Major changes in the global distribution of precipitation, water vapor, and clouds during the El Niño–Southern Oscillation (ENSO) have been documented in many previous studies (e.g., Ropelewski and Halpert 1988; Rasmusson and Wallace 1983; Lau and Chan 1988; Prabhakara et al. 1985). Changes in these atmospheric hydrologic processes may in turn influence the development and evolution of ENSO anomalies via radiation–hydrologic cycle feedback processes. The importance of the interaction of large-scale circulation, hydrologic–radiative processes, and sea surface temperature in leading to interannual and long-term changes in the ocean–atmosphere system has been enunciated by many previous investigators (e.g., Webster 1995; Randall et al. 1989; Ramanathan 1987; Stephens and Webster 1979). However, the relative importance of specific dynamical and/or radiative processes and the temporal and spatial scales in which these processes are most applicable in interpreting observations remain a subject of debate. Given that ENSO-related changes involve all major elements of the global water and energy cycles, and that ENSO has been known to occur at intervals of 2–7 yr without interruption since recorded history of the earth’s climate, it is tempting to suggest that ENSO may serve as a natural surrogate for understanding feedback processes for climate change.
A number of recent studies have pointed out that because of the different roles played by large-scale dynamics versus radiative feedback on different spatial and timescales, there may be difficulties in using interannual or seasonal variations as surrogates for climate change (Lau et al. 1994; Lau et al. 1996; Bony et al. 1995; Lindzen et al. 1995). Moreover, ENSO is known to have significant variations over time (Allan et al. 1991; Gu and Philander 1995) so that analysis based on single or limited number of events may not be representative. Therefore, before inferences can be made from ENSO anomalies to climate change, a more fundamental understanding of the physical nature of changes in hydrologic processes during ENSO is required.
Because of the scarcity of in situ observations over the tropical oceans, changes in global atmospheric hydrologic processes associated with ENSO have always relied on model results and/or satellite observations. Portrayal of atmospheric hydrologic processes based on stand-alone satellite observations or model outputs, while extremely useful and instructive, is often incomplete, and dynamical consistency of the results is not guaranteed. As a result of the recent reanalysis efforts, in situ observations, satellite data, and model characteristics can now be blended through sophisticated four-dimensional data analysis (FDDA) to produce research-quality data for climate diagnostic studies. The FDDA products are free from model changes, since the same model and physical parameterizations are used throughout the reanalysis period. More abundant and better quality-controlled input data are now routinely included in the assimilation system. Furthermore, all the operational centers taking part in the reanalysis project have utilized state-of-the-art physical parameterizations and data assimilation systems as a result of years of continual research and development efforts. While these attributes do not guarantee that the FDDA physical parameters necessarily replicate the real world, they are undoubtedly a vast improvement compared to previous products.
The main thrusts of this paper are to document the three-dimensional structure associated with anomalous atmospheric hydrologic processes during ENSO and to diagnose the role of sea surface temperature (SST) forcing and dynamics–radiation interaction in maintaining these anomalies using numerical experiments. The issue regarding sensitivities of water vapor and cloud forcings to anomalous SST forcing and the use of ENSO as surrogate for climate change based on similar experiments have been reported in Lau et al. (1996). All there sults for this work are based on the Goddard Earth Observing System-Data Assimilation System (GEOS-DAS) FDDA data and experiments with its core general circulation model, the GEOS general circulation model. The data and analysis methodology are discussed in section 2. In section 3, we document the changes in the three-dimensional structure of the temperature, water vapor, and circulation fields during ENSO. Section 4 is devoted to the results of numerical experiments to shed light on the mechanisms of dynamics–radiation interaction in maintaining the anomalous atmospheric hydrologic processes during ENSO. In section 5, we present further discussions and interpretation of main findings. The conclusions are presented in section 6.
2. Data and analysis methodology
We use 9 yr of assimilated model outputs from the GEOS-DAS for the present analysis. The GEOS-DAS products have a horizontal resolution of 2° lat × 2.5° long based on a new incremental update procedure, which assimilates satellite data, upper air, and surface data in finite increments distributed throughout the 6-h assimilation cycle. A detailed description of the data products and merits of the new assimilation procedure can be found in Schubert et al. (1993). Currently, the GEOS reanalysis is available for the period 1985–93. During this period, two warm events (1986–87 and 1991–92) and one cold event (1988–89) occurred. In the present analysis the anomaly fields are defined as the difference formed by the mean of the two warm events and the cold event for December of the first year and January–February of the following year. Recent analysis shows that the GEOS clouds and water vapor fields are systematically overestimated in the upper troposphere in the Tropics and underestimated in the extratropics compared to radiosonde observations (Starr et al. 1995). The current focus on the difference fields partially eliminate the bias present in these hydrologic fields.
3. Results from GEOS-DAS reanalysis
a. Basic hydrologic fields
This section presents a discussion of the response of several basic hydrologic fields to anomalous SST forcing. The results in this section will be used as a benchmark for later discussions. Figures 1a–d show the composite SST anomaly pattern based on the warm minus cold events and the corresponding composite fields of precipitation, total precipitable water, and cloudiness. The anomalous SST is similar to that depicted in typical ENSO SST composites, consisting of a large area of positive anomaly with maximum magnitude up to 3°C over the equatorial central-eastern Pacific and a “horseshoe”-shaped negative pattern located over the maritime continent/western Pacific and the subtropics in both hemispheres. The precipitation pattern shows increased rainfall over the central Pacific, about 20° long to the west of the maximum SST anomaly, with large reduction in rainfall in the South Pacific convergence zone and in the tropical western Pacific. The regions of reduced rainfall generally coincide with areas of the negative SST horseshoe, except over the eastern Pacific intertropical convergence zone near 10°–15°N, 140°–100°W, where the rainfall reduction is found over positive SST anomalies. This is a result of increased subsidence generated by the strong rising motion over the equatorial central Pacific. The anomaly patterns of total water vapor and clouds resemble that of the sea surface temperature and precipitation, respectively. This suggests that the total amount of atmospheric water vapor is strongly controlled by the SST and that the cloudiness variability is closely tied to that of the precipitation. The overall patterns of response of rainfall, water vapor, and cloudiness are similar to those observed during ENSO based on satellite outgoing longwave and microwave radiance (e.g., Rasmusson and Wallace 1983; Prabhakara et al. 1985).
Figure 2 shows the horizontal distribution of tropospheric temperature anomaly at different levels of the atmosphere. It is clear that near the surface, the air temperature anomaly pattern closely resembles that of the underlying SST. However, the anomaly pattern departs increasingly from the SST pattern at increasing height. At 600 mb, the temperature anomaly over the equatorial central Pacific splits into two warm cores away from and straddling the equator, while a cold anomaly develops farther poleward. In the Northern Hemisphere, the temperature anomaly assumes a wave train pattern reminiscent of the well-known Pacific–North America teleconnection during ENSO (Horel and Wallace 1981). In the Southern Hemisphere, a cooling belt encircling the extratropics (30°–60°S) can be found. At 300 mb, the warm core over the central and eastern Pacific expands further in latitude and reaches a maximum value of +3°C in the subtropics of both hemispheres, while the cooling pattern over the northern extratropics loses its wavy characteristics and becomes more zonal. The tropospheric anomalies are consistent with those found by Yulaeva and Wallace (1994) using data from the microwave sounding unit. They are dynamically consistent with large-scale circulation anomalies driven by tropical heating as predicted by linear theory (e.g., Webster 1972; Gill 1980; Lau and Lim 1984). As noted in Yulaeva and Wallace (1994), while linear theory indicates that temperature and height patterns should be located to the west of the heat sources, the observed anomalies are found near the same longitudes as the tropical heating. This suggests that advection by the mean flow and possible nonlinear effects may be important in maintaining the large-scale temperature and vorticity balance.
Most interestingly, there is a reversal in the temperature anomaly about the tropopause near 150 mb. At 100 mb (Fig. 2d), the temperature anomaly pattern resembles a mirror image of the anomaly at 300 mb reflected across the tropopause. The inverse relationship between the temperature anomalies in the upper troposphere and lower stratosphere is also consistent withthe rise of the tropopause observed during ENSO (Gage and Reid 1987). Because of the reversal of the temperature lapse rate above the tropopause, the bulging of the tropopause due to the hydrostatic expansion of the warmer air and the rising motion in the core of the convection pushes the colder upper-tropospheric air into the lower stratosphere thus creating a cold anomaly there. The tropopause movement is evident in Fig. 3, which shows the time series of the 300-mb mean temperature between 10°S and 10°N and the mean temperature difference between 300 mb and the mean temperature between 100 and 70 mb averaged between 10°S and 10°N at 150°W. The former can be used as a proxy for the variation of the mean tropopause height and the latter a measure of the strength of the tropospheric–stratospheric dipole. It is obvious that the rise and fall of the mean tropical tropopause (as indicated by the zonal mean temperature at 300 mb) are strongly correlated with the oscillation of the temperature dipole. During the first half of the record (1985–89), the ENSO signal dominates. In the second half (1990–93), the annual cycle dominates. The temperature dipole seems to develop about one season ahead of the zonal mean temperature in agreement with Yulaeva and Wallace (1994). This is plausible, because the temperature dipole is a direct response to the local heating induced by positive SST, while the zonal mean response represents an indirect effect possibly due to temperature advection by the anomalous wind. Similar temperature reversal between the upper troposphere and lower stratosphere and associated variation in tropopause height were also noted in the semiannual oscillation (van Loon and Jenne 1970).
c. Specific humidity
The anomaly patterns of specific humidity at different levels are shown in Fig. 4. The surface and lower-tropospheric water vapor patterns resemble that of the SST anomaly. Increased (reduced) water vapor near the surface and up to the middle troposphere can be found over the warm (cold) SST anomalies. The largest changes are found below 600 mb. At 600 mb (Fig. 4b), the maximum increase in water vapor is found in the region of heaviest precipitation coinciding with the region of maximum ascent (see Fig. 5c). The water vapor anomaly pattern gradually loses its surface signature at increasing altitude. At 300 mb, the anomaly signal is weak with about one-tenth of the magnitude at 600 mb. Above 300 mb (Fig. 4d), negative water vapor anomalies are found in extended region over the subtropics and extratropics. Noteworthy is that the distribution of water vapor at the upper troposphere and lower stratosphere bears little resemblance to that of the corresponding temperature anomalies. Comparing Figs. 2c and 2d to Figs. 4c and 4d, the moisture anomalies have the same sign in the troposphere and the stratosphere, even though the temperature anomalies have the opposite sign. This is also clear in the temperature and moisture anomalies shown in Figs. 5a,b. Therefore, while the temperature field at all levels and water vapor at low levels are strongly controlled by the lower boundary condition, the upper-level water vapor may be governed by other sources including upward transport in the convective core, evaporation from the upper-level clouds, and advection by the large-scale circulation. It should be noted that while the changes of water vapor in the upper troposphere are small in absolute amount compared to those in the lower and middle troposphere, the radiative effect at the surface is much larger per molecule in the upper troposphere (Arking 1991; Ho et al. 1998). The consequence of the upper-troposphere water vapor radiative feedback in contributing to the observed temperature and circulation anomalies will be discussed later.
d. Meridional cross sections
Figures 5a–d show the meridional-height profile of temperature, specific humidity, vertical motion, and total diabatic heating along 150°W, intersecting the region of maximum temperature signal. The effect of positive SST anomaly in the Tropics is reflected in an overall warming of the tropical troposphere in a broad region within 20°–25° of the equator and below the tropopause near 150 mb. The aforementioned double maxima in temperature in the upper-tropical troposphere and their mirror images in the lower stratosphere are quite obvious. In the extratropics of both hemispheres, the influence of the negative SST extends to about 300 mb. Here, a weak reversal in stratospheric temperature can also be noted. The positive temperature anomalies in the troposphere are associated with the increased latent heating from convection generated over the warm water and negative anomalies due to the upward transfer of the negative SST signal by induced extratropical motions.
The vertical section of the specific humidity anomaly (Fig. 5b) indicates moistening of troposphere from 10°N to 20°S, with maximum response near 400 mb within 10° of the equator. The moistening region matches well with that of the induced rising motion (negative p velocity) and the positive diabatic heating shown in Figs. 5c and 5d, respectively. Increase in moisture up to 70%–100% can be found in this region. Strong compensating drying regions can be found in the subtropics of both hemispheres, coinciding with the descending branch of the local Hadley circulation (see Figs. 5c,d). In the Northern Hemisphere, the drying appears to be quite widespread, extending over much of the extratropical troposphere and stratosphere. Also, the extratropical moisture anomaly pattern does not seem to correspond directly with the ascending and descending branches of the meridional divergent cells shown in Fig. 5c but is thermodynamically consistent with the region of negative extratropical temperature anomaly. The results suggest that thermodynamically controlled moist adiabatic processes may be responsible for the water vapor anomalies in the extratropics.
e. Zonal cross sections
The zonal vertical cross section, averaged along 10°S–10°N for temperature, specific humidity, vertical motion, and the diabatic heating fields are shown in Figs. 6a–d, respectively. The large tropospheric temperature anomaly over the positive SST anomaly in the central and eastern Pacific (120°–180°W) is very pronounced. Equally so is the negative temperature anomaly in the lower stratosphere. The maximum tropospheric temperature anomaly (>3°C) is found near 300 mb and the maximum stratospheric anomaly of the same magnitude is found at about 100 mb. A secondary tropospheric–stratospheric temperature dipole can be found over the African and Indian Ocean sector (0°–90°E), where the SST anomaly is weakly negative. This temperature signal is likely to be induced by remote forcing—that is, as a result of the shift in the climatological Walker circulation forced the positive SST anomaly in the central Pacific. Figures 6b and 6c show clearly that the moistening (drying) of the troposphere is found in the ascending (descending) branches of the anomalous Walker circulation. In the main rising branch of the Walker circulation over the central Pacific, the water vapor increase can be as large as 50% of the mean. Over the western Pacific (100°–140°E) and the Amazon (50°–60°W), the reduction of midtropospheric moisture induced by the anomalous descending motion are also quite substantial. However, the overall moisture content in the tropical troposphere is increased.
Because the east–west cross section is entirely within the Tropics, the anomalies in vertical motion and in water vapor are closely correlated to the total diabatic field shown in Fig. 6d. Here, the dominant signals include increased heating over the central Pacific and reduced heating over the western Pacific and the Amazon. In the diabatic heating vertical section, a double maxima at 500 and 200 mb is noted. A decomposition of the diabatic heating into its basic components—that is, convective heating, shortwave and longwave radiative heating, and turbulent heating—reveals that over the central Pacific, the lower peak at 500 mb can be attributed to latent heating in deep convection (Fig. 7a). It is obvious from comparing Figs. 7a and 6d that latent heating accounts for a large fraction of the total tropospheric heating during ENSO. There is relatively small contribution due to shortwave absorption heating confined to the cloud top (Fig. 7b). The longwave heating appears to make a significant contribution near cloud top and near the surface (Fig. 7c) in the region of enhanced precipitation and cloudiness over the equatorial central Pacific. This heating is related to the trapping of longwave radiation by deep clouds and by shallow clouds near the top of the boundary layer. The longwave heating at about 200 mb accounts for the double heating maximum in the total diabatic heating field. The turbulent heating contribution is entirely negligible (Fig. 7d). In the subsiding branch of the anomalous Walker circulation, the latent heating and longwave heating are both negative. Therefore the positive tropospheric temperature anomalies must be maintained by the adiabatic warming from the descending motion. The above features are clearly dependent on the model cumulus parameterization and on the radiation scheme used. Therefore these profiles should be regarded merely as heating fields that are dynamically consistent with the assimilated temperature and wind fields by the model. As such, their ability to replicate the real atmosphere need to be validated by independent data.
4. Model simulations
The preceding analysis is based on a limited sample of warm and cold events over a short period when other climatic events also took place. For example, the eruption of Mt. Pinatubo in June 1991 may have an impact on the temperature variation during the 1991–92 ENSO. The stratospheric temperature may be affected by the alternating phases of the quasi-biennial oscillation and sudden warmings that occurred during the data period. Hence there may still be some remaining doubt whether or not all the afore-discussed features in the temperature and moisture fields are related to the ENSO anomalous SST forcing. In this section, we will discuss results of numerical experiments using the GEOS-1 GCM to ascertain that these features are indeed forced by SST anomalies and to delineate the role of dynamics–radiation interaction in leading to anomalous hydrologic processes during ENSO.
a. SST-induced anomalies
To obtain the model response to the prescribed SST anomaly, we have carried out a 90-day integration of the GEOS-1 GCM with fixed SST and insolation for climatological January condition. After about 20 simulated days of integration, the model approaches quasi equilibrium. The mean fields for the last 70 days of the integration are used as the control. A second experiment is performed as in the control but with the SST anomaly field (as shown in Fig. 1a) superimposed on the climatological January SST. This experiment contains full physics, including interactive radiation with the hydrologic cycle and shall be designated as FP (full physics) for later reference. The difference between the last 70-day mean of FP and the control experiments will then represent the response of atmosphere to the prescribed anomalous SST forcing. These difference fields will be compared to those shown in the preceding analysis.
Examination of a large number of difference fields have confirmed that major features discussed in the preceding sections are truly due to the ENSO SST forcing. For brevity, only the temperature (Fig. 8) and moisture fields (Fig. 9) are discussed here. The similarity between Figs. 8 and 2 is quite remarkable, both in pattern and in the magnitude of the response. Features such as the upper-tropospheric double temperature maxima near 120°W and the dateline, and the tropospheric–stratospheric dipole are well simulated in FP compared to the GEOS-DAS reanalysis. Significant differences exist between Figs. 8 and 2, particularly in regions remote from the tropical SST anomalies, that is, at upper levels and in the extratropics for both hemispheres. For example, the pattern over the Eurasian continent between 30° and 60°N, at 100 mb is quite different between FP and the reanalysis. This feature may be due to the presence of low-frequency atmospheric transients in the reanalysis and in the model simulation, which are not necessarily related to the ENSO SST forcing.
Comparing Figs. 9 and 4, the similarity in the anomaly pattern of atmospheric water vapor in the lower to middle troposphere between FP and the reanalysis is also quite remarkable. The concentration of the tropospheric moisture source in the equatorial central Pacific covering the entire troposphere and the horseshoe region of widespread drying across the northern and southern Pacific anchoring over the maritime continent are all well simulated. Again, notable differences between the simulation and the reanalysis are found at upper levels and remote from the source region. The simulated moisture field at 100 mb appears to have more contrast and more extensive moist and dry regions than those in the reanalysis, reflecting the effect of an excessive hydrologic cycle. In spite of this, the GEOS-1 GCM simulation is quite successful in reaffirming that the major features described in the previous sections are in direct response to ENSO SST forcing.
b. Hydrologic cycle–radiation interaction
Having established the validity of the stand-alone GEOS-1 climate model in producing ENSO anomalies, we can now proceed to explore the role of interaction between radiation and the hydrologic cycle in producing and maintaining the ENSO anomalies. For this purpose, we have carried out a third experiment, which is identical to FP but differs in that all the radiation fields are decoupled from the hydrologic cycle. Specifically, the radiative forcing terms are kept at the climatological values at every time step as in the control, while the model is forced by the same prescribed ENSO SST anomaly. This experiment shall be referred to as NRA (no radiation anomaly). In NRA, while clouds and water vapor are allowed to change, they are completely passive to the changes in the large-scale circulation. Comparison between NRA and FP will provide an assessment of the role of interaction between radiation and the circulation in maintaining the atmospheric hydrologic anomalies during the ENSO.
c. Meridional cross sections
Figures 10a–d show the 150°W meridional-height cross sections of anomaly fields of temperature, specific humidity for FP minus control and NRA minus control, respectively. The former provides a reference for the simulated ENSO anomalies and the latter indicates the contribution from dynamics alone in generating the anomalies. For brevity, these difference fields shall be referred to as FP and NRA. It also follows that the difference between FP and NRA represents the contribution from dynamics–radiation interaction. Again, the similarity of Figs. 10a and 10b to Figs. 5a and 5b should be noted. As mentioned previously, compared to the reanalysis, the stand-alone model appears to produce deeper tropical convection and therefore strong temperature and moisture anomalies in the upper troposphere and lower stratosphere. When the radiative heating is removed (NRA), the gross patterns in the temperature and moisture patterns are still similar to FP. In the temperature field, the stratospheric–tropospheric dipole remains quite prominent in NRA but with a substantial reduction in amplitude. Notable in NRA is the capping of the negative temperature in the tropical and subtropical stratosphere below 50 mb, above which large warming is found. This is in contrast to FP, where the stratospheric cooling extends up to 20 mb. In NRA, there is a cooling region at 50–70 mb, near 30°–40°N, which is not present in FP. This negative temperature anomaly seems to be correlated with anomalous dryness in the same region (Fig. 10d). In the troposphere, the cooling of the subtropics and midlatitudes is much stronger and extensive in FP than in NRA. In FP (Fig. 10b), the drying and moistening of the atmosphere appear to be confined to well-defined meridional cells that encompass the entire troposphere capped by an elevated tropopause above and near 100 mb. In NRA (Fig. 10d), the moisture anomaly pattern has less definition and contrast, and the vertical scales of the moist and dry areas are much reduced. Also, there appears to be more resident moisture in the upper troposphere and stratosphere at all latitudes in NRA than in FP.
The difference in the temperature and moisture patterns in FP and NRA can be explained in terms of weakening of the SST-induced local meridional overturning due to the suppression of interactive radiative heating. This is illustrated in Figs. 11a,b, which show the meridional profile of the vertical velocity and diabatic heating in the middle troposphere. While the position of the maximum tropical rising motion (Fig. 11a) remain unchanged, the magnitude is reduced substantially as a result of the removal of interactive radiative heating. Also noted is a narrowing of the rising motion to the Tropics, with the region of increased subsidence shifting equatorward. These features are consistent with an enhancement of the local Hadley-type overturning. In the core rising motion of the equatorial region, the rising motion is reduced by more than 40% in NRA compared to FP. This is accompanied by an even more dramatic reduction in the total diabatic field where the maximum diabatic heating rate in the core convective region is reduced from FP (∼8°–9°C day−1) to NR (∼3°–4°C day−1). The reduced local Hadley circulation may be related to changes in stability of the tropical atmosphere induced by radiation-dynamic feedback (see later discussion in section 5). These results are consistent with those reported in previous GCM experiments under more idealized conditions (Randall et al. 1989).
d. Zonal cross sections
The effect of interactive radiation in affecting the equatorial Walker circulation can be seen in Figs. 12a–d, which show the longitude–height cross section of the temperature and moisture anomalies averaged between 10°S and 10°N for FP and NRA, respectively. The pronounced stratospheric–tropospheric temperature reversals across the entire equatorial belt, with centers over the central Pacific (150°W), the Amazon (60°W), and the maritime continent (120°E) are well simulated compared to Fig. 6. Removing dynamics–radiation feedback does not alter the geographic centers of these temperature reversals, but substantially weakens them (Fig. 12c). The suppression of radiative feedback also leads to large warming of the stratosphere above 50 mb, with centers that correspond well with the tropospheric warmings. As a result, the SST-induced stratospheric cooling is mostly capping below 50 mb. In FP, well-defined drying and moistening zones are induced in the equatorial zonal plane with the upper troposphere–lower stratosphere moistening capped by a dry layer near 50–60 mb (Fig. 12b). In NRA (Fig. 12d), the entire troposphere and stratosphere appears to be more moist and the drying regions are much diminished. As in the case for the meridional turning (see Fig. 11), the aforementioned zonal structural changes are related to the spinning up of the Walker circulation by interactive radiation. This is obvious in the zonal profile of the midtropospheric vertical motion and diabatic heating for FP and NR, respectively, shown in Figs. 13a,b. Here, similar to the meridional circulation anomalies shown in Figs. 11a,b, the magnitude of vertical motion is reduced by about 40%–50% in the ascending region when the interaction between radiation and dynamics is excluded. Almost identical signature is found in the zonal profile of diabatic heating (Fig. 13b).
Our results have so far indicated that (a) the characteristic temperature and moisture signals found during ENSO are basically generated by dynamical response of the atmosphere to SST anomalous forcing and (b) the interaction between circulation and radiation tends to enhance the anomalies, maintaining a more vigorous hydrological cycle associated with enhanced Hadley and Walker circulations. We have hinted that induced changes in the stability of the tropical atmosphere may be responsible for the feedback processes. The section is devoted to a more detailed analysis aiming at further understanding of the underpinnings of the dynamic–radiation interaction. Figure 14 shows the vertical profiles of temperature and moisture anomalies for FP and NRA averaged over the tropical region from 10°S to 10°N and from 120°W to the dateline. In the absence of dynamics–radiation feedback (NRA), latent heating from increased deep convection causes the entire troposphere (below 150 mb) to warm up. The heating leads to a rise in the tropopause that brings the tropospheric air into a warmer environment because of the temperature reversal above the tropopause. This is the basic reason for the inverse temperature anomaly in the lower stratosphere. As more heat and moisture are transported upward by vertical ascent in the core of the deep convection, and in the absence of compensating radiative cooling (refer to NRA profile in Figs. 14a and 14b), this moisture and heat are deposited and trapped in the stratosphere. The excess heat gives rise to the large positive temperature anomaly in the upper stratosphere (<50 mb). The effect of radiative–dynamic feedback is to allow increased radiative cooling to space at the top of the atmosphere. This results in net cooling and drying in the entire stratosphere and the upper troposphere, amplifying the dynamically induced anomalies in the lower stratospheric (refer to FP profiles in Figs. 14a and 14b). The cooling and drying near the upper troposphere will make the entire atmospheric column more unstable, favoring more convection. The increased cloudiness and tropospheric moisture leads to further warming of the lower troposphere and cooling above. This is evident in Fig. 15, which shows the vertical profile of longwave and shortwave radiation heating for FP. Note in NRA that there are no radiation anomalies. While the shortwave and the longwave radiation tend to have large compensation, the net effect is a warming of the lower troposphere and cooling of the stratosphere. This provides a positive feedback leading subsequently to a new climate equilibrium with a more vigorous hydrologic cycle as indicated by the enhanced latent heating (FP-NRA) in Fig. 15. Notice that the latent heating difference has a maximum in the midtroposphere and that above 100 mb all the cooling is due to long waves. This new climate equilibrium is manifested in a warmer and more moist troposphere, but a cooler and dryer stratosphere as already evident in the differences in the temperature and moisture profiles of FP relative to NRA (Figs. 14a,b). The effect of the radiation–dynamic feedback to render the tropical atmosphere more unstable locally is also evident in the change in the moist static stability, ∂θe/∂p, where θe is the equivalent potential temperature (see Fig. 16). The more negative values of ∂θe/∂p for FP in the middle troposphere and the lower stratosphere indicate that the atmosphere there is more unstable and therefore more conducive to spontaneous development of deep convection as a result of radiation–dynamics feedback. The region wherethis feedback has the largest impact is near cloud tops at the levels of the tropopause and the lower stratosphere. In contrast, radiative–dynamic feedback appears to have minimal impact on atmospheric moist stability in the lower troposphere, as evident in the nearly identical ∂θe/∂p between FP and NRA in Fig. 16. This is reasonable, because in the lower troposphere, the moist stability is governed by the tight coupling between sea surface temperature and the atmospheric boundary layer.
Using the GEOS-DAS reanalysis data we have documented the basic features of temperature and water vapor response to ENSO SST forcing. We find that while the temperature and moisture anomalies near the surface and in the lower troposphere are thermodynamically coupled to SST anomalies, the largest temperature and moisture responses are found in the upper troposphere and lower stratosphere. These patterns are dynamically generated by anomalous SST forcing. The temperature pattern features a pair of subtropical maxima straddling the equator, consistent with Rossby-type circulation response to tropical heating. Most interestingly, the tropospheric warming is coupled to comparable cooling in the stratosphere over the entire Tropics with strongest coupling in regions of deep convection. The troposphere–stratosphere temperature dipole arises as a result of the reverse climatological temperature gradient in the stratosphere, and the rising of the tropopause due to vertical motion from deep convection thereby bringing the colder tropospheric air from below the tropopause to the warmer lower stratospheric environment. The aforementioned three-dimensional pattern constitutes a very robust structure in the atmospheric temperature response to ENSO SST forcing. Additionally, we find that there is delayed warming of the entire Tropics with about 3–4 months lagging the establishment of the temperature dipole. The moisture distribution indicates large-scale moistening of the entire troposphere in the Tropics and large-scale drying in the subsidence branch of the Hadley and Walker circulations.
Numerical experiments with the GEOS-1 GCM have unraveled a possible role of radiation–dynamic feedback on the maintenance of the temperature and moisture anomalies during ENSO. It is found that while atmospheric dynamics are principally responsible for the generation of temperature and moisture anomalies observed during ENSO, radiative feedback plays a significant role in their enhancement and maintenance. The role played by radiative–dynamics interaction is most important in modifying the moist static stability of the entire atmospheric column, as indicated in the schematic drawing in Fig. 17. In the absence of radiation–dynamics feedback (Fig. 17a), a tropospheric–stratospheric temperature dipole can still be maintained by the bulging of the tropopause due to hydrostatic expansion and vertical motions induced by the SST anomaly. Absent compensating radiative cooling, the upward heat flux from induced deep convection causes an accumulation of heat leading to excessive warming in the upper stratosphere. Dynamics–radiation interaction leads to stronger radiative cooling in the stratosphere and upper troposphere and longwave warming below (Fig. 17b). The net effect is to make the atmospheric column more unstable, favoring more and deeper convection. The enhanced convection reinforces the initial temperature and water vapor anomalies via enhancement of the atmospheric hydrologic cycle. The enhanced hydrologic cycle then leads to more warming and moistening of the middle and lower troposphere, and extends the cooling and drying in the stratosphere to its upper regions. The increased heat transport by convection into the upper troposphere and stratosphere in the new equilibrium climate is compensated by increased radiative cooling, thus removing the excessive heat buildup in the stratosphere.
The present results suggest that the observed large temperature and moisture anomalies in the upper troposphere and lower-to-middle stratosphere during ENSO may be a result of the aforementioned dynamics–radiation interactive processes. In some respect, the appearance of the tropospheric–stratospheric temperature dipole is similar to that found in GCM experiments for CO2 doubling (Held 1993). However, the detailed feedback processes, especially the sensitivity to sea surface warming described in this paper, should not be generalized to global warming scenarios. This is because the radiation–dynamics feedback associated with atmospheric hydrologic processes is strongly dependent on the nature of the large-scale circulation and sea surface temperature anomalies. As indicated by the recent work of Lau et al. (1996) and of Bony et al. (1995), the sensitivities of water vapor and cloud forcings to sea surface temperature are strong functions of the large-scale vertical motion fields. Climate sensitivity estimates based on interannual variability such as ENSO are likely to be very different from those for global climate change. Knowledge of the circulation regimes and dynamics–radiation feedback processes are extremely important in interpreting climate feedback mechanisms for different climate change scenarios.
This work is supported by the Earth Observing System Interdisciplinary Investigation Program and the Physical Climate and Modeling Program of the NASA Mission to Planet Earth Office. The second author, C.-H. Ho, wishes to thank the Korean Research Foundation, which provided partial support for this research.
Allan, R. J., N. Nicholls, P. D. Jones, and I. J. Butterworth, 1991: A further extension of the Tahiti-Darwin SOI, early ENSO events and Darwin pressure. J. Climate,4, 743–749.
Arking, A., 1991: The radiative effect of clouds and their impact on climate. Bull. Amer. Meteor. Soc.,72, 795–813.
Bony, S., J.-P. Duvel, and H. Le Treut, 1995: Observed dependence of water vapor and clear sky greenhouse effect on sea surface temperature: Comparison with climate warming experiments. Climate Dyn.,11, 307–320.
Gage, K. S., and G. C. Reid, 1987: Longitudinal variations in tropical tropopause properties in relation to tropical convection and ENSO events. J. Geophys. Res.,92, 14197–14203.
Gill, A. E., 1980: Some simple solutions of heat-induced tropical circulation. Quart. J. Roy. Meteor. Soc.,106, 447–462.
Gu, D., and S. G. H. Philander, 1995: Secular changes of annual and interannual variability in the Tropics during the past century. J. Climate,8, 864–876.
Held, I. M., 1993: Large-scale dynamics and global warming. Bull. Amer. Meteor. Soc.,74, 228–241.
Ho, C.-H., M.-D. Chou, M. Suarez, K.-M. Lau, and M. Yan, 1998: Comparison of model-calculated and ERBE-retrieved clear-sky outgoing longwave radiation. J. Geophys. Res., in press.
Horel, J. D., and J. M. Wallace, 1981: Planetary scale atmosphere phenomena associated with the Southern Oscillation. Mon. Wea. Rev.,109, 813–829.
Lau, K. M., and H. Lim, 1984: On the dynamics of equatorial forcing of climate teleconnections. J. Atmos. Sci.,41, 161–176.
——, and ——, 1988: Intraseasonal and interannual variations of tropical convection: A possible link between the 40–50 day oscillation and ENSO? J. Atmos. Sci.,45, 506–521.
——, C. H. Sui, M. D. Chou, and W. K. Tao, 1994: An inquiry into the cirrus cloud thermostat effect for tropical sea surface temperature. Geophys. Res. Lett.,21, 1157–1160.
——, C. H. Ho, and M. D. Chou, 1996: Water vapor and cloud feedback over the tropical oceans: Can we use ENSO as a surrogate for climate change? Geophys. Res. Lett.,23, 2971–2974.
Lindzen, R., B. Kirtman, D. Kirk-Davidoff, and E. K. Schneider, 1995: Seasonal surrogate for climate. J. Climate,8, 1681–1684.
Prabhakara, C., D. A. Short, and B. E. Vollmer, 1985: El Niño and atmospheric water vapor: Observations from Nimbus 7 SMMR. J. Climate Appl. Meteor.,24, 1311–1324.
Ramanathan, V., 1987: Atmospheric general circulation and its low frequency variance: Radiative influences. J. Meteor. Soc. Japan,65, 1512–1576.
Randall, D. A., Harshvardhan, D. A. Dazlich, and T. G. Corsetti, 1989: Interactions among radiation, convection and large scale dynamics in a general circulation model. J. Atmos. Sci.,46, 1943–1969.
Rasmusson, E. M., and J. M. Wallace, 1983: Meteorological aspects of the El Niño/Southern Oscillation. Science,222, 1195–1202.
Ropelewski, C., and J. Halpert, 1988: Global and regional precipitation pattern associated with the El Niño–Southern Oscillation. Mon. Wea. Rev.,115, 1606–1626.
Schubert, S. D., R. B. Rood, and J. Pfaendtner, 1993: An assimilated dataset for earth science application. Bull. Amer. Meteor. Soc.,74, 2331–2342.
Starr, D. O’C., B. C. Diehl, A. R. Lare, and B. J. Soden, 1995: Upper tropospheric humidity in the 5-yr GEOS Assimilation. Proc. Workshop on the GEOS-1 Five-Year Assimilation, Washington, DC, NASA, 75–79.
Stephens, G. L., and P. J. Webster, 1979: Sensitivity of radiative forcing to variable cloud and moisture. J. Atmos. Sci.,36, 1542–1556.
van Loon, H., and R. L. Jenne, 1970: On the half-yearly oscillation in the Tropics. Tellus,4, 391–398.
Webster, P. J., 1972: Response of the tropical atmosphere to local, steady forcing. Mon. Wea. Rev.,100, 518–541.
——, 1995: Role of hydrological processes in ocean–atmosphere interactions. Rev. Geophys.,32, 427–476.
Yulaeva, E., and J. M. Wallace, 1994: The signature of ENSO in global temperature and precipitation fields derived from the microwave sounding unit. J. Climate,7, 1719–1736.