## 1. Introduction

Within the past two decades, many studies have shown that changes in the position of the earth’s axis of rotation relative to the earth’s crust are strongly linked to fluctuations in the equatorial atmospheric angular momentum (EAAM) vector. This displacement in the pole position is known as polar motion. On both interannual and intraseasonal time scales, the linear correlation between polar motion and EAAM is typically near a value of 0.6 (e.g., Eubanks et al. 1988; Chao 1993; Nastula 1995). Since polar motion corresponds to fluctuations in the equatorial components of the solid earth’s angular momentum, and because the equatorial angular momentum of the solid earth–atmosphere–ocean system is conserved (excluding tidal influences), this value for the linear correlation implies that both the atmosphere and the ocean must be playing important roles in driving polar motion of the solid earth (e.g., Ponte et al. 1998; Ponte and Stammer 1999; Celaya et al. 1999; Nastula and Ponte 1999). While much research on EAAM has been performed by geodesists, the results from this research have had little impact on the atmospheric science community. Nevertheless, the subject of EAAM is now ripe for research by atmospheric scientists, since recent studies have shown that polar motion and EAAM fluctuations are both driven by well-known large-scale atmospheric dynamical processes (Egger and Hoinka 2000; Feldstein 2003), and because of the remarkable fact that these atmospheric processes alter the location of the earth’s axis by a measurable amount.

**EM**consists of the two components of the atmospheric angular momentum vector that are located within the equatorial plane (the third component is aligned parallel to the earth’s axis; Fig. 1a). The time derivatives in (1) are for the inertial and rotating coordinate systems, indicated by the subscripts “

*i*” and “

*r*,” respectively. The friction, mountain, and equatorial bulge torques are expressed by

**FT**,

**MT**, and

**BT**, respectively. (The definitions for

**EM**and all three torques are presented in the appendix.) The equatorial bulge torque can be understood as being analogous to the mountain torque that arises from the earth’s spheroidal shape (Fig. 1b). Since

**BT**=

**Ω**×

**EM**(Bell 1994), (1) can be rewritten aswhere

_{m}**EM**

_{m}and

**EM**

_{w}are the mass and wind field contributions to

**EM**, respectively [see (A1) and (A2)]. The mass contribution to

**EM**is due to the atmosphere’s rotation with the solid earth’s angular velocity,

**Ω**, and the wind contribution to

**EM**arises from atmospheric motion relative to the earth’s surface. As shown by Eubanks et al. (1988) and Bell (1994), the mass contribution to the equatorial components of

**EM**is about one order of magnitude greater than the wind contribution. In both EAAM and polar motion studies, the two equatorial components of the EAAM vector are typically specified as being aligned with the Greenwich and 90°E meridians.

Recently, Feldstein (2003) examined the dynamical processes that drive EAAM fluctuations in an aquaplanet general circulation model (GCM). In that study, the model had a flat, all-ocean lower boundary with a sea surface temperature field that was both zonally symmetric and symmetric across the equator. The uniformity of the lower boundary ensured that mountain torques were absent and that there were no zonal asymmetries in the model’s climatological flow. One key finding of that study was that the model’s 10-day westward rotation of the EAAM vector (a similar 10-day westward rotation is also seen in the atmosphere; Brzezinski 1987; Egger and Hoinka 2002) relative to the earth’s surface could be understood as arising from the westward propagation of a zonal wavenumber 1, Rossby wave normal mode that is antisymmetric across the equator. The other main finding was that fluctuations in the amplitude of the EAAM vector are associated with equatorial precipitation and wind anomalies showing characteristics typical of a zonal wavenumber 1, equatorial, mixed Rossby–gravity wave. These results led to the hypothesis that the model’s EAAM amplitude fluctuations are ultimately driven by latent heating within the mixed Rossby–gravity wave, and that this latent heating excites poleward-propagating, upper-tropospheric Rossby waves that subsequently induce midlatitude surface pressure anomalies that can alter the EAAM amplitude.

The first goal of this study is to investigate the extent to which the dynamical processes that drive EAAM fluctuations in an aquaplanet GCM (Feldstein 2003) also drive the EAAM fluctuations in the atmosphere. The second goal of this study is to examine the impact of the mountain torque on the EAAM of the atmosphere.

The methodology is described in section 2, followed by an examination of the EAAM budget in section 3. An analysis of the dynamical processes that determine the phase and amplitude of the EAAM vector is presented in sections 4 and 5, respectively, followed by the conclusions in section 6.

## 2. Data and diagnostic techniques

The EAAM dynamics is examined with daily (0000 UTC) National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data. This study will focus on the boreal winter [December–February (DJF)] for the years 1958–97. To investigate the role of tropical convection on driving EAAM amplitude fluctuations, we use outgoing longwave radiation (OLR) data, which are produced by the National Oceanic and Atmospheric Administration. Except for the OLR and mountain torque calculations, all data are truncated at rhomboidal 30 resolution. The analysis of OLR is performed on a 2.5° latitude × 2.5° longitude grid, and the mountain torque is calculated on the NCEP–NCAR triangular 62-degree grid. For this study, unless stated otherwise, all data are temporally unfiltered and the seasonal cycle is subtracted. The seasonal cycle is obtained by applying a 20-day low-pass digital filter to the time mean of each calendar day. This study focuses on anomalies, which are defined relative to the seasonal cycle for the corresponding day. Several additional calculations will be performed with data from the aquaplanet GCM. The purpose of these calculations is to examine the impact of the atmosphere’s zonally inhomogeneous underlying surface.

**i**and

**j**are unit vectors in the inertial coordinate system,

*t*is the fast daily time variable associated with the earth’s rotation at the diurnal frequency, Ω, and

*τ*is a slower time variable associated with amplification and rotation of the EAAM vector relative to the rotating earth. In this study, we will focus on the slower dynamical processes that drive changes in the phase and amplitude of the EAAM vector relative to the rotating earth.

The primary methodology to be used involves the linear regression of different atmospheric variables against the two components of the EAAM vector tendency (i.e., *d*EM_{1}/*dt* and *d*EM_{2}/*dt*) and against the tendency of the amplitude of the EAAM vector, *d*(EM^{2}_{1} + EM^{2}_{2})/*dt*. (The quantities EM_{1} and EM_{2} denote the Greenwich and 90°E components of the EAAM vector.) The former calculations will allow us to examine the processes associated with fluctuations in the phase of the EAAM vector relative to the earth’s surface. In the regression equations, the amplitude of the anomaly is always specified to be equal to one standard deviation. In this study, we will focus on intraseasonal EAAM variability as the majority of the observed variance takes place at this time scale (Salstein and Rosen 1989).

## 3. EAAM budget

We first evaluate the extent to which the EAAM budget is balanced in an inertial reference frame. This is performed by regressing both sides of (1) against the left-hand side (lhs) of that equation, for both the Greenwich and 90°E meridians (Figs. 2a,b). As can be seen, at all lags, there is very close agreement between the EAAM tendency and the equatorial bulge torque. The mountain torque is about an order of magnitude smaller than the equatorial bulge torque, and the friction torque (not shown) is another order of magnitude smaller. These findings indicate that the EAAM budget is dominated by the equatorial bulge torque, as was also shown for the aquaplanet GCM.

To further understand these results for the EAAM budget, we present both the standard deviations for all three torques and also the linear correlations between each of these torques and the corresponding EAAM tendency (see Table 1). It is the product of these two quantities that determines the linear regression coefficients in Fig. 2. The standard deviation values resemble those of Egger and Hoinka (2002). [In contrast, de Viron et al. (1999) found standard deviation values for the mountain torque that are similar to those of the equatorial bulge torque.] The dominance of the equatorial bulge torque in Figs. 2a,b can be understood from these standard deviation and correlation values.

We next examine the EAAM budget in a reference frame rotating with the earth. As for the inertial reference frame, the budget is evaluated by regressing both sides of (2) against the lhs of (2), for the Greenwich and the 90°E meridians. The results of these calculations (Figs. 2c,d) show a moderate degree of balance for the two EAAM components. Although the phase of both sides of (2) tends to be correct, near lag 0, the amplitude of the right-hand side (rhs) is about half that of the lhs. It is possible that inaccuracies in the calculation of the mountain torque may account for these deficiencies in the budget in the rotating reference frame (Egger and Hoinka 2000). Several different methods were used to calculate the mountain torque. This included the three methods suggested by Swinbank (1985) (centered differencing, transformation to log pressure, and explicit calculation of the vertical integral of the pressure gradient force in sigma coordinates) and also a calculation using two-dimensional B-spline interpolation. Each of these methods yielded nearly identical results. Also, since Figs. 2g,h show that the **Ω** × **EM**_{w} and mountain torque terms in (2) undergo rapid variation over a period of a little more than 1 day near lag 0, temporal resolution may also be a factor in the level of balance shown in Figs. 2c,d. To examine this possibility, a digital filter with a 10-day cutoff period was applied to the budget terms (Figs. 2e,f). As can be seen, the filtering of higher frequencies does lead to a much improved balance in the budget.

The regressions of **Ω** × **EM**_{w} and the mountain and friction torques are shown in Figs. 2g,h. As with the aquaplanet GCM, it is found that the **Ω** × **EM**_{w} term is dominant. However, in contrast to the aquaplanet GCM, the mountain torque in the atmosphere is far from being negligible.

## 4. Phase of the EAAM vector

### a. Power spectra

The phase of the EAAM vector is first examined by calculating the power spectra for the two EAAM components (see Fig. 3). As can be seen, the power spectra resemble a first-order autoregressive process with a statistically significant 10-day peak. A 10-day peak was also found in the aquaplanet GCM (Feldstein 2003) and in the atmosphere (Brzezinski 1987; Egger and Hoinka 2002). As with the aquaplanet GCM, lag correlations between EM_{1} and EM_{2} indicate that this 10-day peak corresponds to the westward rotation of the EAAM vector. However, in comparison with the aquaplanet GCM, the 10-day spectral peak in Fig. 3 is both broader and of smaller amplitude, although still well above the 95% a posteriori confidence level.

### b. Rossby wave propagation

The 10-day westward phase propagation is examined by linearly regressing the anomalous surface pressure, *p*′_{s}, against (*d*EM_{1}/*dt*)_{r}, the Greenwich component of the EAAM tendency (see Fig. 4). For this regression, a bandpass Fourier filter is applied to both the surface pressure field and the EAAM tendency. The cutoff periods selected for this filter are 7.5 and 12.9 days, as it is only between these two periods, for both EAAM components, that the power exceeds the 95% a posteriori confidence level (see Fig. 3). The fraction of the intraseasonal variance contributed by this band is 16% for the Greenwich meridian and 28% for the 90°E meridian. As can be seen in Fig. 4, there is a well-organized zonal wavenumber-1 pattern that is antisymmetric across the equator. This pattern propagates westward with a well-defined 10-day period. For the 90°E meridian, an analogous linear regression yields essentially the same results, except for a 90° phase shift in longitude (not shown). These wave propagation characteristics for the observed 10-day EAAM band are very similar to those for the aquaplanet GCM where the analyses was performed with unfiltered data (see Fig. 3 of Feldstein 2003).

Noticeable differences are found between the above bandpassed linear regression and the regression performed with unfiltered anomalous surface pressure and EAAM tendency (Figs. 5 and 6). Compared to the bandpass patterns shown in Fig. 4, the unfiltered wave field is less well organized, and therefore the 10-day period of propagation in the unfiltered data is less apparent. Given the similarity between the above bandpass regression patterns and those for the aquaplanet GCM, it seems likely that the atmospheric mountain torque is generating waves that are shorter than zonal wavenumber 1 with periods outside of the 10-day band (see Figs. 5c, 5f and 6c,f). Furthermore, we also suspect that this contribution from the mountain torque obscures what would otherwise be a more distinct 10-day peak in the EAAM power spectra.

Similar wave propagation characteristics are found for the anomalous, unfiltered 300-mb geopotential height field, *ϕ*′_{300} (for brevity, only the Greenwich component is shown in Fig. 7). The resemblance in spatial structure between the regressed *p*′_{s} and *ϕ*′_{300} fields, especially between lag −2 and lag +2 days, indicates to a fair extent that the wave field associated with this 10-day westward propagation has an equivalent barotropic vertical structure. For both EAAM components, the maximum *p*′_{s} amplitude occurs near lag +2 days (Figs. 4d, 5d, 6d and 7d).

We examine the physical mechanism that accounts for the westward EAAM vector propagation by relating the *p*′_{s} and *ϕ*′_{300} fields to the normal-mode solutions of the shallow-water model equations on the sphere (Eubanks et al. 1988; Bell 1994; Egger and Hoinka 1999), that is, the orthonormal Hough mode solutions to Laplace’s tidal equations [see Longuet-Higgins (1968) and Kasahara (1976, 1980) for details]. Following the notation used in Feldstein (2003), *n* = 2 denotes the first antisymmetric mode, and *n* = 4, the second antisymmetric mode. As in that study, a value of 5.8 km is chosen for the equivalent depth.

Table 2 lists linear correlations between the individual EAAM component time series and the time series obtained by projecting the daily *p*′_{s} and *ϕ*′_{300} fields onto the Hough mode spatial patterns. All of the correlations in Table 2 are statistically significant above the 95% confidence level. The higher-order Hough modes all yield much smaller linear correlations. In addition, it is only the *n* = 2 projection time series that have a strong 10-day peak. These results, which are very similar to those for the aquaplanet GCM, suggest that the westward propagation of the EAAM vector in the atmosphere arises from the Rossby wave propagation of the first antisymmetric, zonal wavenumber-1 normal mode. Overall, however, these correlations are a little smaller than those for the aquaplanet GCM.

### c. Mountain torque

The budget calculations in section 3 showed that the mountain torque plays an important role in determining the phase of the EAAM vector. As implied by the mathematical expression for the mountain torque [see (A5) and (A6)], a nonzero mountain torque occurs only when the zonal wavenumber of the topography and that of the *p*′_{s} field differ by 1. A calculation with various truncations in zonal wavenumber finds that about 80% of the mountain torque is due to interaction between the zonal wavenumber-1 *p*′_{s} and the zonal wavenumber-2 topography.

An inspection of Fig. 2g (lag 0 days) suggests that the mountain torque is causing the rotation rate of the EAAM vector to vary with time. The mountain torque is aligned along the Greenwich meridian at this lag (the 90°E component of the mountain torque is negligible at this lag), and the EAAM vector is close to being parallel to the 90°E meridian at lag 0 days [Fig. 5c; see also (A2)]. Such a phase relationship between the EAAM vector and the mountain torque must lead to an increase in the westward rotation rate of the EAAM vector. This expected increase in the rotation rate is verified from the temporal evolution of the phase of the EAAM vector, arctan(EM_{2}/EM_{1}), where EM_{1} and EM_{2} are obtained by linear regression against (*d*EM_{1}/*dt*)_{r} (Fig. 8a). As can be seen, the phase of the EAAM vector changes most rapidly at lag 0 days. Similar phase changes are seen 2–3 days later for calculations with (*d*EM_{2}/*dt*)_{r} (Fig. 8b). The corresponding phase curves for the aquaplanet GCM are straight (not shown), indicating that in the absence of topography, the westward-rotation rate of the EAAM vector is close to being constant. These differences between the atmosphere and the aquaplanet GCM are consistent with the suggestion that the mountain torque causes the rotation rate of the EAAM vector to fluctuate with time.

### d. Mountain torque spatial structure

To better understand the mountain torque, we also examine its spatial structure at lag 0 days. For this purpose, we regress the integrand in the equation for the Greenwich meridian mountain torque [see (A5)], that is, [*p _{s}*(∂

*h*/∂

*ϕ*) cos

*ϕ*sin

*λ*−

*p*(∂

_{s}*h*/∂

*λ*) sin

*ϕ*cos

*λ*], against the Greenwich meridian EAAM tendency. This is illustrated in Fig. 9, together with the NCEP–NCAR reanalysis topography. As can be seen, there are mountain torque contributions from Greenland, the Rockies, Scandinavia, the Himalayas, the Andes, and Antarctica.

A separate summation over each mountain range in Fig. 9 indicates that the primary contributors to the mountain torque at lag 0 are Antarctica and Greenland, with the contributions from the other regions being about one order of magnitude smaller. It is found that the Antarctic mountain torque is about 1.1 times greater than that from Greenland. These results illustrate the fact that with the exception of Antarctica, even though the local mountain torque is often large, there is substantial internal cancellation in the mountain torque for each region. Calculation of the mountain torque for other lags, when the EAAM tendency is smaller (Fig. 2b), reveals that the mountain torque from other regions can often be comparable to those from Antarctica and Greenland.

## 5. Amplitude of the EAAM vector

We next investigate the dynamical processes that drive fluctuations in the amplitude of the EAAM vector. For this purpose, several different variables are regressed against the tendency of the anomalous EAAM vector amplitude, *d*/*dt*[(EM′_{1})^{2} + (EM′_{2})^{2}], where EM′_{1} and EM′_{2} are deviations from the DJF time-mean values of EM_{1} and EM_{2}. As shown by Egger and Hoinka (2002), the time-mean value of EM_{2} is about two orders of magnitude greater than that of EM_{1}.

The time-lagged regression of the EAAM amplitude against its tendency is illustrated in Fig. 10a. As expected, the EAAM amplitude fluctuates rapidly, attaining its minimum (maximum) 2 days prior to (after) the maximum EAAM amplitude tendency.

**BT**=

**Ω**×

**EM**

_{m}(Bell 1994), the equatorial bulge torque is orthogonal to the mass contribution of the EAAM vector. As a result, only the much smaller wind contribution to the EAAM vector impacts

**BT**·

**EM**. Figure 10b shows that the friction torque is out of phase with the EAAM amplitude, indicating that this torque simply reduces the amplitude of the EAAM vector.

The extent to which (4) is balanced is illustrated in Fig. 10c. As can be seen, the balance is moderate near lag 0, when the tendency of the EAAM amplitude is largest, and better at other lags. However, as with the budgets shown in Fig. 2, application of the digital filter leads to a marked improvement in the balance of the EAAM amplitude budget (not shown).

There are at least three different mechanisms by which the equatorial bulge torque can alter the EAAM amplitude. These mechanisms involve 1) the poleward propagation of a zonal wavenumber-1 Rossby wave from the Tropics into the midlatitudes (Feldstein 2003), 2) wave–wave interaction, and 3) wave–zonal-mean flow interaction. The impact of these three processes, each of which can change the surface pressure field hence the EAAM amplitude, will be examined in this section. Since the friction torque is smaller than the other terms on the rhs of (4), and because it appears to simply dampen the EAAM amplitude, the friction torque will not be further considered in this study.

### a. Poleward Rossby wave propagation

As mentioned in section 1, Feldstein (2003) found that fluctuations in the EAAM amplitude of the aquaplanet GCM are associated with zonal wavenumber-1, equatorial mixed Rossby–gravity waves. In that study, it was suggested the latent heating within these equatorial waves excites the poleward-propagating Rossby waves, and that these waves in turn alter the EAAM amplitude by inducing surface pressure anomalies.

To isolate the wave field associated the EAAM amplitude fluctuations, for each day, the entire flow field is phase shifted to a common longitude. Such a procedure amounts to examining the flow in a coordinate system that follows the zonal wavenumber-1 disturbance. This phase-shifting method is discussed in Feldstein (2003). Very briefly, for each day, a longitude *λ*_{max} is determined, which maximizes the projection of the daily *p*′_{s} field onto the spatial pattern sin*ϕ* cos^{2}*ϕ* cos(*λ* − *λ*_{max}). This pattern corresponds to a linear combination of the mass contribution of the integrands of the EAAM vector [see (A1) and (A2)]. After *λ*_{max} is calculated, the entire *p*′_{s} field is shifted in the zonal direction by an amount that places *λ*_{max} at the same common longitude for each day. We choose this common longitude to be the date line, because although *λ*_{max} is found to occur at all longitudes, the most frequently occurring value for *λ*_{max} is 180°. This phase-shifting technique will result in the DJF, time-mean *p*′_{s} and 300-mb streamfunction, *ψ*′_{300}, having nonzero values (Fig. 11; we examine the 300-mb streamfunction, rather than the 300-mb geopotential height, in order to better isolate the tropical wave pattern).

Figure 12 shows the time evolution of the full anomaly pattern that is obtained by adding the surface pressure, *p*′_{s}, regressed against the EAAM amplitude tendency, to the phase-shifted, DJF, time-mean *p*′_{s} field (Fig. 11a). The resulting evolution shows that as the EAAM amplitude increases (Fig. 10a), the amplitude of the corresponding zonal wavenumber-1 surface pressure pattern also increases. A similar calculation for the *ψ*′_{300} field (not shown) also finds an increase in the amplitude of the zonal wavenumber-1 disturbance.

We next examine the anomalous *ψ*′_{300} regressed against the EAAM amplitude tendency, rather than the full *ψ*_{300} field, in order to search for the presence of equatorial waves (the time-mean *ψ*′_{300} is excluded from this calculation because its amplitude is greater than that of the regressed anomalous *ψ*′_{300} in the Tropics). Figure 13 shows a zonal wavenumber-1 streamfunction pattern that is confined to the Tropics and that tends to be symmetric across the equator (best seen in Fig. 13b). This pattern is typical for an equatorial mixed Rossby–gravity wave. Although less well organized, there is also a zonal wavenumber-1 pattern for the OLR field, which both straddles and is antisymmetric across the equator (this is most clearly seen in Fig. 13b). The relative location and sign of these OLR anomalies resemble both those in Wheeler et al. (2000) and also the precipitation anomalies in Feldstein (2003), which indicates that this OLR pattern is also consistent with an equatorial mixed Rossby–gravity wave. Therefore, as for the aquaplanet GCM, these results suggest that an equatorial mixed Rossby–gravity wave occurs prior to the attainment of the maximum EAAM amplitude. In addition, although not apparent in Fig. 13, a calculation of the momentum flux regressed against the EAAM amplitude tendency for zonal wavenumber 1 reveals an equatorward transport in the midlatitudes of both hemispheres (not shown) when the EAAM amplitude is increasing. These directions for the fluxes imply that there are poleward-propagating Rossby waves prior to the amplification of the EAAM vector.

### b. Wave–wave interactions

To examine the role of wave–wave interactions in changing the amplitude of the EAAM vector, the wavenumber power spectra of both the *p*′_{s} and the *ϕ*′_{300} fields are regressed against the tendency of the EAAM amplitude. For this calculation, the power spectra of these two variables is calculated as a function of zonal wavenumber for each day and for each latitude.

We first examine the regression of the anomalous power spectra for the first eight zonal wavenumbers of the *ϕ*′_{300} field summed over all latitudes (Fig. 14). The power spectra are summed over all latitudes since both hemispheres tend to show similar patterns, with the Northern Hemisphere contribution dominating, and with the anomalous power coming primarily from mid- and high latitudes (not shown). As expected, the power in zonal wavenumber-1 increases immediately after lag 0. Furthermore, as the EAAM amplitude is increasing, the power of zonal wavenumber 2 declines and that of zonal wavenumber 3 becomes larger. These tendencies in the power spectra are consistent with zonal wavenumber 2 giving up energy to both zonal wavenumbers 1 and 3. Opposite tendencies can be seen at earlier and later lags in Fig. 14 when the EAAM amplitude is decreasing.

Similar calculations are also performed for the surface pressure power spectra. The results indicate an increase in power for zonal wavenumber 1 as the EAAM amplitude is increasing, but the changes in the power for all other zonal wavenumbers are very small (not shown). This result indicates that wave–wave interactions are not playing a role at the surface.

For comparison, we also calculate the regressed *ϕ*′_{300} power spectrum for the aquaplanet GCM (not shown). This calculation was not performed in Feldstein (2003). The results indicate that as the EAAM amplitude increases, the power at zonal wavenumbers 3, 4, and 5 declines while that at zonal wavenumber 1 increases.

The above results show that in comparison to the aquaplanet GCM, the wave–wave interactions in the atmosphere involve longer waves. These results allude to a possible indirect role played by the earth’s topography in altering the atmosphere’s EAAM amplitude via the wave–wave interactions, since zonal wavenumbers 1 and 2 may be excited by the topography. Also, since these wave–wave interactions are not observed in the surface pressure power spectra, these results suggest that after the wave–wave interactions take place in the upper troposphere, surface pressure anomalies are induced, as discussed in Feldstein (2003) and in the introduction. Also, a plausible reason for zonal wavenumber 1 being the only wave impacted by the upper-tropospheric wave–wave interactions is that the Rossby penetration depth, *D* = *f L*/*N* (see Hoskins et al. 1985), where *D* is the Rossby penetration depth, *f* is the Coriolis parameter, *L* is the horizontal length scale, and *N* is the buoyancy frequency, is greatest for the largest spatial scales.

### c. Wave–zonal-mean flow interaction

We next consider the impact of wave–zonal-mean flow interaction on the EAAM amplitude. For the zonal wavenumber-1 disturbance associated with EAAM to gain energy baroclinically from the vertical shear of the zonal-mean flow, the eddy heat flux associated with the zonal wavenumber-1 disturbance must be poleward. Calculations reveal that the 850-mb eddy heat flux for zonal wavenumber 1, regressed against the EAAM amplitude tendency, is actually equatorward (not shown). These findings imply that wave–zonal-mean flow interaction actually opposes the amplification of the EAAM vector. However, the 850-mb eddy flux is poleward for zonal wavenumber 2 (not shown). All other waves do not reveal a clear, well-organized heat flux structure. These results suggest that zonal wavenumber 2, which appears to be the source of energy for the wave–wave interaction, gains its energy from the vertical shear of the zonal-mean flow. This energy is then transferred to zonal wavenumbers 1 and 3 via the wave–wave interaction, as discussed in section 5b (Fig. 14). A calculation of the regressed 300-mb eddy momentum fluxes indicates that barotropic wave–zonal-mean flow processes do not play an important role in changing the EAAM amplitude.

Similar calculations were also performed for the aquaplanet GCM. As with the observational data, the eddy heat flux for zonal wavenumber 1 is equatorward, and the eddy momentum flux did not play an important role. There were no coherent eddy heat fluxes for other zonal wavenumbers. These differences between the observations and the aquaplanet GCM suggest that the topography may initially be exciting the zonal wavenumber-2 disturbance in the atmosphere, and that poleward eddy heat fluxes would then further amplify this growth.

### d. Mountain torque

The influence of the mountain torque on the EAAM amplitude is examined by regressing both the *p*′_{s} field and the integrands of (A5) and (A6) against the EAAM amplitude tendency. This calculation is performed without the phase shifting of the previous section, otherwise the impact of the mountain torque would be obscured. The regression of the (A5) and (A6) integrands (not shown) indicates that the primary contributors are again Antarctica and Greenland, with much smaller contributions coming from other regions.

## 6. Conclusions

This investigation compares the dynamics of EAAM fluctuations in the atmosphere with those from an aquaplanet GCM (Feldstein 2003), and it examines the impact of the mountain torque on EAAM. The following summarizes the main similarities:

- The atmospheric wave field associated with the 10-day westward rotation of the EAAM vector corresponds to the westward propagation of the first antisymmetric, zonal wavenumber-1 normal mode of the atmosphere.
- EAAM amplitude fluctuations are associated with a tropical disturbance that resembles equatorially trapped, zonal wavenumber-1 mixed Rossby–gravity waves. These equatorial waves are accompanied by the poleward propagation of Rossby waves into midlatitudes.
- Wave–wave interaction in the midlatitude upper troposphere drive EAAM amplitude fluctuations.
- Wave–zonal-mean flow interaction dampens the amplitude of the EAAM vector.

The latter three processes alter the EAAM amplitude via the equatorial bulge torque.

A number of differences were also found between the EAAM dynamics of the aquaplanet GCM and that of the atmosphere. Compared to the aquaplanet GCM, in the atmosphere, the 10-day spectral peak of both EAAM components is weaker and broader, and the resemblance with normal modes is lower. These results were attributed to the mountain torque causing the phase speed of the zonal wavenumber-1 disturbance to fluctuate about its 10-day value. Mountain torques also appear to cause the surface pressure field to take on a less coherent spatial structure than in the aquaplanet GCM. In addition, the wave–wave interactions that drive the EAAM amplitude fluctuations involve mostly planetary waves in the atmosphere, while in the aquaplanet GCM, both planetary- and synoptic-scale waves play a role. These differences hint at the possibility that topography may be exciting the waves that are involved in this wave–wave interaction.

Mountain torque calculations for the boreal winter showed that Antarctica and Greenland are the most prominent topographic features on the earth’s surface for driving EAAM phase and amplitude fluctuations.

One important remaining question involves the determination of the relative strength of the different mechanisms. It needs to be quantified as to whether it is the poleward Rossby wave propagation, wave–wave interaction, or the driving by mountain torques, that is most important for changing the EAAM amplitude. Another open question is the difference between boreal and austral winter intraseasonal EAAM variability. Because Northern Hemisphere processes were dominant in this boreal winter study, especially the wave–wave interactions, it would not be too surprising if the EAAM characteristics of the austral winter would be quite different.

## Acknowledgments

This research was supported by National Science Foundation through Grants ATM 0224870 and 0514034. I would like to thank Dr. Sukyoung Lee for her beneficial conversations and comments. The comments of two anonymous reviewers were also very helpful. I would like to thank the NOAA/Climate Diagnostics Center for providing me with the NCEP–NCAR reanalysis and OLR datasets.

## REFERENCES

Bell, M. J., 1994: Oscillations in the equatorial components of the atmosphere’s angular momentum and torques on the earth’s bulge.

,*Quart. J. Roy. Meteor. Soc.***120****,**195–213.Brzezinski, A., 1987: Statistical investigations on atmospheric angular momentum functions and on their effects on polar motion.

,*Manuscr. Geod.***12****,**268–281.Celaya, M. A., , J. M. Wahr, , and F. O. Bryan, 1999: Climate-driven polar motion.

,*J. Geophys. Res.***104****,**12813–12829.Chao, B. F., 1993: Excitation of earth’s polar motion by atmospheric angular momentum variations, 1980–1990.

,*Geophys. Res. Lett.***20****,**253–256.de Viron, O., , C. Bizouard, , D. Salstein, , and V. Dehant, 1999: Atmospheric torque on the Earth and comparison with atmospheric angular momentum variations.

,*J. Geophys. Res.***104****,**4861–4875.Egger, J., , and K. P. Hoinka, 1999: The equatorial bulge, angular momentum and planetary wave motion.

,*Tellus***51A****,**914–921.Egger, J., , and K. P. Hoinka, 2000: Mountain torques and the equatorial components of global angular momentum.

,*J. Atmos. Sci.***57****,**2319–2331.Egger, J., , and K. P. Hoinka, 2002: Equatorial components of global atmospheric angular momentum: Covariance functions.

,*Quart. J. Roy. Meteor. Soc.***128****,**1137–1157.Eubanks, T. M., , J. A. Steppe, , J. O. Dickey, , R. D. Rosen, , and D. A. Salstein, 1988: Causes of rapid motions of the Earth’s pole.

,*Nature***334****,**115–119.Feldstein, S. B., 2003: The dynamics associated with equatorial atmospheric angular momentum in an aquaplanet GCM.

,*J. Atmos. Sci.***60****,**1822–1833.Hoskins, B. J., , M. E. McIntyre, , and A. W. Robertson, 1985: On the use and significance of isentropic potential vorticity maps.

,*Quart. J. Roy. Meteor. Soc.***111****,**877–946.Kasahara, A., 1976: Normal modes of ultralong waves in the atmosphere.

,*Mon. Wea. Rev.***104****,**669–690.Kasahara, A., 1980: Effect of zonal flows on the free oscillations of a barotropic atmosphere.

,*J. Atmos. Sci.***37****,**917–929.Longuet-Higgins, M. S., 1968: The eigenfunctions of Laplace’s tidal equations over a sphere.

,*Philos. Trans. Roy. Soc. London***262A****,**511–607.Nastula, J., 1995: Short periodic variations of polar motion and hemispheric atmospheric angular momentum excitation functions in the period 1984–1992.

,*Ann. Geophys.***13****,**217–225.Nastula, J., , and R. M. Ponte, 1999: Further evidence for oceanic excitation of polar motion.

,*Geophys. J. Int.***139****,**123–130.Ponte, R. M., , and D. Stammer, 1999: Role of ocean currents and bottom pressure variability on seasonal polar motion.

,*J. Geophys. Res.***104****,**23393–23409.Ponte, R. M., , D. Stammer, , and J. Marshall, 1998: Oceanic signals in observed motions of the Earth’s pole of rotation.

,*Nature***391****,**476–479.Salstein, D. A., , and R. D. Rosen, 1989: Regional contributions to the atmospheric excitation of rapid polar motions.

,*J. Geophys. Res.***94****,**9971–9978.Swinbank, R., 1985: The global atmospheric angular momentum balance inferred from analyses made during the FGGE.

,*Quart. J. Roy. Meteor. Soc.***111****,**977–992.Wheeler, M., , G. N. Kiladis, , and P. J. Webster, 2000: Large-scale dynamical fields associated with convectively coupled equatorial waves.

,*J. Atmos. Sci.***57****,**613–640.

## APPENDIX

### Equations for Torques

_{1}and EM

_{2}are the Greenwich and 90°E components of the equatorial angular momentum vector,

**EM**, respectively. In (A1) and (A2),

*p*is the surface pressure;

_{s}*u*and

*υ*are the zonal and meridional winds relative to the earth’s surface, respectively;

*λ*,

*ϕ*, and

*p*are the longitude, latitude, and pressure, respectively;

*R*is the earth’s radius; and

*g*is the gravitational acceleration. The first (second) term on the right-hand side of (A1) and (A2) is the mass (wind) contribution to EM

_{1}and EM

_{2}.

_{1}, BT

_{2}, MT

_{1}, MT

_{2}, FT

_{1}, and FT

_{2}are the Greenwich meridian and 90°E components of the equatorial bulge, mountain, and friction torques. The variables

*τ*and

_{λ}*τ*are the frictional stresses in the zonal and meridional directions, respectively, and

_{ϕ}*h*is the topographic height.

Std devs of the equatorial bulge, mountain, and friction torques for the Greenwich and 90°E meridians, denoted by BT_{1}, MT_{1}, FT_{1}, and BT_{2}, MT_{2}, FT_{2}, respectively. Linear correlation values are shown for each of the torques with the corresponding component of the EAAM vector tendency.

Linear correlations between the EAAM component time series and the time series generated by projecting the daily *p*′_{s} and *ϕ*′_{300} fields onto the Hough mode spatial patterns. Subscripts 1 and 2 denote the Greenwich and 90°E meridians, respectively, and (*n* = 2) and (*n* = 4) indicate the first two antisymmetric Hough modes.