Dispersion of carbon dioxide plumes in African woodland: implications for host-finding by tsetse flies

In Zimbabwe, high-resolution (10 Hz to an accuracy of +/- 0.1 p.p.m.) measurements were made of atmospheric and host-produced CO2 in tsetse habitats during the dry season. The diel structure of atmospheric CO2 concentrations is bimodal, with a minimum at approximately 16.00 hours and maxima at approximately 05.00 hours and 20.00 hours, respectively. The background CO2 noise is greater in densely vegetated riverine woodland than in leafless, deciduous (mopane) woodland. Variation in atmospheric CO2 concentrations is correlated with decreasing wind speed and increasing thermal stability. Consequently, the background noise during the day is greatest in riverine woodland during early morning and late afternoon, when winds are typically light and stable, and thermal inversion conditions are developing. Measurements were made of CO2 at 8-64 m downwind from natural (two cattle) or synthetic sources (4-20 L min(-1) CO2). The signal from the sources appears as fluctuations above threshold (approximately 355-362 p.p.m.), in the form of intermittent 'bursts' of CO2. The strength, duration and intermittency of the signals attributable to these sources declines with source strength and distance from the source. In riverine woodland, approximately 50% of all bursts are 0.1 s long, and 10% are > 2.0 s long. Carbon dioxide signals from equivalent sources are stronger in riverine woodland than mopane. Carbon dioxide dispensed at rates of 4-20 L min(-1) is detected up to 64 m downwind of the source but peaks are typically < 10 p.p.m. above threshold. Consequently, host-CO2 signals are obscured during periods of large fluctuations in atmospheric CO2. These results suggest that CO2 is detectable, at least in some circumstances, at tens of metres downwind and hence dispels the notion that its action is limited to that of a short-range attractant.