On a Weakening of the Walker Circulation

by Ray Pierrehumbert and Rasmus Benestad

Second article of our 3-part series on atmospheric circulation and global warming

In Part I we outlined some general features of the tropical circulation, and discussed ways in which increases in anthropogenic greenhouse gases might affect El Niño. Now we take up the question of how global warming might affect the quasi-steady east-west overturning circulation known as the Walker Circulation. The Walker circulation affects convection and precipitation patterns, the easterly Trade Winds, oceanic upwelling and ocean biological productivity; hence, changes in this circulation can have far-reaching consequences. It also provides the background state against which El Niño events take place, and so changes in the Walker circulation should form an intrinsic part of thinking about how global warming will affect El Niño. In a paper that recently appeared in Nature, Vecchi, Soden, Wittenberg, Held, Leetmaa and Harrison present intriguing new results which suggest that there has already been a weakening of the Walker circulation in the past century, and that the observed changes are consistent with those expected as a response to increases in anthropogenic greenhouse gases. The discussion in Vecchi et al. also raises some very interesting issues regarding the way the hydrological cycle might change in a warming world.

1. The main result of the paper

Vecchi et al. compared the observed trend in the Walker circulation between 1861 and 1992 to that yielded by simulations from the GFDL CM2 general circulation model, run with and without anthropogenic forcing. The comparison was done on the basis of surface pressure, because the humble barometer is a simple instrument which, with low technology, can nonetheless yield very accurate results; hence there are good long term instrumental records from the logs of intrepid tropical mariners. Instrumentation for accurately measuring winds only came along later, to say nothing of instrumentation for monitoring convection patterns or the subtle circulation aloft. Surface pressure provides a good proxy for the Walker circulation because, near the Equator, winds tend to flow from regions of high pressure to regions of low pressure, under the acccelerating action of pressure gradient forces. In the Walker circulation, the low pressure is in the West Pacific and the high pressure is in the East Pacific. This gradient strengthens the Easterly Trade Winds to the east of the rising branch (above the low pressure cell) and counters the Easterly Trades on the west side of the low pressure cell, weakening them or even turning them into westerlies. (See the Walker cell sketch in Part I ) .

The following figure shows maps of the observed and modeled pressure changes between 1861 and 1992. The observations (upper left panel) show an increase of pressure in the Western Pacific and a decrease of pressure in the Eastern Pacific, indicating a weakening of the east-west pressure gradient associated with the Walker circulation. The model simulations (upper right panel) driven by all known climate forcings over the period in question show a very similar pattern of weakening. The bottom two panels demonstrate that this weakening is due entirely to the anthropogenic forcings — greenhouse gas increases offset by sulfate aerosol effects. The simulations shown are the mean of an ensemble of five simulations of the period starting with slightly different initial conditions.

Most press reports summarized this result as a "weakening of the Trade Winds" in response to global warming. As a description, that’s not too bad, given that the indicated trend in the Walker circulation does indeed lead to a weakening of the Trades over most of the Pacific. However, the Trade Winds are primarily caused by the Hadley circulation, and are only modulated by the Walker circulation, so it is more precise to think of this result as indicating a change in strength of the Walker circulation.

2. Circulations and the hydrological cycle in a warming world

One of the things that makes the findings of Vecchi et al. especially interesting is that they are consistent with some rather robust theoretical arguments linking the strength of circulations to certain aspects of changes in precipitation and water vapor. These arguments are explained in more detail in Held and Soden 2006 (preprint available here). The argument begins by noting that the Clausius-Clapeyron equation, predicts a strong increase of boundary layer water vapor content with temperature (about 7% per degree of warming); the increase of low level water vapor with temperature is not controversial, since oceanic boundary layers are in contact with their moisture source and stay rather near saturation. Observations tend to support the expected increase (e.g. Wentz and Schabel Nature 403, January 2000. Some of the observational references given by Vecchi et al. in support of the increase actually deal more with the extratropics than the tropics, but the general principle is not seriously in question.) One might then expect that there would be a precipitation increase in proportion to the increase in water vapor content. However, it has been known since the earliest general circulation simulations by Manabe that as the Earth warms in response to increasing CO₂, the precipitation increases much more slowly than Clausius-Clapeyron would suggest — typically only 2-3% per degree of warming. Because latent heat release in the course of precipitation must be balanced in the global mean by infrared radiative cooling of the troposphere (over time scales at which the atmosphere is approximately in equilibrium), it is sometimes argued that radiative constraints limit the rate at which precipitation can increase in response to increasing CO₂. This argument is stated, for example, in Allen and Ingram.and repeated in Vecchi et al. The argument isn’t actually as firm a constraint as generally believed, since the infrared radiative cooling of the atmosphere is affected by the temperature difference between air and the underlying surface, which can adjust to accommodate any amount of evaporation Nature wants to dump into the atmosphere (as shown in Pierrehumbert 1999 ("Subtropical water vapor…" available here)). This is why single-column radiative convective models can show stronger increases of precipitation with temperature, even approaching the Clausius-Clapeyron limit. However, the relatively weak increase in precipitation with temperature seen in general circulation models is robust across models, suggesting that with suitable additional conditions the argument given in Allen and Ingram can be made to work.

Taking the slow increase of precipitation with temperature as a given, the more rapid increase of boundary layer humidity implies that the rate of transport of moisture from the boundary layer to higher levels where it rains out must go down. One way to do this is to decrease the strength of large scale circulations like the Walker and Hadley circulations.

If this argument seems obscure, here’s an analogy that may prove helpful. Suppose you live on a tropical island where water must be brought in buckets to your hut by the local authorities. The size of the buckets (which are always full) stands in for the boundary layer water vapor content. You need the water because you raise pigs, which are very temperature sensitive; to keep them comfortable you need to throw a certain amount of water over them each day. The amount needed per day is determined by the temperature, via aspects of pig physiology we need not go into. The amount of water per day you need to dump on your pigs stands in for precipitation. The standard bucket is one gallon, and at a normal temperature you need to throw four gallons per day over your pigs, meaning you have to rouse yourself four times per day and go out to empty your buckets. Now (parbleu!) global warming strikes, and it is two degrees warmer. By the physiology of pigs, you now need to dump eight gallons of water per day over your pigs. However the government has gone overboard and passed the Clausius-Clapeyron law, which mandates that each pig farmer now gets four-gallon buckets when the temperature gets two degrees warmer. That means that you now only need to get up twice a day (i.e. half as often) to throw water over your pigs. The rate at which you have to go dump water over the pigs (oh happy pigs!), which also equals the rate at which the local authorities must come refill your buckets, is analogous to the strength of the moisture-transporting atmospheric circulation.

Note that the above argument only shows that the rate of moisture exchange between the boundary layer and the free troposphere should decrease. This does not prove that the large scale circulation itself must decrease, for the moisture exchange consists of a small scale convective mass flux as well as a portion due to the large scale circulation. It is only the combination of the two that must become more sluggish. Even if the net rate of moisture exchange were to remain fixed, the Walker circulation could still become stronger or weaker, if the circulation reorganized itself to put less or more of the exchange in the form of small scale convective motions. The factors governing this partitioning remain to be elucidated.

3. Caveats and other viewpoints

The biggest caveat leaps out at the reader upon examining the upper panel of Figure 3 in Vecchi et al. In this figure it is evident that the observed trend in pressure gradient is almost entirely due to a precipitous drop in the late 1970s , which persisted through most of the 1990s. This shift coincides with an apparent 1976-77 climate shift in the character of ENSO, the attribution of which to global warming has been much debated. It’s not evident why the smooth trend in 20th century climate forcing should give rise to such an abrupt shift, and indeed the individual members of the model ensemble do not show a clearly analogous shift. Comparison of individual model runs with the observations is further complicated by the large decadal variability of the simulations. It remains disconcerting that the whole trend appears to rest on a meager handful of anomalous individual El Niño events. On the other hand, there is no reason to believe that the Walker circulation should change smoothly as a function of climate forcings; perhaps the potential for change builds up over many years, and manifests itself all of a sudden, in the fashion of an avalanche. Backing up this speculation with equations is a challenge for the future. Another decade or two of data will greatly clarify the situation.

Reconciling the picture in Vecchi et al with other analyses of climate change in the late 20th century also poses some difficulties. In particular Cane et al (Cane, M. A. et al. (1997). "Twentieth-Century Sea Surface Temperature Trends." Science 275: 957-960) have suggested that the upwelling of cold water in the Eastern Pacific provides a kind of thermostat which keeps the Eastern waters from warming as much as the Western warm pool waters. (Their result must be treated with some caution, since it doesn’t enforce the top of atmosphere balance, and should disappear in the long term after the water tapped for upwelling begins to warm; still the idea has a lot of merit in the transient warming situation we are now in.). This suggests an intensification of east-west sea surface temperature gradients, which ordinarily ought to yield a strengthening in the Walker circulation. In fact, Cane et al (1997) argue that the tendency toward increased SST gradient is precisely what is seen if one uses a robust trend analysis to decrease sensitivity of the trend analysis to outliers such as the very large 1982/1983 El Nino event (this event, and the equally large 1997/1998 El Nino event, greatly influence the estimate of a weakening trend of the Walker circulation in Vecchi et al). Hoerling and Kumar (Hoerling, M. and A. Kumar (2003). "The Perfect Ocean for Drought." Science, 299: 691-694 ) find suggestions of a similar pattern to that argued by Cane et al. Then too, it should be remembered that Vecchi et al present results only for one model, whereas there is evidence that the anthropogenic changes in tropical circulation can be model-dependent (Collins et al (2001), Climate Dynamics, 17: 61-81).

So, this paper probably shouldn’t be seen as a smoking gun for global warming, nor should confirmation of the results in the paper be seen as a crucial test for global warming theory in general. For the most part, the somewhat speculative nature of the results are in the nature of the data itself, and will only be resolved by another decade of observations. Nonetheless, Vecchi et al provides an important harbinger of what may be in store for the Tropics, reminding us once more that there is more to climate than just temperature and precipitation. We will be watching future trends in the Walker circulation with bated breath.