Understanding the Madden–Julian oscillation

Precipitation in the tropics organizes on multiple spatial and temporal scales. At the time scale of a day, precipitation is predominantly determined by the diurnal cycle in solar radiation, the effects of which vary across different landscapes and climate regimes. At longer time scales, changes in the atmospheric circulation related to Earth’s seasonal cycles determine the distribution of precipitation.

Before the advent of satellite observations, however, no cycles at the time scale of several days to weeks were known to modulate precipitation. For example, if you were an observer located at Darwin, Australia, you would notice that some weeks were rainier than others. But because the variations are not exactly cyclical, you, the observer on the ground, would perceive them as random.

In a letter published by the American Meteorological Society in 1963, Jule Charney, one of the founding fathers of modern meteorology, noted that large-scale motions in the tropics are characterized by small spatial and temporal variability in temperature and pressure. He also noted that flow in the tropics must be quasi-nondivergent—in other words, the tropical wind field is dominated by rotational motion. He suggested that the circulations must be driven by latent heat release (energy released during condensation and freezing of water) within smaller-scale regions of ascent (thunderstorms).

Observations of cloud tops from geostationary satellites revealed widespread cloudiness within the tropics that are connected to two regions: the latitudinally narrow Intertropical Convergence Zone (ITCZ), which circles the equator, and the Indo-Pacific Warm Pool, which extends from East Africa to French Polynesia. If you were to inspect an individual picture from those satellites, the tropics would appear similar to a pot of boiling water. Just as bubbles continuously appear and burst inside the pot, so does convection burst and then dissipate. You would probably conclude that there is hardly any organized meteorological structures in those regions other than the occasional tropical cyclone.

That conclusion is mistaken, however. Extensive analysis in the 1970s and 1980s revealed that convection is often organized in space and time in the form of waves. Many of those waves exhibit characteristics that closely resemble solutions to an atmospheric model analyzed in 1966 by Taroh Matsuno. Because of their connection with deep convection, the waves subsequently became known as convectively coupled equatorial waves. Indeed, wave motion dominates variations in rainfall at the waves’ time scales of several days to weeks.

However, not all waves in the tropics follow the solutions derived by Matsuno. In one of the first comprehensive studies of radiosonde (weather balloon) data across the tropics, Roland Madden and Paul Julian identified in 1971 an oscillation that modulates wind and precipitation across the Indo-Pacific Warm Pool with a robust spectral peak in the 40–50 day time scale. A followup study in 1972 using sounding data across the equatorial belt found that this wave is a manifestation of a planetary-scale wind field, with overturning circulations in the equatorial plane tens of thousands of kilometers long. Convection and precipitation are enhanced within the rising branch of the circulation while convection is suppressed in the sinking branch. Madden and Julian found that the wave slowly propagates eastward at a rate of about 5 m/s. The observed characteristics of this wave-like disturbance, since named the Madden–Julian oscillation (MJO), could not be explained by any existing theoretical framework.

Subsequent studies, together with the development of global climate models, further characterized the MJO. The oscillation’s horizontal structure in the lower troposphere resembled the wave response to an equatorial heat source—one that is dominated by latent heat released from condensation of water vapor into liquid water droplets in large thunderstorm clusters. The response is manifested as a region of equatorial easterly wind anomalies to the east that is roughly in phase with low surface pressure. A pair of cyclonic gyres that flank a narrow region of equatorial westerly winds characterizes the region to the west of the center of convection.

The MJO’s horizontal structure suggests that the interaction between large-scale motions and heating associated with deep convection (thunderstorms) plays a critical role in its dynamics. However, whereas several mechanisms were proposed to explain the coupling between large-scale motions and convection, none were proven capable of reproducing the MJO. That failure prompted many authors to consider whether other important physical processes could be central to MJO dynamics.

Subsequent studies in tropical convection found that regions of precipitation in the tropics are highly correlated in space and time with positive anomalies in tropospheric water vapor. In other words, convection in the tropics is largely confined to the regions where the higher water vapor content is found. Analysis of satellite-derived water vapor and precipitation found a similar correlation between water vapor content and convection in the MJO. Could water vapor’s role in maintaining convection in the MJO yield any insight into its propagation mechanisms?

Reliable tools exist that can help us further understand how thermodynamic processes and large-scale motions interact. Among the tools are radar, lidar, and other ground-based observations, satellite-derived products from polar orbiting satellites, and state-of-the art reanalysis products such as the Interim Reanalysis of the European Center of Medium Range Weather Forecast (ECMWF), or ERA-Interim for short. A reanalysis can be understood as a “reforecast” of a past atmospheric state, incorporating observations from satellite and ground-based products in order to provide the best possible representation of the atmosphere of that past state.

Although perhaps not as accurate as a direct observation, reanalysis products have the advantage that they are evenly distributed in space and time. They also have global coverage in multiple layers of the atmosphere, including data-sparse regions like the tropical oceans. Those advantages make the use of reanalyses ideal for studying a large-scale, low-frequency tropical phenomena such as the MJO.

In the presence of surface friction, low-level convergence is observed in the region of easterlies (westward flow) to the east while divergence is observed to the west, resulting in a vertical velocity profile that is shallow on the eastern end of the region of enhanced rainfall and elevated on the western end. If you were an observer located somewhere near the equator and saw an MJO event propagate across, you would initially see a preponderance of shallow cumulus and cumulus congestus clouds when surface convergence begins (though you would still see all sorts of clouds). At that phase of the MJO cycle, water vapor is increasing in the lower troposphere above the tops of the shallowest cumulus clouds.

As days pass by, the anomalous moisture field deepens, and tall thunderstorm clouds that bring heavy precipitation dominate the cloud population. Relative humidity in the mid troposphere is rapidly increasing at this point. As mid-tropospheric relative humidity and lower-tropospheric water vapor reach their peak, the thunderstorms become wider and cover a larger area of the sky.

Immediately after moisture and precipitation reach their peaks, equatorward wind anomalies in the troposphere above the boundary layer—in conjunction with equatorial westerlies—begin to dry the moist region in the lower troposphere. The wind anomalies are accompanied by surface divergence along with downward motion near the surface. That anomalous motion is part of the aforementioned wave response to an equatorial heat source and causes the relative humidity anomalies to be confined to the upper troposphere. At this time, precipitation begins to taper off and becomes dominated by weaker precipitation from large regions of thick, stratiform clouds (though all cloud types would still be present).

The above description of the MJO from reanalysis highlights the complexity of processes across different scales—from clouds to moisture to large-scale wave dynamics—all of which must be well represented in order to accurately simulate the MJO’s structure. Improved representation of cloud processes and their interaction with moisture may be crucial to our understanding of MJO initiation, maintenance and propagation, yet no existing theoretical framework for MJO dynamics incorporates all those processes.

There may also be other important processes involving radiation or multi-scale interactions that we do not yet understand. Despite all these uncertainties, increased interest in the topic, observations, field experiments, high-quality reanalysis products, and improved MJO representation in climate models all provide hope that future progress in this field will lead to greater predictability and understanding of what is arguably one of the most important problems in tropical meteorology.

Ángel Adames is a PhD candidate in the department of Atmospheric Sciences at the University of Washington in Seattle. His research lies in the area of tropical meteorology with a focus on understanding the structure and propagation of the Madden–Julian oscillation.