The actual flows of air through the Macrotermes mound are measurable using tracer gas methods. The basic technique is a so-called tracer pulse-chase experiment. This involves injecting a bolus of tracer gas (the pulse), in this instance 500 mls of 1% propane in air into some location within the mound or nest. Very sensitive combustible gas sensors are then placed around the mound and nest to detect where and when and how much of the tracer appears (the chase). By placing an array of these combustible gas sensors at many locations around the mound, a fairly complete picture of the movements of air can be assembled. The photograph to the right shows a mound outfitted combustible gas sensors and injection points for a pulse-chase experiment.
These measurements reveal a very complicated pattern of air flow that bears little resemblance to the simple circulatory flow that Martin Lüscher's thermosiphon model would suggest.
How does air actually flow from the nest through the mound?
Air is mostly drawn upward in the surface conduits
Air is drawn mostly out the mound's upper and downwind surfaces
Air injected into the surface conduit does not return to the nest
The surface conduits constitute a well-mixed space stirred by winds
Air is distributed from the chimney to all the surface conduits, then redistribuited by wind
Air flow through the mound is very complex, as is revealed by the pulse-chase experiment diagrammed to the right.
Here, tracer gas was injected into the nest, and combustible gas sensors were arrayed throughout the mound and nest.
Wind speeds and directions are indicted by the blue traces and by the arrow cards. Position of each sensor is indicated by a heading in degrees, and is marked on the compass card that accompanies each trace.
There is clearly a very complicated pattern of air flow. Tracer gas soon appears in the chimney, but fluctuates with the tidal flows of air in the chimney. Tracer gas appears in certain of the surface conduits (109 upper and 339 upper) within ten minutes. The appearance of tracer in the lower conduits is much slower, 40 to 60 minutes after the injection.
One could interpret this pattern as supporting the thermosiphon model, because it appears to indicate a downward flow of air (upper conduit -> lower conduit). This appears to not be the case, as shown here. Rather, the pattern is due to shifting speeds and direction of the wind. Tracer appears only slowly in some upper conduits, like 172 upper, and this is correlated with changes of wind direction. Also, if air were flowing as the thermosiphon model suggests it should, downward flow in the surface conduit should eventually bring tracer into the lateral air space of the nest. It does not show up there in appreciable quantities: the small quantities that do appear are probably due to lateral diffusion from the nest.
In short, tracer pulse-chase measurements indicate that air flows in the mound are strongly driven by wind, and do not appear to circulate in the pattern that Martin Lüscher's model suggests it should.
In the thermosiphon model, air should flow downward through the surface conduits. If air flows within the mound are driven by wind, on the other hand, air should be drawn mostly upward through the surface conduits. Pulse-chase measurements favors the latter.
When tracer is injected into the nest (left), it appears first and most strongly in the upper surface conduit (orange trace, leftt). It appears a bit later and more weakly in the lower surface conduit (green trace, left). Although this time lag could be interpreted as a downward movement of tracer, in this case, it is mostly an artifact of sensor placement. The upper sensor in this experiment (orange dots, left) was positioned roughly downwind, while the lower sensor (green dots, left) was positioned more lateral/leeward to the winds. In all likelihood, the upper and lower sensors were placed in different surface conduits. The time lag does not represent a sequential movement of tracer-laden air from the upper part of the mound to the lower, but different times of arrival in different surface conduits that differ in how strongly wind is driving air flow.
The effect of sensor placement is shown more clearly when the upper and lower sensors are positioned the same with respect to winds (right). Now, the upper and lower sensors are probably in the same surface conduit. After tracer tracer is injected into the nest, the tracer appears simultaneously at both sensors, not sequentially as would be expected for a thermosiphon flow. Air from the nest is being drawn by wind from the nest, to the chimney, from where it is distributed according to the surface pressures drawing it up. Those pressures will be strongest at the upper surfaces and will draw tracer-laden air most strongly there. This is precisely what the tracer measurements show.
The wind-induced suction pressures over the mound surface are strongest downwind / lateral to the wind direction. When tracer gas is injected into the nest, one should therefore see the tracer emerge preferentially at those locations. This is precisely the case (below).
Virtually no tracer emerges from the mound's upwind surfaces. Nearly all emerges from surfaces lateral to the wind, and downwind, most strongly at angles of +/- 90 to 120 degrees with respect to the wind direction. This is the case both for average emission of tracer (a and b, below), and total emission of tracer (c and d, below).
The thermosiphon model predicts that air is returned from the surface conduits back to the nest. This is not the case. Tracer injected into a surface conduit (orange trace, below) close to the ground does not appear at any time in the nest (brown trace, below). The reason tracer does not appear in the nest, of course, is that air in the surface conduits is mostly drawn upward by the wind-induced suction pressures at the mound surface.
Even though each surface conduit is separate from all the others, they are connected below by the extensive reticulum of the lateral connectives. This means that the surface conduits constitute a single 'well-mixed space', in which air in one surface conduit can readily move to others. This is shown by a pulse-chase experiments where tracer is injected into one surface conduit, and seeing where it ends up. In these experiments, there is one sensor in the conduit where tracer was injected, and another in a conduit on the opposite side of the mound. The locations of the surface conduits were chosen with respect to the prevailing winds that day, so that one is situated at the mound's upwind surface, while the other is placed downwind. However, shifting winds through the experiment did not always preserve this orientation, with interesting results.
Tracer injected into a mound's 'upwind' surface conduit is very quickly redistributed to downwind conduits.
In the experiment diagrammed to the right, the upwind surface conduit is situated at a heading of 264 degrees, and the downwind conduit is positioned at a heading of 84 degrees. Both surface conduit sensors were located about two meters above ground level. Winds were steady out of the northwest, as indicated by the wind rosette and did not change appreciably through the experiment, as indicated by the compass cards. When tracer is injected into the upwind conduit, it is not detectable at the injection site, but appears within two minutes at the downwind conduit. Clearly, tracer-laden air in the upwind conduit is quickly driven to the mound's downwind surfaces by the wind-induced pressure field over the mound surface.
At the same time, shifting winds can move tracer throughout the array of surface conduits.
This is seen in the experiment diagrammed below and to the right. Here, the arrangement if sensors and injection point are similar to the experiment described above. Initially, tracer was injected into an 'upwind' conduit situated at a heading of 104 degrees. An additional sensor was
located 'downwind' at 292 degrees. Winds were variable out of the east.
At first, the pattern of tracer movement is as expected: no tracer appears at the upwind point of injection (green trace, right), but appears rapidly at the downwind side (red trace, right), distributed there by the wind. However, roughly 30 minutes into the measurements, the wind shifts to northerly, and the orientation of the sensors with respect to wind shifts with it. From 30 to 40 minutes, both sensors are now lateral to the wind, and tracer begins to appear at the formerly 'upwind' sensor. As the wind shifts back to easterly between 40-48 minutes, the sensors are brought back to their original orientations with respect to wind. Tracer concentration then declines in the 'upwind sensor again. At 50 minutes, the wind shifts to the northwesterly direction, almost reversing the orientation of the sensors: the formerly 'downwind' sensor is now close to upwind and vice-versa. Tracer again 'sloshes' within the surface conduits to the now-downwind sensor.
Clearly, the surface conduits consitute a well-mixed space that is driven strongly by wind.
Even though the surface conduits constitute a well-mixed space, air appears to be distributed equally from the chimney to all surface conduits.
In the experiment graphed at the right, tracer gas was injected into the chimney, and monitored with sensors at two surface conduits: 'upwind' (70 degrees, green trace) and 'downwind' (300 degrees, red trace). Shortly after injection, tracer appears at both the upwind and downwind sensors. Following this, though, the tracer is quickly redistributed away from the upwind sensor to the downwind sensor.
The distribution from the chimney to all surface conduits may be due to two factors: