1. Introduction
Ocean–atmosphere coupling plays a fundamental role for the mean state and variability in the tropical Pacific. Both dynamical coupling via surface winds, affecting the ocean circulation, and thermodynamic coupling via the surface heat fluxes have a strong influence on the SST and upper ocean temperature (e.g., Lloyd et al. 2009; Bayr et al. 2021). Despite great progress over the recent years, most coupled general circulation models (CGCMs) still exhibit significant errors in the mean state in the tropical Pacific sector and its interannual variability associated with El Niño–Southern Oscillation (ENSO) (e.g., Bellenger et al. 2014; Wang et al. 2014; Planton et al. 2021; Guilyardi et al. 2020).
A long-standing problem in CGCMs is the cold equatorial Pacific (EP) SST bias (hereafter cold bias), which manifests as an excessive and overly narrow cold tongue extending far west into the warm pool (Davey et al. 2002; Guilyardi et al. 2009a, 2020). The cold bias substantially disturbs the atmospheric mean state as well as the atmospheric response to SST anomalies, with far-reaching consequences for the local and global climate (Bayr et al. 2018, 2019b). The double intertropical convergence zone (ITCZ) pattern is another notable bias shared by CGCMs, which exhibits excessive precipitation off the equator, insufficient precipitation along the equator, and a zonally oriented rainband in the southern tropical Pacific (Mechoso et al. 1995; Lin 2007). The equatorial precipitation deficit is tightly linked to the cold bias through the Bjerknes feedback in models participating in phase 5 of the Coupled Model Intercomparison Project (CMIP5) (Li and Xie 2014).
CGCMs with a cold bias also tend to exhibit erroneous ENSO dynamics and properties (Guilyardi et al. 2009a). Along with the cold bias, a too westward-located rising branch of the Walker circulation weakens the amplifying wind feedback and the damping heat-flux feedback and hampers simulated ENSO dynamics due to error compensation (Bayr et al. 2019a, 2020; Ding et al. 2020). Moreover, CGCMs with a reduced cold bias simulate the seasonal phase locking of ENSO and ENSO asymmetry more realistically (Wengel et al. 2018; Bayr et al. 2021).
Recent studies also suggest that the cold bias may also explain why current climate models are unable to simulate the observed SST cooling and strengthening of the Pacific Walker circulation in recent decades (Seager et al. 2019, 2022). Linked to the cold tongue bias, CMIP5 models generally exhibit higher relative humidity and lower wind speed, both of which contribute to a too high sensitivity of the SSTs in the region to rising greenhouse gases and hence overwhelm the effect of upwelling (Seager et al. 2019). The cold bias also seems to introduce uncertainty to global warming projections of ENSO-amplitude change, as climate models with a small cold bias and realistic ENSO asymmetry agree on increasing ENSO amplitude under global warming (Cai et al. 2021), while there is still a large disagreement among all CMIP5 and CMIP6 models (Beobide-Arsuaga et al. 2021).
Further, the atmospheric teleconnections between the tropical Pacific and the other tropical ocean basins as well as to the midlatitudes may benefit by alleviating the cold bias (Alexander et al. 2002; Yeh et al. 2018; Timmermann et al. 2018). For example, in CMIP5 models, the cold bias leads to a westward shift in the atmospheric response to ENSO over the North Pacific and a significant underestimation of the ENSO-related precipitation anomalies over California (Bayr et al. 2019b). Therefore, addressing the cold bias is essential for improving not only tropical but also extratropical climate variability.
There are several systematic errors shared by the CGCMs, originating in the equatorial regions that may contribute to the cold bias (Vannière et al. 2013, 2014). Lin (2007) found that excessive latent heat and insufficient shortwave radiation flux along with overly strong trade winds occur over the EP in stand-alone atmosphere model simulations with prescribed observed SSTs. Stronger equatorial upwelling of cold water forced by too intense trade winds leads to the creation of a cold bias (Guilyardi et al. 2009b). Both wind and SST biases are amplified by coupled ocean–atmosphere interaction related to the Bjerknes feedback (Li and Xie 2014; Mauritsen et al. 2012). Some studies trace the cold bias to deficiencies in the representation of atmospheric convection. A revised convection scheme can largely eliminate the cold bias and subsurface water temperature biases, and the deficit in equatorial precipitation by strengthening negative shortwave flux–SST feedback and positive evaporation–SST feedback (Song and Zhang 2009). Other studies suggest that the cold bias is caused by too weak surface heating due to weak cloud responses to atmospheric ascent (Sun et al. 2003; Ferrett et al. 2018).
Regions off the equator can affect the EP SST through the atmospheric bridge in association with the atmospheric Rossby waves and the oceanic bridge through thermocline subduction (McCreary and Lu 1994; Liu and Yang 2003). Among these factors, easterly wind biases over both the equatorial and off-equatorial regions play an important role in generating the cold bias (Burls et al. 2017). The subtropical cells, which consist of subtropical subduction, equatorward advection of cool subsurface water, and upwelling at the equator, are largely driven by the wind stress (McCreary and Lu 1994). Thus, wind biases over the off-equatorial region could affect the equatorial SST through the subtropical gyres (Thomas and Fedorov 2017). Some studies have supported this hypothesis, demonstrating that in CGCMs, the east–west EP-SST gradient is tightly linked to the meridional cloud-albedo gradient via advection by the subtropical gyres (Burls and Fedorov 2014; Burls et al. 2017).
While biases in the atmosphere model may largely explain the cold bias, the ocean model also contributes. Overly strong subsurface mixing has been proposed to cause the cold bias (Moum et al. 2013). Underestimated biogeochemical attenuation of penetrative shortwave radiation may cause spurious cooling in the eastern EP (Murtugudde et al. 2002). In addition to thermodynamic processes, the simulated spatial SST distribution is very sensitive to ocean heat transport (Zhang and Song 2010). Eddy-resolving ocean models improve the simulation of small-scale ocean processes to a large extent and hence the simulation of tropical Pacific mean climate and its variability (Small et al. 2014).
So far, efforts to reduce the cold bias have generally focused on tuning parameters of convection schemes, albedo of low stratus clouds, or vertical mixing schemes over the equatorial and off-equatorial Pacific (Gordon et al. 2000; Richards et al. 2009; Guilyardi et al. 2009b; Burls et al. 2017; Bayr et al. 2018). Here, we explore and quantify the influence of an improved atmospheric circulation on the cold bias in a CGCM. To what extent can improved equatorial and off-equatorial atmospheric surface forcing affect the cold bias and by which mechanism?
We perform a set of dedicated experiments with the Kiel Climate Model (KCM; Park et al. 2009) forced by reanalysis wind stress. The individual contributions of surface wind biases in different regions to the cold bias are assessed. Further, the response of the surface winds to a biased SST in different regions is investigated with stand-alone atmosphere model experiments.
The paper is organized as follows: Section 2 describes the datasets, model, experimental setup, and the methods used in this study. The results of the model simulations are presented in section 3. In section 3a, the SST and wind biases and their relation are described in the KCM and in the CMIP5/CMIP6 models. The influence of the surface wind biases over different regions on the cold bias is explored in section 3b with a set of coupled experiments. The mechanisms behind the cold bias are described in section 3c. The response of the surface winds to SST biases in different regions is investigated in section 3d with a set of stand-alone atmosphere model experiments. A summary and discussion are presented in section 4.
2. Data, model, and method
a. Observational and reanalysis datasets
To quantify the biases in the coupled and uncoupled experiments, the model results are compared to reanalysis data and where possible to observations. To obtain the mean-state SST biases, the monthly version of the Hadley Centre Sea Ice and Sea Surface Temperature dataset (HadISST v1.1; Rayner et al. 2003) averaged over the period 1982–2017 is used. Zonal-mean upper-ocean temperature is compared against the average over 1975–2004 from Volume I of the World Ocean Atlas (Locarnini et al. 2019). The wind biases are mainly estimated by comparing the model data to the 10-m winds from the National Centers for Environmental Prediction Climate Forecast System Reanalysis (NCEP/CFSR; Saha et al. 2010) averaged over the period 1982–2017. Wind biases relative to ERA-Interim and ERA5 reanalyses are considered as well (Dee et al. 2011; Hersbach et al. 2018). Surface sensible and latent heat fluxes are taken from the Objectively Analyzed Air–Sea Fluxes (OAFlux) dataset (Yu et al. 2008) averaged over the period 1983–2017. Net surface shortwave and longwave radiative flux and cloud cover are taken from the third-generation, high-resolution version of the International Satellite Cloud Climatology Project flux product (ISCCP-FH; Zhang and Rossow 2023) available over the period 1983–2017. Interannually varying monthly NCEP/CSFR surface wind stress spanning 1982–2017 is employed to force the coupled model. We choose the HadISST and the NCEP/CFSR to estimate the SST and wind biases, as the HadISST exhibits negligible differences relative to the Optimum Interpolation SST (not shown), to which the NCEP/CFSR wind stresses are strongly relaxed.
b. Model and experimental setup
In this study, coupled and stand-alone atmosphere model experiments are conducted with the KCM and its atmospheric component, respectively. The KCM is a fully coupled atmosphere–ocean–sea ice general circulation model, which consists of the atmosphere model ECHAM5 (Roeckner et al. 2003) and the ocean–sea ice model NEMO (Madec 2008). ECHAM5 is used with a T42 horizontal resolution (∼2.8° × 2.8°) and with 19 vertical levels. The horizontal resolution of NEMO is approximately 2° (ORCA2 grid), with a meridional refinement of ∼0.5° near the equator and 31 vertical levels.
We first consider a present-day control integration of the KCM (hereafter KCMCTR). The KCMCTR is initialized from the Levitus climatology and integrated for 110 years. Concentrations of greenhouse gases used in the KCMCTR are representative of the mid-1980s. Specifically, the CO2 concentration amounts to 348 ppm (parts per million), the CH4 concentration to 1650 ppb (parts per billion), and the N2O concentration to 306 ppb. The monthly output of the last 36 years is used for analysis. The HadISST dataset considered is over the period of 1982–2017. Owing to different radiative forcings, the simulated global mean SST in KCMCTR tends to differ from HadISST. To enable a fair comparison, we compute the relative SST bias by subtracting the area averaged SST over the tropical Pacific (20°S–20°N, 120°E–70°W). Studies suggest that the relative SSTs depict the relation between SSTs and tropical atmospheric convection better than the absolute SSTs (He et al. 2018; Izumo et al. 2019). The KCM, like many other coupled climate models, suffers from significant SST and 10-m wind biases over the tropical and extratropical Pacific (Fig. 1). Figures 1a and 1d depict the relative SST and the 10-m wind speed bias in the KCMCTR, respectively, which both exhibit a banded structure in the zonal direction. Three regions where cold SST biases coexist with positive wind speed biases are highlighted by the blue boxes in Fig. 1. We define these regions as the northern Pacific (NP; 15°–38°N, 130°E–130°W), equatorial Pacific (EP; 5°S–5°N, 130°E–100°W), and southern Pacific (SP; 35°–15°S, 160°E–90°W).
Relative sea surface temperature (SST) bias (°C) in the (a) KCMCTR, (b) CMIP5, and (c) CMIP6 multimodel mean relative to the HadISST v1.1 averaged over the period 1982–2017. The relative SST bias is computed by subtracting the area averaged SST over the tropical Pacific (20°S–20°N, 120°E–70°W). Also shown are 10-m winds (vectors; m s−1) and wind speed biases (shading; m s−1) in the (d) KCMCTR, (e) CMIP5, and (f) CMIP6 multimodel mean relative to the NCEP/CSFR reanalysis averaged over the period 1982–2017. Dashed boxes indicate the equatorial Pacific (EP; 5°S–5°N, 130°E–100°W), northern Pacific (NP; 15°–38°N, 130°E–130°W), and southern Pacific (SP; 35°–15°S, 160°E–90°W), respectively.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
We performed a set of coupled experiments with the KCM and a set of stand-alone experiments with the ECHAM5 (Table 1). Four coupled sensitivity experiments have been conducted, in which the interannually varying NCEP/CSFR monthly wind stress is prescribed over the NP, EP, SP, and globally (hereafter NP-Wind, EP-Wind, SP-Wind, and Global-Wind, respectively). In each sensitivity experiment, the wind stress forcing, spanning 1982–2017, is repeated three times. The SST bias, especially over the equatorial Pacific, fully develops in the first two forcing cycles consistent with Vannière et al. (2014). The monthly output from the last forcing cycle is taken for analysis. We examine with this set of experiments the influence of realistic wind stress forcing on the SST biases.
Experimental setup for the coupled and stand-alone atmospheric experiments.
The second set of experiments consists of an atmosphere-only control simulation (hereafter ECHAM5CTR) and three uncoupled sensitivity experiments. In the ECHAM5CTR, monthly SST and sea ice (HadISST) from 1982 is used as forcing during the initial three model years. Thereafter, the HadISST over the period 1983–2017 is applied. In the atmospheric sensitivity experiments, the HadISST forcing over the NP, EP, and SP regions is completely replaced by the biased SSTs from the KCMCTR (hereafter ECHAM5Biased-NP-SST, ECHAM5Biased-EP-SST, and ECHAM5Biased-SP-SST, respectively) such that the full impact of SST biases on the wind stress can be assessed. Each of these experiments is integrated for 38 years with the last 36 years used for analysis. Nonetheless, there are limitations of this approach. A spurious SST gradient is generated at the boundary where the biased SSTs are implemented, which may slightly affect the local response of wind stress to SST biases. Additionally, the temporal variability in the coupled model may be different from the observations, which creates some uncertainty in the simulated response of surface winds. We therefore define the response as annual changes in surface winds averaged over the last 36 years. This set of uncoupled experiments helps us to investigate the response of the surface winds to the cold SST biases in the NP, EP, and SP regions separately.
c. Methods
The term Qresidual is the residual term that includes all processes that cannot be computed, such as mesoscale eddies and vertical and horizontal mixing, as well as possible errors in other terms. Contributions from high-frequency processes are also included in this term. We evaluate the annual-mean changes in the surface heat flux and ocean heat transport averaged over one forcing cycle (i.e., 36 years) as the contributions to the SST biases. As the annual SST biases averaged in the last forcing cycle reach a quasi-steady state, the heating rate δQ is approximately equal to zero. Thus, the residual term Qresidual can be estimated from the imbalance between the net surface heat flux Qnet and the resolved oceanic heat transport Qdyn.
A constant MLD of 50 m is chosen in the heat budget analysis, which corresponds to the average MLD in the equatorial Pacific. A 20-m shallower or deeper MLD would have a negligible impact on the results. Further, this MLD is sufficiently large compared to the penetrative depth of shortwave radiation estimated using satellite ocean-color data (Murtugudde et al. 2002).
3. Results
a. Equatorial and off-equatorial bias in the coupled experiments
Figure 1 shows the relative SST and absolute 10-m wind speed biases in the Pacific with respect to HadISST and the NCEP/CFSR reanalysis, respectively, over the period 1980–2005 for the KCMCTR, the multimodel mean of 26 CMIP5 and 33 CMIP6 models. A list of the CMIP5 and CMIP6 models considered in this study can be found in Fig. 2, and we use here all CMIP models for which the required variables are available. In the KCMCTR, there are significant cold biases in the EP and in the subtropical regions especially in the Northern Hemisphere (Fig. 1a). The relative SST in the KCMCTR is lower by 1.12°C relative to HadISST averaged over the EP (5°S–5°N, 130°E–100°W). Large 10-m wind speed biases, which can be largely attributed to the zonal wind component, go along with the SST biases over the aforementioned three regions (Fig. 1d). Additionally, the KCMCTR as well as the CMIP5 and CMIP6 models exhibits large warm SST biases along the Peruvian and Californian coasts (Fig. 1). These warm biases can be caused by erroneous meridional winds leading to reduced upwelling (Harlaß et al. 2018) and/or underestimated formation of low-level stratus clouds (Mechoso et al. 1995). Corrected wind stress reduces the SST bias along the Californian coast but less clearly along the Peruvian coast (Fig. 3). The physical mechanisms giving rise to the coastal warm SST biases, however, are beyond the scope of this study.
Relationships between the relative sea surface temperature (SST) bias and 10-m winds in the KCMCTR, CMIP5, and CMIP6 models. Scatterplot of (a) 10-m wind speed (m s−1) and (b) 10-m zonal wind (m s−1) over the equatorial Pacific (EP; 5°S–5°N, 130°E–100°W) vs local SST bias (°C) for the CMIP5 and CMIP6 models. Scatterplot of 10-m zonal wind over the (c) NP and (d) SP vs 10-m zonal wind over the EP. Red and blue numbers represent the CMIP5 and CMIP6 models, respectively. The red, blue, and black solid (dashed) lines indicate the significant (insignificant) linear regression for the CMIP5, CMIP6, and all CMIP models. The linear correlation coefficient r is given in the upper or lower right corner, with no, one, or two stars indicating an insignificant correlation or a significant correlation at the 90% or 95% confidence level using the Pearson correlation coefficient, respectively.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
Sea surface temperature (SST) responses (°C) in (a) the Global-Wind, (b) NP-Wind, (c) EP-Wind, and (d) SP-Wind experiments relative to KCMCTR averaged over the period 1982–2017. Hatching indicates that the differences are significant at the 95% confidence level using Student’s t test. (e) SST bias over various domains in KCMCTR and four sensitivity experiments with respect to the HadISST v1.1 averaged over the period of 1982–2017.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
The SST and wind biases over these regions, especially over the EP, persist throughout the year with also seasonal variations (Figs. S1 and S2 in the online supplemental material). The entanglement between the annual-mean biases and its seasonal cycle would lead to a very complex analysis and is beyond the scope of this study. We focus in the following on the annual-mean SST biases.
We calculate from the CMIP5 and CMIP6 historical simulations the biases over the period 1980–2005 (Taylor et al. 2012; Eyring et al. 2016). The cold SST and 10-m wind biases in CMIP5 and CMIP6 (Figs. 1b–f) somewhat resemble the banded pattern in the KCMCTR, albeit they are slightly smaller—possibly due to bias compensation among different models (Figs. S3–S6). Major differences between the KCMCTR and the CMIP5 and CMIP6 models are observed in wind speed biases, where the CMIP5 and CMIP6 models largely agree with each other, pointing toward different physical processes underlying the biases in the KCMCTR and in the CMIP models. The ERA-Interim and ERA5 reanalyses are used to estimate 10-m wind biases in the KCM, CMIP5, and CMIP6 models (Fig. S7), which are of the same sign over most of the tropical Pacific.
Figures 2a and 2b show that there is a statistically significant relationship between the equatorial SST bias and the local surface wind speed (10-m zonal surface wind) with a correlation of 0.74 (0.71) considering models all together (KCM, CMIP5, and CMIP6), suggesting an important role of dynamical atmosphere–ocean coupling in amplifying the cold tongue bias in the CMIP models (Li and Xie 2014). We note that the cold tongue bias in the KCMCTR is larger than that in most historical simulations by the CMIP models. This could be due to relatively low horizontal resolution and/or the tuning of convective parameters in the KCMCTR. Additionally, there exist positive correlations between the equatorial zonal wind bias and the off-equatorial zonal wind biases in the subtropical gyres. The CMIP5 models generally depict stronger equatorial linkages to the zonal winds in the subtropics than the CMIP6 models (Figs. 2c,d). As shown by the KCM and all considered CMIP models, the correlation between the EP zonal wind bias and the NP zonal wind bias amounts to r = 0.41, which is comparable to the correlation with its southern counterpart of r = 0.44 (Figs. 2c,d). Both correlation coefficients are significant at the 95% confidence level against the Pearson correlation coefficient. The individual influence of the zonal and meridional components is assessed through two complementary experiments (not shown). Their analysis shows that the influence of the wind biases on the EP SST bias is dominated by the zonal wind bias, and the meridional wind bias appears to be less crucial.
b. Response of SSTs to the surface winds in the coupled experiments
The above analysis suggests that the equatorial cold bias has a local origin involving dynamic coupling with the zonal wind but also remote contributions from the northern and southern subtropical gyres. To verify this hypothesis, we perform sensitivity experiments with the KCM suffering from similar biases as that observed in many CMIP models.
In these sensitivity experiments, the KCM is forced by NCEP/CFSR reanalysis wind stress over different regions. Figure 3 depicts the SST responses relative to the KCMCTR averaged over the period 1982–2017. When the interannually varying reanalysis wind stress is prescribed globally, the SSTs in the EP region warm by 1.14°C corresponding to a complete reduction of the cold bias (Figs. 3a,e). The differences between the Global-Wind run and observations are not statistically significant (Fig. S8). This suggests that the equatorial cold bias can be reduced by improving wind stress. When the reanalysis wind stress is only applied over the NP (Fig. 3b), EP (Fig. 3c), and SP regions (Fig. 3d), the cold bias in the equatorial Pacific is reduced by 0.14°C (12%), 0.59°C (52%), and 0.26°C (23%), respectively (Fig. 3e). Overall, the wind stress biases over the three regions together explain 87% of the cold bias in the EP region assuming linearity of the response. We conclude that the biased wind stress over the EP (Fig. 1d) is the major source of the equatorial cold bias in the KCMCTR (Figs. 3a,c). However, wind biases over the subtropical regions cannot be ignored when discussing the origin of the equatorial cold bias in the KCMCTR (Figs. 3b,d). The wind biases in the SP region exhibit a stronger influence on the equatorial cold bias relative to that in the NP region.
Vertical sections of the zonally averaged upper-ocean temperatures and velocities averaged over 130°E–100°W from the KCMCTR are shown in Fig. 4a. The vertical velocities have been multiplied by 104 for better visibility. Pronounced Ekman cells, forced by the easterly wind stress at the equator, exist in both hemispheres, with upwelling at the equator and divergent meridional velocities approximately in the upper 50 m giving rise to the minimum in the upper-ocean temperatures at the equator. Subsurface subtropical waters in the KCMCTR suffer from marked cold biases, especially between 10° and 20°S in the Southern Hemisphere (Fig. 4b). The cold subsurface biases, connecting the subtropics with the equatorial upwelling, suggest that the equatorial cold bias in the KCMCTR partly originates in the subtropics. Figures 4c–g depict the temperature and current responses in the sensitivity experiments relative to the KCMCTR. The temperatures in the upper 80 m of the EP become warmer in all experiments, with the exception of the EP-Wind run in which the warming only extends down to about 60 m. Warming in the EP is strongest at the surface in the Global-Wind and EP-Wind runs (Figs. 4c,e). The oceanic current changes suggest that the warming in the top 80 m (Global-Wind) and top 60 m (EP-Wind) is linked to the weakening of the Ekman cells (tropical cells), which is associated with a reduced upwelling of cold subsurface waters to the surface. In the NP-Wind and SP-Wind runs, the warming in the equatorial region is much weaker and uniform over the upper 80 m and is suggested to mainly result from the significantly reduced cold subsurface biases around 15°N (Fig. 4d) and 15°S (Fig. 4f). As shown in Global-Wind and NP-Wind, correcting the wind stress in the NP region tends to amplify the local SST bias in NP, which is mainly attributed to increased equatorward horizontal transport of cold water and locally strengthened upwelling (Figs. 4c,d). The subsurface cold bias in the Southern Hemisphere subtropics slightly intensifies in the NP-Wind, which may explain the smaller equatorial warming compared to the SP-Wind (Fig. 4g).
(a) The vertical profile of the zonal-mean ocean temperature (shading; °C) and velocities (vectors; m s−1) over 130°E and 100°W in KCMCTR. The vertical current is multiplied by 10 000 for better visuality. (b) The vertical profile of bias (shading) in the zonal-mean ocean temperature over 130°E and 100°W in KCMCTR with respect to the World Ocean Atlas 2018 (WOA18). Contours show the zonal-mean temperature for the WOA18. Green boxes indicate the subsurface in the northern Pacific (NPSub; 12°–19°N, 130°E–100°W, 100–180 m), upper layers in the equatorial Pacific (EPUpper; 5°S–5°N, 130°E–100°W, 0–80 m), and the subsurface in the southern Pacific (SPSub; 10°–18°S, 130°E–100°W, 120–200 m). Responses of ocean temperature (shading; °C) and current (vectors; m s−1) in (c) Global-Wind, (d) NP-Wind, (e) EP-Wind, and (f) SP-Wind relative to KCMCTR over the period 1982–2017. Hatching indicates that the differences are significant at the 95% confidence level using Student’s t test. (g) Temperature biases over the three domains [as shown by the green boxes in (b)] in KCMCTR and four sensitivity experiments with respect to the WOA18.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
To obtain more insight into the ocean circulation changes, we calculate the wind stress curl (Fig. 5) and the barotropic streamfunction (Fig. 6) in the different experiments, where the climatologies of the KCMCTR are shown in Figs. 5a and 6a. Differences of wind stress and wind stress curl in the coupled sensitivity experiments are shown in Figs. 5b–e, which is a forcing by definition of the experimental setup. In the off-equatorial region between 10° and 25°N, enhanced equatorward oceanic mass transport is driven by the negative changes wind stress curl in Global-Wind (Fig. 5b). Similarly, enhanced yet weaker equatorward mass transport is forced in the South Pacific counterpart (Fig. 5b). Additionally, weakened subtropical gyres driven by local wind forcing transport reduced warm water from the tropics to the central areas of the NP and SP basin in the Global-Wind run, as indicated by the decreased meridional gradient in the barotropic streamfunction (Fig. 6b), thereby contributing to the pronounced cooling in the subtropical Pacific, especially in the Northern Hemisphere (Fig. 3). Near the equator between 5°S and 10°N, equatorial currents are weakened owing to the direct Ekman transport driven by substantially reduced meridional winds in addition to the positive wind stress curl (Fig. 5b). As a result, reduced cold water is exported from the cold tongue into the warm pool (Fig. 6b), thereby warming the SSTs in the EP. The oceanic circulation changes in NP-Wind and SP-Wind approximately depict the response in Global-Wind in the North and South Pacific, respectively (Figs. 5c,e and 6c,e). Notably, in the off-equatorial region, biases in surface winds and wind stress curl also reduce in the EP-Wind, especially around 15°N and 15°S, indicating remote atmospheric responses to the alleviated equatorial SST bias (Fig. 5d).
(a) Climatological mean wind stress (blue vectors; N m−2) and wind stress curl (shading; N m−3) in KCMCTR. Difference between prescribed wind stress (vectors) and wind stress curl (shading) in the (b) Global-Wind, (c) NP-Wind, (d) EP-Wind, and (e) SP-Wind experiments and simulated wind stress and wind stress curl KCMCTR for the period 1982–2017. Contours in (b)–(e) show the climatological mean wind stress curl for the respective experiments. The contour interval is 0.4 × 10−7 N m−2.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
(a) Climatological mean barotropic streamfunction (PSI; Sv; 1 Sv ≡ 106 m3 s−1) in KCMCTR. In the meridional direction (from south to north), the increasing (decreasing) streamfunction stands for eastward (westward) volume transport. Responses of barotropic streamfunction (Sv) in the (b) Global-Wind, (c) NP-Wind, (d) EP-Wind, and (e) SP-Wind experiments relative to KCMCTR for the period 1982–2017. The contours in (b)–(e) show the climatological mean PSI for the respective experiments. The contour interval is 10 Sv.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
c. Mixed layer heat budget in the coupled experiments
In this section, we quantify the contributions of ocean dynamical processes and surface heat flux to the SST responses in the sensitivity experiments with the KCM. Figure 7a depicts the time-averaged ocean dynamical heating (color shading) and the horizontal ocean currents (vectors) averaged over the upper 50 m in the KCMCTR. The ocean dynamics tend to cool the SST in the EP and to warm the off-equatorial Pacific in the KCMCTR. Specifically, the zonal, meridional, and vertical components of the ocean dynamical heating (Figs. S9a,b and S10a) each contribute to the cooling in the EP, where the zonal component is largest at the equator, the meridional component just off the equator, and the vertical component in the very east. Figures 7b–e depict the responses of the ocean dynamical heating and the horizontal ocean currents in the sensitivity experiments. The cooling effect by the ocean dynamics in the EP decreases in the four sensitivity experiments, thereby supporting SST warming relative to the KCMCTR (Fig. 3). The dynamical cooling averaged over the EP weakens by 42.5, 6.8, 27.8, and 5.8 W m−2 in the Global-Wind, NP-Wind, EP-Wind, and SP-Wind runs, respectively (Fig. 8). The response of the ocean dynamical heating in the EP-Wind is 65% of that in the Global-Wind, which is comparable to the ratio (52%) of the reduced SST bias in the EP-Wind (0.59°C) and Global-Wind (1.14°C). Consistently, the responses of the ocean dynamical heating in the NP-Wind and SP-Wind are much weaker relative to the EP-Wind (Figs. 7c,e). In terms of individual contributions, the zonal and meridional components of the ocean dynamical heating are comparable in magnitude and much larger than the vertical component (Fig. 8). The response of the zonal dynamical heating component is strongest west of the date line (Figs. S9c,e,g,i), while that of the meridional component is more zonally uniform in the EP with a minimum at the equator (Figs. S9d,f,h,j). The vertical ocean dynamical heating component only contributes close to the South American coast and in the eastern EP and is relatively noisy (Fig. S10).
(a) Climatological mean ocean dynamical heating (shading; W m−2) and ocean horizontal current (vectors; m s−1) averaged in the upper 50 m in KCMCTR. Responses of ocean dynamical heating (shading; W m−2) and ocean horizontal current (vectors; m s−1) in (b) Global-Wind, (c) NP-Wind, (d) EP-Wind, and (e) SP-Wind relative to KCMCTR averaged over the period 1982–2017.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
Responses of heat budget in the mixed layer over the equatorial Pacific (EP as defined in Fig. 1) in the coupled sensitivity experiments relative to KCMCTR. The terms Qu, Qυ, and Qw are the zonal, meridional, and vertical dynamical heating components; Qnet represents net surface heat flux and Qresidual is a residual term, which includes all processes that cannot be calculated with monthly output, as well as the possible errors in the other terms.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
The changes in the zonal and meridional ocean dynamical heating components can be mainly attributed to the eastward and equatorward responses of the horizontal upper-ocean currents (Figs. S9c–j). The eastward change in the zonal currents reduces the cold-water transport from the cold tongue toward the western Pacific warm pool, while the equatorward response of the meridional currents reduces the export of warm water to the off-equatorial region. We note that the response of the vertical dynamical heating in the EP-Wind is small but negative (Fig. 8), which represents a cooling influence on the SST, especially in the cold tongue region (Fig. S10d). This cooling effect is linked to the stronger vertical temperature stratification overriding the effect of the weaker upwelling (Fig. 4e). Unresolved oceanic processes as included by the residual term play an important role in damping SST biases (Fig. 8). It is noted that the actual mixed layer in the equatorial Pacific is deeper than 50 m in the west and shallower in the east. A constant MLD used in the heat budget analysis may thus create some uncertainty in the heat budget analysis, which is partly represented by the residual term.
Next, we examine the net surface heat flux Qnet and cloud-cover responses in the coupled experiments. Figures 9a and 10a present the long-term averages of Qnet and cloud cover simulated in the KCMCTR. Strong positive (downward) Qnet associated with relatively low cloudiness occurs in the EP. Compared to the OAFlux observations, excessive downward Qnet, especially associated with too large surface solar radiation (not shown), is linked to a reduced cloud cover over this region (Figs. 9b and 10b). This suggests that the biased Qnet and cloud cover in the KCMCTR can be attributed to the compensation of the cold SST bias by the cloud-radiative effect. The responses of Qnet and cloud cover in the coupled sensitivity experiments are shown in Figs. 9c–f and 10c–f. Downward Qnet is reduced over the EP and increases over the subtropics (Fig. 9c). This is accompanied by cloud cover changes of opposite signs (Fig. 10c), suggesting that the thermodynamic coupling counteracts the SST biases over the tropical Pacific. Therefore, prescribed surface wind stress not only reduces SST biases but also alleviates erroneous surface heat fluxes and cloud clover. The largest changes in Qnet and cloud cover are observed over the western EP in the KCMCTR, as well as in the NP-Wind (Figs. 9d and 10d), EP-Wind (Figs. 9e and 10e), and SP-Wind (Figs. 9f and 10f).
(a) Climatological mean net surface heat flux (Qnet; W m−2) in KCMCTR. (b) The Qnet bias relative to the OAFLUX averaged over the period of 1983–2017. Green contours indicate the uncertainty of Qnet biases, defined as one standard deviation of monthly biases. Response of Qnet in (c) Global-Wind, (d) NP-Wind, (e) EP-Wind, and (f) SP-Wind relative to KCMCTR averaged over the period of 1983–2017. Hatching indicates that the differences are significant at the 95% confidence level using Student’s t test.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
As in Fig. 9, but here for the cloud cover (%). Contours in (b)–(f) show the climatological mean wind stress curl for (b) the ISCCP, (c) Global-Wind, (d) NP-Wind, (e) EP-Wind, and (f) SP-Wind. The contour interval is 8%.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
d. Origin of wind bias
The question that still remains is the origin of the biased surface winds and their interaction with the SST. To address this question, we conducted a set of stand-alone experiments with the atmospheric component of the KCM, the ECHAM5 atmospheric general circulation model (AGCM).
Figure 11a shows near-surface (10-m) wind speed in the ECHAM5CTR forced by HadISST. There is a pronounced wind speed bias relative to the NCEP/CFSR reanalysis that is mainly caused by too strong easterlies over the central equatorial, northern, and southern subtropical Pacific (Fig. 11b). In contrast to ECHAM5CTR, wind speed biases are even stronger over the eastern EP and central subtropical NP and SP in KCMCTR (Fig. 1d). Differences in the pattern and intensity of wind biases between ECHAM5CTR and KCMCTR arise mainly from the ocean–atmosphere coupling.
The 10-m winds (vectors; m s−1) and wind speed (shading; m s−1) in (a) ECHAM5CTR spanning from 1982 to 2017. (b) Biases in 10-m winds and wind speed relative to the NCEP/CFSR reanalysis averaged over the period 1982–2017, respectively. Differences in 10-m winds and wind speed in (c) ECHAM5Biased-NP-SST, (d) ECHAM5Biased-EP-SST, and (e) ECHAM5Biased-SP-SST relative to ECHAM5CTR. Hatching indicates the differences are statistically significant at the 95% confidence level using Student’s t test.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
In ECHAM5CTR, the wind bias averaged over the NP is 1.39 m s−1 and comparable to the bias of 1.70 m s−1 in KCMCTR (Fig. 12). On the other hand, the wind speed bias over the EP and SP is smaller in EHCAM5CTR than that in KCMCTR (Fig. 12). When the biased SST simulated in KCMCTR is specified over the NP, the local wind speed bias increases over the eastern subtropical NP and remains basically unchanged over the EP and SP (Figs. 11c and 12). In ECHAM5Biased-EP-SST, the surface winds over several regions (i.e., the western equatorial Pacific, central and eastern subtropical NP, and eastern subtropical SP) intensify in response to the too cold equatorial SSTs (Fig. 11d). The averaged 10-m wind speed bias over the NP, EP, and SP regions strengthens to 1.75, 1.12, and 1.22 m s−1, respectively, which is comparable to that in the KCMCTR (Fig. 12). This indicates an important role of the EP-SST bias for changing the local and subtropical wind biases from the pattern in ECHAM5CTR toward that in the KCMCTR. Similar to the influence of the SST bias in the NP region, the biased SST in the SP region slightly strengthens the local bias over the central subtropical SP, with little impact over the NP and EP (Figs. 11e and 12).
The 10-m wind bias over different domains in KCMCTR and the stand-alone atmospheric experiments with respect to the NCEP/CFSR reanalysis averaged over the period of 1982–2017. Red error bars indicate the uncertainty of 10-m wind biases, defined as one standard deviation of monthly wind biases.
Citation: Journal of Climate 36, 24; 10.1175/JCLI-D-22-0874.1
We can infer from the uncoupled AGCM experiments that significant 10-m wind speed biases are already present in the atmosphere model and that they are amplified and expanded in the coupled mode with large differences between the three regions. Over the NP, most of the wind bias in the KCMCTR is due to the atmosphere model, and this bias is slightly amplified in the coupled mode mainly due to the SST biases in the NP and EP (increase by 20% from 1.39 m s−1 in ECHAM5CTR to 1.70 m s−1 in KCMCTR). Over the eastern subtropical SP, the wind bias is also mainly due to the atmosphere model, but the amplification over the central and eastern SP by coupling is larger than over the NP, mainly caused by the SST biases in the SP and EP (increasing by 64% from 0.61 m s−1 in ECHAM5CTR to 1.01 m s−1 in KCMCTR). Over the EP, the local coupling is the largest contributor to the wind bias (increasing by 160% from 0.55 m s−1 in ECHAM5CTR to 1.45 m s−1 in KCMCTR).
4. Summary and discussion
A control integration of the KCM (KCMCTR), like many CMIP5 and CMIP6 models, exhibits pronounced cold SST biases over the northern tropical Pacific (NP), equatorial Pacific (EP), and southern tropical Pacific (SP). Over these regions, the models also suffer from strong easterly surface wind biases. In the CMIP5 and CMIP6 historical simulations, we find that the SST bias is strongly related to the easterly wind strength over the EP. Furthermore, the zonal surface wind strength over the EP is correlated with those over the NP and SP.
The origin of biases in SST over the Pacific and their relationship to the near-surface (10-m) winds are investigated by means of dedicated experiments with the KCM and stand-alone experiments with its atmospheric component. When the NCEP/CFSR reanalysis wind stress is specified globally in the coupled model, the cold SST bias over the EP is completely reduced relative to the KCMCTR. Additional coupled sensitivity experiments are conducted to investigate the influence of the surface wind biases over selected regions on the cold SST bias over the EP. The excessively strong easterlies over the EP are the major contributor to the cold SST bias over the EP, with the wind biases over the NP and SP having a much smaller influence, which, however, cannot be ignored. When reanalysis wind stress is specified over the EP, the cold SST bias there is alleviated by 52%, when specified over the NP by 12%, and when specified over the SP by 23%.
In the KCMCTR, the cold SST bias over the EP is dominated by dynamical cooling due to too strong zonal westward and meridional poleward currents and too strong upwelling. The sensitivity experiments with the KCM show that reanalysis wind stress when only specified over the EP weakens the local equatorial currents, thereby reducing considerably the cold SST bias over the region. Therefore, the influence of processes in the vertical direction is much weaker compared to the horizontal processes. Reanalysis wind stress specified over the NP and SP also alleviates the cold SST bias over the EP but to a much lesser extent. Analysis of surface heat flux and cloud cover reveals that in contrast to the dynamical coupling by winds, the thermodynamic coupling counteracts the cold SST bias over the EP.
The origin of the overly strong easterly wind stress in the three selected regions is investigated by a set of stand-alone experiments with the ECHAM5, the atmospheric component of the KCM. The results suggest that the atmosphere model by itself simulates a large wind bias over the NP and smaller biases over the EP and SP, which amount to 82%, 38%, and 60% of those in the KCMCTR, respectively. The cold SST bias over the EP substantially enhances the wind biases over all three considered regions, while the SST biases over the NP and SP regions mainly support local amplification of the wind bias. Thus, the cold SST bias over the EP is mainly due to the (coupled) Bjerknes feedback and further enhanced by too cold subtropical SSTs. Previous studies identified two pathways for subtropical SSTs to influencing equatorial SST: one major pathway, reaching down to a depth of about 600 m, which follows the recirculation near the western boundary of the Pacific basin and then shoals toward the equator and Equatorial Undercurrent. The other connects the equator with off-equatorial regions through the shallow overturning cells in the central and eastern Pacific (Gu and Philander 1997; Thomas and Fedorov 2017; Burls et al. 2017).
It is noted that the experiments EP-Wind, NP-Wind, and SP-Wind together explain 87% of the SST bias over the EP assuming linearity of the response, which is 13% less relative to the Global-Wind experiment. The contributions from other tropical basins (i.e., the tropical Atlantic and the tropical Indian Ocean) may also be important to the cold bias over the EP (Cvijanovic and Chiang 2013; Meehl et al. 2021). As many CMIP models, the KCMCTR suffers from a weaker Atlantic meridional overturning circulation (AMOC; Wang et al. 2014) relative to present-day estimates that could lead to a cold SST bias over the northern tropical Atlantic. In the ECHAM5 atmosphere model, local cold SST biases contribute to overly strong surface winds over the northern tropical Atlantic that is as part of a westward shift of the Walker circulation (Svendsen et al. 2014), which in turn could influence surface winds and SST over the EP. Pacemaker experiments that assimilate warming SST trend over the Indian Ocean favorably yield cooling SST trends of the equatorial Pacific via the Walker circulation (Mochizuki et al. 2016). Remote forcing from the Atlantic and the Indian Ocean, however, appears not to play a crucial role in the KCM, as most of the SST biases over the tropical Pacific can be explained by local processes.
Large uncertainties in the observational surface heat fluxes, especially over the off-equatorial region and over the eastern equatorial Pacific, limit the assessment for the responses of surface heat fluxes to the SST biases. The changes of surface heat fluxes in the sensitivity experiment are comparable to the observational uncertainty (Fig. 9). Similarly, the response of surface winds to the biased SSTs as simulated in the stand-alone atmosphere-model experiments is weaker than the uncertainties in the NCEP/CFSR reanalysis (Fig. 12). The reasons are twofold. First, large uncertainties exist in the NCEP/CFSR surface winds. Second, we evaluate the wind biases in the coupled and stand-alone atmosphere-model experiments using the same definition of the EP, NP, and SP regions. Yet the bias pattern in the coupled experiment differs from the atmosphere-only experiment (Figs. 1 and 11).
In this study, we use the wind stress from the NCEP/CFSR reanalysis as a forcing in partially coupled experiments. We note that ocean–sea ice models can be quite sensitive to the atmospheric forcing dataset, which causes significant differences in the simulated ocean circulation (e.g., Hunke and Holland 2007; Lindsay et al. 2014). Thus, our results may be dependent on the applied wind stresses. On the other hand, the intercomparison of atmospheric reanalysis products shows that they are comparable due to similar physical assumptions and assimilation of common observations (Chaudhuri et al. 2013; Stopa and Cheung 2014). Therefore, a different choice of the atmospheric forcing dataset may not necessarily change the main conclusion of this study that the SST bias over the EP is linked to the equatorial and subtropical surface wind biases, both of which being amplified through ocean–atmosphere interaction. This hypothesis is supported by another set of partially coupled experiments with the KCM forced by the ERA-Interim wind stresses (Fig. S11). EP SST biases in the simulations forced by the ERA-Interim winds are similar yet relatively larger than those forced by the NCEP/CFSR winds.
Suffering from limitations such as a constant MLD in the heat budget analysis and relatively different wind bias pattern in the KCM, our results still provide implications for improving some CGCMs. As indicated by our stand-alone atmosphere model experiments, the surface wind biases are largely internal to the atmosphere model with the patterns and intensities of biases being influenced by coupling especially over the EP. This is consistent with the study of Harlaß et al. (2018), who investigated the warm tropical Atlantic SST bias. Thus, one way to reduce tropical SST biases in CGCMs is to improve the atmospheric components. Roberts et al. (2009) reported that increasing the horizontal resolution in the atmosphere model improves the simulation of surface winds and SST in the Hadley Centre Global Environmental Model version 1. Jung et al. (2012) suggested that shorter time steps rather than increased horizontal resolution in the atmosphere model help to improve the near-surface winds over the tropical Pacific as well as the mean meridional atmospheric circulation in the tropics. This study also implies that improving the surface winds in the atmosphere model holds a large potential to enhance the simulation of tropical Pacific mean climate in CGCMs.
Acknowledgments.
The suggestions of the three anonymous referees greatly improved an earlier version of the manuscript. We thank Joakim Kjellsson for his helpful comments and discussions. Z. S. was supported by the National Natural Science Foundation of China (Grant 42206206) and the Natural Science Foundation of Guangdong Province (Grant 2022A1515010917). Y. Z. was sponsored by the China Scholarship Council (CSC) and supported by the Fundamental Research Funds for the Central Public Welfare Scientific Institutes (Grant PM-zx703-202104-091). T. B. was supported by the Deutsche Forschungsgemeinschaft (DFG) project “Influence of Model Bias on ENSO Projections of the 21st Century” through Grant 429334714, W. P. by Institute for Basic Science (IBS-R028-D1), and A. R. by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie Grant agreement 101026271. The integration with the Kiel Climate Model (KCM) was conducted at the Computing Centre of Kiel University (CAU).
Data availability statement.
CMIP data were retrieved from the Earth System Grid Federation (https://esgf-node.llnl.gov/projects/cmip6/) and the KCM output is archived at the Zenodo repository (https://zenodo.org/deposit/7376944#). Observational datasets required for the analysis are available freely at the locations cited in the text.
REFERENCES
Alexander, M. A., I. Bladé, M. Newman, J. R. Lanzante, N.-C. Lau, and J. D. Scott, 2002: The atmospheric bridge: The influence of ENSO teleconnections on air–sea interaction over the global oceans. J. Climate, 15, 2205–2231, https://doi.org/10.1175/1520-0442(2002)015<2205:TABTIO>2.0.CO;2.
Bayr, T., M. Latif, D. Dommenget, C. Wengel, J. Harlaß, and W. Park, 2018: Mean-state dependence of ENSO atmospheric feedbacks in climate models. Climate Dyn., 50, 3171–3194, https://doi.org/10.1007/s00382-017-3799-2.
Bayr, T., C. Wengel, M. Latif, D. Dommenget, J. Lübbecke, and W. Park, 2019a: Error compensation of ENSO atmospheric feedbacks in climate models and its influence on simulated ENSO dynamics. Climate Dyn., 53, 155–172, https://doi.org/10.1007/s00382-018-4575-7.
Bayr, T., D. I. V. Domeisen, and C. Wengel, 2019b: The effect of the equatorial Pacific cold SST bias on simulated ENSO teleconnections to the North Pacific and California. Climate Dyn., 19, 3771–3789, https://doi.org/10.1007/s00382-019-04746-9.
Bayr, T., D. Dommenget, and M. Latif, 2020: Walker circulation controls ENSO atmospheric feedbacks in uncoupled and coupled climate model simulations. Climate Dyn., 54, 2831–2846, https://doi.org/10.1007/s00382-020-05152-2.
Bayr, T., A. Drews, M. Latif, and J. Lübbecke, 2021: The interplay of thermodynamics and ocean dynamics during ENSO growth phase. Climate Dyn., 56, 1681–1697, https://doi.org/10.1007/s00382-020-05552-4.
Bellenger, H., E. Guilyardi, J. Leloup, M. Lengaigne, and J. Vialard, 2014: ENSO representation in climate models: From CMIP3 to CMIP5. Climate Dyn., 42, 1999–2018, https://doi.org/10.1007/s00382-013-1783-z.
Beobide-Arsuaga, G., T. Bayr, A. Reintges, and M. Latif, 2021: Uncertainty of ENSO-amplitude projections in CMIP5 and CMIP6 models. Climate Dyn., 56, 3875–3888, https://doi.org/10.1007/s00382-021-05673-4.
Burls, N. J., and A. V. Fedorov, 2014: Simulating Pliocene warmth and a permanent El Niño-like state: The role of cloud albedo. Paleoceanography, 29, 893–910, https://doi.org/10.1002/2014PA002644.
Burls, N. J., L. Muir, E. M. Vincent, and A. Fedorov, 2017: Extra-tropical origin of equatorial Pacific cold bias in climate models with links to cloud albedo. Climate Dyn., 49, 2093–2113, https://doi.org/10.1007/s00382-016-3435-6.
Cai, W., and Coauthors, 2021: Changing El Niño–Southern Oscillation in a warming climate. Nat. Rev. Earth Environ., 2, 628–644, https://doi.org/10.1038/s43017-021-00199-z.
Chaudhuri, A. H., R. M. Ponte, G. Forget, and P. Heimbach, 2013: A comparison of atmospheric reanalysis surface products over the ocean and implications for uncertainties in air–sea boundary forcing. J. Climate, 26, 153–170, https://doi.org/10.1175/JCLI-D-12-00090.1.
Cvijanovic, I., and J. C. H. Chiang, 2013: Global energy budget changes to high latitude North Atlantic cooling and the tropical ITCZ response. Climate Dyn., 40, 1435–1452, https://doi.org/10.1007/s00382-012-1482-1.
Davey, M., and Coauthors, 2002: STOIC: A study of coupled model climatology and variability in tropical ocean regions. Climate Dyn., 18, 403–420, https://doi.org/10.1007/s00382-001-0188-6.
Dee, D. P., and Coauthors, 2011: The ERA‐Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
DiNezio, P. N., A. C. Clement, G. A. Vecchi, B. J. Soden, B. P. Kirtman, and S.-K. Lee, 2009: Climate response of the equatorial Pacific to global warming. J. Climate, 22, 4873–4892, https://doi.org/10.1175/2009JCLI2982.1.
Ding, H., M. Newman, M. A. Alexander, and A. T. Wittenberg, 2020: Relating CMIP5 model biases to seasonal forecast skill in the tropical Pacific. Geophys. Res. Lett., 47, e2019GL086765, https://doi.org/10.1029/2019GL086765.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Ferrett, S., M. Collins, and H.-L. Ren, 2018: Diagnosing relationships between mean state biases and El Niño shortwave feedback in CMIP5 models. J. Climate, 31, 1315–1335, https://doi.org/10.1175/JCLI-D-17-0331.1.
Gordon, C. T., A. Rosati, and R. Gudgel, 2000: Tropical sensitivity of a coupled model to specified ISCCP low clouds. J. Climate, 13, 2239–2260, https://doi.org/10.1175/1520-0442(2000)013<2239:TSOACM>2.0.CO;2.
Gu, D., and S. G. H. Philander, 1997: Interdecadal climate fluctuations that depend on exchanges between the tropics and extratropics. Science, 275, 805–807, https://doi.org/10.1126/science.275.5301.805.
Guilyardi, E., A. Wittenberg, A. Fedorov, M. Collins, C. Wang, A. Capotondi, G. J. van Oldenborgh, and T. Stockdale, 2009a: Understanding El Nino in ocean–atmosphere general circulation models: Progress and challenges. Bull. Amer. Meteor. Soc., 90, 325–340, https://doi.org/10.1175/2008BAMS2387.1.
Guilyardi, E., P. Braconnot, F.-F. Jin, S. T. Kim, M. Kolasinski, T. Li, and I. Musat, 2009b: Atmosphere feedbacks during ENSO in a coupled GCM with a modified atmospheric convection scheme. J. Climate, 22, 5698–5718, https://doi.org/10.1175/2009JCLI2815.1.
Guilyardi, E., A. Capotondi, M. Lengaigne, S. Thual, and A. T. Wittenberg, 2020: ENSO modeling: History, progress, and challenges. El Niño Southern Oscillation in a Changing Climate, Geophys. Monogr., Vol. 253, Amer. Geophys. Union, 199–226, https://doi.org/10.1002/9781119548164.ch9.
Harlaß, J., M. Latif, and W. Park, 2018: Alleviating tropical Atlantic sector biases in the Kiel climate model by enhancing horizontal and vertical atmosphere model resolution: Climatology and interannual variability. Climate Dyn., 50, 2605–2635, https://doi.org/10.1007/s00382-017-3760-4.
He, J., N. C. Johnson, G. A. Vecchi, B. Kirtman, A. T. Wittenberg, and S. Sturm, 2018: Precipitation sensitivity to local variations in tropical sea surface temperature. J. Climate, 31, 9225–9238, https://doi.org/10.1175/JCLI-D-18-0262.1.
Hersbach, H., and Coauthors, 2018: Operational global reanalysis: Progress, future directions and synergies with NWP. ERA Rep. Series 27, 65 pp., https://doi.org/10.21957/tkic6g3wm.
Hunke, E. C., and M. M. Holland, 2007: Global atmospheric forcing data for Arctic ice-ocean modeling. J. Geophys. Res., 112, C04S14, https://doi.org/10.1029/2006JC003640.
Izumo, T., J. Vialard, M. Lengaigne, and I. Suresh, 2019: Relevance of relative sea surface temperature for tropical rainfall interannual variability. Geophys. Res. Lett., 47, e2019GL086182, https://doi.org/10.1029/2019GL086182.
Jung, T., and Coauthors, 2012: High-resolution global climate simulations with the ECMWF model in Project Athena: Experimental design, model climate, and seasonal forecast skill. J. Climate, 25, 3155–3172, https://doi.org/10.1175/JCLI-D-11-00265.1.
Li, G., and S.-P. Xie, 2014: Tropical biases in CMIP5 multimodel ensemble: The excessive equatorial Pacific cold tongue and double ITCZ problems. J. Climate, 27, 1765–1780, https://doi.org/10.1175/JCLI-D-13-00337.1.
Lin, J.-L., 2007: The double-ITCZ problem in IPCC AR4 coupled GCMs: Ocean–atmosphere feedback analysis. J. Climate, 20, 4497–4525, https://doi.org/10.1175/JCLI4272.1.
Lindsay, R., M. Wensnahan, A. Schweiger, and J. Zhang, 2014: Evaluation of seven different atmospheric reanalysis products in the Arctic. J. Climate, 27, 2588–2606, https://doi.org/10.1175/JCLI-D-13-00014.1.
Liu, Z., and H. Yang, 2003: Extratropical control of tropical climate, the atmospheric bridge and oceanic tunnel. Geophys. Res. Lett., 30, 1230, https://doi.org/10.1029/2002GL016492.
Lloyd, J., E. Guilyardi, H. Weller, and J. Slingo, 2009: The role of atmosphere feedbacks during ENSO in the CMIP3 models. Atmos. Sci. Lett., 10, 170–176, https://doi.org/10.1002/asl.227.
Locarnini, R.A., and Coauthors, 2019: Temperature. Vol. I, World Ocean Atlas 2018. A. Mishonov, Technical Editor, NOAA Atlas NESDIS 81, 52 pp., https://www.ncei.noaa.gov/sites/default/files/2020-04/woa18_vol1.pdf.
Madec, G., 2008: NEMO ocean engine. Note du Pole modélisation de l’Institut Pierre-Simon Laplace 27, 401 pp., https://epic.awi.de/id/eprint/39698/1/NEMO_book_v6039.pdf.
Mauritsen, T., and Coauthors, 2012: Tuning the climate of a global model. J. Adv. Model. Earth Syst., 4, M00A01, https://doi.org/10.1029/2012MS000154.
McCreary, J. P., Jr., and P. Lu, 1994: Interaction between the subtropical and equatorial ocean circulations: The subtropical cell. J. Phys. Oceanogr., 24, 466–497, https://doi.org/10.1175/1520-0485(1994)024<0466:IBTSAE>2.0.CO;2.
Mechoso, C. R., and Coauthors, 1995: The seasonal cycle over the tropical Pacific in coupled ocean–atmosphere general circulation models. Mon. Wea. Rev., 123, 2825–2838, https://doi.org/10.1175/1520-0493(1995)123<2825:TSCOTT>2.0.CO;2.
Meehl, G. A., and Coauthors, 2021: Atlantic and Pacific tropics connected by mutually interactive decadal-timescale processes. Nat. Geosci., 14, 36–42, https://doi.org/10.1038/s41561-020-00669-x.
Mochizuki, T., M. Kimoto, M. Watanabe, Y. Chikamoto, and M. Ishii, 2016: Interbasin effects of the Indian Ocean on Pacific decadal climate change. Geophys. Res. Lett., 43, 7168–7175, https://doi.org/10.1002/2016GL069940.
Moum, J. N., A. Perlin, J. D. Nash, and M. J. McPhaden, 2013: Seasonal sea surface cooling in the equatorial Pacific cold tongue controlled by ocean mixing. Nature, 500, 64–67, https://doi.org/10.1038/nature12363.
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, https://doi.org/10.1175/1520-0442(2002)015<0470:EOPROT>2.0.CO;2.
Park, W., N. Keenlyside, M. Latif, A. Ströh, R. Redler, E. Roeckner, and G. Madec, 2009: Tropical Pacific climate and its response to global warming in the Kiel climate model. J. Climate, 22, 71–92, https://doi.org/10.1175/2008JCLI2261.1.
Planton, Y. Y., and Coauthors, 2021: Evaluating climate models with the CLIVAR 2020 ENSO metrics package. Bull. Amer. Meteor. Soc., 102, E193–E217, https://doi.org/10.1175/BAMS-D-19-0337.1.
Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rpwell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108, 4407, https://doi.org/10.1029/2002JD002670.
Richards, K. J., S.-P. Xie, and T. Miyama, 2009: Vertical mixing in the ocean and its impact on the coupled ocean–atmosphere system in the eastern tropical Pacific. J. Climate, 22, 3703–3719, https://doi.org/10.1175/2009JCLI2702.1.
Roberts, M. J., and Coauthors, 2009: Impact of resolution on the tropical Pacific circulation in a matrix of coupled models. J. Climate, 22, 2541–2556, https://doi.org/10.1175/2008JCLI2537.1.
Roeckner, E., and Coauthors, 2003: The atmospheric general circulation model ECHAM5, Part I: Model description. Max Planck Institute for Meteorology Rep. 349, 140 pp., https://pure.mpg.de/rest/items/item_995269_4/component/file_995268/content.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1058, https://doi.org/10.1175/2010BAMS3001.1.
Seager, R., M. Cane, N. Henderson, D.-E. Lee, R. Abernathey, and H. Zhang, 2019: Strengthening tropical Pacific zonal sea surface temperature gradient consistent with rising greenhouse gases. Nat. Climate Change, 9, 517–522, https://doi.org/10.1038/s41558-019-0505-x.
Seager, R., N. Henderson, and M. Cane, 2022: Persistent discrepancies between observed and modeled trends in the tropical Pacific Ocean. J. Climate, 35, 4571– 4584, https://doi.org/10.1175/JCLI-D-21-0648.1.
Small, R. J., and Coauthors, 2014: A new synoptic scale resolving global climate simulation using the Community Earth System Model. J. Adv. Model. Earth Syst., 6, 1065–1094, https://doi.org/10.1002/2014MS000363.
Song, X., and G. J. Zhang, 2009: Convection parameterization, tropical Pacific double ITCZ, and upper-ocean biases in the NCAR CCSM3. Part I: Climatology and atmospheric feedback. J. Climate, 22, 4299–4315, https://doi.org/10.1175/2009JCLI2642.1.
Stopa, J. E., and K. F. Cheung, 2014: Intercomparison of wind and wave data from the ECMWF Reanalysis Interim and the NCEP Climate Forecast System Reanalysis. Ocean Modell., 75, 65–83, https://doi.org/10.1016/j.ocemod.2013.12.006.
Sun, D.-Z., J. Fasullo, T. Zhang, and A. Roubicek, 2003: On the radiative and dynamical feedbacks over the equatorial Pacific cold tongue. J. Climate, 16, 2425–2432, https://doi.org/10.1175/2786.1.
Svendsen, L., N. G. Kvamstø, and N. Keenlyside, 2014: Weakening AMOC connects equatorial Atlantic and Pacific interannual variability. Climate Dyn., 43, 2931–2941, https://doi.org/10.1007/s00382-013-1904-8.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498, https://doi.org/10.1175/BAMS-D-11-00094.1.
Thomas, M. D., and A. V. Fedorov, 2017: The eastern subtropical Pacific origin of the equatorial cold bias in climate models: A Lagrangian perspective. J. Climate, 30, 5885–5900, https://doi.org/10.1175/JCLI-D-16-0819.1.
Timmermann, A., and Coauthors, 2018: El Niño–Southern Oscillation complexity. Nature, 559, 535–545, https://doi.org/10.1038/s41586-018-0252-6.
Vannière, B., E. Guilyardi, G. Madec, F. J. Doblas-Reyes, and S. Woolnough, 2013: Using seasonal hindcasts to understand the origin of the equatorial cold tongue bias in CGCMs and its impact on ENSO. Climate Dyn., 40, 963–981, https://doi.org/10.1007/s00382-012-1429-6.
Vannière, B., E. Guilyardi, T. Toniazzo, G. Madec, and S. Woolnough, 2014: A systematic approach to identify the sources of tropical SST errors in coupled models using the adjustment of initialised experiments. Climate Dyn., 43, 2261–2282, https://doi.org/10.1007/s00382-014-2051-6.
Wang, C., L. Zhang, S.-K. Lee, L. Wu, and C. R. Mechoso, 2014: A global perspective on CMIP5 climate model biases. Nat. Climate Change, 4, 201–205, https://doi.org/10.1038/nclimate2118.
Wengel, C., M. Latif, W. Park, J. Harlaß, and T. Bayr, 2018: Seasonal ENSO phase locking in the Kiel climate model: The importance of the equatorial cold sea surface temperature bias. Climate Dyn., 50, 901–919, https://doi.org/10.1007/s00382-017-3648-3.
Yeh, S.-W., and Coauthors, 2018: ENSO atmospheric teleconnections and their response to greenhouse gas forcing. Rev. Geophys., 56, 185–206, https://doi.org/10.1002/2017RG000568.
Yu, L., X. Jin, and R. A. Weller, 2008: Multidecade global flux datasets from the Objectively Analyzed Air-sea Fluxes (OAFlux) project: Latent and sensible heat fluxes, ocean evaporation, and related surface meteorological variables. OAFlux Project Tech. Rep. OA-2008-01, 64 pp.
Zhang, G. J., and X. Song, 2010: Convection parameterization, tropical Pacific double ITCZ, and upper-ocean biases in the NCAR CCSM3. Part II: Coupled feedback and the role of ocean heat transport. J. Climate, 23, 800–812, https://doi.org/10.1175/2009JCLI3109.1.
Zhang, Y., and W. B. Rossow, 2023: Global radiative flux profile data set: Revised and extended. J. Geophys. Res. Atmos., 128, e2022JD037340, https://doi.org/10.1029/2022JD037340.