1. Introduction
The mid-Holocene climate is characterized by pronounced changes in the insolation (Fig. 1), primarily due to precession but also obliquity, and robust climate changes during that time constitute a stringent test for climate theory and coupled climate models. A strong case for reduced ENSO variance during the mid-Holocene has been made from combining several sources in different locations, including coral, ocean, and lake sediment records; these have been reviewed in several previous papers [e.g., Cane et al. (2000) and Brown et al. (2008); a concise and current summary can be found in Roberts (2007, chapter 1)]. A novel study by Koutavas et al. (2006) well illustrates the changes to the eastern equatorial Pacific sea surface temperatures (SST) during this time: by measuring the δ18O of individual G. ruber in an ocean sediment core—effectively a random sampling of SST at various points in time—they were able to estimate the range in SST in mid-Holocene climate compared to the present (Fig. 2). While the mean SST did not change appreciably (Mg/Ca SST measurements suggest a cooling of ∼0.5°C), the variance was reduced by around 50%. They attribute the reduction to equivalent contributions by both ENSO and the annual cycle; while it is impossible to infer this breakdown directly from their data, their interpretation for significant ENSO reduction is supported by New Guinean coral records, which suggest a reduction to the ENSO variance by up to 80% (Tudhope et al. 2001).
We have leads on why this reduction occurs, but solution to the problem remains elusive. The poor performance of our current climate models in simulating the reduction parallels this lack of knowledge: previously reported fully coupled model simulations of the mid-Holocene generally achieve some reduction to the ENSO variance, but fail to do so at the level suggested by the paleodata (Liu et al. 2000; Zheng et al. 2008; Brown et al. 2006). We examined the mid-Holocene ENSO variance changes in the current-generation fully coupled models that participate in the Paleoclimate Modelling Intercomparison Project II (PMIP2) (Braconnot et al. 2007). Of the seven models that we examined, four exhibit mild reductions and one had a substantial increase; only one model shows a substantial (>50%) reduction to the ENSO variance (Table 1).1
A number of hypotheses have been proposed to explain the ENSO behavior. The most influential is the study by Clement et al. (2000), where they forced the Cane–Zebiak ENSO model (Zebiak and Cane 1987) with mid-Holocene insolation and find both a reduction to the annual cycle amplitude and ENSO variance. The Cane–Zebiak model is an anomaly model of the tropical Pacific only, and lacks many of physical processes that may affect ENSO during the mid-Holocene, in particular the effect of climate changes occurring outside the tropical Pacific on the tropical Pacific climate. The model’s simplicity, however, allowed Clement et al. to propose a specific hypothesis for mid-Holocene ENSO, making two particular points: the tropical Pacific is capable of directly responding to insolation changes, and the ENSO variance changes are a consequence of changes to the tropical Pacific mean state.
However, insolation changes during the mid-Holocene are seasonally large and global in extent, so the entire global climate adjusts to the insolation forcing. For ENSO, the relevant question is whether or not climate changes external to the tropical Pacific can influence ENSO in any way by this global adjustment. Liu et al. (2000) argued that, indeed, it may happen through two pathways: First, the intensification of the Asian summer monsoon increases the tropical Pacific trades, resulting in a steeper east–west equatorial SST gradient; the cooler conditions in the equatorial Pacific in turn suppresses the coupled interactions that grow an emerging El Niño in the early summer. Second, the equatorial thermocline weakens as a result of subduction of warmer southern subtropical Pacific waters, the latter being a consequence of the mid-Holocene insolation forcing. Altered seasonal forcing of the tropical Pacific by, for example, an increased Asian monsoon may also affect ENSO through the nonlinear mechanism of frequency entrainment (Chang et al. 1994; Liu 2002; Pan et al. 2005). More recently, Roberts (2007) argued that a cooler early Holocene tropical Pacific climate is responsible for the reduced ENSO activity, through reduction to the strength of the tropical ocean–atmosphere coupling.
The main goals of this paper are to (i) argue that the changes driving ENSO during the mid-Holocene are teleconnected from the outside, rather than a direct response of the tropical Pacific to mid-Holocene insolation, and (ii) advance an alternate hypothesis of mid-Holocene ENSO reduction from a reduction in stochastic forcing originating from the North Pacific. The basis of our claim comes from analysis and idealized simulations of an intermediate coupled model—an atmospheric general circulation model (AGCM) coupled to a reduced-gravity ocean model—that simulates present-day ENSO well and, when put through mid-Holocene insolation forcing, achieves significant reduction in both variance of the cold tongue annual cycle and interannual variability, comparable to that seen in paleoproxy data.
The larger implication of our study is highlighting the effect of the assumed ENSO paradigm on the mid-Holocene ENSO problem. There are two broad camps: one ascribes to a “nonlinear, weakly chaotic” paradigm where the ENSO system is self-sustaining without the need for external drivers, drawing its energy from the basic state. This is how the Cane–Zebiak model works and the Clement et al. (2000) hypothesis is, by implication, embedded in this paradigm. The other is a “linear, externally forced” paradigm where the ENSO system is damped and cannot sustain its own activity unless externally forced; this view was first articulated by Penland and Sardeshmukh (1995) and elaborated by, among others, Chang et al. (1996) and Thompson and Battisti (2000, 2001). Since models ascribing to either paradigm simulate modern-day ENSO with reasonable fidelity, one cannot discriminate between them, hence the dichotomy in the views of ENSO. The model that we use (section 2) turns out to be sensitive to external forcing and, hence, likely to be in the latter category. What we, in effect, show is that a linear, externally forced ENSO model is equally capable of explaining mid-Holocene ENSO reduction, so the mid-Holocene observation cannot (yet) discriminate between the two ENSO paradigms.
2. Model
Our model is the Community Climate Model, version 3.6 (CCM3), running at T42 resolution coupled to a 1.5-layer reduced-gravity ocean (RGO) with a resolution of 1° latitude × 2° longitude (hereafter referred to as the CCM3–RGO). The CCM3 is a widely used atmospheric model consisting of a spectral dynamical core and full physics package including radiation, convection, boundary layer, and a diagnostic treatment of clouds, as well as a prescribed land surface and sea ice (Kiehl et al. 1998). The 1.5-layer reduced-gravity ocean model had been extensively used for simulating coupled tropical ocean processes; the version used in this model is same as the one documented in Chang (1994) except that ours uses a time-fixed but spatially varying mixed layer depth estimated from Levitus (1982) data, and the variation of entrained subsurface temperature is parameterized in terms of variation of thermocline depth using a multivariate linear relationship (Fang 2005).
The model that we use is modified from the original setup in these ways. First, we replace the original monthly varying flux adjustment with a time-invariant flux correction, which allows the model ocean more latitude in responding to seasonal variations, at the expense of a slight degradation in simulating the SST climatological annual cycle. Second, since our climate changes require global SST adjustments, we extended the reduced-gravity ocean from 30°N/S out to the extratropics 80°N/S. Ocean points poleward of 80°N/S are handled by the mixed layer ocean built into the original CCM3. Since the RGO cannot handle the outcropping of thermocline at high latitudes, we set thermocline depth to mixed layer thickness wherever the former is shallower than the latter. In such circumstances, the surface currents and entrainment velocity will be calculated by the momentum and mass conservation equations. Finally, as the vertical entrainment is considered less important than the surface heat flux in the heat budget of the mixed layer ocean outside the tropics, we ignore the vertical entrainment term of the thermodynamic equation of the mixed layer poleward of 30°N/S. This way, we can retain the original relationship between the thermocline depth anomaly and the temperature of the water vertically entrained into the mixed layer, as used in the RGO.
The end result is a model that is capable of simulating the climatological mean state and annual cycle with fidelity, as well as the important tropical interactions that generate the leading modes of tropical Pacific interannual–decadal variations. It lacks some important physics that fully coupled models possess: in particular, sea ice feedbacks and deep ocean circulation. However, the climate components that we include have reasonably short adjustment time scales, which means that we only need to integrate a few decades for the model climate system to equilibrate. Furthermore, the model is sufficiently moderate in its use of computational resources, allowing us to do both long simulations and a suite of sensitivity tests—necessary to fully explore our problem.
Results of a 300-yr control simulation with present-day conditions (we refer to this simulation in the text as the “standard control”) show that the model reproduces the tropical Pacific mean state and annual cycle very well (Fig. 3, left); this is not surprising because of the flux correction and anomaly coupling employed but, since we do employ a time-independent correction, the annual cycle is simulated to some extent. Tropical Pacific variability is captured surprisingly well: in particular, the first three modes of a maximum covariance analysis (MCA) (Bretherton et al. 1992) on model SST and surface winds over the central and eastern Pacific resembles that observed in both spatial and temporal characteristics—the first mode in particular being ENSO and the second being the Pacific meridional mode (PMM) (see Fig. 4). Our model ENSO suffers from too much variance in the biennial band, and a SST anomaly (SSTA) that extends too far westward in the equatorial Pacific, typical of other models.2 However, the model clearly simulates ENSO with reasonable fidelity.
3. Simulation of the mid-Holocene
The same model was run for 249 years under mid-Holocene orbital conditions (6000 BP, hereafter 6K), taking the same values of eccentricity (0.018 682), obliquity (24.105°), and angular precession (0.87°) as used by the PMIP2 project for the mid-Holocene (we refer to this simulation in the text as the “standard 6K” or “standard mid-Holocene” simulation). All other parameters and boundary conditions were kept the same as the control. The mean SST in the eastern equatorial Pacific does not change significantly compared to the control, but does show a reduction in the range of the monthly Niño-3 index (SST averaged over 5°S–5°N, 150°–90°W) (Figs. 5 a,b). These are consistent with the SST measurements of Koutavas et al. (2006) for the mid-Holocene (cf. Fig. 5c with Fig. 2). The reduced range of Niño-3 is contributed both by reduction to the annual cycle amplitude and interannual variation: the total variance in monthly Niño-3 is reduced by 38.8% and ENSO variance (as measured by the 2–7-yr bandpassed Niño-3) is reduced by 39.8% (Table 2a). In fact, the variance reduces across the entire interannual-to-decadal part of the spectrum (Fig. 7, discussed below). By using 50-yr idealized simulations in which we separately apply the precession and obliquity components of the mid-Holocene insolation, the reduction was found to be virtually all from the precession change. (Note that in assessing significance of ENSO variance changes, we compare 50-yr simulations to the PDF of the variance of overlapping 50-yr segments of the 300-yr control run Niño-3; the lowest 5% of this PDF have variance lower than 0.156 K2, and we use this threshold to assess significance.)
The combination of reduced ENSO variance and reduced annual cycle amplitude significantly narrows the range of values that the cold tongue SST occupies, and in particular the cold tongue does not get as cold or as warm as it does in the present day. These appear consistent with various paleorecord measures of mid-Holocene ENSO activity, in particular the reduction of extreme warm ENSO events as measured by clastic laminate in an Ecuadorian lake record (Rodbell et al. 1999) and the warmer minimum cold tongue SSTs, suggested by Clement et al. (2000) as an interpretation of the survival of tropical mollusks in coastal Peru during that time period (Sandweiss et al. 1996). Our mean state and annual cycle changes are also qualitatively similar to what Clement et al. found in the Cane–Zebiak model when applied with mid-Holocene forcing.
Clement et al. (2000) called for reduction to the annual cycle and ENSO variance through changes in the tropical Pacific mean state caused directly by tropical insolation forcing changes wrought by the orbital changes. To assess this idea in our framework, we ran the CCM3–RGO for 50 years with mid-Holocene insolation applied within the deep tropical Pacific only (15°S–15°N, 120°–280°E), leaving the insolation outside the region the same as for the control simulation. Our model’s ENSO variance did not change significantly—a 9.7% reduction to the 2–7-yr bandpassed Niño-3 (Table 2b). A repeat of this idealized simulation, but expanding the tropical insolation changes to 27°S–27°N, reduces the variance further, but only marginally (Table 2c). An analogous simulation but with 6K orbital conditions applied outside the deep tropical Pacific showed, on the other hand, significant reduction (41.6%) to the variance of ENSO (Table 2d). We conclude from these that nonlocal changes to the insolation drive the 6K ENSO changes.
Furthermore, we find that teleconnected changes determine much of the annual cycle changes in the equatorial Pacific, though in this case the direct effect of mid-Holocene insolation on the tropical Pacific is also important. Using the change in the variance of the full Niño-3 as a quantitative measure of the annual variation, about half of the reduction in the cold tongue annual cycle can be accounted for by direct insolation forcing and the other half by teleconnection (cf. Table 2b and 2d). The reduction to the cold tongue annual cycle by the direct insolation can be readily explained as a thermodynamic response to the mid-Holocene insolation change, which increased over the equator when the cold tongue is in its cold phase (Fig. 3, right panel).
For equatorial Pacific climate fields determined strongly by the dynamics, teleconnected influences tend to dominate the changes in the 6K simulations. Figure 6 shows annual cycle changes to the equatorial surface zonal wind (left column), surface meridional wind (middle column), and temperature of the waters entrained into the mixed layer (right column) comparing the standard 6K simulation changes (top row) with changes in the simulation where 6K insolation is applied only in the tropical Pacific (middle row) and changes where 6K insolation is applied everywhere outside the tropical Pacific (bottom row; the bottom row of Fig. 6 will be referenced in the next section). For the equatorial zonal wind, a dominant feature in the standard 6K simulation are the reduced easterlies during boreal spring throughout the central and eastern equatorial Pacific, followed by increased trades in the western Pacific in the boreal summer. For the equatorial surface meridional wind, there are increased southward flow in the boreal winter and early boreal spring in the central-western Pacific, but relatively small changes in the other times of the year. These behaviors are largely reproduced by the simulation imposing 6K insolation outside the tropical Pacific, but are not reproduced in the simulation with 6K insolation imposed inside the tropical Pacific.
Since tropical surface winds determine the thermocline and, hence, the subsurface temperature structure, it is not surprising that the same holds when looking at changes to the entrainment temperatures into the mixed layer (Fig. 6, bottom column). The major features in the subsurface temperature anomalies in the standard 6K simulation are the decreased subsurface temperatures in the western Pacific throughout the year and increased temperatures in the latter half of the year in the central-eastern Pacific. These features are essentially reproduced in the simulation with 6K insolation applied outside the tropical Pacific. While the seasonal cycle of subsurface temperature is decoupled from the seasonal cycle of the equatorial Pacific SST (e.g., Gu et al. 1997), it may act to change longer-term variations (interannual and longer) of SST.
Given these results, it is reasonable to assume that the model’s mid-Holocene ENSO changes (annual cycle and ENSO) originate from outside the tropical Pacific.
4. Nonlocal influences on the tropical Pacific
We first look to previously proposed hypotheses of ENSO change through nonlocally induced changes to the tropical Pacific mean state and annual cycle to explain our result. From the analysis of a coupled model simulation of the mid-Holocene, Liu et al. (2000) had proposed that a strengthening summer Asian monsoon can drive decreased ENSO activity through its effect of strengthening the tropical Pacific trades and driving a colder cold tongue, which helps suppress ENSO activity. Despite the presence of a stronger Asian summer monsoon in the model, the central and eastern Pacific trades do not show a mean strengthening in the boreal summer (see Fig. 6, top-left panel), and the cold tongue is, if anything, slightly warmer during that time. Liu et al. also proposed that changes to the subduction of waters that make up the equatorial Pacific thermocline might affect ENSO; however, we are unable to assess this mechanism because of the simplified physics of our ocean model.
A relatively new idea is that ENSO variance can decrease as a consequence of larger-amplitude annual forcing on the tropical Pacific, through the mechanism of frequency entrainment (Chang et al. 1994; Liu 2002; Pan et al. 2005). Seasonality in the tropical Pacific is generated nonlocally, coming about through interhemispheric contrasts that drive pressure gradients and, hence, winds over the equatorial Pacific. Mid-Holocene insolation can drive larger monsoonal flows and hence provide larger seasonal forcing to the tropical Pacific. Pan et al. (2005) explored the frequency entrainment hypothesis by imposing larger zonal mean wind seasonal amplitude in the Cane–Zebiak model for ENSO (the Cane–Zebiak model is an anomaly model that requires the basic state to be imposed), finding that its ENSO decreases with increased annual cycle forcing, in agreement with the predictions of frequency entrainment.
We find that frequency entrainment is unlikely to be a significant factor in influencing ENSO changes in our simulations. This is confirmed by two observations. First, the annual cycle forcing on the model tropical Pacific does not appear to increase in the 6K simulations. While the amplitude of the annual cycle of meridional wind forcing in both the western and central equatorial Pacific increases significantly (Fig. 6, cf. the top panel for the anomalous V with the climatological annual cycle for V in the bottom panel), the relevant forcing on the equatorial ocean dynamics is with the zonal winds. The zonal wind annual cycle amplitude changes are however mixed, with increases in the eastern Pacific but decreases in the central Pacific [see Fig. 6, again, for U; we also note that we do not get the uniformly increased annual cycle amplitude in the zonal winds that Pan et al. (2005) assumed in their study]. The ocean dynamical response—as measured by the subsurface temperature—shows a strong weakening in the annual cycle amplitude in the eastern Pacific.
The second observation arguing against frequency entrainment relates to changes in the frequencies of the ENSO response. In general, frequency entrainment would tend to “pull” the ENSO oscillations toward the annual frequency, so with a larger annual cycle forcing ENSO variations should oscillate at higher frequencies. We compare power spectra of the control and mid-Holocene Niño-3 (Fig. 7). Rather than showing a redistribution of variance, it shows reductions across the board from multidecadal to annual periods, including pronounced declines in the 2–7-yr ENSO band. We conclude from these that there is no clear signature of frequency entrainment operating.
5. Pacific meridional mode and reduced ENSO variance
An alternative hypothesis for ENSO links its activity to external forcing (Chang et al. 1996; Thompson and Battisti 2000, 2001), the source of which may include variations to the Asian monsoon and the Madden–Julian oscillation. More recently, Vimont et al. (2003b) argued that forcing originates from the North Pacific wintertime midlatitude atmospheric variations that communicate to the tropical Pacific through a “seasonal footprinting mechanism” (Vimont et al. 2001), whereby North Pacific trade wind variations associated with the midlatitude variability leave a SST signature in the subtropical eastern North Pacific during boreal spring that, in turn, forces a tropical response during boreal summer.
This subtropical–tropical response has since been linked to a broader class of variations intrinsic to tropical ITCZ–cold tongue variations, the so-called meridional mode (Chiang and Vimont 2004; a brief description and index for the PMM is available online at http://www.cdc.noaa.gov/Timeseries/Monthly/PMM/). The Pacific meridional mode is independent of ENSO and shares similar characteristics and underlying physics to the well-known Atlantic meridional mode (aka the interhemispheric mode or tropical Atlantic dipole) that is found in the tropical Atlantic. The source of the PMM comes from wintertime variations in the northern subtropical trades: in one phase, the wintertime northern subtropical trades are weakened and evaporation reduced, making the underlying SSTs warmer. The north subtropical SSTs persist into the boreal summer when it forces equatorial wind stress anomalies, eliciting an equatorial response. A recent model study by Chang et al. (2007) shows that the PMM can mechanistically trigger ENSO events. Warm subtropical SST anomalies generated by the PMM in turn drives equatorial wind anomalies, and these in turn are thought to force ocean equatorial Rossby waves that increase the upper-ocean heat content and initiate an El Niño event (Vimont et al. 2003b). A rigorous understanding of the mechanism of PMM forcing on ENSO is still to be developed, but the statistical and coupled model evidence thus far presented is suggestive.
The model used in our study is essentially the same as that used in Chang et al. (2007), so, by implication, our model should be sensitive to changes in the PMM-type external forcing. Can the hypothesis of PMM forcing of ENSO be applicable to understanding our model’s mid-Holocene response? A simple lag correlation between the 300-yr control simulation December–February Niño-3 anomalies with SSTA of the prior March–May season shows a distinct PMM signature in the North Pacific, with maximum correlation in the northeastern equatorial Pacific of around 0.4 (Fig. 8); this is the largest correlation at any location and demonstrates that the PMM is a viable precursor of ENSO events in this model. Interestingly, the pattern in the Pacific also resembles the “stochastic optimal” pattern of ENSO, identified by Penland and Sardeshmukh (1995, their Fig. 6a), with same-sign loadings over the tropical eastern North Pacific (the PMM signature) and tropical eastern South Pacific. We produce a measure for PMM activity by projecting the PMM SSTA pattern found in the MCA analysis (Fig. 4, left panel of the middle row; the black solid box delineates the region onto which the SSTA pattern is projected) onto the SSTA of the control and 6K simulations; the reduction in the variance of the PMM index averaged over its peak months of March–May in the 6K simulation is 40%.
To discount the possibility that the PMM reduction may have been a consequence rather than a cause of the reduction in ENSO activity, we repeated the control and 6K simulations with the same model, but with the temperature of the subsurface water entrained into the ocean mixed layer by upwelling set to climatology; this means that variations in thermocline depth no longer influence SST, undermining the fundamental mechanism of ENSO. We hereafter refer to these simulations as the no-ENSO simulations. The first MCA mode for the no-ENSO simulations is the PMM (not shown), consistent with it being a mode of variability independent of ENSO and confirming that the PMM is the dominant mode of variation in the tropical Pacific beyond ENSO. We find also that the PMM activity (as measured by the variance of the March–May-averaged PMM index) in the 6K no-ENSO simulations decreases by 38% compared to the control simulation.
We performed an additional check on the PMM forcing through idealized simulations that artificially constrain all components of the surface fluxes in the north subtropical Pacific region (7°–20°N) to the appropriate monthly mean climatology. Because SST anomalies in that region now cannot be generated through surface fluxes, it should suppress PMM activity. We performed a control simulation and a 6K simulation, using the constraint (hereafter called the no-PMM runs). The climatologies of the no-PMM simulations are virtually indistinguishable from their standard counterparts (not shown); furthermore, the PMM variability becomes effectively muted in these simulations. The ENSO variance of the control no-PMM simulation is reduced by about half from the standard control, as would be expected if the PMM drives ENSO (Table 2e). In comparing the control no-PMM ENSO activity with that for the 6K no PMM, the 6K no-PMM ENSO variance is now slightly larger than that for the control no-PMM simulation (Table 2f)! On the other hand, the annual cycle amplitude of the cold tongue in the 6K no-PMM run is reduced from the control no PMM, similar to the standard simulations. We interpret this as further evidence for the PMM influence on mid-Holocene ENSO.
We further examined PMM activity in the PMIP2 mid-Holocene simulations. To do this, we performed a similar MCA analysis on tropical Pacific SST and winds as we did for the observations and CCM3–RGO model output (see Fig. 4, and section 2). Of the seven models that we checked (listed in Table 1), only three (CCSM, MIROC, and MRI) produce PMM behavior similar to observations; the other models either do not show PMM behavior, have a PMM that is too weak, or have incorrect seasonality. The change to the variance in March–May PMM activity in these three “good” models is shown in the last column of Table 1. It shows that the PMM activity in all three models is reduced, though by only by a few percent for the CCSM. More suggestively, the reduction in PMM activity correlates with the reduction in the 2–7-yr bandpassed Niño-3 activity, as one might expect if PMM drives ENSO. While this result is suggestive, unfortunately three models is too small a sample size to make any conclusions.
The question remains as to what leads to the reduction in PMM during the mid-Holocene. A strong source of North Pacific trade wind variations, though not the only source, is from intrinsic wintertime North Pacific midlatitude atmospheric variability associated with the North Pacific Oscillation (NPO) (Rogers 1981; Vimont et al. 2003a), characterized spatially by a dipole in sea level pressure with one pole in the central-eastern midlatitude North Pacific around 37°N, 160°W; the other source is centered around western Alaska. We check whether North Pacific wintertime atmospheric variability is decreasing in the 6K runs. We characterize the dominant mode of North Pacific wintertime interannual atmospheric variability through an empirical orthogonal function (EOF) of the wintertime sea level pressure, similar to the procedure followed by Vimont et al. (2002). The sea level pressure is averaged over December–March (DJFM) and the EOF is calculated for the correlation matrix over grid points within 15°–90°N, 110°–270°E. A cosine weighting is used to account for the variation of gridpoint area with latitude. The first EOF, using the control no-ENSO simulation output, is the NPO (not shown); this result is similar to that obtained by Vimont et al. (2002) for the CSIRO model, except that the variance explained (∼32%) is slightly higher in our model runs. We estimate the NPO activity of individual runs by projecting their DJFM-averaged sea level pressure anomalies onto this sea level pressure pattern. We find that the NPO activity for the standard 6K simulation is reduced by ∼50% compared to the standard control simulation. However, the NPO activity in the 6K no-ENSO simulation is reduced only by 20%, compared with its control.
Our result suggests that, while the reduction in NPO activity is likely to contribute significantly to the PMM reduction during 6K, this is not the whole story: there are complicating factors. There are likely to be other sources of stochastic forcing that alter the PMM variance in the 6K simulations; also, ENSO may feed back on the NPO activity. As will be shown in section 7, the wintertime westerlies in the subtropical North Pacific strengthen during the mid-Holocene, which would modify the upper-tropospheric potential vorticity gradients supporting the Rossby wave propagation out of the equatorial Pacific; this could potentially allow ENSO to affect the NPO.
6. External forcing of mid-Holocene ENSO
In this section, we show further evidence for external forcing in general to be implicated in the reduction to mid-Holocene ENSO and independent of the assumption of PMM influence on ENSO.
a. Decrease in atmospheric noisiness over the tropical Pacific
The mid-Holocene simulations display a general reduction in atmospheric variability in the tropical Pacific (the western half in particular) compared to the present-day simulations. Linear stochastically forced models of ENSO associate the ENSO activity directly to the level of stochastic forcing: for example, Thompson and Battisti (2001) show that the responses to individual monthly perturbations in their stochastically forced ENSO model sum up to give the total contribution to the model Niño-3 (their Fig. 9). We show that, in general, the variance of stochastic forcing in the mid-Holocene simulations appear to be reduced on daily time scales over the ENSO growth seasons of boreal spring and summer.
Figure 9 shows the change to the atmospheric stochastic forcing in the 6K no-ENSO simulation compared to the control no-ENSO simulation over the tropical Pacific basin, examining the variance of daily output of two variables: sea level pressure as a measure of the mechanical forcing and latent heat flux as a measure of the thermal forcing. We use the no-ENSO simulations in this analysis to emphasize that these are not a consequence of the ENSO response. Variance in daily sea level pressure is generally reduced over the western equatorial Pacific throughout the ENSO growth period of April through October (Fig. 9a), indicating a reduction in the mechanical forcing of the tropical ocean. This reduced daily variance also occurs with zonal wind stress to some extent, though not as convincingly as for the sea level pressure (not shown). There is also reduced variance in daily latent heat flux over the western equatorial Pacific and in the ENSO growth, though more pronounced in the boreal summer (Fig. 9b). This pattern of reduced variance is similarly exhibited in the net surface flux (not shown).
The analysis thus suggests that reduced stochastic forcing of the atmosphere during boreal spring and summer over the western tropical Pacific may be responsible for the reduced ENSO variance seen in the mid-Holocene simulations. In the other months, in particular January and February, stochastic forcing appears to have increased, complicating our story. However, it can be argued that these are the months when the ENSO system is least responsive to stochastic forcing (e.g., Penland and Sardeshmukh 1995).
b. Role of ENSO stability
Our hypothesis of the PMM significant forcing relies on the assumption that ENSO lies in the stable regime: indeed, our model ENSO appears to act that way. An unstable ENSO may act in a very different fashion to external forcing, as pointed out in Penland and Sardeshmukh (1995). We explore this question by altering the stability of our model ENSO system. We achieve this by changing the relationship between thermocline depth changes and the temperature of the upwelled water, through altering the coefficient that linearly links the two quantities. We reran both the control and 6K simulations with a coefficient 0.5, 1.5, and 2 times its usual value so that the model ENSO is more stable in the first case and less stable in the latter two. The results are shown in Table 3; the “take-home” result is that more stable configurations increased the percentage of ENSO variance reduction (as measured by the 2–7-yr bandpassed Niño-3) and a more unstable configuration resulted in significantly less variance reduction.
A plausible interpretation of our results is that, under an unstable ENSO regime, the model ENSO “cares less” about stochastic forcing and, so, is less susceptible to reduced PMM forcing during 6K. This interpretation is consistent with a finding by Seager et al. (2004), but coming from the other direction: when they applied PMM-like external forcing on the Cane–Zebiak ENSO model (which is self-sustaining), they found that it had little to no impact on the forecast skill of the model. Our experiments are not perfect however, as the standard model is tuned to give a reasonable rendition of ENSO and altering the coupling strength changes the character of the simulated ENSO—in particular, the simulated ENSO amplitude changes significantly with the coupling strength. This limits how much we can take away from this exercise. However, it does allow us to make the general point that the reduction in ENSO variance is likely to depend on the stability characteristics of the model ENSO system; furthermore, it suggests that a more stable ENSO regime is more able to give a larger ENSO variance reduction in the mid-Holocene.
We finally note that, unlike the ENSO variance, there is a pronounced reduction in the annual cycle amplitude in all cases (as indicated by the reduction in the full Niño-3 variance, see Table 3). This indicates that the reduction in the annual cycle of the tropical Pacific cold tongue is not notably affected by changes to the ENSO stability. This apparent dichotomy between changes in ENSO and changes to the annual cycle may be the hallmark of a linear stochastic model of ENSO. In a nonlinear framework, one might expect the Bjerknes feedback to affect both the ENSO and the annual cycle at the same time (e.g., Dijkstra and Neelin 1999).
7. Boreal wintertime climate change over the Pacific sector
What is it about the boreal wintertime climate of the mid-Holocene that reduces the PMM forcing? We do not have an answer for this, but for completeness we briefly summarize boreal wintertime climate changes over the Pacific sector that may be related to this phenomenon. There are pronounced boreal wintertime climate changes over the tropical and North Pacific sector during the mid-Holocene, including a deepened Aleutian low (Fig. 10a), reduced midlatitude westerlies, and an increased subtropical jet in the upper troposphere (Fig. 10d). The changes to the upper-level westerlies appear consistent, thermal-wind-wise, with the altered meridional SST gradient there: an intense region of colder SSTs over the western midlatitude Pacific strengthens the gradient in the subtropics and reduces them in the midlatitudes (Fig. 10b). There is also a horseshoe-like warmer SST over the eastern part of the basin. With tropical precipitation, there is a reduction in the northern branch of the Pacific ITCZ and the intensification of the south Pacific convergence zone (SPCZ) (Fig. 10c).
In analyzing our various idealized model simulations where we applied mid-Holocene orbital conditions but were restricted to limited spatial regions, all of the simulations that had significant reduction of ENSO variance also exhibited the North and tropical Pacific boreal wintertime climate changes described above. It suggests, though does not conclude, that the ENSO variance change is somehow tied to these features of Pacific boreal wintertime climate change.
8. Summary and discussion
We argue for reduction to the external forcing of El Niño–Southern Oscillation (ENSO), wrought by Pacific-wide climate changes to mid-Holocene conditions, as a viable explanation for the observed reduction in ENSO activity during that time. We come to this conclusion through comprehensive analysis of an intermediate model—a full atmospheric general circulation model coupled to a reduced-gravity ocean model—that achieves significant reduction to ENSO variance to mid-Holocene orbital conditions. Our argument is based on several lines of reasoning:
Mid-Holocene ENSO changes in this model are not driven by mean-state changes driven by insolation changes in the deep tropical Pacific, but rather from mid-Holocene insolation applied outside of the deep tropical Pacific.
This means that mid-Holocene climate changes outside the tropical Pacific are responsible for the ENSO variance change, and teleconnection effects have to be invoked to bring this influence into the tropical Pacific ENSO region.
Previously proposed hypotheses for mid-Holocene ENSO change from nonlocal influences, including tropical Pacific mean-state changes induced by a stronger Asian monsoon and increased annual cycle forcing bringing about ENSO reduction through frequency entrainment, do not appear to be operating.
A recent hypothesis postulates the Pacific meridional mode (PMM) as an external driver of ENSO (Chang et al. 2007). PMM activity in the mid-Holocene simulation is reduced at a level comparable to the ENSO variance reduction. The PMM changes are caused directly by the mid-Holocene insolation changes and are not a consequence of the ENSO variance change.
There is a reduction in the North Pacific midlatitude atmospheric activity as expressed by the North Pacific Oscillation (NPO); the NPO is thought to be a driver of the PMM. In general, there appears to be a reduction in atmospheric stochastic forcing over the western tropical Pacific during the ENSO growing seasons of boreal spring and summer.
The magnitude of mid-Holocene ENSO variance reduction appears to depend on how receptive the model ENSO system is to external forcing. Altering the stability of our model ENSO—by changing the relationship between thermocline depth changes and the temperature of the entrained subsurface water—led to changing sensitivity of the model ENSO such that a more unstable system lead to less reduction in the ENSO variance, and vice versa.
Related to our hypothesis, a strong suggestion coming out of our study is that it is very unlikely that the tropical Pacific ENSO system can be viewed to respond directly to mid-Holocene orbital forcing and isolated from climate changes teleconnected from outside the tropical Pacific, as the hypothesis of Clement et al. (2000) requires. One cannot obtain the correct tropical Pacific mean state or variability changes to the mid-Holocene with just the insolation forcing applied to the tropical Pacific. In retrospect, given that insolation, and climate, changes during the mid-Holocene are global and that the tropical Pacific climate is intimately connected to the global climate, it would be remarkable indeed if ENSO turns out not to be related to the global adjustments that climate makes to mid-Holocene orbital conditions.3
An outstanding question that our study does not answer is the link between North Pacific climate changes, midlatitude atmospheric variations, and the PMM. While we do not have a lead on how the PMM is linked to the others, the modeled mid-Holocene North Pacific wintertime climate changes (see Fig. 10) have striking resemblance to large-scale climate changes in today’s climate associated with the changes to “midwinter suppression” activity (Nakamura 1992; Nakamura et al. 2002). Midwinter suppression is an observed phenomenon where North Pacific midlatitude transient eddy activity is reduced in the middle of winter, despite large-scale conditions that supposedly favor it. This will be subject of a forthcoming work.
One may wonder about the value of a study that advances yet another mechanism of mid-Holocene change to ENSO variance and, indeed, why this mechanism should be taken more seriously than any of the others. Our credibility comes in part from our coupled model: On one hand, by incorporating a full atmosphere our model is physically more complete than the Cane–Zebiak class of ENSO model on which many mechanistic studies of mid-Holocene ENSO are based (e.g., Clement et al. 2000; Pan et al. 2005); in particular, it simulates the climate changes to mid-Holocene insolation rather than having it prescribed, as in the case of the Cane–Zebiak model. On the other hand, our model is certainly less sophisticated than the fully coupled models, in particular having drawbacks associated with using a 1.5-layer reduced-gravity ocean. It means that we are unable to simulate the subsurface ocean climate and all the associated phenomena that could result from it (e.g., Gu and Philander 1997); the ocean requires a flux correction to simulate the present climate correctly, which has associated dangers to it; we do not properly simulate the dynamical effects of the midhigh latitude ocean; and finally, we do not simulate changes to sea ice formation. Our model does have some distinct advantages, however: the main ones being that it is capable of simulating the ENSO reduction at levels comparable to that observed, unlike many current-generation fully coupled models, and that it is able to simulate the annual cycle and interannual variability, including the PMM, with fidelity. Taken together, the correspondence of the model simulation with the observed climate lends credibility to our model study.
Why there are still several hypotheses for mid-Holocene ENSO, and why coupled model ENSO responses are not consistent across the various models. While we do not have a definitive answer, our experience suggests that the level of idealization of the climate model physics may play a significant role: in the absence of climate influences external to the tropical Pacific, the Cane–Zebiak model is sensitive to mean-state changes, but when the tropical Pacific dynamics becomes incorporated in the larger global climate physics, the driving of ENSO by external forcing swamps the influence of the mean-state changes. By the same reasoning, our own results may be vulnerable to the CCM3–RGO’s lack of deep ocean physics.
The larger implication of this study, however, is demonstrating a viable hypothesis for mid-Holocene ENSO reduction that follows the “linear, externally forced” paradigm of ENSO. We are not alone in this: a recent study by Roberts (2007) also does the same, though his proposed mechanism is quite different. He demonstrated though detailed analysis of a linearly stable version of the Cane–Zebiak model that the early Holocene cooling of the tropical Pacific mean-state SST can reduce the coupling strength between the atmosphere and ocean and, hence, explain the reduction to the ENSO activity. In particular, he did not need to invoke reduced stochastic forcing in his model to get the weaker ENSO activity. Our choice of ENSO model, therefore, may have predetermined the results that we obtained. This conclusion is philosophically consistent with a recent model study of mid-Holocene ENSO by Brown et al. (2008), who concluded that model ENSO behavior is more sensitive to internal model physical parameterizations than to climate forcings.
So, with what are we left? When we started this work, the hope was that the observed reduction in mid-Holocene ENSO activity, arguably the most robust ENSO constraint coming out of paleoclimate studies, would be able to constrain in some way the possibilities of the ENSO physics. However, we now think that there are too many viable ways to fulfill the observed mid-Holocene constraint. It follows that robust ENSO observations at other time periods when different climate forcings acted—in particular the Last Glacial Maximum—are necessary in order for paleoclimate observations to usefully contribute to our understanding of ENSO dynamics.
Acknowledgments
We thank Mai Nguyen for her work on the ENSO variance calculations with the PMIP2 simulations, and David Battisti for useful conversations. Comments by Julien Emile-Geay and an anonymous reviewer greatly improved the presentation of the paper. We acknowledge support from NSF ATM-0438201 and the Comer Science and Education Foundation (to JCHC and YF), and NSF ATM-99007625 and NOAA Grant NA16GP1572 (to PC). The PMIP2 output was provided by several international modeling groups, and collected and archived by the Laboratoire des Sciences du Climat et de l’Environnement. The PMIP2/MOTIF Data Archive is supported by CEA, CNRS, the EU project MOTIF (EVK2-CT-2002-00153) and the Programme National d’Etude de la Dynamique du Climat (PNEDC). More information is available online (http://pmip2.lsce.ipsl.fr/ and http://motif.lsce.ipsl.fr/).
REFERENCES
Braconnot, P., and Coauthors, 2007: Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial Maximum—Part 1: Experiments and large-scale features. Climate Past, 3 , 261–277.
Bretherton, C. S., C. Smith, and J. M. Wallace, 1992: An intercomparison of methods for finding coupled patterns in climate data. J. Climate, 5 , 541–560.
Brown, J., M. Collins, and A. Tudhope, 2006: Coupled model simulations of mid-Holocene ENSO and comparisons with coral proxy records. Adv. Geosci., 6 , 29–33.
Brown, J., M. Collins, A. Tudhope, and T. Toniazzo, 2008: Modelling mid-Holocene tropical climate and ENSO variability: Towards constraining predictions of future change with palaeo-data. Climate Dyn., 30 , 19–36.
Cane, M. A., A. C. Clement, M. K. Gagan, L. K. Ayliffe, and A. Tudhope, 2000: ENSO through the Holocene, depicted in corals and a model simulation. CLIVAR Exchanges, Vol. 5, No. 1, International CLIVAR Project Office, Southampton, United Kingdom, 3–7.
Chang, P., 1994: A study of the seasonal cycle of sea-surface temperature in the tropical Pacific Ocean using reduced gravity models. J. Geophys. Res., 99 , 7725–7741.
Chang, P., B. Wang, T. Li, and L. Ji, 1994: Interactions between the seasonal cycle and the Southern Oscillation—Frequency entrainment and chaos in a coupled ocean-atmosphere model. Geophys. Res. Lett., 21 , 2817–2820.
Chang, P., L. Ji, H. Li, and M. Flugel, 1996: Chaotic dynamics versus stochastic processes in El Niño Southern Oscillation in coupled ocean-atmosphere models. Physica D, 98 , 301–320.
Chang, P., Y. Fang, R. Saravanan, L. Ji, and H. Seidel, 2006: The cause of the fragile relationship between the Pacific El Niño and the Atlantic Niño. Nature, 443 , 324–328.
Chang, P., L. Zhang, R. Saravanan, D. J. Vimont, J. C. H. Chiang, L. Ji, H. Seidel, and M. K. Tippett, 2007: Pacific meridional mode and El Niño-Southern oscillation. Geophys. Res. Lett., 34 , L16608. doi:10.1029/2007GL030302.
Chiang, J. C. H., and D. J. Vimont, 2004: Analogous Pacific and Atlantic meridional modes of tropical atmosphere–ocean variability. J. Climate, 17 , 4143–4158.
Clement, A. C., R. Seager, and M. A. Cane, 2000: Suppression of El Niño during the mid-Holocene by changes in the Earth’s orbit. Paleoceanography, 15 , 731–737.
Clement, A. C., A. Hall, and A. J. Broccoli, 2004: The importance of precessional signals in the tropical climate. Climate Dyn., 22 , 327–341.
Dijkstra, H. A., and J. D. Neelin, 1999: Coupled processes and the tropical climatology. Part III: Instabilities of the fully coupled climatology. J. Climate, 12 , 1630–1643.
Fang, Y., 2005: A coupled model study of the remote influence of ENSO on tropical Atlantic SST variability. Ph.D. thesis, Texas A&M University, 93 pp.
Gu, D. F., and S. G. H. Philander, 1997: Interdecadal climate fluctuations that depend on exchanges between the tropics and extratropics. Science, 275 , 805–807.
Gu, D. F., S. G. H. Philander, and M. J. McPhaden, 1997: The seasonal cycle and its modulation in the eastern tropical Pacific Ocean. J. Phys. Oceanogr., 27 , 2209–2218.
Kiehl, J. T., J. J. Hack, G. B. Bonan, B. A. Boville, D. L. Williamson, and P. J. Rasch, 1998: The National Center for Atmospheric Research Community Climate Model: CCM3. J. Climate, 11 , 1131–1149.
Koutavas, A., P. B. Demenocal, G. C. Olive, and J. Lynch-Stieglitz, 2006: Mid-Holocene El Niño-Southern Oscillation (ENSO) attenuation revealed by individual foraminifera in eastern tropical Pacific sediments. Geology, 34 , 993–996.
Levitus, S., 1982: Climatological Atlas of the World Ocean. NOAA Prof. Paper 13, 173 pp. and 17 microfiche.
Liu, Z. G., 2002: A simple model study of ENSO suppression by external periodic forcing. J. Climate, 15 , 1088–1098.
Liu, Z. Y., J. Kutzbach, and L. X. Wu, 2000: Modeling climate shift of El Niño variability in the Holocene. Geophys. Res. Lett., 27 , 2265–2268.
Nakamura, H., 1992: Midwinter suppression of baroclinic wave activity in the Pacific. J. Atmos. Sci., 49 , 1629–1642.
Nakamura, H., T. Izumi, and T. Sampe, 2002: Interannual and decadal modulations recently observed in the Pacific storm track activity and East Asian winter monsoon. J. Climate, 15 , 1855–1874.
Pan, A. J., Q. Y. Liu, and Z. Y. Liu, 2005: Periodic forcing and ENSO suppression in the Cane-Zebiak model. J. Oceanogr., 61 , 109–113.
Penland, C., and P. D. Sardeshmukh, 1995: The optimal growth of tropical sea surface temperature anomalies. J. Climate, 8 , 1999–2024.
Roberts, W. H., 2007: An investigation into the causes for the reduction in the variability of the El Niño-Southern Oscillation in the early Holocene in a global climate model. Ph.D. thesis, University of Washington, 163 pp.
Rodbell, D. T., G. O. Seltzer, D. M. Anderson, M. B. Abbott, D. B. Enfield, and J. H. Newman, 1999: An ∼15,000-year record of El Niño-driven alluviation in southwestern Ecuador. Science, 283 , 516–520.
Rogers, J. C., 1981: The North Pacific Oscillation. J. Climatol., 1 , 39–57.
Sandweiss, D. H., and Coauthors, 1996: Geoarchaeological evidence from Peru for a 5000 years BP onset of El Niño. Science, 273 , 1531–1533.
Seager, R., A. R. Karspeck, M. A. Cane, Y. Kushnir, A. Giannini, A. Kaplan, B. Kerman, and J. Velez, 2004: Predicting Pacific decadal variability. Earth’s Climate: The Ocean-Atmosphere Interaction, C. Wang et al., Eds., Amer. Geophys. Union, 105–120.
Thompson, C. J., and D. S. Battisti, 2000: A linear stochastic dynamical model of ENSO. Part I: Model development. J. Climate, 13 , 2818–2832.
Thompson, C. J., and D. S. Battisti, 2001: A linear stochastic dynamical model of ENSO. Part II: Analysis. J. Climate, 14 , 445–466.
Tudhope, A. W., and Coauthors, 2001: Variability in the El Niño-Southern oscillation through a glacial-interglacial cycle. Science, 291 , 1511–1517.
Vimont, D. J., D. S. Battisti, and A. C. Hirst, 2001: Footprinting: A seasonal connection between the tropics and mid-latitudes. Geophys. Res. Lett., 28 , 3923–3926.
Vimont, D. J., D. S. Battisti, and A. C. Hirst, 2002: Pacific interannual and interdecadal equatorial variability in a 1000-yr simulation of the CSIRO coupled general circulation model. J. Climate, 15 , 160–178.
Vimont, D. J., D. S. Battisti, and A. C. Hirst, 2003a: The seasonal footprinting mechanism in the CSIRO general circulation models. J. Climate, 16 , 2653–2667.
Vimont, D. J., J. M. Wallace, and D. S. Battisti, 2003b: The seasonal footprinting mechanism in the Pacific: Implications for ENSO. J. Climate, 16 , 2668–2675.
Zebiak, S. E., and M. A. Cane, 1987: A model El Niño-Southern Oscillation. Mon. Wea. Rev., 115 , 2262–2278.
Zheng, W., P. Braconnot, E. Guilyardi, U. Merkel, and Y. Yu, 2008: ENSO at 6ka and 21ka from ocean–atmosphere coupled model simulations. Climate Dyn., 30 , 745–762.
Top-of-atmosphere changes to insolation during the mid-Holocene (6000 yr BP), as applied to our model simulations. The contour interval is 5 W m−2, and positive values (unshaded) are directed toward the earth.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Top-of-atmosphere changes to insolation during the mid-Holocene (6000 yr BP), as applied to our model simulations. The contour interval is 5 W m−2, and positive values (unshaded) are directed toward the earth.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Top-of-atmosphere changes to insolation during the mid-Holocene (6000 yr BP), as applied to our model simulations. The contour interval is 5 W m−2, and positive values (unshaded) are directed toward the earth.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
The 18O of individual G. ruber (open circles) from core V21–30. Black triangles are down-core 18O data from the same site measured on bulk G. ruber samples. Solid circles with error bars indicate mean and standard deviation of pooled individuals from late (n = 93) and mid-Holocene (n = 96). Standard deviation of mid-Holocene pool (σ = 0.34) is 30% less than late Holocene (σ = 0.48), and total variance (σ2) is 50% less, significant at 99% confidence. Gray shading marks predicted 18O ranges due to seasonal and ENSO variations. [From Kartaras et al. (2006).] See Koutavas et al. (2006) for details.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
The 18O of individual G. ruber (open circles) from core V21–30. Black triangles are down-core 18O data from the same site measured on bulk G. ruber samples. Solid circles with error bars indicate mean and standard deviation of pooled individuals from late (n = 93) and mid-Holocene (n = 96). Standard deviation of mid-Holocene pool (σ = 0.34) is 30% less than late Holocene (σ = 0.48), and total variance (σ2) is 50% less, significant at 99% confidence. Gray shading marks predicted 18O ranges due to seasonal and ENSO variations. [From Kartaras et al. (2006).] See Koutavas et al. (2006) for details.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
The 18O of individual G. ruber (open circles) from core V21–30. Black triangles are down-core 18O data from the same site measured on bulk G. ruber samples. Solid circles with error bars indicate mean and standard deviation of pooled individuals from late (n = 93) and mid-Holocene (n = 96). Standard deviation of mid-Holocene pool (σ = 0.34) is 30% less than late Holocene (σ = 0.48), and total variance (σ2) is 50% less, significant at 99% confidence. Gray shading marks predicted 18O ranges due to seasonal and ENSO variations. [From Kartaras et al. (2006).] See Koutavas et al. (2006) for details.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
(left) Model seasonal cycle of Pacific SST averaged from 5°S to 5°N (the annual mean value is subtracted). (right) Difference between the 6K and control-simulated seasonal cycle of Pacific SST. The contour interval is 0.3 K, and negative values are shaded. Note that in the eastern equatorial Pacific, the 6K changes are out of phase with the mean seasonal cycle, implying a reduction to the cold tongue seasonality.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
(left) Model seasonal cycle of Pacific SST averaged from 5°S to 5°N (the annual mean value is subtracted). (right) Difference between the 6K and control-simulated seasonal cycle of Pacific SST. The contour interval is 0.3 K, and negative values are shaded. Note that in the eastern equatorial Pacific, the 6K changes are out of phase with the mean seasonal cycle, implying a reduction to the cold tongue seasonality.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
(left) Model seasonal cycle of Pacific SST averaged from 5°S to 5°N (the annual mean value is subtracted). (right) Difference between the 6K and control-simulated seasonal cycle of Pacific SST. The contour interval is 0.3 K, and negative values are shaded. Note that in the eastern equatorial Pacific, the 6K changes are out of phase with the mean seasonal cycle, implying a reduction to the cold tongue seasonality.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Modes 1–3 of a MCA analysis of SST and surface winds over the central-eastern tropical Pacific (32°N–21°S, 175°E–95°W). (right) Control simulation and (left) the observed detrended SST. The calculation of the observed fields is similar to the one done in Chiang and Vimont (2004), except that the cold tongue index (a measure of ENSO activity) is not removed from the dataset prior to analysis. The contour interval is 0.1 K (the zero contour is not plotted) and negative anomalies are shaded. Wind vectors with speed less than 0.2 m s−1 are not plotted. The box in the left-hand center row, corresponding to the North Pacific north of 5°N and west of 150°E, delimits the pattern used to derive a projection of the Pacific meridional mode in the simulations (see section 5).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Modes 1–3 of a MCA analysis of SST and surface winds over the central-eastern tropical Pacific (32°N–21°S, 175°E–95°W). (right) Control simulation and (left) the observed detrended SST. The calculation of the observed fields is similar to the one done in Chiang and Vimont (2004), except that the cold tongue index (a measure of ENSO activity) is not removed from the dataset prior to analysis. The contour interval is 0.1 K (the zero contour is not plotted) and negative anomalies are shaded. Wind vectors with speed less than 0.2 m s−1 are not plotted. The box in the left-hand center row, corresponding to the North Pacific north of 5°N and west of 150°E, delimits the pattern used to derive a projection of the Pacific meridional mode in the simulations (see section 5).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Modes 1–3 of a MCA analysis of SST and surface winds over the central-eastern tropical Pacific (32°N–21°S, 175°E–95°W). (right) Control simulation and (left) the observed detrended SST. The calculation of the observed fields is similar to the one done in Chiang and Vimont (2004), except that the cold tongue index (a measure of ENSO activity) is not removed from the dataset prior to analysis. The contour interval is 0.1 K (the zero contour is not plotted) and negative anomalies are shaded. Wind vectors with speed less than 0.2 m s−1 are not plotted. The box in the left-hand center row, corresponding to the North Pacific north of 5°N and west of 150°E, delimits the pattern used to derive a projection of the Pacific meridional mode in the simulations (see section 5).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
(a), (b) Histogram of the distribution of monthly Niño-3 (SST averaged over 5°S–5°N, 150°–90°W) for the (a) control and (b) mid-Holocene simulation. The bin size is 0.2 K, and y values are reported as a fraction of the total (=1). The control histogram was constructed from a 300-yr simulation, whereas the mid-Holocene histogram was from a 249-yr simulation. (c) Random sampling of the monthly Niño-3 of the control and mid-Holocene simulations, plotted in the format of Fig. 2. There are 98 points chosen at random for the control Niño-3, and 96 for the mid-Holocene. The solid line with the circle in the middle shows ±1 std dev about the mean, but computed using all Niño-3 points of the respective simulations (and not the sample).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
(a), (b) Histogram of the distribution of monthly Niño-3 (SST averaged over 5°S–5°N, 150°–90°W) for the (a) control and (b) mid-Holocene simulation. The bin size is 0.2 K, and y values are reported as a fraction of the total (=1). The control histogram was constructed from a 300-yr simulation, whereas the mid-Holocene histogram was from a 249-yr simulation. (c) Random sampling of the monthly Niño-3 of the control and mid-Holocene simulations, plotted in the format of Fig. 2. There are 98 points chosen at random for the control Niño-3, and 96 for the mid-Holocene. The solid line with the circle in the middle shows ±1 std dev about the mean, but computed using all Niño-3 points of the respective simulations (and not the sample).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
(a), (b) Histogram of the distribution of monthly Niño-3 (SST averaged over 5°S–5°N, 150°–90°W) for the (a) control and (b) mid-Holocene simulation. The bin size is 0.2 K, and y values are reported as a fraction of the total (=1). The control histogram was constructed from a 300-yr simulation, whereas the mid-Holocene histogram was from a 249-yr simulation. (c) Random sampling of the monthly Niño-3 of the control and mid-Holocene simulations, plotted in the format of Fig. 2. There are 98 points chosen at random for the control Niño-3, and 96 for the mid-Holocene. The solid line with the circle in the middle shows ±1 std dev about the mean, but computed using all Niño-3 points of the respective simulations (and not the sample).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Changes to the seasonal cycle in the equatorial Pacific over months (y axis) and longitude (x axis): (left) 992-mb zonal wind difference between the standard 6K simulation (from top to second from bottom) from the control, for the simulation with 6K insolation applied in the tropical Pacific only, and for the simulation with 6K insolation applied outside the tropical Pacific; (middle) for 992-mb meridional winds; and (right) the subsurface temperature anomaly in the reduced gravity ocean model. (bottom) The climatological annual cycle changes (i.e., subtracting out the annual mean from the monthly means) for the respective fields. In all cases, the fields are averaged between 5°S and 5°N. Shaded values are negative, and contour intervals are indicated on the figure title. The mid-Holocene changes in the equatorial Pacific are dominated by teleconnections from outside the tropical Pacific.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Changes to the seasonal cycle in the equatorial Pacific over months (y axis) and longitude (x axis): (left) 992-mb zonal wind difference between the standard 6K simulation (from top to second from bottom) from the control, for the simulation with 6K insolation applied in the tropical Pacific only, and for the simulation with 6K insolation applied outside the tropical Pacific; (middle) for 992-mb meridional winds; and (right) the subsurface temperature anomaly in the reduced gravity ocean model. (bottom) The climatological annual cycle changes (i.e., subtracting out the annual mean from the monthly means) for the respective fields. In all cases, the fields are averaged between 5°S and 5°N. Shaded values are negative, and contour intervals are indicated on the figure title. The mid-Holocene changes in the equatorial Pacific are dominated by teleconnections from outside the tropical Pacific.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Changes to the seasonal cycle in the equatorial Pacific over months (y axis) and longitude (x axis): (left) 992-mb zonal wind difference between the standard 6K simulation (from top to second from bottom) from the control, for the simulation with 6K insolation applied in the tropical Pacific only, and for the simulation with 6K insolation applied outside the tropical Pacific; (middle) for 992-mb meridional winds; and (right) the subsurface temperature anomaly in the reduced gravity ocean model. (bottom) The climatological annual cycle changes (i.e., subtracting out the annual mean from the monthly means) for the respective fields. In all cases, the fields are averaged between 5°S and 5°N. Shaded values are negative, and contour intervals are indicated on the figure title. The mid-Holocene changes in the equatorial Pacific are dominated by teleconnections from outside the tropical Pacific.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Power spectrum of Niño-3 for the 300-yr standard control (solid line) and 249-yr 6K (dashed line) simulations, using the Thompson multitaper method with seven discrete prolate spheroidal sequence tapers. The gray shading is the chi-squared 95% confidence limits for the control Niño-3 spectrum, and the ENSO 2–7-yr band is indicated by the dashed box. The spectrum was calculated using the “pmtmPH.m” Matlab subroutine (developed by P. Huybers, available online at http://www.people.fas.harvard.edu/~phuybers/Mfiles/index.html).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Power spectrum of Niño-3 for the 300-yr standard control (solid line) and 249-yr 6K (dashed line) simulations, using the Thompson multitaper method with seven discrete prolate spheroidal sequence tapers. The gray shading is the chi-squared 95% confidence limits for the control Niño-3 spectrum, and the ENSO 2–7-yr band is indicated by the dashed box. The spectrum was calculated using the “pmtmPH.m” Matlab subroutine (developed by P. Huybers, available online at http://www.people.fas.harvard.edu/~phuybers/Mfiles/index.html).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Power spectrum of Niño-3 for the 300-yr standard control (solid line) and 249-yr 6K (dashed line) simulations, using the Thompson multitaper method with seven discrete prolate spheroidal sequence tapers. The gray shading is the chi-squared 95% confidence limits for the control Niño-3 spectrum, and the ENSO 2–7-yr band is indicated by the dashed box. The spectrum was calculated using the “pmtmPH.m” Matlab subroutine (developed by P. Huybers, available online at http://www.people.fas.harvard.edu/~phuybers/Mfiles/index.html).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Correlation of the wintertime [December–February (DJF)] Niño-3 index with SST anomalies in the prior March–May in the 300-yr control simulation. The contour interval is 0.1, and negative values are shaded; the zero contour is not plotted. Strongest correlations are in the tropical northeastern Pacific (r > 0.4).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Correlation of the wintertime [December–February (DJF)] Niño-3 index with SST anomalies in the prior March–May in the 300-yr control simulation. The contour interval is 0.1, and negative values are shaded; the zero contour is not plotted. Strongest correlations are in the tropical northeastern Pacific (r > 0.4).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Correlation of the wintertime [December–February (DJF)] Niño-3 index with SST anomalies in the prior March–May in the 300-yr control simulation. The contour interval is 0.1, and negative values are shaded; the zero contour is not plotted. Strongest correlations are in the tropical northeastern Pacific (r > 0.4).
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Change in atmospheric stochastic forcing of the tropical Pacific. Change to the monthly mean variance of daily (a) sea level pressure and (b) latent heat flux between the mid-Holocene and control no-ENSO simulations: (left) values averaged over 15°S–15°N and negative values shaded; (right) values averaged over the western Pacific 15°S–15°N, 120°–190°E. Solid lines are the control simulation values and dashed are for the 6K simulation derived from 20 years of daily output.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Change in atmospheric stochastic forcing of the tropical Pacific. Change to the monthly mean variance of daily (a) sea level pressure and (b) latent heat flux between the mid-Holocene and control no-ENSO simulations: (left) values averaged over 15°S–15°N and negative values shaded; (right) values averaged over the western Pacific 15°S–15°N, 120°–190°E. Solid lines are the control simulation values and dashed are for the 6K simulation derived from 20 years of daily output.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Change in atmospheric stochastic forcing of the tropical Pacific. Change to the monthly mean variance of daily (a) sea level pressure and (b) latent heat flux between the mid-Holocene and control no-ENSO simulations: (left) values averaged over 15°S–15°N and negative values shaded; (right) values averaged over the western Pacific 15°S–15°N, 120°–190°E. Solid lines are the control simulation values and dashed are for the 6K simulation derived from 20 years of daily output.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Boreal wintertime (DJF) climate anomalies, 6K standard simulation minus control: (a) sea level pressure (contour interval 1 mb); (b) surface temperature (contour interval 2 K), here the mean anomaly areally averaged from 60°S to 60°N is removed from the anomaly field prior to plotting in order to emphasize the spatial variation of the temperature anomaly; (c) precipitation (contour interval 1 mm day−1); and (d) 250-mb zonal winds (contour interval 2 m s−1). In all panels, negative values are shaded and the zero contour is not plotted. The anomaly fields are similar if the no-ENSO simulations are used instead, indicating that these changes are independent of mid-Holocene ENSO changes.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Boreal wintertime (DJF) climate anomalies, 6K standard simulation minus control: (a) sea level pressure (contour interval 1 mb); (b) surface temperature (contour interval 2 K), here the mean anomaly areally averaged from 60°S to 60°N is removed from the anomaly field prior to plotting in order to emphasize the spatial variation of the temperature anomaly; (c) precipitation (contour interval 1 mm day−1); and (d) 250-mb zonal winds (contour interval 2 m s−1). In all panels, negative values are shaded and the zero contour is not plotted. The anomaly fields are similar if the no-ENSO simulations are used instead, indicating that these changes are independent of mid-Holocene ENSO changes.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Boreal wintertime (DJF) climate anomalies, 6K standard simulation minus control: (a) sea level pressure (contour interval 1 mb); (b) surface temperature (contour interval 2 K), here the mean anomaly areally averaged from 60°S to 60°N is removed from the anomaly field prior to plotting in order to emphasize the spatial variation of the temperature anomaly; (c) precipitation (contour interval 1 mm day−1); and (d) 250-mb zonal winds (contour interval 2 m s−1). In all panels, negative values are shaded and the zero contour is not plotted. The anomaly fields are similar if the no-ENSO simulations are used instead, indicating that these changes are independent of mid-Holocene ENSO changes.
Citation: Journal of Climate 22, 4; 10.1175/2008JCLI2644.1
Percentage change in the variance of the full Niño-3, the 2–7-yr bandpassed Niño-3, and the variance of the March–May (MAM) PMM index between the preindustrial control run and mid-Holocene ENSO simulations as archived by the PMIP2. The bandpass was performed using a Butterworth filter. Boldface values show an increase to the mid-Holocene variance. For the PMM variance, we first removed the possible influence of ENSO activity by linearly regressing out the Niño-3 index averaged over the previous DJF, from the model MAM SST anomalies, prior to extracting the PMM index. The seven models are the Community Climate System Model (CCSM), ECHAM5, Fast Ocean Atmosphere Model (FOAM), Goddard Institute for Space Studies (GISS), Meteorological Research Institute Coupled General Circulation Model (MIROC), Model for Interdisciplinary Research on Climate (MRI), and the Hadley Centre coupled model, version 3 (HadCM3).
Percent change in the variance of the full Niño-3 (second column) and the 2–7-yr bandpassed Niño-3 (third column) for various runs of the CCM3–RGO as described in the first column. The bandpass was performed using a Butterworth filter. Boldface values show increase to the variance.
Changes to Niño-3 variance in 6K simulations, but using configurations of the model where the slope of the linear relationship between the thermocline depth anomaly and temperature of entrained water was altered. A coupling strength of 1 is the standard model, and a larger coupling strength means a larger variation in the temperature of entrained water for a given thermocline depth anomaly. In computing the change to the Niño-3 variance for each coupling strength, we first simulated the appropriate control; for example, for a coupling strength of 2, we first executed a 50-yr control simulation, and then a mid-Holocene simulation, and the change to the ENSO variance is then computed from the two simulations.
Our PMIP2 results are somewhat inconsistent with Zheng et al. (2008), who also calculated the change to ENSO amplitude in several PMIP2 simulations. We apply a 2–7-yr bandpass to the Niño-3 index, which Zheng et al. does not do; this may be part of the reason for the difference. In any case, the PMIP2 ENSO variance calculation is a minor point for our purposes, and does not affect our main argument.
In other coupled models, the too-far westward extension of the ENSO SST anomaly is linked with a similar extension of the mean cold tongue. In our model, this too-far extension of the ENSO SST anomaly occurs despite the fact that the mean cold tongue is “simulated” correctly.
In a later paper, Clement et al. (2004) also arrives at a similar conclusion in exploring the global tropical response to precessional forcing.
