Spontaneous clumping of tropical cloudsGOES west geostationary satellite image from a few days ago), you’ll notice that convective clouds (the tall thunderstorm clouds associated with strong circulations) tend to be organized into clusters. This clustering ranges from features such as squall lines and tropical cyclones to planetary scale phenomena such as the Madden-Julian Oscillation. A large fraction of tropical cloudiness and rainfall is associated with these organized clusters, so it is important that we understand how tropical clouds form clusters.
I’ve spent most of the last five years working on one specific type of cloud organization called “self-aggregation.” Self-aggregation is the tendency of tropical clouds to spontaneously clump together, solely due to interactions between the clouds and the surrounding environment. It occurs in numerical model simulations of an imaginary patch of the tropical atmosphere, in which the model starts off with the same temperature and moisture everywhere over a uniform, fixed temperature ocean surface. To see an example of one of these simulations, watch the movie below. The white shading indicates clouds and the colors indicate rainfall. Make sure to watch the whole thing; the most exciting part is about 1 minute into the movie.
Quickly after the simulation begins, clouds and rainfall start to form. In the beginning of the simulation, the clouds occur randomly throughout the domain. Sometimes it stays like this forever, but in this case, in some areas of the domain the air starts to get drier and in other areas the air starts to get moister. As this happens, the clouds and precipitation are increasingly confined to the moistest regions, and help them become even moister. Eventually, all the clouds and rainfall can become confined to a single cluster: this is self-aggregation. This transition from random to clustered clouds is not just a reorganization; the average state of the whole domain is greatly altered, which makes the process of self-aggregation potentially important for regulating tropical climate.
From my work and the work of several other research groups, we know that physical processes that dry the drier air columns and moisten the moister air columns are critical for self-aggregation. It is a bit like the rich get richer and the poor get poorer, but regarding humidity, not money. But there are still a lot of things we don’t know about the specifics, one of which is whether the mechanisms that cause self-aggregation are the same across different types of model setups. This is the topic of a new study I did in collaboration with Timothy Cronin, a postdoc at Harvard University, which is titled “Self-aggregation of convection in long channel geometry” and was recently accepted for publication by the Quarterly Journal of the Royal Meteorological Society.
In our study, we perform simulations with a “cloud-resolving” model, which has a high enough resolution to explicitly resolve clouds without having to approximate them like global climate models do. We use a long, narrow 3D channel domain; a horizontal slice of this domain looks like a bowling alley. After letting the model run for a few dozen days, we find pretty dramatic results: The convection organizes into bands perpendicular to the long-axis of the channel, which have high relative humidities and contain most of the clouds and precipitation, and alternate with bands of dry, mostly clear air. This is self-aggregation, and you can watch a mesmerizing animation of it here. In order to fit the long channel into one movie, its length is divided into quarters which are then plotted one beneath the other, where each segment is wrapped left-to-right as if it were lines of text.
The separation of convection into multiple moist and dry bands occurs whether we set the sea surface temperature to be very cold (we tried as cold as 7°C — 44.6°F) or very warm (we went as high as 37°C — 98.6°F), which is a new result. A big advantage of the channel geometry is we actually get multiple bands of clouds, unlike previous studies that used square domains and only ever got one cluster. This means that we can define the distance between cloudy, rainy regions and therefore determine the length scale of self-aggregation. This length scale is on the order of 1,000 km, although it does vary with temperature (generally smaller at higher temperature). Now that we know what the length scale is, we can work towards determining what controls it, and in the paper we propose one theory based on a simple model of moistening of the air near the surface as it moves away from the center of a dry region. I won’t try to explain the simple model here because it explains some, but not all, of our results, so this particular question remains largely unanswered.
We have a better handle on the physical mechanisms that cause the self-aggregation into moist and dry bands. We apply an analysis framework introduced in one of my previous papers to quantify the contributions of different processes to self-aggregation. We quantify these processes by describing them as “feedbacks,” which is when something changes in a system and a process responds to that change in a way that enhances the original change (a positive feedback) or counteracts the original change (a negative feedback). In climate science, the term “feedback” is often used with regards to the global surface temperature, but that is not what we mean here. When we refer to “feedbacks,” we are referring to feedbacks on self-aggregation. In our framework, a positive feedback is one that pushes the model towards aggregation by amplifying humidity anomalies and causing the spatial variance of humidity to increase (a bigger difference between the humidity in the dry and moist regions).
Which feedbacks are the most important for self-aggregation, for making the dry regions drier and the moist regions moister? As others have, we find that feedbacks involving surface fluxes of heat and evaporation from the ocean surface to the atmosphere and feedbacks involving radiative heating generally favor aggregation. Since this is similar to the results found in other studies, it suggests that the aggregation we have in our channel model is the same phenomenon as in the other studies where the details were different — an important step to figuring out if self-aggregation is relevant to the real world.
How do these feedbacks work? In the moist regions, there are stronger winds, which cause stronger fluxes of heat from the ocean to atmosphere. Also in the moist regions, there are more clouds and water vapor, which increase the amount of solar radiation absorbed by the atmosphere and reduce how much the atmosphere cools via longwave radiation. These processes tend to further increase the humidity of the moist regions and therefore all act as a positive feedback on self-aggregation. However, these feedbacks don’t always work the same way; the surface flux feedbacks only help aggregation at the very beginning of the simulation, and in simulations at very cold temperatures the radiative feedbacks behave differently because of the effects of clouds. On the other hand, advection by the circulation (the winds moving temperature and moisture around) nearly always opposes aggregation, which is different than what some other studies have found. The cool thing about self-aggregation is that all these feedbacks are internal interactions between the clouds, moisture, and radiation; nothing is forcing the clouds to cluster, they do so on their own.
While we are making progress on understanding self-aggregation, there is still a lot of work to do to figure out the relevance of self-aggregation physics to real world phenomena. Are the radiative-convective feedbacks that cause this spontaneous clumping of convection important for the formation of tropical cyclones (a problem I am currently working on), or the growth of the Madden-Julian Oscillation? How do the physical mechanisms I described above play out in the real world, where there are complicating factors?
However, the impact of self-aggregation might be much broader than its role in these specific (but very important!) phenomena. Clouds and humidity play a big role in determining how easily the atmosphere warms in response to a forcing like increased greenhouse gas concentrations. Self-aggregation changes the amount of clouds and humidity, so it could affect this process (in model simulations like the ones I described here, the domain overall is drier when the clouds are aggregated rather than randomly distributed). We need to figure out if self-aggregation really happens in nature, and if so, how it affects climate sensitivity. This is one of the central questions of the WCRP Grand Challenge on Clouds, Circulation and Climate Sensitivity, and is being actively studied by researchers across the world.