Special Investigation: Something's Up Down Under (pt.2: The MJO)

In my previous post, I began to investigate tropical cyclones that last a remarkably long time over Australia by using a sample of eighteen such storms over the past ten years. First I arranged the storms by the month of the year they occurred in, and found that there is a tendency for these storms to occur near the ends of the tropical cyclone season. However, for several reasons, this correlation was not considered very strong. Next, I compared the timing of the sample storms to the state of the El Nino/Southern Oscillation (ENSO). I found a rather strong correlation between the storms and La Nina conditions, and a weak correlation to El Nino conditions. This data was quite clear and straight forward, so I considered this to be more scientifically sound than the time of year analysis. Now I will continue my investigation by looking into the Madden-Julian Oscillation (MJO), which has a profound impact on Australia.



The Madden-Julian Oscillation (MJO)

The MJO is essentially a pulse of anomalous weather that travels eastward along the equator on the timescale of 30 to 60 days or so. It is most pronounced over the Indian Ocean and the "Marine Continent" (the islands and seas surrounding Indonesia and adjacent countries). To a large extent, the pulse dies out in the central to eastern Pacific.
Several different atmospheric and oceanic parameters are affected by the MJO. The active phase of the MJO is identified by an area of enhanced convection (cumulus and cumulonimbus clouds), which are often discernible on satellite imagery. Along with the increase in cloudiness comes an increase in precipitation. Often behind the active phase is a region of noticeably clearer air and decreased precipitation. Due to the variation in cloud cover, the sea surface temperature (SST) also varies due to the MJO. Clearer air ahead of the active phase allows more sunlight to reach the surface, thus SSTs rise. The convection associated with the MJO active phase removes heat from the ocean through evaporation, which transports latent heat away from the surface. Therefore, the western side of the active phase typically has cooler than average SSTs. Finally, the wind direction is also affected by the MJO. The increase in convection in the active phase of the MJO is caused in part to convergence, that is, winds from opposite directions meet and the only direction to travel is up. This means that around the western edge of an active phase, air is moving with a eastward component. This is a major departure from normal as the tropics are typically dominated by trade winds, which have a westward component. Below is an animation of a well defined MJO active phase. It forms just off the coast of Africa and slowly makes its way across the Indian Ocean. When it reaches the Maritime Continent, the cloud cover increases, allowing its movement to become clearer. As it nears the eastern edge of the image it stalls and begins to die out.

 

How does one identify the location of the active phase? For years there was no single agreed upon definition, until a phase space diagram was developed that takes into account cloudiness and wind anomalies. The result is a diagram that splits the tropics into eight different phases, each designated by a number. Phases 1 and 2 are the Indian Ocean, 3 and 4 are the Maritime Continent, 5 and 6 are the Western Pacific Ocean, and 7 and 8 are the Western Hemisphere and Africa. Therefore, all one has to do is look for a date along the line on the diagram to find out roughly where the MJO active phase was at that time. The further the line is from the center, the stronger the active phase is. If the line is within the circle in the middle of the diagram, the MJO is very weak, or might not exist at all at that time. Below is the phase space diagram that includes the active phase in the above animation. One can see how it has the strongest signal in the eastern Indian Ocean and the Western Pacific Ocean before rapidly weakening. Since these diagrams are largely a result of wind anomalies, the strongest signal might not be at a time when the satellite shows the most convection.

To compare my sample storm times to MJO events I conducted two analyses: one was to see if there was any correlation to how strong the active phase was and the other was to where the active phase was. To gauge the intensity of the pulse, I divided the phase space diagrams into four concentric circles. The innermost circle is the one already plotted on the diagram, thus any cases where the date landed in the circle received a 0 for strength. The other three circles were evenly spaced and represented strengths of 1, 2, and 3. When this data is plotted, the graph below is produced:

Quite obvious is the absence of any cases in which the active phase was a 3. Furthermore, the vast majority occurred when the active phase was weak (1) or nonexistent (0). It is pretty well established that frequency of tropical cyclone formation in general is increased by the active phase of the MJO. Therefore, it seems that there may be some correlation between these long-lived storms and the MJO, but only when the active phase is weak. It is likely that too much of the conditions that breed these storms end up being a detriment.

For the active phase location at the time of these storms, I simply grouped the storms into the eight phases on the diagram. For the cases in which the active phase was indiscernible (in the center circle) I assigned them a "phase" of 0. Below is a plot of the results:

Clearly, many of the cases occurred during a 0 "phase", which makes sense given all of the cases that had a strength of 0. The peak of the rest of the distribution occurs at phase 5, which roughly correlates to the center of the active phase being over eastern Australia. The second highest counts were phase 4 (roughly western Australia) and phase 6 (around 160E). Also notable is the lack of any cases in phases 1 and 2, which is roughly the western two-thirds of the Indian Ocean.
The reasoning behind all this left me stumped until I came across a short article from 1998 on the MJO. Two diagrams in particular caught my attention, the first was this:

This shows the basic cross-section of the MJO. Note how ahead of the active phase convection is suppressed and the SSTs are cooler. This would help explain the lack of cases when the active phase is in phase 1 or 2. The warmer SSTs just ahead of the active phase supports the existence of sample storms while it is in phase 3.
The other diagram was a bit more complex:

This shows the wind pattern associated with the MJO when the active phase is near 120E and when it is near 150E. 120E is essentially the eastern portion of phase 4 while 150E is very close to the center of phase 5. A close look at the top figure shows a little loop at lower levels, just to the southwest of the deep convection. The "C" inside the loop stands for cyclonic, which is clockwise in the southern hemisphere. This cyclonic flow could provide the "spin" that allows the sample storms to last so long. In the bottom figure, the loop has changed shape and now lags behind the convection a little, thus it is still providing "spin" that aids these systems. Furthermore, at upper levels in both of the figures is strong anticyclonic wind, positioned roughly above the low level cyclonic flow. Anticyclonic flow directly above tropical cyclones allows for vigorous outflow which helps sustain them. It therefore seems likely that these flow patterns are very beneficial to long-lived storms.

When comparing the two data analyses, it appears that the correlation to the location of the active phase is far more useful. Although the active phase strength correlation was pretty straight forward, one must keep in mind that those strength counts apply to each storm, regardless of whether the active phase is near it, or on the other side of the globe. As far as the location correlation is concerned, it is important to remember that the phases in the phase space diagrams are only approximate, and therefore can only very loosely be assigned actual longitude values. Overall, of the three factors considered thus far, the MJO appears to have to largest role in the cause of long lived Australian tropical cyclones.

One final, very intriguing correlation that arose seems to support the validity of some of these analyses. It is well established that the MJO has the tendency to be strongest during the Northern Hemisphere winter. Keeping that in mind, my analysis seems to indicate that the storms in question do not tend to form under strong MJO events. This would imply that the storms would occur least often in the middle of winter. Amazingly, this shows up on the time analysis I made in part one, which showed that most of the storms occurred during the beginning or end of winter, not at its height. Therefore, these two correlations should be taken with more weight than previously thought.


NEXT TIME: I'll finish the investigation with a look at small scale factors.