalies is included in the τinter model because of its importance to ENSO evolution, we use a seasonally nondependent (P − E)inter model for simplicity, in which the SVD analysis is performed on all time series data irrespective of season. From the consideration of the sequence of the singular values (Fig. 3) and the reconstruction testing of the FWF anomaly fields from SST anomalies, the first two leading SVD modes are retained in the empirical model for having reasonable amplitude in the simulation. Thus, given a SST anomaly, the atmospheric FWF response can be calculated from the empirical model.
Figure 2c shows the FWF anomalies reconstructed using the empirical FWF model from the given SST anomalies (Fig. 2a). The model very well captures large-scale interannual P − E variability associated with ENSO evolution. For example, the spatial structure represents the pattern of large-scale FWF variability at the mature phase of El Niño or La Niña events, with largest anomalies over the ITCZ in the central and eastern tropical Pacific. Figure 5 presents the horizontal distribution of the standard deviation of interannual FWF variability and the percentage explained by the empirical model with the first two SVD modes retained. The amplitude of the reconstructed FWF anomalies (Fig. 2c) is comparable to the original field (Fig. 2a), with the reconstructed variance being more than 70% in the central and eastern tropics. Thus, most of the variance can be captured by the SVD-based model with the first two modes retained. This indicates that the first two SVD modes can be sufficient for recovering reasonable strength of the FWF variability. However, the simulated FWF anomalies are somewhat weaker, smoother, and less noisy, indicating that the selected SVD modes effectively act as a low-pass filter.
c. Coupling procedure
As shown in Fig. 1, the layer OGCM is coupled to two empirical submodels for the interannual variability of wind stress and FWF, while an advective AML model is used to estimate sea surface heat fluxes (Seager et al. 1995; Murtugudde et al. 1996). In the context of the HCM, atmospheric climatological forcing fields to the OGCM are specified from observations; they include τclim, Pclim, solar radiation, cloud, and wind speed. The climatological E field (Eclim) is estimated using the AML model from simulated SSTclim fields of the OGCM.
The coupling between the atmospheric components and the OGCM is as follows. At each time step, the OGCM calculates SSTs, which are averaged to obtain daily mean fields. The corresponding large-scale interannual SST anomalies (SSTinter) are obtained relative to its SSTclim fields that are predetermined from the spinup OGCM run forced by observed τclim fields. The two interannual anomaly fields [τinter and (P − E)inter] are calculated using the corresponding empirical submodels from the SST anomalies. The τinter and (P − E)inter fields are added to their corresponding prescribed τclim and (P − E)clim fields to force the OGCM. The (P − E)inter and τinter fields are updated every day from the corresponding large-scale SSTinter fields. In the coupled simulation, heat flux is calculated using the AML model from the OGCM SST and is updated every time step. The heat flux and freshwater flux are used to calculate the buoyancy flux (QB) to force a mixed layer model which is explicitly embedded in the OGCM (Chen et al. 1994).
The coupled experiment is initiated from the OCGCM spinup run, with an imposed westerly wind anomaly for eight months. Evolution of anomalous conditions thereafter is determined solely by coupled interaction in the system; as shown in Zhang et al. (2006), the model can sustain an interannual oscillation with about a 4-yr period. Then, all sensitivity experiments for this paper start with the same initial conditions that are chosen arbitrarily from a fully coupled, long-term run, denoted as year 24.
In addition, as examined previously by numerous studies (e.g., Barnett et al. 1993; Syu et al. 1995), coupled behavior of interannual variability in the tropical Pacific depends on the so-called relative coupling coefficient (ατ); that is, the wind stress anomalies calculated using the empirical model from SSTinter anomalies can be multiplied by a scalar parameter before being added to the climatological wind stress fields to drive the OGCM. Several tuning experiments have been performed with varying values of ατ to examine their effects on the coupled interannual variability. As shown in Zhang et al. (2006), the HCM with ατ = 1.2 produces a sustained interannual variability with an oscillation period of 4 yr. In this paper, we choose ατ = 1.2 for all experiments shown below. Similarly, interannual FWF anomalies calculated using the empirical model can also be multiplied by a scalar parameter (αFWF), which represents the strength of anomalous FWF forcing. As shown in the reconstruction (Figs. 2b,c), the empirical model with αFWF = 1.0 produces a reasonable FWF variability in the tropical Pacific (but apparently the amplitude is weaker as compared with the original data). In this paper, different values of αFWF will be used to quantify the effect of anomalous FWF forcing on modulating interannual variability in the HCM.
3. A simulation with interannually varying FWF forcing—A standard run
The HCM will be used to examine the effects of FWF forcing on interannual variability in the tropical Pacific. When the FWF forcing is included in the HCM (Fig. 1), the total FWF exchange between the atmosphere and ocean can be written as FWF = (P − E)clim+ αFWF(P − E)inter, in which its climatological part, (P − E)clim, is specified (the Pclim is from observation; the Eclim is estimated from simulated SSTclim fields using the advective AML model). Its anomalous part, (P − E)inter, is calculated using the SVD-based empirical model from interannual SST anomalies. The coefficient, αFWF, represents the strength of the anomalous FWF forcing. This interannual (P − E)inter forcing simulation with αFWF =1 is referred to as a standard run.
a. The overview of simulated interannual variability
Basically, the HCM with αFWF = 1 can produce quite well the mean ocean climatology and its interannual oscillations with about a 4-yr period. Examples are shown in Figs. 6 –8 for total fields of SST, SSS, and MLD from this simulation. Climatological features in the region include the warm pool in the west and the cold tongue in the east (Fig. 6). Seasonally, large variations in SST are in the eastern equatorial Pacific: a warming takes place during the spring and a cooling occurs during the fall. Fresh waters are located in the far western equatorial Pacific, and saline waters in the central basin, with a front near the date line (Fig. 7). The mixed layer is deep in the west but shallow in the east (Fig. 8).
Interannual variability is associated with ENSO events, which are predominantly determined by the coupling among SST, winds, and the thermocline. Large longitudinal displacements are clearly evident of the warm/fresh pool in the west and the cold tongue in the eastern equatorial Pacific (Figs. 6 and 7). For example, during El Niño, the cold tongue shrinks in the east; warm waters in the west extend eastward along the equator, with the 26°C isotherm of SST being located east of 130 °W. During La Niña, the cold tongue develops anomalously strongly in the east and expands westward along the equator while the warm pool retreats to the west, with the 25°C isotherm of SST being located west of 150°W. SSS has largest variability around the eastern edge of the warm pool near the date line (Fig. 7). Associated with ENSO events, the SSS front also moves back and forth along the equator. For example, during El Niño, a freshening occurs in the western and central basin, accompanied by its extension eastward beyond the date line. In the central and western equatorial Pacific, the mixed layer is anomalously deep during La Niña but shallow during El Niño (Fig. 8), respectively.
b. An analysis of freshwater flux and buoyancy flux
Parameters influenced directly by FWF in the ocean include SSS and buoyancy flux (QB) fields. The former is a state variable of the ocean which is controlled by the conservation equation of salt and is an important variable to determine the oceanic density field, which in turn influences the upper-ocean stability and the vertical mixing at the base of the mixed layer. The latter, QB, acts as a forcing field at the ocean–atmosphere interface that, together with heat flux and wind, controls the evolution of MLD which affects the entrainment of subsurface cold water into the mixed layer.
Model outputs are used to analyze these fields to understand the FWF-induced direct effects on MLD, an important indicator for the strength of the mixing and entrainment in the equatorial Pacific. Figure 9 shows the climatological distribution of the two QB components and its total field along the equator from the standard run. Their units are all expressed in 10−6 kg s−1 m−2, but can be correspondingly converted to the commonly used units for heat flux and FWF (e.g., 1.0 × 10−6 kg s−1 m−2 is equivalent to 13.3 W m−2 for heat flux in Fig. 9a or to 3.4 mm day−1 for P − E in Fig. 9b).
In the western tropical Pacific, there is a large surplus of P over E on average, with a net gain of freshwater into the ocean (Fig. 9b). The (P − E)clim field changes sign from being positive in the west to being negative in the east with the zero line crossing around the date line. To the east, E exceeds P by over 10 mm month−1, with a net loss of freshwater out of the ocean. The QT (Fig. 9a) and QS (Fig. 9b) fields are mostly of opposite sign in the eastern equatorial Pacific, but are of the same sign in the western regions. In the east, the amplitude of QT is about one order larger than that of QS; thus, the climatological QB field is dominated by QT, with a small offset by QS. In the western and central tropical Pacific, the amplitudes of QT and QS fields are more or less on the same order; thus, QS can be a significant contributor to QB as well.
Figure 10 illustrates the interannual anomalies along the equator for these buoyancy flux components. Since the magnitudes of QT and Qs anomaly fields are on the same order, their variations both contribute to interannual QB variability in the equatorial Pacific. As seen, large interannual variations in QT and QS fields are coherently associated with SST anomalies produced by ENSO cycles. During El Niño, warm SST anomalies exist in the central and eastern equatorial Pacific, which are accompanied by a negative QT anomaly in the central basin (Fig. 10a) but a positive QS anomaly (Fig. 10b). (The anticorrelation between SST and QT showing damping effect of surface heat flux on SST has been pointed out by historical data analyses, e.g., Wang and McPhaden 2001.) Because of the dominance of QT over QS, a warming is associated with a net negative QB anomaly (Fig. 10c; an anomalous loss of QB out of the ocean). During La Niña, cold SST anomalies are associated with a positive QT anomaly in the central basin but a negative QS anomaly. The resultant QB anomaly is positive (an anomalous gain of QB into the ocean) due to the dominant QT contribution. Thus, a negative correlation exists between interannual variations in SST and QB and in QB and QS, but a positive correlation exists between those in QB and QT. Figure 11 illustrates the horizontal distribution of the standard deviation for interannual variability of the buoyancy flux components simulated from the standard run. Large variability centers of QT and QS are located in the central basin. Although the interannual variability of QB is dominated by QT in the tropical Pacific, the amplitude of QS can be more than 30% that of QT in the western and central regions. So, FWF can also be a significant contributor to QB in these regions. Since variations in QT and QS are out of phase, their contributions to QB are compensated for each other. Because of the dominance of QT over QS, the sign of interannual QB anomalies goes with that of QT (Figs. 10a,c)—it is negative during El Niño but positive during La Niña. Thus, QS is generally representing a modulation on QB during ENSO cycles. As a result, the compensating effect of the FWF forcing on QB leads to a reduction in variability of QB, with less negative during El Niño and less positive during La Niña, respectively.
c. Relationships among interannual anomaly fields
Interannual anomalies of various atmospheric and oceanic fields along the equator are shown in Figs. 10, 12, and 13a from the standard run. The evolution and the phase relationships among SST, winds, and thermocline have been extensively described in observations and modeling studies before (e.g., Zhang and Levitus 1997; Zhang and Zebiak 2004). For example, in terms of spatial structure, the largest variability regions of wind stress (Fig. 12b) and SSS (Fig. 12c) are in the central basin, while large SST anomalies (Fig. 12a) are located in the central and eastern equatorial Pacific. Variations in SSS show a dominant standing pattern (Fig. 12c) concentrated in the central basin near the date line. Furthermore, variations in SST and surface wind are nearly in phase temporally (Figs. 12a,b), while variations in the thermocline depth (or sea level) in the west have a phase lead relative to SST anomalies in the east (figures not shown). The coupled variability from this HCM simulation is the mixture of the SST mode and thermocline mode as discussed by Neelin and Jin (1993).
As has been shown in the SVD analysis (Fig. 4), FWF exhibits large interannual variations over the tropical Pacific, with a nonlocal positive correlation with SST during ENSO cycles (Figs. 10b and 12a). The ENSO-induced anomalous FWF is expected to have direct effects on SSS. As shown in Fig. 12c, the largest variability center of SSS is located in the central basin near the date line, where FWF variability is also large (e.g., Fig. 10b). During El Niño, SST is anomalously warm in the central and eastern basin (Fig. 12a); SSS is anomalously low in the western and central basin (Fig. 12c). The direct effect of the positive FWF anomaly (an anomalous FWF into the ocean) in the western and central basin is to reduce SSS. Correspondingly, the waters become fresher in the central basin. The freshening of the surface layer tends to stabilize the upper ocean and depress the mixing of heat at the base of the mixed layer. These oceanic processes lead to a warming in the surface layer. During the evolution into La Niña phase, SST is anomalously cold; SSS becomes high in the central basin. As the FWF perturbations change sign to be negative mainly because of the deficit of P in the central basin, the negative FWF anomaly acts to increase SSS and destabilize the upper layer and thus enhance the mixing of heat, which in turn reinforces the cooling during La Niña. Thus, interannual variations in FWF tend to enhance temporal variability of SSS, with their negative correlation during ENSO cycles.
Interannual FWF anomalies, through its contribution to QB, also have direct effects on MLD and thus the entrainment of subsurface water into the mixed layer (ML). The OGCM is equipped with an explicit bulk mixed layer model to directly calculate MLD (Chen et al. 1994). Figure 13a shows interannual variations of MLD along the equator from the standard run. A seesaw pattern is evident in zonal direction. During El Niño (La Niña), the ML is anomalously deep (shallow) east of about 150°W, but shallow (deep) to the west. In the western and central regions, the ENSO-related anomalous FWF forcing tends to reinforce these ENSO-related patterns of the MLD variability. Because of its out-of-phase relationships with variations in QT (Fig. 10a) and the compensation for QB, interannual QS anomalies (Fig. 10b) act to reduce the amplitude of QB variability during ENSO cycles (Fig. 10c). During El Niño, the positive FWF anomaly (Fig. 10b) tends to reduce the amplitude of the negative QB anomaly (Fig. 10c), which acts to reduce the depth of ML in the central region (Fig. 13a), leading to less entrainment of subsurface water into the mixed layer. These processes further act to increase SST. During La Niña, the ML is anomalously deep in the central basin. The negative FWF anomaly tends to reduce the positive QB anomalies in the central basin, which leads to an increase in MLD, acting to enhance the entrainment of subsurface cold water into the mixed layer. These processes in turn lead to a cooling of the surface layer. Thus, the MLD variations in the central basin are negatively correlated with the FWF anomalies. Since the changes in MLD induced by the FWF forcing in the central basin are of the same sign with those produced by ENSO, the anomalous FWF forcing tends to enhance temporal variability of MLD during ENSO cycles. This accordingly modulates the entrainment of subsurface cold water into the mixed layer and influence SSTs. The phase relationships among these anomaly fields indicate a positive feedback between FWF and SST in the coupled system.
4. Sensitivity experiments
We have performed two more HCM experiments using the identical OGCM that is coupled to the same SVD-based atmospheric wind stress and P − E models, but with differing αFWF values to represent the strength of anomalous FWF forcing. In this section, results from these HCM simulations will be further analyzed and compared to each other to illustrate the impacts of anomalous FWF forcing.
a. A simulation with climatological FWF forcing
An experiment is performed in which only the climatological FWF forcing is included in the HCM (αFWF = 0.0), referred to as the (P − E)clim run. Examples are shown in Figs. 6 –8 for the simulated total fields and in Figs. 14 and 13b for the simulated anomaly fields. Basic features of interannual variability are quite similar to those in the αFWF = 1.0 run, including SSS (Fig. 7b) and MLD (Fig. 8b) fields. This indicates that, to the first order, interannual variability in the tropical Pacific is driven predominantly by large-scale SST–wind coupling. As indicated, SSS and MLD still have large interannual variability in the western and central basin when interannual FWF forcing is excluded in the HCM.
While the exclusion of anomalous FWF forcing in the HCM has not caused a qualitative change in the simulated interannual variability, it does have a quantitative effect as compared with the standard run with αFWF = 1.0. This can be clearly seen not only in fields of SSS (Fig. 7) and MLD (Fig. 8) but also in those of SST (Fig. 6), winds (e.g., Fig. 14b), and other fields. Since FWF has large variations in the western and central basin, large effects are expected on SSS there. As compared with the standard (P − E)inter run (Fig. 7a), for example, SSS in the central basin from the climatological FWF run has much weaker temporal variability (Fig. 7b). Also, the east–west migration of the SSS front is less strong. Furthermore, the exclusion of anomalous FWF forcing results in a weaker variability of SST in the tropical Pacific, while being less pronounced in the zonal migration (Fig. 6). Clearly, excluding anomalous FWF forcing in the HCM leads to the simulated interannual variability that is weaker, including wind anomalies in the central basin (e.g., Figs. 12b and 14b). As a result, the amplitude of ENSO cycles is reduced.
Figure 15 further shows the difference fields along the equator between the two HCM simulations with the interannual and climatological FWF forcings. The net effects of anomalous FWF forcing in the (P − E)inter run are represented clearly in these difference fields mostly affected by FWF: SSS (Fig. 15a), MLD (Fig. 15b), SST (Fig. 15c), and heat flux (Fig. 15d). Coherent and systematic differences are evident in these two runs. For example, during El Niño, a positive perturbation of FWF in the central basin directly decreases SSS (a negative difference), which stabilizes the mixed layer and suppresses the vertical mixing. In the equatorial region, the weakened mixing leads to a positive SST difference (Fig. 15c). The positive FWF anomaly also results in a reduced negative QB anomaly at the ocean surface (Fig. 10c). Since MLD is determined by the ratio of wind generation of turbulent kinetic energy to the buoyancy flux in regions of net positive (downward) buoyancy flux (Fig. 9c), the increased anomalous buoyancy flux (a less negative QB anomaly due to the positive FWF anomaly) leads to a decrease in MLD in the western and central equatorial Pacific (Fig. 15b), with less entrainment of subsurface cold water into the mixed layer. This effect on MLD through QB also contributes to a positive difference in SST (Fig. 15c).
Similarly, during La Niña, a negative FWF anomaly in the central basin (i.e., a net FWF out of the ocean) leads to an increase in SSS and a reduced positive QB anomaly (a decrease in the anomalous buoyancy flux into the ocean). The former acts to increase the oceanic density fields, which destabilize the upper ocean and enhance the vertical mixing at the base of the mixed layer. The latter acts to increase MLD in the central basin, with more entrainment of cold subsurface water into the mixed layer. All these related processes result in a decrease in SST and thus a negative difference in SST (Fig. 15c).
In addition, the difference fields in the heat flux (Fig. 15d) are out of phase with those in SST; a negative (positive) difference in heat flux is associated with a positive (negative) difference in SST (Fig. 15c). This indicates that the net air–sea heat flux is representing a negative feedback to changes in SST, acting to damp SST variability (always reducing SST anomalies). Thus the positive (negative) differences in SST induced by the FWF forcing during El Niño (La Niña) should be attributed to oceanic processes as described above, with heat flux being simply representing a response to the SST changes.
Clearly, the difference fields of SSS, MLD, and heat flux in the two simulations are negatively correlated with those of FWF. The difference fields induced by anomalous FWF forcing during El Niño and La Niña tend to be of the same sign with interannual anomalies originally generated by ENSO cycles. In particular, the SST changes induced by the FWF forcing are enhancing the SST anomalies associated with El Niño and La Niña events. As a result, the oceanic processes induced by anomalous FWF forcing are acting in such a way to reinforce interannual variability in the coupled ocean–atmosphere system of the tropical Pacific.
b. An enhanced FWF forcing simulation
Next, we consider an enhanced FWF forcing run with αFWF = 2.0. Some examples of the results are highlighted in Figs. 13c, 16, and 17. As expected, interannual anomalies are stronger than those with αFWF = 1.0. For example, during El Niño, a larger positive FWF anomaly in the central basin (Fig. 17b) leads to a greater freshening of the mixed layer (Fig. 16c), which stabilizes the surface layer and reduce the cooling effects of the vertical mixing; the resultant less negative QB anomaly (Fig. 17c) also leads to a mixed layer that is shallower in the central basin (Fig. 13c), with less entrainment of subsurface water. These processes lead to a larger increase in SST in the central basin (Fig. 16a), acting to reinforce the warm SST anomaly produced by El Niño. As a result, the SST variability in the central-eastern basin is increased significantly (Fig. 16a). Similarly, during La Niña, the net effect of a larger negative FWF anomaly (Fig. 16b) tends to increase SSS (Fig. 16c) and MLD (Fig. 13c) in the central and western regions. These processes act to enhance more strongly the cold SST anomalies (Fig. 16a). Thus, an enhanced FWF forcing during ENSO cycles tends to increase more the temporal variability of SSS, MLD, and SST, leading to a stronger positive feedback in the coupled system. All these are in line with analyses described above.
It is interesting to note that the enhanced FWF forcing (Fig. 17b), as represented by doubling the coupling coefficient for (P − E)inter in the αFWF = 2.0 run, leads to an increase in variability of SST, SSS, and MLD but a decrease in that of QB (Fig. 17c). As analyzed above, variations in QB are attributed to those in QT and QS both, with its positive correlation with QT, but negative with QS. The advective AML model allows for interactive adjustment of the calculated heat flux to changes in SSTs, thus being able to realistically represent its feedback with MLD and SSTs (Seager et al. 1995; Murtugudde et al. 1996). As seen here, doubling the FWF feedback induces a larger SST variability (Fig. 16a), which, at the same time, leads to an increase in heat flux variability (Fig. 17a). As QT increases (Fig. 17a), its effects on QB increases, which would lead to an increase in QB. The results in Fig. 17 indicate that, although QT and QS both increase as a result of the enhanced FWF forcing, the amplitude of the QB variability does not increase as with QT; instead, the net contributions of QS and QT to QB result in a reduction in variability of QB (Fig. 17c), which influences MLD and the entrainment in the equatorial Pacific. The fact that the amplitude of QB variability decreases because of an increased contribution of QS (rather than increases as with an increased heat flux effect) indicates an active role the FWF forcing plays in modulating QB. In particular, the reduced negative (positive) QB anomaly during El Niño (La Niña) acts in such a way to enhance SST variability in the coupled system (Fig. 17a). Clearly, the FWF forcing is playing a different role from heat flux forcing, with the former acting to drive a change in SST, while the latter being representing a passive response to the SST change.
Here, the experiment with doubling the FWF-related feedback coefficient (αFWF) is for illustrative purpose to demonstrate the effect of its positive feedback not only on the ocean state, but also on other feedbacks in the system. For example, when the feedback coefficient for (P − E)clim is doubled, the heat flux forcing (which is dominated by evaporative coupling) is also changed (an increase in its variability). Since these forcing and feedbacks can all affect each other, the differences in the simulated interannual variability from these runs present a net effect of all related feedbacks in the coupled system, including the Bjerknes feedback and SST–heat flux feedback. The results from the enhanced (P − E)inter run as compared with other runs also have implication for understanding global warming effect since it is representing an accelerated hydrological cycle in the earth system.
5. An FWF-induced positive feedback
In this paper, three experiments have been made to examine the effect of anomalous FWF forcing. It has been seen that the inclusion of FWF forcing acts to enhance interannual variability in a coupled ocean–atmosphere system. The larger the anomalous FWF forcing, the stronger the interannual variability. Analyses of these runs indicate a positive feedback induced by anomalous FWF forcing in the coupled ocean–atmosphere context. In this section, the results of these three experiments are further contrasted to delineate the positive SST–FWF feedback more clearly.
The oceanic processes involved have been represented in the difference fields (Fig. 15), the major fields mostly affected by FWF forcing. A heat budget analysis has been preformed to understand the mechanism by which FWF-induced positive feedback is enhancing SST variability. As demonstrated above and other studies, air–sea heat flux provides a negative feedback to SST variability. Thus, the enhanced SST variability due to the inclusion of the FWF forcing must be attributed to oceanic processes. The primary mechanism proposed is a reduction of the vertical mixing and entrainment of subsurface water into the mixed layer in the central Pacific during warm events, which act to amplify the warming tendency on interannual time scales. Conversely, mixing and entrainment is enhanced during cold events. These are confirmed in Fig. 18 exhibiting the sum of the vertical mixing and advection terms at 160°W on the equator for the climatological run and enhanced (P − E)inter run, respectively. When the FWF feedback is included, these processes act to reduce the cooling effect on SST during El Niño (e.g., more warming in the years 2035–36), but to increase the cooling effect on SST during La Niña (e.g., more cooling in the years 2037–38). Note again that these differences, induced directly by the FWF feedback, should be viewed as a net effect of various feedbacks in the coupled system. More detailed heat budget analyses will be presented elsewhere.
The effects of anomalous FWF forcing on some selected variables are quantified in Table 1. As analyzed above, the relationships among interannual anomaly fields indicate a positive effect of FWF forcing on SST during ENSO cycles. In the standard (P − E)inter run (αFWF = 1.0), the positive feedback between SST and FWF is included in the HCM simulation. The FWF anomalies induce additional ocean processes in such a way to reinforce the warming during El Niño and cooling during La Niña. The enhanced SST anomalies further increase interannual variability in the coupled system. When the positive SST–FWF relationship is not included as in the climatological FWF forcing run (αFWF = 0.0), these additional oceanic effects that could be induced by the FWF forcing are disabled, and there is no positive feedback. As a result, the simulated interannual variability in the αFWF = 0.0 run is weakened as compared with that in the αFWF = 1.0 run. When the anomalous FWF forcing is enhanced as represented in the αFWF = 2.0 run, its obvious direct effects are to increase the temporal variability of SSS and MLD in the central basin. Also, it increases the compensating effect of QS on QT, with a net reduction in QB variability. All these processes tend to cause more warming during El Niño and more cooling during La Niña, which acts to reinforce SST variability during ENSO cycles. As the positive feedback is exaggerated, a stronger interannual variability emerges. Clearly, the oceanic processes induced by the FWF forcing and the related feedback act in such a way to enhance the strength of ENSO cycles.
Figure 19 further exhibits the standard deviation (std dev) of SSS and SST along the equator from the three experiments. These experiments indicate that SSS and SST variability is proportional to the strength of the feedback between SST and FWF. The stronger the feedback between SST and FWF, the stronger the interannual variability in SSS and SST. As further quantified in Table 1, the std dev of Niño-3 SST (Niño-4 SSS) anomalies is 0.76°C (0.16 psu) in the standard αFWF = 1.0 run, 0.92°C (0.28 psu) in the αFWF = 2.0 run, and 0.67°C (0.11 psu) in the αFWF = 0.0 run. Relative to the standard αFWF = 1.0 run, these values are representing an increase of 21% (75%) in the αFWF = 2.0 run but a reduction of 12% (31%) in the αFWF = 0.0 run. Also, the std dev of zonal wind stress at the Niño-4 site is 0.19 dyn cm−2 (0.019 Pa) in the αFWF = 1.0 run; it increases to 0.23 dyn cm−2 (0.023 Pa) in the αFWF = 2.0 run (an increase by 21%) but reduces to 0.16 dyn cm−2 (0.016 Pa) in the αFWF = 0.0 run (a reduction by 16%). Thus, a significant fraction of the SSS and SST variability can be attributed to anomalous FWF forcing.
6. Summary and discussion
Most previous modeling studies have focused on the roles of atmospheric forcing components of winds and heat flux; FWF forcing and its related salinity role in coupled climate variability have not received much attention. Furthermore, the effects of FWF forcing have been examined mostly in forced ocean-only modeling studies. In this work, the impacts of FWF forcing on salinity and interannual variability are examined in a hybrid coupled ocean–atmosphere context in which climatological atmospheric forcing fields of wind and FWF are specified, while their anomaly parts can be added on or removed separately or collectively.
In this work, an empirical model is developed to specifically relate the (P − E)inter field to SST anomalies using a singular value decomposition (SVD) analysis of their historical data. Results indicate that this empirical FWF model can realistically simulate the nonlocal FWF response to SST anomalies. This submodel for (P − E)inter is then embedded into our previous HCM of the tropical Pacific climate system (Zhang et al. 2006) to represent the effect of anomalous FWF forcing.
We have designed various experiments using the HCM with differing strengths of anomalous FWF forcing. Three cases are considered. In a standard simulation, the climatological and anomalous FWF fields are both taken into account [i.e., FWF = (P − E)clim + (P − E)inter]. The constructed HCM can well reproduce interannual variability associated with ENSO in the tropical Pacific. Two more sensitivity experiments are then performed using the HCM with the climatological forcing only [i.e., FWF = (P − E)clim] and an enhanced FWF forcing run [i.e., FWF = (P − E)clim + 2.0(P − E)inter].
Interannual variability in the tropical Pacific is predominantly driven by wind-induced feedback; simulations with climatological FWF forcing still show large interannual variability of SSS and SST in the tropical Pacific with basic feature unchanged. This indicates that wind forcing is of primary importance for dynamics of interannual variability in the tropical Pacific climate system. However, a significant effect can be seen, arising from anomalous FWF forcing. Quantitatively, taking the (P − E)inter run as a standard, the SST variance at the Niño-3 site can be reduced by about 12% in the climatological FWF forcing run, but enhanced by 21% in the enhanced FWF forcing run; the variances for SSS and zonal wind stress at the Niño-4 site are reduced by about 31% and 16% in the climatological FWF run, but enhanced by 16% and 75% in the enhanced FWF forcing run, respectively. Thus, anomalous FWF forcing can modulate interannual variability in a substantial way.
Processes responsible for the effects are analyzed. A positive correlation exists between interannual variations in SST and FWF during ENSO cycles, which, as indicated by Fig. 4, is nonlocal. Behind this statistical relationship is a positive feedback between SST and FWF in the coupled ocean–atmosphere system of the tropical Pacific. Large FWF anomalies in the western and central Pacific directly affect SSS and also MLD through its contribution to buoyancy flux, and are thus able to modify the stability of the upper layer and the mixing and entrainment of subsurface cold waters into the mixed layer in the equatorial Pacific. These oceanic processes act to influence SST, which in turn feeds back to the atmosphere. Furthermore, this feedback can modulate the strength of the Bjerknes feedback. As a result, the ENSO cycle is enhanced because of the inclusion of anomalous FWF forcing in a coupled ocean–atmosphere model. In contrast, heat flux provides a negative feedback, acting to damp interannual variability.
These results are useful not only for a physical understanding of the FWF-induced effects on the ocean dynamics and ENSO, but also for improvements in coupled models. As clearly evident in Figs. 19b and 14a, our HCM produces unrealistic spatial distribution of interannual standard deviation of SST along the equator, with its maximum around the date line and second maximum in the eastern Pacific (between 120° and 100°W). This is different from observations, which show almost monotonically increased interannual standard deviation toward the east along the equator (e.g., Fig. 2a). The weakness of SST variability in the eastern tropical Pacific is a common bias in many coupled models. Sensitivity experiments with different strengths of FWF feedback indicate a clear effect on SST simulation. For example, a stronger FWF feedback leads to an enhanced SST variability in the eastern equatorial Pacific (Figs. 12a, 14a, and 16a). Thus, FWF forcing can be a clear source for model biases in SST simulation in the tropical Pacific and thus needs to be taken into account adequately, which provides the potential for model improvement in coupled models. However, including FWF feedback has not improved the unrealistic spatial distribution of interannual SST along the equator. Actually, the similarities among three curves in Fig. 19b imply the missing FWF and associated feedback is not the main source for such unrealistic SST distribution. Such unrealistic SST distribution cannot just be neglected and needs to be investigated more carefully for its source. Similarly, this positive feedback has not been included in most previous ICMs and HCMs used for ENSO simulation and prediction. The simple empirical model we have constructed in this paper can be applied to these simplified coupled ocean–atmosphere models to take into account the anomalous FWF forcing. The inclusion of this atmospheric forcing component is expected to improve SST simulations and prediction in the coupled system. Moreover, the work presented here has demonstrated the need for a rigorous diagnosis of the hydrological cycle in coupled ocean–atmosphere models.
Additional modeling studies are under way. In this paper, we intend to check the impact of FWF-induced feedback through running three experiments for the coupled ocean–atmosphere system of the tropical Pacific. Three feedbacks between SST–wind, SST–heat flux, and SST–FWF are all active and can affect one another. The large differences induced by the inclusion of the FWF forcing represent a net effect of all these related feedbacks in the system, not only directly but also indirectly. Further experiments are needed to isolate the direct effect of the SST–FWF feedback and its sole contribution to SST by disabling other feedbacks. For example, in the three experiments presented above, both climatological run (αFWF = 0.0) and enhanced run (αFWF = 2.0) receive different wind stress and heat flux forcing from the standard run (αFWF = 1.0). What if the same wind stress and heat flux forcing derived from the standard run is also applied to the other two sensitivity runs? By doing so, only “direct” SST–FWF feedback is kept active and can be focused on. In addition, some other feedbacks are still not taken into account in our HCM, such as the wind–evaporation–SST feedback, the SST–solar radiation feedback, bioclimate feedback (e.g., Ballabrera-Poy et al. 2007), and tropical instability wave (TIW)-induced wind feedback (e.g., Zhang and Busalacchi 2008). The demonstrated positive feedback induced by FWF forcing needs to be combined with these feedbacks to examine their net effects on ENSO simulations in the coupled ocean–atmosphere context. Also, we only demonstrate here the effects on simulations of interannual variability; its effects on ENSO prediction and predictability need to be examined further.
Acknowledgments
We thank Drs. J. Carton, D. G. DeWitt, P. Arkin, B. Huang, S. Zebiak, P. Chang, L. Wu, S.-P. Xie, Ballabrera-Poy, and E. Hackert for their comments. The authors wish to thank anonymous reviewers for their numerous comments that helped to improve the original manuscript. We appreciate the assistance from Drs. R. Murtugudde and D. Chen for their expert advice in using the OGCM, and from Dr. M. Tippett at IRI for providing the ECHAM ensemble mean wind stress and P and E data. This research is supported in part by an NSF Grant ATM-0727668 and NASA Grants NCC5374, NAG512246, and NNX08AI76G.
REFERENCES
Ballabrera-Poy, J., R. G. 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 Pacific. J. Climate, 20 , 353–374.
Barnett, T. P., M. Latif, E. Kirk, and E. Roeckner, 1991: On ENSO physics. J. Climate, 4 , 487–515.
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 Pacific sea surface temperature with a hybrid coupled ocean–atmosphere model. J. Climate, 6 , 1545–1566.
Bjerknes, J., 1969: Atmospheric teleconnections from the equatorial Pacific. Mon. Wea. Rev., 97 , 163–172.
Boyer, T. P., J. I. Antonov, S. Levitus, and R. Locarnini, 2005: Linear trends of salinity for the World Ocean, 1955–1998. Geophys. Res. Lett., 32 , L01604. doi:10.1029/2004GL021791.
Carton, J. A., 1991: Effect of seasonal surface freshwater flux on sea surface temperature in the tropical Atlantic Ocean. J. Geophys. Res., 96 , 12593–12598.
Chang, P., L. Ji, and R. Saravanan, 2001: A hybrid coupled model study of tropical Atlantic variability. J. Climate, 14 , 361–390.
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.
Cooper, N. S., 1988: The effect of salinity on tropical ocean models. J. Phys. Oceanogr., 18 , 697–707.
Cronin, M. F., and M. J. McPhaden, 1999: Diurnal cycle of rainfall and surface salinity in the western Pacific warm pool. Geophys. Res. Lett., 26 , 3465–3468.
Delcroix, T., S. Cravatte, and M. J. McPhaden, 2007: Decadal variations and trends in tropical Pacific sea surface salinity since 1970. J. Geophys. Res., 112 , C03012. doi:10.1029/2006JC003801.
Eckert, C., and M. Latif, 1997: Predictability of a stochastically forced hybrid coupled model of El Niño. J. Climate, 10 , 1488–1504.
Fedorov, A. V., R. C. Pacanowski, S. G. H. Philander, and G. Boccaletti, 2004: The effect of salinity on the wind-driven circulation and the thermal structure of the upper ocean. J. Phys. Oceanogr., 34 , 1949–1966.
Gent, P., and M. A. Cane, 1989: A reduced gravity, primitive equation model of the upper equatorial ocean. J. Comput. Phys., 81 , 444–480.
Hackert, E. C., A. J. Busalacchi, and R. Murtugudde, 2001: A wind comparison study using an ocean general circulation model for the 1997–1998 El Niño. J. Geophys. Res., 106 , (C2). 2345–2362.
Harrison, E. F., P. Minnis, B. Barkstorm, and G. Gibson, 1993: Radiation budget at the top of the atmosphere. Atlas of Satellite Observations Related to Global Change, R. J. Gurney, J. L. Foster, and C. Parkinson, Eds., Cambridge University Press, 19–38.
Harrison, M. J., A. Rosati, B. J. Soden, E. Galanti, and E. Tziperman, 2002: An evaluation of air–sea flux products for ENSO simulation and prediction. Mon. Wea. Rev., 130 , 723–732.
Huang, B., and V. M. Mehta, 2004: Response of the Indo-Pacific warm pool to interannual variations in net atmospheric freshwater. J. Geophys. Res., 109 , C06022. doi:10.1029/2003JC002114.
Huang, B., and V. M. Mehta, 2005: Response of the Pacific and Atlantic oceans to interannual variations in net atmospheric freshwater. J. Geophys. Res., 110 , C08008. doi:10.1029/2004JC002830.
Huang, R. X., 1993: Real freshwater flux as a natural boundary condition for salinity balance and thermohaline circulation forced by evaporation and precipitation. J. Phys. Oceanogr., 23 , 2428–2446.
Kessler, W. S., 1999: Interannual variability of the subsurface high salinity tongue south of the equator at 165°E. J. Phys. Oceanogr., 29 , 2038–2049.
Lagerloef, G. S. E., 2002: Introduction to the special section: The role of surface salinity on upper ocean dynamics, air-sea interaction and climate. J. Geophys. Res., 107 , 8000. doi:10.1029/2002JC001669.
Levitus, S., J. Antonov, J. Wang, T. L. Delworth, K. W. Dixon, and A. J. Broccoli, 2001: Anthropogenic warming of earth’s climate system. Science, 292 , 267–270.
Lukas, R., 2001: Freshening of the upper thermocline in the North Pacific subtropical gyre associated with decadal changes of rainfall. Geophys. Res. Lett., 28 , 3485–3488.
Lukas, R., and E. Lindstrom, 1991: The mixed layer of the western equatorial Pacific Ocean. J. Geophys. Res., 96 , 3343–3357.
Luo, Y., L. M. Rothstein, R-H. Zhang, and A. J. Busalacchi, 2005: On the connection between South Pacific subtropical spiciness anomalies and decadal equatorial variability in an ocean GCM. J. Geophys. Res., 110 , C10002. doi:10.1029/2004JC002655.
Maes, C., 2000: Salinity variability in the equatorial Pacific Ocean during the 1993–98 period. Geophys. Res. Lett., 27 , 1659–1662.
Maes, C., J. Picaut, and S. Belamari, 2002: Salinity barrier layer and onset of El Niño in a Pacific coupled model. Geophys. Res. Lett., 29 , 2206. doi:10.1029/2002GL016029.
McCreary, J. P., and D. L. T. Anderson, 1991: An overview of coupled ocean–atmosphere models of El Niño–Southern Oscillation. J. Geophys. Res., 96 , (Suppl.). 3125–3150.
Miller, J. R., 1976: The salinity effect in a mixed layer ocean model. J. Phys. Oceanogr., 6 , 29–35.
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.
Neelin, J. D., 1990: A hybrid coupled general circulation model for El Niño studies. J. Atmos. Sci., 47 , 674–693.
Neelin, J. D., and F-F. Jin, 1993: Modes of interannual tropical ocean–atmosphere interaction—A unified view. Part II: Analytical results in the weak-coupling limit. J. Atmos. Sci., 50 , 3504–3522.
Reason, C. J. C., 1992: On the effect of ENSO precipitation anomalies in a global ocean GCM. Climate Dyn., 8 , 39–47.
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 für 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 the subtropical and tropical Pacific Ocean. J. Phys. Oceanogr., 28 , 322–343.
Schiffer, R. A., and W. B. Rossow, 1985: ISCCP Global Radiance Set: A new resource for climate research. Bull. Amer. Meteor. Soc., 66 , 1498–1505.
Schmitt, R. W., P. S. Bogden, and C. E. Dorman, 1989: Evaporation minus precipitation and density fluxes for the North Atlantic. J. Phys. Oceanogr., 19 , 1208–1221.
Schneider, N., 2000: A decadal spiciness mode in the tropics. Geophys. Res. Lett., 27 , 257–260.
Schneider, N., and T. P. Barnett, 1995: The competition of freshwater and radiation in forcing the ocean during El Niño. J. Climate, 8 , 980–992.
Seager, R., M. Blumenthal, and Y. Kushinir, 1995: An advective atmospheric mixed layer model for ocean modeling purposes: Global simulation of surface heat fluxes. J. Climate, 8 , 1951–1964.
Smith, T. M., and R. W. Reynolds, 2004: Improved extended reconstruction of SST (1854–1997). J. Climate, 17 , 2466–2477.
Sprintall, J., and M. Tomczak, 1992: Evidence of the barrier layer in the surface layer of the tropics. J. Geophys. Res., 97 , (C5). 7305–7316.
Syu, H-H., J. D. Neelin, and D. Gutzler, 1995: Seasonal and interannual variability in a hybrid coupled GCM. J. Climate, 8 , 2121–2143.
Vialard, J., and P. Delecluse, 1998a: An OGCM study for the TOGA decade. Part I: Role of salinity in the physics of the western Pacific fresh pool. J. Phys. Oceanogr., 28 , 1071–1088.
Vialard, J., and P. Delecluse, 1998b: An OGCM study for the TOGA decade. Part II: Barrier-layer formation and variability. J. Phys. Oceanogr., 28 , 1089–1106.
Waliser, D. E., B. Blanke, J. D. Neelin, and C. Gautier, 1994: Shortwave feedbacks and El Niño–Southern Oscillation: Forced ocean and coupled ocean–atmosphere experiments. J. Geophys. Res., 99 , 25109–25125.
Wang, W., and M. J. McPhaden, 2001: Surface layer temperature balance in the equatorial Pacific during the 1997–98 El Niño and 1998–99 La Niña. J. Climate, 14 , 3393–3407.
Weller, R. A., and S. P. Anderson, 1996: Surface meteorology and air-sea fluxes in the western equatorial Pacific warm pool during the TOGA Coupled Ocean-Atmosphere Response Experiment. J. Climate, 9 , 1959–1990.
Xie, P., and P. Arkin, 1995: An intercomparison of gauge observations and satellite estimates of monthly precipitation. J. Appl. Meteor., 34 , 1143–1160.
Xie, S-P., and S. G. H. Philander, 1994: A coupled ocean-atmosphere model of relevance to the ITCZ in the eastern Pacific. Tellus, 46A , 340–350.
Yang, S., K-M. Lau, and P. S. Schopf, 1999: Sensitivity of the tropical Pacific ocean to precipitation-induced freshwater flux. Climate Dyn., 15 , 737–750.
Yeager, S. G., and W. G. Large, 2007: Observational evidence of winter spice injection. J. Phys. Oceanogr., 37 , 2895–2919.
Yu, L., and R. A. Weller, 2007: Objectively analyzed air–sea heat fluxes for the global ice-free oceans (1981–2005). Bull. Amer. Meteor. Soc., 88 , 527–539.
Zebiak, S. E., and M. A. Cane, 1987: A model El Niño–Southern Oscillation. Mon. Wea. Rev., 115 , 2262–2278.
Zhang, R-H., and S. Levitus, 1997: Interannual variability of the coupled tropical Pacific Ocean–atmospheric system associated with the El Niño–Southern Oscillation. J. Climate, 10 , 1312–1330.
Zhang, R-H., and S. E. Zebiak, 2004: An embedding method for improving interannual variability simulations in a hybrid coupled model of the tropical Pacific ocean–atmosphere system. J. Climate, 17 , 2794–2812.
Zhang, R-H., and A. J. Busalacchi, 2005: Interdecadal changes in properties of El Niño in an intermediate coupled model. J. Climate, 18 , 1369–1380.
Zhang, R-H., and A. J. Busalacchi, 2008: Rectified effects of tropical instability wave (TIW)-induced atmospheric wind feedback in the tropical Pacific. Geophys. Res. Lett., 35 , L05608. doi:10.1029/2007GL033028.
Zhang, R-H., T. Kagimoto, and S. E. Zebiak, 2001: Subduction of decadal North Pacific 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 Niño simulation and prediction. Geophys. Res. Lett., 30 , 2012. doi:10.1029/2003GL018010.
Zhang, R-H., S. E. Zebiak, R. Kleeman, and N. Keenlyside, 2005: Retrospective El Niño forecast using an improved intermediate coupled model. Mon. Wea. Rev., 133 , 2777–2802.
Zhang, R-H., A. J. Busalacchi, and R. G. Murtugudde, 2006: Improving SST anomaly simulations in a layer 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) in decadal modulation of ENSO. J. Climate, 21 , 674–704.
A schematic diagram illustrating an HCM for the tropical Pacific ocean–atmosphere system, consisting of an OGCM and a simplified atmospheric model, whose forcing fields to the ocean include three components: wind stress, freshwater flux, and heat flux. The total wind stress (τ) consists of prescribed climatological wind stress (τclim) from observations and its interannual anomalies (τinter) associated with large-scale SST anomalies (SSTinter); the total FWF, represented by P − E, consists of prescribed climatological freshwater flux [(P − E)clim] and its interannual anomalies [(P − E)inter]; heat flux (HF) is calculated using the Seager et al. (1995) advective AML model. Empirical submodels for τinter and (P − E)inter fields are constructed using an SVD analysis. Buoyancy flux (QB) is calculated from the heat flux and freshwater flux to force a mixed layer model, which is embedded in the OGCM. Climatological SST (SSTclim) fields are specified from a spinup run of the OGCM forced by observed climatological atmospheric fields.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Anomaly fields along the equator during the period 1980–96: (a) observed SST from Reynolds et al. (2002), (b) P − E from the ensemble mean of 24-member ECHAM4.5 AGCM simulations, and (c) P − E constructed using the SVD-based empirical model from the SST anomalies shown in (a). The contour interval is 0.5°C in (a) and 2 mm day−1 in (b) and (c).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Singular values of modes 1–10 from the SVD analysis that is based on the covariance matrix calculated from time series of observed SST and modeled P − E anomaly fields as shown in Figs. 2a,b, respectively.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Spatial patterns derived from the SVD analysis for the first pair of singular vectors of (a) SST and (b) P − E fields, and (c) the time series associated with the first SVD mode. The contour interval is 0.2 in (a) and (b).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
(a) Horizontal distribution of the std dev for interannual (P − E)inter variability obtained from the ensemble ECHAM4.5 simulation and (b) the percentage of total (P − E)inter variance explained by the empirical model with the first two leading SVD modes retained. The contour interval is 0.5 mm day−1 in (a) and 10% in (b).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Total SST fields along the equator simulated from the HCM with (a) the interannual FWF forcing and (b) the climatological FWF forcing. The contour interval is 1.0°C.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
The same as in Fig. 6 but for total SSS fields. The contour interval is 0.1 psu.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
The same as in Fig. 6 but for the depth of the mixed layer. The contour interval is 5 m.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Climatological distribution of buoyancy flux along the equator simulated from the HCM with the interannual FWF forcing: (a) the heat flux part (QT), (b) the freshwater flux part (QS), and (c) the total field (QB). The contour interval is 0.8 × 10−6 kg s−1 m−2 in (a) and (c), and 0.4 × 10−6 kg s−1 m−2 in (b).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Anomalies along the equator of buoyancy flux simulated from the HCM with the interannual FWF forcing: (a) the heat flux part (QT), (b) the freshwater flux part (QS), and (c) the total field (QB). The contour interval is 0.5 × 10−6 kg s−1 m−2.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Horizontal distribution of the std dev for interannual variability of buoyancy flux simulated from the HCM with the interannual FWF forcing: (a) the heat flux part, (b) the freshwater flux part, and (c) the total field. The contour interval is 0.3 × 10−6 kg s−1 m−2.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Anomalies along the equator simulated from the HCM with the interannual FWF forcing: (a) SST, (b) zonal wind stress, and (c) SSS. The contour interval is 0.5°C in (a), 0.1 dyn cm−2 in (b), and 0.1 psu in (c).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Anomalies of the MLD along the equator simulated from the HCM with (a) the interannual FWF forcing, (b) the climatological FWF forcing, and (c) the enhanced interannual FWF forcing. The contour interval is 4 m.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
The same as in Fig. 12 but for the climatological FWF forcing.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
The difference fields along the equator between the HCM simulations with the interannual and climatological FWF forcings: (a) SSS, (b) MLD, (c) SST, and (d) heat flux. The contour interval is 0.05 psu in (a), 2 m in (b), 0.5°C in (c), and 5 W m−2 in (d).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
The same as in Fig. 12 but for the enhanced interannual FWF forcing.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
The same as in Fig. 10 but for the enhanced interannual FWF forcing.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Time series of the mixed layer heat budget (shown only for the sum of vertical mixing and advection terms) at 160°W on the equator during one El Niño and La Niña cycle for the climatological (line with solid circles) and the enhanced (line with open circles) FWF forcing runs. The units are °C s−1 × 10−6.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
Equatorial distributions of the std dev of interannual variability for (a) SSS and (b) SST, with the interannual FWF forcing, the climatological FWF forcing, and the enhanced interannual FWF forcing. The std dev is calculated from the HCM simulations for 16 yr from 2024 to 2039.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2543.1
The std dev of some selected anomaly fields from the HCM simulations with the climatological (αFWF = 0.0), standard (αFWF = 1.0), and enhanced (αFWF = 2.0) FWF forcings. Shown at the Niño-4 region (5°N–5°S, 160°E–150°W) are SSS, SST, MLD, zonal wind stress (τ), and buoyancy flux (QB) and its heat flux part (QT) and freshwater flux part (QS). Also shown in the last two rows are SST for the Niño-1+2 (0°–10°S, 90°–80°W) and Niño-3 (5°N–5°S, 150°–90°W) regions. The units are psu for SSS, °C for SST, m for MLD, dyn cm−2 for τ, and 10−6 kg s−1 m−2 for QB, QT, and QS.