Theor Appl Climatol DOI 10.1007/s00704-013-1048-0
ORIGINAL PAPER
Blunt ocean dynamical thermostat in response of tropical eastern Pacific SST to global warming Soon-Il An & Seul-Hee Im
Received: 4 May 2013 / Accepted: 6 November 2013 # Springer-Verlag Wien 2013
Abstract Using an intermediate ocean–atmosphere coupled model (ICM) for the tropical Pacific, we investigated the role of the ocean dynamical thermostat (ODT) in regulating the tropical eastern Pacific sea surface temperature (SST) under global warming conditions. The external, uniformly distributed surface heating results in the cooling of the tropical eastern Pacific “cold tongue,” and the amplitude of the cooling increases as more heat is added but not simply linearly. Furthermore, an upper bound for the influence of the equatorially symmetric surface heating on the cold tongue cooling exists. The additional heating beyond the upper bound does not cool the cold tongue in a systematic manner. The heat budget analysis suggests that the zonal advection is the primary factor that contributes to such nonlinear SST response. The radiative heating due to the greenhouse effect (hereafter, RHG) that is obtained from the multi-model ensemble of the Climate Model Intercomparison Project Phase III (CMIP3) was externally given to ICM. The RHG obtained from the twentieth century simulation intensified the cold tongue cooling and the subtropical warming, which were further intensified by the RHG from the doubled CO2 concentration simulation. However, the cold tongue cooling was significantly reduced and the negative SST response region was shrunken toward the equator by the RHG from the quadrupled CO2 concentration simulation, while the subtropical warming increased further. A systematic RHG forced experiment having the same spatial pattern of RHG from doubled CO2 concentration simulation with different amplitude of forcing revealed that the ocean dynamical response to global warming tended to enhance the cooling in the tropical eastern Pacific by virtue
S.
of meridional advection and upwelling; however, these cooling effects could not fully compensate a given RHG warming as the external forcing becomes larger. Moreover, the feedback by the zonal thermal advection actually exerted the warming over the equatorial region. Therefore, as the global warming is intensified, the cooling over the eastern tropical Pacific by ODT and the negative SST response area are reduced.
1 Introduction Different from the high latitude region (Manabe and Stouffer 1980; Moritz et al. 2002), the tropical region is less sensitive to changes in external forcing. The relatively stable condition of the tropical ocean through the earth's history is primarily due to oceanic thermal inertia as well as strong thermodynamical and dynamical thermostats (e.g., Newell, 1979; Ramanathan and Collins, 1991, 1992; Clement et al., 1996; Seager and Murtugudde, 1997; Liu and Huang, 1997; Li et al., 2000; Xie et al., 2010; An et al., 2012). In particular, the evaporative cooling effectively regulates the sea surface temperature (SST) in the warm oceans such as the equatorial western Pacific, i.e., “warm pool” (Newell 1979; Wallace 1992; Fu et al. 1992), where the warmest SST in the world ocean is routinely recorded. On the other hand, the oceanic dynamical cooling strongly regulates the equatorial eastern Pacific SST (also called the cold tongue temperature) to maintain on average, a few degree colder SST compared to the tropical western Pacific, which is called the “oceanic dynamical thermostat” (ODT) (Clement et al. 1996; Xie et al. 2010). The increase in the concentration of greenhouse gas in the atmosphere enhances the downward long-wave radiation to the earth surface, thereby obviously leading to increase the SSTs in both the warm pool and cold tongue regions. Theoretically, the warming trend in the warm pool region
S.-I. An, S.-H. Im
should be weaker than that in the cold tongue region because the deeper mixed layer and the warmer surface temperature in the warm pool are responsible for a larger heat capacity and effective evaporative cooling, respectively. Furthermore, the blocking mechanism of incoming solar radiation through increasing cirrus cloud cover with increasing warm pool temperature also acts to reduce the warming trend of warm pool (Ramanathan and Collins 1991). However, the observed SST warming trend for the last 100 years tends to be larger for the warm pool than the cold tongue (Cane et al. 1997; Karnauskas et al. 2009; Compo and Sardeshmukh 2010; An et al. 2012). This La Niña-like response is generally understood to be a result of the upwelling cooling over the cold tongue region overcompensating for the radiative surface warming, namely ODT mechanism. As a result, the greenhouse effect initially causes a stronger warming trend over the warm pool region than over the cold tongue region (Clement et al. 1996; Seager and Murtugudde 1997; Fang and Wu 2008; An et al. 2012). Clement et al. (1996) introduced ODT mechanism. Using Zebiak–Cane coupled ocean–atmosphere model (Zebiak and Cane 1987), they demonstrated that a uniformly imposed surface heating is almost balanced by equatorial ocean upwelling cooling in the eastern equatorial Pacific so that ocean dynamics could regulate the tropical climate. Seager and Murtugudde (1997) confirmed the heat balance shown by Clement et al. (1996) using the ocean general circulation model, and they further argued that the sensitivity of the surface fluxes to SST perturbations increases with latitude so that the uniform heating results in a smaller SST change in the subtropics than in the equatorial region indicating a zonally symmetric response. However, since the surface current associated with mean upwelling in the eastern equatorial Pacific moves heat poleward, the SST change in the eastern pacific is smaller than in the western Pacific (Seager et al. 1988; Clement et al. 1996). Fang and Wu (2008) also confirmed the results of Clement et al. (1996) using a fully coupled GCM with a similar experimental design to Clement et al. (1996) but with zonally symmetric heating. They further showed that the imposed heating intensifies the zonal SST gradient in the equatorial Pacific, which leads to intensification of equatorial upwelling in the eastern Pacific through Bjerkness feedback (Bjerkness 1969). As the warming further progresses, the delayed warming of the ocean subsurface gradually reduces the ocean stratification (An et al. 2008) and reduces the efficiency of the ODT as well. This is possible because the weaker ocean stratification results in the less trapping of momentum in the ocean mixed layer provided by wind stress forcing, and thus, the currents and upwelling in the ocean mixed layer become weaker as the ocean stratification is reduced. Therefore, the net thermal damping may decline, as shown by Xie et al. (2010), and the warming of the cold tongue could possibly exceed the warming of the warm pool.
By analyzing the Climate Model Intercomparison Project Phase III (CMIP3) data subjected to a future global warming scenario, An et al. (2012) demonstrated that the warmer climate could lead a regime shift in the tropical Pacific from a La Niña-like to El Niño-like response to global warming because the efficiency of the ODT decreased as the global warming progressed. However, the primary cause of the eventual weakening of the ODT effect predicted by global coupled GCMs (Vecchi and Soden 2007) has not yet been determined. In the past, the efficiency of the ODT was diagnosed using the intermediate coupled model (Clement et al. 1996) and ocean general circulation model (Seager and Murtugudde, 1997). They tested the SST response to the spatially uniform forcing with just the ocean or coupled with a simple atmosphere model and concluded a La Niña-like response to the uniform heating. Fang and Wu (2008) further demonstrated a La Niña-like response to the imposed zonally symmetric surface heating with the equatorial maximum that mimics the global warming. On the contrary, Kug et al. (2011) argued that the non-uniform atmospheric forcing associated with global warming could lead to an El Niño-like response through modifying the equatorial trade wind. Kug et al. (2011) also stated that even a uniformly warmed ocean surface generates non-uniform atmospheric forcing, possibly because of the nonlinearity of SST-induced precipitation, which in turn weakens the equatorial trade wind and reduces the east–west SST gradient by suppressing the equatorial upwelling over the eastern Pacific. In addition, Liu (1997) and Liu and Huang (1997) proposed that the zonal SST gradient over the equatorial Pacific is regulated by an upper bound that is determined by the latitudinal difference of the equilibrium SST and the dynamic coupling strength of the atmosphere–ocean coupled system. In other words, the latitudinal, non-uniform heating may influence the upper bound of the zonal SST gradient (e.g., Liu et al. 2005; Fang and Wu 2008; Kug et al. 2011). Therefore, it is natural to raise the following questions: Can the ODT strictly regulate the tropical Pacific SST to nonuniform radiative forcing? Is the cooling effect over the tropical eastern Pacific by ODT linearly proportional to the external heating intensity? In this study, we investigated the role of ODT in regulating the tropical Pacific SST under a global warming scenario using an intermediate ocean–atmosphere coupled model (ICM). Section 2 of this paper introduces the data and ICM utilized. We computed the ICM's response first to idealized, radiative heating at the ocean surface (Section 3) and then to radiative heating due to a future global warming scenario (Section 4). Finally, Section 4 discusses the sea surface temperature (SST) budget, which quantifies the effect of the ODT on the SST. The summary and discussion of results of this study are presented in Section 5.
Role of ODT in regulating SST under global warming conditions
were computed using the ocean model forced by the basic state surface winds. Details can be found in Zebiak and Cane (1987).
2 Data and intermediate coupled model (ICM) 2.1 Data from the CMIP3 Simulations from a total of five models from the World Climate Research Program Coupled Model Intercomparison Project Phase III (WCRP CMIP3; https://esg.llnl.gov:8443/ index.jsp) were utilized in this study, including the GFDL_ CM2.0, GFDL_CM2.1, MIROC3.2 (medres), MPI_ ECHAM5, and MRI_CGCM2.3.2a models. We used the twentieth century (20C3M), CO2 doubling (2CO2), and CO2 quadrupling (4CO2) scenario simulations. The twentieth century simulation included data from 1850 to 2000 AD, which were simulated under climate forcing caused by greenhouse gases, insolation, volcanic eruption, and aerosols. The doubling and quadrupling CO2 scenario simulations computed the projected response as a 1 % per year increase in the CO2 concentration. Both simulations were started at 348 ppmv and then increased to 696 ppmv for the doubling experiment and to 1,392 ppmv for the quadrupling experiment, respectively, and then, they were integrated continuously with a fixed CO2 concentration rate. 2.2 Intermediate ocean–atmosphere coupled model The model utilized in this study was the “Cane–Zebiak” (CZ) model (Cane et al. 1986; Zebiak and Cane 1987), which has been used numerous times for El Niño-Southern Oscillation (ENSO) studies and predictions. The atmospheric component is expressed as a steady-state, linear shallow water equation in an equatorial beta-plane. The ocean model has the fixed-depth mixed layer where SST anomalies are determined and the dynamical upper ocean is expressed as a linear shallow water equation. Ocean currents are computed by a combination of the geostrophic component and the Ekman component. The SST is determined using the horizontal advections in the mixed layer, vertical advection caused by the upwelling, and the Newtonian cooling comprehending all of the thermodynamical processes. SST tendency equation is as follows: ∂T 0 ∂T 0 ∂ ðT m þ T 0 Þ ∂T 0 ∂ ðT m þ T 0 Þ − vm ¼ − um − u0 − v0 ∂x ∂y ∂t ∂x ∂y ∂T 0 ∂ ðT m þ T 0 Þ − aT 0 ; − fM ðwm þ w 0 Þ−M ðwm Þg −M ðwm Þ ∂z ∂z
ð1Þ
where T is the SST, u and v are the zonal and meridional mixed layer currents, respectively, and w is the upwelling velocity. The subscript m and the superscript prime indicate the prescribed climatological mean quantity and anomaly, respectively. The value of M(w ) is w when w is positive, otherwise the value of M(w ) is 0. The basic states for SST and surface winds were obtained from monthly climatological data, and the basic states for the current and upwelling velocity
3 Idealized experiments In order to compute the SST response, the external fixed radiative forcing (Q ) was given in the whole model domain, and then, the model was integrated for 1,000 years for each experiment. Practically, we add the Q term into the right-hand side of Eq. 1 and take an average for the last 500 years. This long-term average was assumed to be a steadystate response for a given forcing because the difference among each 100-year segment average taking from the last 500 years data is quite small compared to the 500year average. From Eq. 1, the steady-state solution of the model can be assumed to be 0 Q ∂T 0 0 ∂ðT m þ T Þ ¼ um þu ρC p H ∂x ∂x 0 ∂T ∂ðT m þ T 0 Þ þ v0 þ vm ∂y ∂y ∂T 0 ∂ðT m þ T 0 Þ ; þ fM ðwm þ w0 Þ−M ðwm Þg þ M ðwm Þ ∂z ∂z þ½aT 0 þ½R
ð2Þ where Q is the external forcing; each term within the [ ] is the temporal average for 500 years; the first three groups are the zonal, meridional advections, and upwelling, respectively; the fourth group is the thermodynamical damping term (i.e., Newtonian cooling); and the last term is the residual. A residual value may be attributed to the asymmetric oscillatory feature of the ICM, but this value was negligible. The constant thermodynamical damping rate, a(=(125 days)−1) is primarily caused by evaporative cooling and depends on the mixed layer depth. Therefore, the value of the damping rate may not be a constant in general. However, we assumed this value to be constant in order to simplify our study, since our main focus was to determine the dynamical response. In order to quantify a simplest possible response, we assumed that the radiative external forcing was uniformly constant over the entire domain, i.e., Q (x , y )=Q 0 =constant. Actually an almost identical experiment using the same ICM was completed by Clement et al. (1996). As shown in Fig. 1, the uniform heating over the entire tropical Pacific results in the cooling of the cold tongue and the warming of the rest of the Pacific. Warming centers are zonally elongated along the subtropical region and the east–west contrast along the equator intensifies as the surface heating increases. Our results indicated that cold tongue cooling is primarily due to the
S.-I. An, S.-H. Im Fig. 1 a Annual mean SST response to the uniform external surface heating (Q =4 W/m2) obtained from the intermediate coupled model. b Equatorial band (5°S–5°N) averaged annual mean SST response to varied uniform external surface heating. Units are in degree Celsius. The results for 0 (control), 1, 2, 4, and 8 W/m2 are indicated by the solid, longshort-dashed, dotted, short dashed, and dot-dot-dashed lines, respectively
a
b
ocean dynamical response, i.e., ODT, with respect to external surface heating, and therefore, we reproduced the results of previous studies by Clement et al. (1996). However, the model does not linearly respond to external surface heating (see Fig. 1b) as indicated by the SST change for the doubled heating from 2 to 4 W/m2 being 1 order of magnitude less than the SST change for the doubled heating either from 4 to 8 W/m2 or from 1 to 2 W/m2, indicating a possible nonlinear response in amplitude. The values of each term in Eq. 2, which correspond to the experiments noted above and represent the heat budget averaged over the Niño-3 region (5°S–5°N, 150°–90°W), are presented in Table 1. It should be noted that the positive and negative values for the heat budget terms, i.e., the right-hand side of Eq. 2, indicate cooling and warming of the ocean mixed layer, respectively. The surface heating is largely balanced over the cold tongue region by two horizontal advections and about a quarter of the heating is compensated for by the upwelling. Therefore, the net heat flux becomes negative (downward flux), indicating the negative SST anomaly. Interestingly, the negative SST anomaly is primarily due to overcompensation by the sum of the horizontal advections and not by the upwelling. These results are dependent on the background state and the air–sea coupling because the SST tendency is modified depending on the background state as inferred from Eq. 2. The surface heating of the warm pool is not fully compensated for by the dynamical cooling, and therefore, an upward heat flux is necessary, resulting in a positive SST anomaly. For the second idealized experiment, we modified the forcing function so that the spatial structure was in the
meridional direction. The resulting forcing function is represented by the following equation: y 2 Qðx; yÞ ¼ Q0 W =m2 exp − ⋅ ð3Þ 10 O
where Q 0 is constant and y is a latitudinal grid in degrees. The maximum value of Q(x, y) in Eq. 3 is found along the equator and the value decreases exponentially toward the poles. This mimics the spatial pattern of the radiative heating due to greenhouse effect (see Fig. 5). As shown in Fig. 2, the equatorial structure of the SST response is similar to previous results (Fig. 1). However, the off-equatorial response is noticeably decreased due to weak surface heating and the cold tongue expands further toward the poles. The cooling rate of the cold tongue increases again as the external surface heating increases and the nonlinear response recorded in the previous experiment was also observed. In order to examine the model sensitivity to the meridional gradient of the external heating, the forcing function was further modified to the following equation: y 2 2 Qðx; y; λÞ ¼ Q0 W =m exp − ð4Þ λ ; where λ is an e-folding scale, and thus, the meridional slope of the external heating is inversely proportional to λ. We performed a series of simulations for different λ and Q 0 values. The response patterns obtained from this series of experiments looked similar to Fig. 2, and, therefore, we focused on the change in the cold tongue temperature. Figure 3 shows the Niño-3 SST (5°S–5° N and 150–90°W) obtained from the ICM simulation for a given λ and Q 0 value. As seen in Fig. 3,
Role of ODT in regulating SST under global warming conditions Table 1 Heat budget of each term of Eq. 2 averaged over the Niño-3 region obtained from the idealized experiment and associated with Fig. 1. ZADV, MADV, UP, HF, R indicate the zonal advection, meridional advection, vertical advection, surface heat flux, and residual value, respectively. Positive Q indicates warming, while positive value of ZADV, MADV, UP, HF, and R indicates cooling of ocean surface. Negative values are opposite Q [W/m 2 ]
ZADV MADV UP [W/m 2 ] [W/m 2 ] [W/m 2 ]
HF [W/m 2 ]
R SSTa [W/m 2 ] [°C]
2 4 8
1.98 2.25 5.71
−1.85 −2.14 −4.24
0.02 0.00 0.03
1.06 3.39 5.22
0.79 0.50 1.28
−0.40 −0.46 −0.91
when the external heating is weak, the cooling of the cold tongue is almost linearly proportional to the given external heating regardless of the meridional gradient of Q. However, as the external heating becomes stronger (>6 W/m2), the cold tongue SST becomes less sensitive to changes in the external heating. This was especially true for a large meridional gradient of Q (i.e., λ <15°); however, the cold tongue SST was still sensitive with a small meridional gradient of Q (i.e., λ <15°). These results indicate that a saturation level may exist for the cold tongue temperature (Liu, 1997). In addition, the largest cold tongue temperature change (i.e., Niño-3 SST anomaly) (approximately −0.9 °C) was obtained when the value of λ was between approximately 19° and 25° and the value of Q 0 was approximately 8.5 W/ m2. These results indicate that a nonlinear response of this system is more pronounced depending on the meridional structure of the external forcing.
Fig. 2 As in Fig. 1, but for the latitudinal dependent Q indicated in Eq. 3
a
b
We analyzed the heat budget for these sensitivity experiments in order to find the cause of the nonlinear behavior of the model response. Figure 4 illustrates the zonal, meridional, and vertical advections over the Niño-3 region in the Q 0 vs. λ, which correspond to the first three groups in Eq. 2, respectively. As seen in Fig. 4, the meridional and vertical advections both depend on the value of Q 0 but not λ, except for higher values of Q 0. The value of λ influences the SST response for higher Q 0 values, but the heating pattern of these two variables does not resemble the SST response pattern (Fig. 3). On the other hand, the zonal advection pattern is similar to the SST response pattern, indicating that the nonlinear response of the SST pattern to global warming may be attributed to zonal advection. This conclusion will be further confirmed in the following section.
4 Actual radiative forcing experiments In the previous section, we discussed the results of an experiment testing the sensitivity of the ICM to idealized surface heating. In this section, we show the model responses to the actual radiative forcing due to greenhouse effect obtained from the CMIP3 climate change scenario experiment. The radiative heating by the greenhouse effect (hereafter, RHG) was computed by subtracting the outgoing long-wave radiation at the top of the atmosphere from the upward long-wave radiation at the surface, and the corresponding radiative forcing was assumed to be the backward radiation to the surface. Furthermore, the net backward radiation of each scenario experiment was
S.-I. An, S.-H. Im
a
Fig. 3 Annual mean Niño-3 SST response to a given external surface heating in Q–λ domain. See details in text. Units are in degree Celsius
computed by subtracting the backward radiation of the preindustrial experiment. The surface heating from the backward radiation caused by the net greenhouse effect and the corresponding SST response were computed using the ICM and are shown in Fig. 5. Same as the idealized experiment, here we integrate the model for 1,000 years and take an average for the last 500 years as a final result. The backward radiation for the 20C3M experiment is an order of 1, but the backward radiation for the 2CO2 experiment is 1 order of magnitude greater. The backward radiation for the 4CO2 experiment is approximately double the backward radiation for the 2CO2 experiment. The backward radiation increases almost linearly as the CO2 concentration increases, but the SST response does not simply follow the radiative forcing pattern. For the 20C3M experiment, the SST response pattern is similar to that of the idealized experiment shown in Fig. 1. This result may be due to the fact that the heating rate is rather uniform. The cold tongue becomes colder for the 2CO2 experiment and the off-equatorial SST gets warmer, so the meridional SST gradient is intensified. For the 4CO2 experiment, the cooling over the cold tongue narrows and weakens, while the off-equatorial warming is even further intensified compared to that of the 2CO2 experiment. The backward radiation associated with the greenhouse effect is much dominant over the equatorial region than over the offequatorial region and it intensifies as the CO2 concentration increases. The SST response increases as the backward radiation increases over the off-equatorial region. This generally linear response implies that the thermodynamical thermostat primarily regulates the off-equatorial SST. On the other hand, the rather nonlinear response of the equatorial eastern Pacific SST infers that the ODT plays an important role in regulating the SST over that region. It appears that the cold tongue
b
c
Fig. 4 Zonal (a), meridional (b), and vertical (c) advections associated with the Niño-3 SST anomaly shown in Fig. 3. See details in text. Units are in watts per square meter
response to the radiative forcing for the 4CO2 experiment reaches or passes its upper bound, and this result is in agreement with the results of previous idealized experiments.
Role of ODT in regulating SST under global warming conditions
a
e
b
f
c
g
d
h
Fig. 5 Ensemble of warming rate by greenhouse gases (left; watts per square meter) obtained from the ensembles of CMIP3 models for a 20C3M, b 2CO2, c 4CO2, and the corresponding SST response to the ensemble warming rate (right; degree Celsius) obtained from an intermediate coupled model for e 20C3M, f 2CO2, g 4CO2. d Equatorial band
(5S–5N) averaged warming rate of a (solid line), b (long dashed line), and c (dashed line). h As in d but for SST response. Scales for d and h are watts per square meter and degree Celsius. Contour intervals for a, b, and c are 0.5, 3.0, and 5.0 W/m2, respectively, and those for e, f, and g are 0.2, 0.5, and 0.5 °C, respectively. Negative values are shaded
The heat budget averaged over the Niño-3 region for the experiment discussed above is presented in Table 2. As seen in Fig. 5, the Niño-3 SST anomaly for the 2CO2 experiment is colder than that for the 20C3M experiment. However, as the CO2 concentration levels increase further, the anomaly becomes smaller. Interestingly, as the radiative heating (or CO2 concentration) increases from 2CO2 to 4CO2, both the meridional and vertical advections increase [especially the vertical advection (upwelling cooling)], but the zonal advection
decreases. Therefore, the results of the RHG experiment are consistent with those of the previous idealized experiment in the sense of that the nonlinear SST response is related to changes in the zonal advection. Note that in the idealized experiment, the zonal advection was saturated as the external heating increases, while in the RHG experiment, the zonal advection decreased. Finally, we computed the SST response to radiative forcing, which was the same as that from the 2CO2 experiment,
Table 2 As in Table 1, but for those results associated with Fig. 5. RHG indicates the radiative heating due to the greenhouse effect
Experiment
RHG [W/m 2 ]
ZADV [W/m 2 ]
MADV [W/m 2 ]
UP [W/m 2 ]
HF [W/m 2 ]
R [W/m 2 ]
SSTa [°C]
20C3M 2CO2 4CO2
1.75 19.83 42.15
1.65 7.79 6.64
0.96 10.74 16.70
0.71 4.88 18.89
−1.57 −3.64 −0.33
0.00 0.06 0.25
−0.34 −0.78 −0.058
S.-I. An, S.-H. Im
but we changed its amplitude in order to test the importance of the forcing intensity in driving the nonlinear behavior of the SST response. As shown in Fig. 6, the cold tongue amplitude gradually increases as the amplitude of the forcing increases and reaches its maximum amplitude at 75 % of the original forcing. Then, the cold tongue amplitude gradually decreases as the amplitude of the forcing increases. The area of the negative SST anomaly also gradually narrows as the forcing increases. On the other hand, the off-equatorial SST response increases as the forcing increases (see Fig. 7). These results support the applicability of the previous idealized experiment to the realistic case. The heat budget shown in Table 3 further confirms the importance of the nonlinearity associated with the zonal advection. Both the meridional and vertical advections monotonically increase as the forcing increases, while the zonal advection reaches its maximum at 125 % of the original forcing and then decreases. To reveal why the cooling rate of cold tongue is reduced and the negative SST anomaly response area over the eastern
tropical Pacific is narrowed as the global warming increases, we represent the meridional distribution of each dynamical thermal advection term. As seen in Fig. 8, over the out of the equatorial area, all dynamical cooling effects (i.e., zonal, meridional, and vertical advections) are small regardless the external radiative forcing, and thus, the thermodynamical cooling should be balanced with the external heating through the increasing of the surface temperature. Although there is the gradual intensification of both the meridional, especially slightly off the equator (~3°N and 3°S–10°S), and vertical advections over the equator (2°S–2°N), but these cold advections do not fully compensate the increasing external heating. Therefore, the negative SST response area is gradually narrowing as the radiative forcing increases. Furthermore, once the positive SST anomaly appears, then the zonal advection reinforces the positive SST tendency (e.g., 3°N–5°N) (Fig. 8a), and thus, this feedback further squeezes the negative SST anomaly area and reduces the cooling tendency. Note that consistent results for the meridional distribution of heat
a
e
b
f
c
g
d
h
Fig. 6 a–g Annual mean SST response to the fixed pattern of the warming rate for 2CO2 experiment in the ensemble of CMIP3 models with different amplitudes as indicated in each panel. h Equatorial band (5S–5N) averaged value (unit: degree Celsius) of a (open square), b
(open circle), c (open triangle), d (open diamond), e (closed square), f (closed circle), and g (closed triangle). See details in text. Contour intervals are 0.5 °C. Negative values are shaded
Role of ODT in regulating SST under global warming conditions
Fig. 7 As in Fig. 6 but for latitudinal distribution of a band averaged SST anomaly over 150–90°W for amplitude coefficient of 0.25 (open square), 0.5 (open circle), 0.75 (open triangle), 1.0 (open diamond), 1.25 (closed square), 1.5 (closed circle), and 2.0 (closed triangle). Units are in degree Celsius
budget are also found in the previous RHG experiment (not shown here).
5 Summary and discussion In order to investigate the effect of the ODT, the tropical Pacific SST response to global warming was computed using the ICM. The main reason to use this model is that the SST in
Table 3 As in Table 1, but for those results associated with Fig. 6. RHG indicates the radiative heating due to the greenhouse effect
0.25 0.5 0.75 1.0 1.25 1.5 2.0
this model is primarily regulated by the ODT due to simple formulations of both the atmosphere model and the surface heat flux. The uniformly constant external heating produces a La Niña-like SST response. As the external heating increases, the cold tongue is also intensified, but not monotonically. The non-uniform heating with the Gaussian shape centered at the equator also results in a La Niña-like pattern of SST. However, the cold tongue temperature becomes less sensitive to changes in external heating when the meridional gradient of the external heating pattern becomes rather large. This nonlinear response of the cold tongue cooling with respect to changes in the external forcing is controlled by zonal advection. The atmospheric warming caused by greenhouse gases was computed using the scenario experiments of the CMIP3, including the 20C3M, 2CO2, and 4CO2 experiments, and then, the computed heating were entered into the ICM as an external forcing. This external forcing has the shape of Gaussian distribution with its maximum value at the equator, which resembles the results from the non-uniform idealized experiment. The positive external heating over the surface computed in the 20C3M, 2CO2, and 4CO2 experiments always leads to a negative SST anomaly over the equatorial eastern Pacific and a positive SST anomaly over the offequatorial Pacific and the tropical western Pacific. The cooling over the cold tongue was more intense in the 2CO2 experiment than in the 20C3M experiment. The SST response induced by the radiative forcing in the 4CO2 experiment was significantly smaller than that in the 20C3M experiment. These ICM experiments may be applicable for understanding the differing response across climate models. The heat budget analysis of the RHG experiment confirmed that the nonlinear response of the cold tongue to the external surface heating is related to changes in the zonal advection as previously proposed from the idealized experiments. Over the off-equatorial region, the dynamical cooling is so small that the external heat forcing is mainly balanced by the thermodynamical cooling, and thus, SST almost linearly increases as the forcing increases. Over the equatorial region,
RHG [W/m 2 ]
ZADV [W/m 2 ]
MADV [W/m 2 ]
UP [W/m 2 ]
HF [W/m 2 ]
R [W/m 2 ]
SSTa [°C]
4.96 9.92 14.88 19.83 24.79 29.75 39.67
2.80 5.40 6.70 7.79 8.93 8.84 7.06
3.15 5.68 8.77 10.74 13.18 15.53 18.04
1.54 2.39 3.31 4.88 6.31 8.35 15.63
−2.54 −3.59 −3.95 −3.64 −3.56 −2.91 −1.24
0.01 0.04 0.05 0.06 −0.07 −0.06 0.18
−0.55 −0.77 −0.85 −0.78 −0.77 −0.62 −0.26
S.-I. An, S.-H. Im
a
b
c
Fig. 8 As in Fig. 7 but for a zonal advection, b meridional advection, and c vertical advection. Scale for x-axis is watts per square meter. Positive (negative) values correspond to the cooling (warming) SST tendency
the upwelling over the equator and the meridional advection over the slight off the equator increase as the external heat forcing increases, but they cannot fully compensate the external heating. Therefore, the positive SST anomaly is induced over the equatorial region too, and moreover, this perturbation is intensified by the zonal advection. The model used in this study does not take into account possible changes in the temperature at the thermocline depth, which may respond to variations in the SST. This is because the atmospheric response to variations in the SST in turn modifies the thermocline structure, and in addition, the subsurface ocean temperature associated with the thermocline is known to respond to global warming with a time delay of more than a decade (An et al., 2008). In other words, variations in the subsurface ocean temperature are expected to modify the efficiency of the ODT by affecting the upwelled water temperature. As one possible example regarding this issue, Seager and Murtugudde (1997) computed thermocline temperature and concluded that the thermocline adjustment is not sufficient to cancel the thermostat mechanism. However, there is a limitation in a direct comparison between theirs and ours because they use a forced ocean general circulation model. Furthermore, we use the same background states in our experiment regardless of the external forcing. Since change in the external forcing leads to different climate states, it is a crude assumption. It is expected that changing the climate state may change the efficiency of ODT. To what extent the efficiency of ODT changes as the climate state changes is not easily investigated because they are dynamically linked. Another caveat is that the buoyancy forcing was not taken into account in ICM. Since the dynamical thermal advection controls ocean surface properties, the resultant thermodynamical feedback associated with buoyancy forcing is expected. However, because the wind forcing is more important than buoyancy forcing in terms of controlling of the upper ocean storage (Wang et al. 2003), our experiment
may be valid with the first-order approximation. Although our choice of the damping rate belongs to a reasonable range, in order to check the model sensitivity, we performed the same experiment with strong thermal damping (α = 1/60 day−1) and found that the nonlinear response was rather reduced (not shown). Further increase of the linear damping rate might reduce the nonlinear properties of the model response even further. This is because the additional heat is mainly balanced to the linear damping ([αT ′]½αT in Eq. 2). Thus, to compute more actuate response of SST to global warming, both dynamical and thermodynamical feedback including atmospheric cloudradiation feedback such as shortwave damping must be taken into account. On the whole, the effect of the ODT tested in this study is somewhat exaggerated than it would be in nature due to the simple formulation of the atmospheric model and the thermodynamical process. Nevertheless, we confirmed that the range of the SST that can be regulated by the ODT may be limited depending on the pattern and amplitude of the external forcing. This limitation is expected to occur under higher CO2 concentrations as shown in An et al. (2012). Finally, it should be noted that by contrast with An et al. (2012), this study presents quantitative efficiencies of ODT due to not only upwelling cooling but also horizontal thermal advection, and by using ICM, ODT is separately investigated from other regulation processes such as atmospheric processes. In addition to such detailed investigation, we found that the blunt ODT in the equatorial eastern Pacific under the higher CO2 concentration condition is related to the feedback by the zonal thermal advection. Acknowledgments We would like to thank two supportive reviewers, and J. Choi who helped us implement the intermediate coupled model. This work was supported by the National Research Foundation of Korea Grant funded by the Korean Government (NRF-2009-C1AAA001-20090093042).
Role of ODT in regulating SST under global warming conditions
References An SI, Kug JS, Ham YG, Kang IS (2008) Successive modulation of ENSO to the future greenhouse warming. J Climate 21:3–21 An SI, Kim JW, Im SH, Kim BM, Park JH (2012) Recent and future sea surface temperature trends in tropical Pacific warm pool and cold tongue regions. Clim Dyn 39:1373–1383 Bjerknes J (1969) Atmospheric teleconnections from the equatorial Pacific. Mon Weather Rev 97:163–172 Cane MA, Zebiak SE, Dolan SC (1986) Experimental forecasts of El Nino. Nature 321:827–832 Cane MA, Clement AC, Kaplan A, Kushnir Y, Pozdnyakov D, Seager R, Zebiak SE, Murtugudde R (1997) Twentieth-century sea surface temperature trends. Science 275:957–960 Clement AC, Seager R, Cane MA, Zebiak SE (1996) An ocean dynamical thermostat. J Climate 9:2190–2196 Compo GP, Sardeshmukh PD (2010) Removing ENSO-related variations from the climate record. J Climate 23:1957–1978 Fang C, Wu L (2008) The role of ocean dynamics in tropical Pacific SST response to warm climate in a fully coupled GCM. Geophy Res Lett 35, L08703. doi:10.1029/2007GL033097 Fu R, Del Genio AD, Rossow WB, Liu WT (1992) Cirrus-cloud thermostat for tropical sea surface temperature tested using satellite data. Nature 358:394–397 Karnauskas KB, Seager R, Kaplan A, Kushnir Y, Cane MA (2009) Observed strengthening of the zonal sea surface temperature gradient across the equatorial Pacific ocean. J Climate 22:4316–4321 Kug JS, Sooraj KP, Jin F-F, Ham Y-G, Kim D (2011) A possible mechanism for El Nino-like warming in response to the future greenhouse warming. Int J Climatol 31:1567–1572 Li T, Hogan TF, Chang C-P (2000) Dynamic and thermodynamic regulation of ocean warming. J Atmos Sci 57:3353–3365 Liu Z (1997) Oceanic regulation of the atmospheric walker circulation. Bull Am Meteorol Soc 78:407–412
Liu Z, Huang B (1997) A coupled theory of tropical climatology: warm pool, cold tongue, and Walker circulation. J Climate 10: 1662–1679 Liu Z, Vavrus S, He F, Wen N, Zhong Y (2005) Rethinking tropical ocean response to global warming: the enhanced equatorial warming. J Climate 18:4684–4700 Manabe S, Stouffer RJ (1980) Sensitivity of a global climate model to an increase of CO2 concentration in the atmosphere. J Geophys Res 85: 5529–5554 Moritz RE, Bitz CM, Steig EJ (2002) Dynamics of recent climate change in the Arctic. Science 297:1497–1502 Newell RE (1979) Climate and the ocean. Amer Sci 67:405–416 Ramanathan V, Collins W (1991) Thermodynamic regulation of ocean warming by cirrus clouds deduced from observations of the 1987 El Nino. Nature 351:27–32 Ramanathan V, Collins W (1992) Thermostat and global warming. Nature 357:649–650 Seager R, Murtugudde R (1997) Ocean dynamics, thermocline adjustment, and regulation of tropical SST. J Climate 10: 521–534 Seager R, Zebiak SE, Cane MA (1988) A model of the tropical Pacific sea surface temperature climatology. J Geophys Res 93:1265–1280 Vecchi GA, Soden BJ (2007) Global warming and the weakening of the tropical circulation. J Climate 20:4316-4340 Wallace J (1992) Effect of deep convection on the regulation of tropical sea surface temperature. Nature 357:230–231 Wang D, Wang J, Wu L, Liu Z (2003) Relative importance of wind and buoyance forcing for interdecadal regime shift in the Pacific Ocean. Sci In China (series D) 46:417–427 Xie S-P, Deser C, Vecchi GA, Ma J, Teng H, Wittenberg AT (2010) Global warming pattern formation: sea surface temperature and rainfall. J Climate 23:966–986 Zebiak SE, Cane MA (1987) A model El Nino-Southern Oscillation. Mon Weather Rev 115:2262–2278