1. Introduction

El Niño–Southern Oscillation (ENSO) events are characterized by prominent changes to the tropical climate system. Under ENSO warm phase (or El Niño) conditions, widespread warming of both the troposphere and surface is evident across the Pacific as well as throughout the rest of the tropical latitudes (hereafter the “remote Tropics”) and is accompanied by a spatially complex distribution of precipitation anomalies. In the Pacific positive precipitation anomalies occur in the vicinity of the equator from the date line eastward to the coastline of South America and are attributed to the presence of positive sea surface temperature anomalies (SSTAs) in the central and eastern Pacific (Quinn et al. 1978; Ropelewski and Halpert 1987). Throughout the remote Tropics, on the other hand, widespread negative precipitation anomalies are characteristic during El Niño. Both observational and modeling studies have noted the occurrence of regional precipitation deficits during El Niño in the Amazon basin (Zeng and Neelin 1999), northeastern Brazil (Hastenrath and Heller 1977), the tropical Atlantic (Giannini et al. 2001), the Sahel region of West Africa (Janicot et al. 2001), and the South Asian monsoon region (Webster and Yang 1992; Torrence and Webster 1999).

Although the features of the ENSO tropical teleconnection have been widely documented, the mechanisms underlying the remote ENSO teleconnection, especially the precipitation response, are not completely understood. The leading paradigm for El Niño–forced precipitation changes over the tropical belt is the “anomalous Walker circulation,” the zonal circulation pattern that results from zonal displacements in convection over the tropical Pacific (Kidson 1975; Saravanan and Chang 2000). The conceptual framework of the anomalous Walker circulation is often attributed to Gill (1980), who argued for a simplification of the vertical structure of the convectively driven circulation in the Tropics. Assuming a single baroclinic mode, Gill reduced the tropical dynamics to a system of linearized, damped shallow-water equations on an equatorial beta plane. Gill derived analytical solutions for the circulations driven by anomalous imposed diabatic forcing, including the structure of forced Kelvin and Rossby waves and the associated subsidence field away from the forcing region. In the context of the ENSO tropical teleconnection, the reduced precipitation throughout the remote Tropics during El Niño is attributed to the increase of subsidence forced by anomalous uplift over the warm Pacific SSTA (Wu and Newell 1998; Klein et al. 1999; Su et al. 2001; Su and Neelin 2002).

While conceptually powerful, the Gill model has significant shortcomings when applied to the tropical ENSO teleconnection. Most significant is its neglect of moist thermodynamics, radiation, and turbulent fluxes that ultimately determine the climate (including precipitation) response to El Niño. As the Gill framework is essentially a dynamical construct, it only predicts precipitation changes to the extent that these may be inferred from changes to subsidence. A characteristic feature of the Gill solution to an idealized equatorially symmetric heating source is the relatively spatially smooth nature of the subsidence field away from the forcing region (cf. Fig. 1d of Gill 1980), implying that subsidence is relatively uniformly distributed over the remote Tropics. However, a cursory examination of the anomalous tropical pressure velocity field during either El Niño or La Niña conditions (e.g., see Su and Neelin 2002) shows that the anomalous subsidence field exhibits a nontrivial spatial structure (Fig. 1). During El Niño events (Fig. 1b), negative (ascent) anomalies prevail over the central and eastern Pacific. Outside of this region, there is a tendency for positive (descent) anomalies, but the structure is rather heterogeneous. The distribution of anomalous subsidence appears to be linked to the mean convective conditions of the remote tropical region and, in fact, points to the important role of moist thermodynamic feedbacks. Indeed, recent studies (e.g., Su et al. 2001; Chiang and Sobel 2002, hereafter CS02; Su and Neelin 2002; Neelin et al. 2003; Su et al. 2004) have recognized the significant role of moist thermodynamics and have begun to approach the tropical ENSO teleconnection with a moist thermodynamic emphasis.

The motivation and approach for our current study derives from CS02, who sought to highlight the role of moist thermodynamics by simplifying the horizontal dynamics of the teleconnection problem. CS02 assumed that the effect of horizontal tropical dynamics is to spread the tropospheric temperature warming caused by El Niño uniformly across the tropical latitudes, invoking the “weak temperature gradient” (WTG) approximation (e.g., see Sobel and Bretherton 2000). Since the Coriolis effect is small near the equator, the tropical troposphere is unable to maintain significant horizontal temperature gradients. CS02 proceeded by imposing tropospheric temperature anomalies within a single-column model (SCM) equipped with parameterizations of moist convection, radiation, and boundary layer physics and coupled to a passive thermal ocean. With the tropospheric temperature profile imposed, the climate response (including subsidence) of the remote Tropics is determined simply by thermodynamic energy balance constraints. Among the notable results of CS02 were a relatively simple explanation for the remote tropical surface warming during El Niño (further explored in Chiang and Lintner 2005) and a proposed mechanism for remote tropical precipitation anomalies based on the disequilibrium between the free tropospheric and boundary layer moist static energies.

We extend the work of CS02 by investigating the ENSO response of the entire tropical atmosphere using a set of SCMs connected through the WTG requirement. Sobel and Bretherton (2000) showed that an intermediate level complexity model [the Quasi-equilibrium Tropical Circulation Model (QTCM); see section 2] modified in this manner was able to reproduce a reasonable tropical climatology; we use the same model and a similar procedure to model perturbations to tropical climate during El Niño. Since the WTG approximation by itself constrains only gradients in tropical tropospheric temperature, we determine the absolute level of tropospheric temperature through a mass balance requirement that the net divergence anomaly summed over the tropical belt is zero; similar approaches to “close” the problem have previously been used in more idealized models employing WTG, for example, Shaevitz and Sobel (2004). The mass balance constraint follows from the assumption that Pacific region ascent (descent) anomalies associated with El Niño (La Niña) are mostly compensated for by descent (ascent) elsewhere in the Tropics, an aspect of the anomalous Walker circulation that can be justified on the basis of the tropical waveguide properties of equatorial atmospheric dynamics (Gill 1980). An examination of the observed divergence anomalies associated with ENSO supports this view (see section 3). In what follows, we will refer to the constraints on the anomalous tropical tropospheric temperature and divergence fields collectively as the “WTG framework.”

We will demonstrate that the gross features of the teleconnection—for example, the tropospheric temperature and precipitation response averaged over the remote tropical region—appear to be well simulated within the WTG framework, as are the spatial details of the precipitation response in the tropical Pacific. By contrast, the spatial details of the remote tropical precipitation response are less adequately simulated; we suggest that this disagreement may result from the neglect of features such as horizontal temperature gradients. The organization of the remainder of this paper is as follows. After introducing the models and methodology and providing some justification for the WTG approach in sections 2 and 3, we briefly document the principal features of the mean precipitation field simulated by the standard and WTG versions of the QTCM (section 4). An in-depth discussion of the teleconnected responses to El Niño (and La Niña) forcing conditions in both model versions follows in section 5. We then present results of perturbative simulations with increased atmospheric greenhouse gas loading (section 6); the purpose of these simulations is to demonstrate further the applicability of the WTG approach as well as to highlight the distinct influences of temperature and large-scale subsidence changes during El Niño. In section 7, we briefly explore the impact of anomalous temperature gradients on the characteristics of the response to El Niño conditions. We conclude with a summary of our principal findings and a discussion of the implications of our study for the understanding of ENSO tropical teleconnectivity (section 8).

2. Models and methodology

a. The Quasi-equilibrium Tropical Circulation Model

The QTCM is an intermediate level complexity model of the tropical atmosphere developed by the Climate Systems Interactions group at the University of California, Los Angeles (Neelin and Zeng 2000; Zeng et al. 2000). The QTCM incorporates all relevant dynamics and physical parameterizations that are necessary to determine the tropical climate state, that is, atmospheric dynamics, convection, clouds, radiation, and turbulent fluxes. Our use of QTCM is motivated by the ease with which it can be modified to employ the WTG framework (as we show below); moreover, the QTCM contains all of the important physics (including moist thermodynamics) that we assume to be of importance for the teleconnection problem. The QTCM has been used previously to investigate many aspects of tropical climate, including monsoons (Chou et al. 2001) and tropical ENSO teleconnections (Su et al. 2001; Su and Neelin 2002).

Employing constraints obtained from quasi-equilibrium convective theory (Brown and Bretherton 1997; Neelin and Zeng 2000), the QTCM represents regions of deep convection with a limited number of vertical degrees of freedom: the model version used here has a single vertical structure function for temperature and moisture and two for velocity. In particular, the temperature and water vapor equations are represented as

View Expanded

View Expanded

where T and q are temperature and humidity, respectively; ω is the divergence (or vertical velocity); and M_s and M_q are the dry static energy and moisture stratification, respectively (see Neelin and Zeng 2000 for further details). The terms Q_T and Q_q are, respectively, forcing functions for the temperature and humidity fields

View Expanded

View Expanded

with adv_T (adv_q) and dfs_T (dfs_q) denoting horizontal advection and diffusion of temperature (humidity), R the total radiation absorbed by the atmospheric column, H the sensible heat flux, E the latent heat flux, and P the convective precipitation rate.

b. The WTG version of QTCM

The WTG version of the QTCM was constructed following Sobel and Bretherton (2000) by “disconnecting” the vertical columns of the standard QTCM, thereby recasting the QTCM as a set of SCMs. Interactions between columns are restricted to advection of moisture and temperature by the imposed mean circulation as the horizontal circulation field is fixed to an appropriate climatology. We impose upon the set of SCMs the climatological mean tropospheric temperature field with its full complement of 2D spatial variability; by applying such a temperature profile, realistic climatologies of mean climate fields like precipitation were obtained (see section 4). In place of the prognostic temperature equation (1), the WTG version of the QTCM uses a diagnostic vertical velocity field ω (the “diabatic vertical velocity”):

View Expanded

where the operator 〈. . .〉 denotes areal-mean averaging over a specified region and M_s is taken to be spatially invariant. The last term on the right-hand side of (5) represents a predetermined areal mean of ω; its value depends upon the region over which the average is computed and, for the simulations described here, was obtained from a control integration of the model (see section 3). We stress that the diabatic vertical velocity as determined by Eq. (5) is distinct from the vertical velocity appearing in Eqs. (1) and (2); the latter is defined in terms of the divergence of (prognostically determined) horizontal momentum.

The modifications encapsulated in Eq. (5) are essentially identical to Sobel and Bretherton (2000). Now, because our interest is the anomalous climate response to El Niño (or other perturbative forcing scenarios), we also wish to provide the QTCM with the flexibility of determining a self-consistent anomalous temperature “offset” as a means of adjusting to the perturbation. Following Shaevitz and Sobel (2004), who solved a similar system but with only two columns, we obtain solutions subject to a uniform temperature change of the form

View Expanded

where k is a scale factor the value of which (here, taken to be 0.1) was selected to obtain stable solutions. Note that because of the arbitrariness of k, Eq. (6) should be regarded simply as an algorithm designed to obtain a self-consistent steady solution rather than a true time tendency. At equilibrium, when 〈T ′〉 → const and 〈Q_T〉 → M_s 〈ω₀〉, the choice of k is irrelevant; and, by construction, 〈ω〉 = 〈ω₀〉 for all times.

Our approach treats the temperature field as two distinct components, namely an imposed, heterogeneous mean field and an adjustable, homogeneous offset. As such, our application of WTG only applies to perturbations about the tropospheric temperature climatology and not to the climatology itself—this is not unreasonable since our interest is in perturbations of the tropical climate to ENSO. It should be noted that we implemented WTG fully by requiring a spatially homogeneous mean tropospheric temperature field; while the full WTG approach did not perform as well in simulating the mean climate, qualitatively similar anomalous climate behavior was obtained with the full WTG approach. We also point out that, strictly speaking, the application of WTG does not require temperature gradients to vanish. Rather, the approximation involves imposing the limit of zero-valued temperature gradients in the thermodynamic equation [Eqs. (1) and (2)] only. Temperature gradients also appear in the horizontal momentum equations, but WTG places no constraints on temperature gradients as they influence momentum. However, our application of WTG effectively assumes a zero-valued anomalous horizontal circulation field, so the issue of nonzero temperature gradients in the momentum equations is neglected. In simpler applications of WTG (such as the two-column model of Shaevitz and Sobel 2004), it is, in fact, possible to infer a posteriori a circulation consistent with WTG by invoking continuity; the implementation of a WTG-consistent circulation for a model setup with N > 2 columns is less straightforward.

3. Justification for the WTG approach

Justification for the weak temperature gradient constraint has been thoroughly argued in previous studies [e.g., see Sobel and Bretherton (2000) and references therein for the tropical mean state, and CS02 for the ENSO problem], so we do not go into the details here. To assess the validity of the mass balance constraint—that perturbations in ascent induced by El Niño are essentially entirely compensated for by subsidence within the tropical band—we present composite analysis results for the annual-mean 500-mb pressure velocity field, obtained from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis dataset and averaged over selected domains in the Tropics for the period 1950–2002 (Table 1).

Consider the vertical velocity field averaged over 20°S–20°N. In both the Pacific and remote regions, the areal-mean pressure velocity is negative, corresponding to ascending motion in the rising branch of the Hadley circulation. During El Niño phases, there is an ∼5% increase in ascent (from −0.96 × 10⁻² Pa s⁻¹ to −1.01 × 10⁻² Pa s⁻¹) in the tropical Pacific that is compensated almost entirely by an increase in the subsidence (i.e., the magnitude of the remote-averaged pressure velocity decreases from 0.37 × 10⁻² Pa s⁻¹ to 0.30 × 10⁻² Pa s⁻¹) over the remote Tropics so that the mean ascent over the entire Tropics is relatively unchanged. A similar picture, but with Pacific and remote regional anomalies of opposing sign, emerges for La Niña conditions. Thus, we conclude that the vertical velocity field, when averaged over the entire Tropics (at least for a reasonable range of latitudes), is relatively insensitive to ENSO phase. Note that, for an averaging interval of 30°S–30°N, the average of the vertical velocity computed over all longitudes is much closer to zero, reflecting the cancellation of the ascending and descending branches of the Hadley circulation in the meridional direction.

In a related vein, we note the pronounced tendency of observed remote tropical precipitation anomalies to anticorrelate with the equatorial Pacific anomalies. Over the period 1979–99, for example, time series of the aggregate, monthly mean Pacific and remote tropical precipitation anomalies from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) dataset are anticorrelated beyond the 99.9% level (r = −0.30). Given the association between ascent (subsidence) and enhanced (reduced) precipitation, an anomalous zonal dipole behavior in precipitation is not surprising. Similar regionally antiphased rainfall anomalies are also evident in models. For example, a multidecadal (1950–2000) integration of the QTCM with imposed Pacific SSTs exhibits strongly pronounced antiphasing between its Pacific and remote regional average precipitation anomalies (r = −0.93); other models [including, for example, the National Center for Atmospheric Research (NCAR) Community Climate Model version 3.10 (CCM3; see Kiehl et al. (1998)] produce similar behavior, although the degree of anticorrelation is variable.

4. Simulated precipitation climatologies

We first briefly document the principal features of the precipitation climatologies for the WTG and standard versions of the QTCM [see Sobel and Bretherton (2000) for a more thorough discussion of the mechanisms underlying the establishment of the mean state precipitation climatology in the single-column version of QTCM]. Global maps of the mean precipitation fields for the control runs of both the standard and WTG QTCM frameworks, averaged over the period when the simulations have equilibrated (months 61–96), are presented in Figs. 2a and 2b, respectively. Given the simplifications inherent in the WTG version of QTCM, the distribution and magnitudes of the WTG precipitation climatology match the standard QTCM climatology remarkably well. Focusing on the tropical regions, both versions exhibit centers of deep convection over the Amazon basin, Africa, and the equatorial Indian Ocean/Maritime Continent/western Pacific regions. For the continental convection zones, peak monthly mean rainfall rates of up to 14 mm day⁻¹ are evident, with rainfall rates reaching upward of 20 mm day⁻¹ near the Maritime Continent.

There are however some differences between the simulated climatologies: in particular, the WTG QTCM simulates slightly higher precipitation rates (typically 2 mm day⁻¹) in the regions of strongest mean convection and lower rates on the margins of the strongest convection zones. In other words, the climatological convection field in the WTG QTCM tends to be spatially more sharply defined than in the standard version. Such differences likely arise from the lack of interactive high frequency coupling between thermodynamics and dynamics (i.e., transient humidity and temperature fluxes) that would tend to “smear” the edges of deep convective regions. Despite these differences, the overall correspondence between the climatologies should permit reasonable comparisons of the anomalous climate response to ENSO in the two QTCM versions.

5. Standard and WTG QTCM simulations of ENSO forcing

To produce the El Niño forcing response in both the standard and WTG versions of the QTCM, the SSTA field corresponding to the observed anomalies for January 1998—the peak month of the 1997/98 El Niño (the “El Niño of the century”)—was added to January SST climatology in the Pacific forcing region. (Note that the SSTAs were linearly suppressed to zero between 30°–20°S and 20°–30°N, 110°–125°E.) For the WTG QTCM, the diabatic vertical velocity field and uniform temperature offsets were computed via Eqs. (5) and (6), respectively. In these equations, a meridional averaging interval of 30°S–30°N was employed; the use of different averaging intervals (e.g., 20°S–20°N) was found to yield qualitatively similar results. As an additional test of the applicability of the WTG framework, we also performed simulations using La Niña phase SSTA. To facilitate comparison to the El Niño scenario, we used as the imposed perturbation forcing the SSTA field from the El Niño simulation but with its polarity reversed. Actual La Niña events, such as the 1989 event, were simulated with qualitatively similar behavior obtained.

6. Understanding the impacts of tropospheric temperature and subsidence changes during El Niño

We now try to understand the origins of the WTG El Niño response in the remote Tropics by attempting to separate the relative influences of the tropospheric temperature warming and subsidence on the climate. Although the El Niño–induced remote tropical increases in tropospheric temperature and subsidence are causally related and difficult to separate (as adiabatic compression associated with subsidence gives rise to warming), from a precipitation perspective, the effects of tropospheric temperature and subsidence may impact convection through different pathways—that is, tropospheric temperature through stabilization of the atmospheric column and subsidence through divergence of column moisture [see Eq. (7) below and the related discussion]. Within the WTG framework we can attempt to assess the relative roles of these two influences.

Apart from the potential to diagnose the mechanisms of the ENSO tropical teleconnection, consideration of the “pure” tropospheric temperature influence of ENSO separate from the subsidence effect is also of potential interest for determining the tropical rainfall response to increased tropospheric greenhouse gas loading, as both increasing tropospheric greenhouse gas concentrations and El Niño are associated with Tropics-wide tropospheric temperature increases. In fact, because of the warming effect common to greenhouse gases and El Niño, the latter has been suggested as a proxy for the near-surface effects of global warming (Soden 1997). Our basic conclusion here is that the direct effects of temperature and subsidence have significantly different impacts on the remote rainfall response to ENSO, so the value of ENSO as a global warming proxy is questionable.

To investigate the pure tropospheric temperature influence, an increased GHG scenario (denoted “ΔGHG+”) was simulated in each model framework using a 50-m MLD slab ocean model at all oceanic grid points. The CO₂ concentration was increased (at the initiation of the perturbation simulation) from its control value of 330 ppmv to a new uniform value everywhere; the level of perturbation CO₂ forcing was chosen to yield the same mean equilibrium level of anomalous tropospheric warming as in the WTG El Niño simulation. (The level of perturbation CO₂ required to achieve the desired temperature change, 0.78 K, was found to be approximately 1.46 times the standard concentration, or 483 ppmv.) For the WTG ΔGHG+ simulation, the zero divergence anomaly constraint was again enforced over 30°S–30°N.

Anomalous precipitation fields simulated under the ΔGHG+ scenario for the standard and WTG versions of the QTCM are illustrated in Fig. 6. The anomalous rainfall distributions are quite similar between the standard and WTG simulations of the ΔGHG+ scenario (cf. Figs. 6a,b to Figs. 6c,d). In fact, the remote tropical averages of various climate parameters for the two QTCM versions subject to enhanced GHG forcing are found to agree to within a percent (Table 3). The overall agreement between the standard and WTG versions of QTCM may reflect the fact that the forcing induced by increased GHG loading is, to lowest order, spatially uniform, so the dynamical responses tied to pressure gradients (which our WTG framework does not model) are negligible. In terms of the response to the greenhouse gas forcing itself, our results are consistent with those previously reported by Neelin et al. (2003) and Chou and Neelin (2004) using the QTCM. We find that extensive regions of positive precipitation anomalies are observed to occur in the vicinity of the regions of strongest mean convection and drying in marginal convection regions adjacent to the main convective centers, most notably along the eastern coastline of South America, the western coastline of Africa, and across the north and south tropical Indian Ocean.

Neelin et al. (2003) and Chou and Neelin (2004) ascribed the precipitation response to anomalous GHG forcing to the action of two related mechanisms dubbed “upped ante” and “rich-get-richer.” Under the first of these, the increase in tropospheric temperature arising from increased GHG is thought to increase the thermal barrier for convection by stabilizing the troposphere. Deep convective zones, characterized by convergent mean flow, can overcome the upped ante for convection by outcompeting marginal zones for moisture so that the convectively rich become convectively richer. Some elements of these mechanisms (e.g., the increase in precipitation over equatorial Africa) are discernible in the El Niño simulations discussed in section 5.

We now compare the climate responses of the WTG El Niño and ΔGHG+ simulations over the remote tropical region, given that both forcing scenarios exhibit the same uniform tropospheric temperature warming but that they differ in the presence or lack of forced large-scale subsidence. The intriguing result of this comparison is that, while the aggregate remote tropical precipitation anomalies are of opposite sign for the WTG El Niño and ΔGHG+ forcing scenarios, the spatial patterns of the anomalous remote tropical precipitation fields clearly resemble one another, as can be seen by direct comparison of Figs. 6d and 3d. These results suggest that the operation of distinct influences associated with the remote tropical tropospheric temperature and large-scale subsidence increases during El Niño give rise to, respectively, a similar spatial distribution and a dissimilar level of gross-scale anomalous precipitation relative to the ΔGHG+ forcing case. In other words, the presence of a (uniform) temperature increase in the WTG El Niño simulation imparts a GHG-like structure to the spatial distribution of precipitation anomalies over the remote tropical region. However, the presence of a forced large-scale descent anomaly—which is required to balance anomalous ascent in the Pacific—produces the remote-averaged offset in the level of anomalous precipitation.

We can demonstrate the aforementioned aspects of the anomalous precipitation behavior more clearly using a binned averaging procedure to examine changes to the precipitation field as a function of mean-state precipitation. Our motivation for analyzing the data in this way follows from the observation that the sign of the precipitation anomalies in the WTG (or standard) ΔGHG+ scenario appears to depend on the mean-state precipitation, changing from negative (for marginally convective regions) to positive (for deep convective regions). By contrast, for El Niño conditions, the anomalous remote tropical precipitation response appears to be (at least in the standard QTCM) relatively insensitive to the mean precipitation state, that is, negative precipitation anomalies are geographically widespread.

Binned average El Niño (solid lines) and ΔGHG+ (dashed lines) precipitation histograms for the standard and WTG QTCM appear in Figs. 7a and 7b, respectively. Here, the anomalies have been averaged over mean precipitation bins of width 2 mm day⁻¹ and aggregated over the entire remote tropical region (20°S–20°N, 70°W–110°E). The results have been further averaged over the final 36 months of the simulations, with the error bars representing the 2σ level of each bin average. Comparing the results for the standard QTCM (Fig. 7a), it can be seen that the El Niño histogram is negative for nearly all mean precipitation values while the ΔGHG+ histogram is negative below ∼6 mm day⁻¹ and positive above, clearly revealing the distinct nature of the anomalous precipitation responses for the two forcing scenarios.

For the WTG simulations (Fig. 7b), the ΔGHG+ histogram (dashed line) essentially lies on top of the standard ΔGHG+ histogram. By contrast, the standard and WTG El Niño histograms (Fig. 7b) are rather dissimilar, as the latter exhibits a more GHG-like structure: the precipitation anomalies of the WTG El Niño simulation transition from negative to positive with increasing mean-state precipitation but with an offset that depresses the entire histogram relative to the WTG ΔGHG+ result. Although fairly constant for all mean values, the offset is largest at intermediate mean precipitation rates (i.e., 2–6 mm day⁻¹). Again, it is likely that the differences between the standard and WTG QTCM El Niño simulations are attributable to effects neglected by the WTG approach, such as anomalous spatial gradients in temperature.

So, how does subsidence lead to reduced rainfall, and how does this fit with the role of tropospheric temperature? The QTCM derives precipitation from a simple Betts–Miller convective parameterization (Betts and Miller 1986) in which the precipitation rate is determined by the local relaxation of tropospheric temperature to a reference temperature profile, T_c:

View Expanded

Here τ_c is the convective adjustment time scale, which is fixed at 2 h, and q_s is the saturation specific humidity (a function of temperature only). It is clear from (7) that an increase in tropospheric temperature for fixed T_c will decrease precipitation. However, T_c, which depends on the specific and saturation specific humidities, is also free to adjust to the perturbative forcing, so it is the relative change of tropospheric temperature and the reference temperature profile that is integral to setting the sign (and magnitude) of the anomalous precipitation response.

The tropospheric drying (moistening) effect of subsidence (uplift) acts to reduce (increase) the column humidity through column divergence (convergence) of moisture and, therefore, lowers (raises) the Betts–Miller reference temperature profile. Examination of a longitudinal profile of atmospheric humidity in the WTG QTCM El Niño and ΔGHG+ scenarios (Fig. 8) demonstrates the linkage between vertical motion and humidity. The humidity change of the WTG ΔGHG+ simulation (triangles) is observed to be longitudinally quasi-uniform; the localized regional maxima in the humidity profile correspond to regions of climatological deep convection, where precipitation and vertical motion are increased. On the other hand, the humidity profile for the WTG El Niño simulation (squares) is sharply peaked in the central and eastern Pacific where strong anomalous convection, locally forced by the imposed SSTA, is present. Outside of the tropical Pacific, the humidity profile of the El Niño simulation lies below the level of the ΔGHG+ simulation, suggesting the suppression of specific humidity by mean subsidence over the remote Tropics as a whole in the El Niño simulation. We conclude that an important effect of the anomalous divergence induced by El Niño conditions in the Pacific region is the establishment and maintenance of a lower level of remote tropospheric humidity than that which occurs in the presence of the temperature-only forcing associated with perturbed GHG.

Although the generalized reference temperature profile is a nonlinear function of temperature because of the nonlinear temperature dependence of saturation specific humidity, for our purposes it is illustrative to linearize perturbations of (7) as

View Expanded

where k is a nonconstant scale factor. It should be noted that this approximation is reasonable when applied to regions where the mean state is convecting and to averages over sufficiently large spatial domains. According to (8), an increase in humidity during El Niño will tend to offset or partially compensate an increase in temperature—it is this behavior that is the basis of the upped-ante and rich-get-richer mechanisms of Neelin et al. (2003) and Chou and Neelin (2004). In the context of the WTG El Niño and ΔGHG+ simulations described here, we note that the temperature changes are identical, so the difference in the level of anomalous precipitation change over the remote Tropics between the El Niño and ΔGHG+ scenarios is tied to the difference in humidity: while both scenarios experience a rise in remote region tropospheric humidity, the smaller gross-scale remote tropical humidity increase in the El Niño case leads to a negative anomaly in precipitation.

7. Influence of anomalous temperature gradients in the WTG framework

One potential source of the poor representation of the spatial distribution of El Niño–induced precipitation anomalies in the WTG simulations is the assumption of no anomalous temperature gradients. Given that some tropospheric temperature gradients may develop in response to El Niño forcing (see Figs. 4g and 4h), it is of interest to determine what impact the addition of anomalous temperature gradients to the WTG simulation may have on the climate response to ENSO. To do this, we simply added to the WTG El Niño forcing scenario the spatial structure of the anomalous temperature field as simulated by the standard QTCM. Before adding the temperature gradient anomalies, a mean temperature anomaly corresponding to the areal average over the 30°S–30°N strip used to compute the temperature offset was first removed such that the areal-mean of the imposed temperature gradient perturbation was zero.

The ad hoc inclusion of anomalous temperature gradients produces an anomalous precipitation response in the Pacific (Fig. 9a) that more closely matches the standard simulation (Fig. 3a) than does the WTG El Niño simulation lacking anomalous temperature gradients (Fig. 3c). In particular, the zone of positive precipitation anomalies in the central and eastern Pacific is reduced in extent and magnitude with the inclusion of an anomalous temperature gradient field. The near-equatorial negative precipitation anomaly is also reduced in magnitude and extent, while the north-flanking negative anomaly is enhanced. Some improvements are also evident over the remote region (Fig. 9b): in particular, there is a more pronounced penetration of negative anomalies into the interior of South America.

Of course, not all regions show improvements in the precipitation response with the addition of anomalous temperature gradients. Over Africa and the Indian Ocean, the precipitation response in Fig. 9b still essentially resembles the WTG simulation lacking anomalous gradients. However, temperature gradients are relatively weak over this portion of the remote Tropics, suggesting that something other than spatial structure of the temperature field may contribute to the anomalous precipitation behavior. We speculate that circulation effects not modeled in the WTG framework, particularly wind speed changes to the latent heat flux, may impart some of the structure to the standard simulation precipitation field, particularly over the eastern Indian Ocean basin, although an understanding of the details of these effects is beyond the scope of the present study.

8. Conclusions

In this paper, we explored the reorganization of the tropical climate during El Niño using a weak temperature gradient (WTG) framework, with an emphasis on the mechanisms giving rise to the ENSO remote tropical precipitation teleconnection. The fundamental assumptions at the heart of this approach are 1) a spatially homogeneous perturbation temperature change and 2) an anomalous mass conservation constraint applied to the entire tropical belt. In the QTCM, the intermediate-level complexity model used in this study, application of the WTG assumptions simplifies the model by eliminating the prognostic equations for horizontal momentum and replacing the prognostic equation for temperature with a diagnostic equation for divergence and a single anomalous temperature tendency equation closed by the mass balance constraint.

ENSO simulations performed with the QTCM modified according to the WTG framework were found to simulate an appropriate level of tropospheric temperature change as well as a reasonable spatial distribution of precipitation anomalies in a gross-scale sense, demonstrating that the WTG framework can account for the basic features of the tropical response to ENSO forcing conditions. We showed additionally that the WTG QTCM realistically simulates the climate response to enhanced greenhouse gas forcing, at least in comparison to the standard version of the same model. We further used the WTG framework to show how the differential precipitation responses to GHG and El Niño forcing may be tied to fundamental differences in the nature of the forcing, that is, temperature only for GHG and temperature with large-scale-forced subsidence for El Niño.

Despite the agreement of the gross-scale El Niño precipitation deficits in the standard and WTG versions of the QTCM, the regional-scale features of the anomalous precipitation (and surface temperature) field in the remote Tropics manifest some marked contrasts. For example, while the standard QTCM version was found to simulate a widespread reduction of precipitation across the remote region in response to El Niño forcing conditions, the WTG version yielded a more GHG-like pattern of enhanced precipitation over deep convective zones and reduced precipitation along the margins of these zones. The addition of anomalous temperature gradients to the WTG simulation somewhat improved the spatial agreement between the WTG and QTCM precipitation anomaly fields.

Significantly, we suggest that the WTG approach outlined here may provide the underpinnings of a conceptual framework for adjustment of the tropical climate to ENSO, in particular the remote tropical response. The canonical conceptual framework of the ENSO teleconnection, the Gill (1980) model, has proven successful in many ways, especially in terms of understanding the surface and upper-tropospheric atmospheric circulation anomalies during El Niño and the gross changes to the zonal Walker circulation given the anomalous convective heating associated with warm SSTA. However, the Gill model lacks the thermodynamics necessary to understand what the climate response of the remote Tropics is beyond that which can be inferred from the linkages to anomalous subsidence predicted by the dynamical framework. The WTG framework presented here, on the other hand, gives a precise interpretation of how the remote tropical climate responds under El Niño forcing: the climate (and hence anomalous subsidence) responds in such a way that spatial gradients in anomalous tropospheric temperature are negligible. The WTG framework thus naturally lends itself to the nonuniform distribution of anomalous subsidence (as was seen for the observations in Fig. 1), since the level of subsidence consistent with a change to tropospheric temperature in any one vertical column of the QTCM depends on the mean climate of that column, the particular suite of climate feedbacks experienced by the column (with the exception of those processes involving the nondivergent component of the anomalous circulation, which was not modeled), and connections to neighboring columns (through, e.g., moisture fluxes).

Of course, a conceptual framework is only as good as the assumptions underlying it. The simulations outlined in this study demonstrate that while the gross features of the tropical adjustment to ENSO are reasonably simulated—that is, mean tropical tropospheric temperature changes, precipitation responses over the remote Tropics, and the zonal and meridional shifts in convection over the tropical Pacific—the regional-scale spatial features of the remote tropical precipitation and surface temperature anomalies are not well captured. The simulations in which we imposed horizontal tropospheric temperature gradients (section 7) suggest that some of these regional-scale discrepancies arise from the assumption of vanishing tropospheric temperature gradients. Indeed, zonal gradients in tropospheric temperature anomalies are quite apparent in the observational data during ENSO warm phases (cf. Fig. 6 of Yulaeva and Wallace 1994), and the climate impact associated with such gradients may be significant. A determination of the anomalous zonal tropospheric temperature gradients in the standard QTCM simulation compared to a similar simulation in a full 3D atmospheric general circulation model (e.g., CCM3) reveals that the horizontal gradients produced by the QTCM may be somewhat too large, such that the influence of temperature gradients may be overemphasized in the QTCM simulations. In a future study, we aim to investigate the dependence of the tropical ENSO teleconnection on the magnitude of horizontal temperature gradients.

Finally, we briefly comment on the applicability of using El Niño as a proxy for understanding the impacts of global warming, as suggested in previous studies. It is tempting to make an analogy between El Niño– and GHG-induced global warming as both are characterized by increased tropospheric temperatures—indeed, Neelin et al. (2003) and Chou and Neelin (2004) used the warming of the troposphere as a starting point for their upped-ante and rich-get-richer mechanisms for tropical drought regions found in the QTCM under global warming and some El Niño teleconnection regions. Our WTG simulations of the GHG and El Niño scenarios—in which different remote tropical precipitation responses were obtained despite the fact that the tropospheric temperature anomalies were identical in both simulations—suggest that this analogy may be misleading. In particular, we have noted the impact of large-scale, El Niño phase subsidence (induced by anomalous Pacific region convection) on the remote tropical atmospheric humidity field, with subsidence-induced drying of the troposphere during El Niño associated with a reduction of remote-averaged precipitation. In the GHG scenario, where large-scale subsidence does not accompany the temperature increase, the opposite gross-scale precipitation response is observed.

While we do not yet have a complete understanding of the nature of the differences between El Niño and global warming (particularly in terms of causality), it suffices to say that how the tropospheric temperature changes are generated is important—whether temperature changes through the atmospheric greenhouse effect or through energy redistributions by vertical circulations appears to influence the climate response to the forcing. One possible approach to understanding the differences between the El Niño and GHG forcing scenarios is through examination of the moist static energy budget (H. Su 2005, personal communication). In equilibrium, the combination of the moisture and temperature equations leads to a diagnostic equation for vertical pressure velocity as a function of the combined forcing terms for temperature and humidity [the Q_T and Q_q of Eqs. (1) and (2)]. Hence, the difference in the vertical velocity between the GHG and El Niño cases can be tied to the difference of one or more of the individual terms of Q_T and Q_q. Under suitable simplifications—for example, areal averaging and the assumption of weak temperature gradients—it may be possible to associate the differential climate responses to greenhouse gas increases and El Niño to a single-component process (like changes to the top of the atmosphere radiative balance). Such an analysis will be undertaken in a future study.