Odour movement, wind direction, and the problem of host‐finding by tsetse flies

ABSTRACT. Movement of host odour was modelled in natural tsetse habitats with smoke and ultra‐light 7‐cm‐long wind vanes; the speed and direction of the air movements were analysed from video recordings thereof. Wind of <1 ms‐1 did not move in straight lines, since large packets of air (>10 m across) often changed direction together. The rate of this change of direction (meander) correlated negatively with windspeed. In open woodland with a shrubby understorey (in which windspeed was reduced by a factor of >5 from that above the canopy, to ax 0.3 m s‐1), this wind meander fell by 2d̀ s‐1 change of direction for each 0.1 m s‐1 increase in windspeed (r2=0.96). Over open ground without shrub cover, the meander fell by 0.5d̀ s‐1 per 0.1 m s‐1 increase in windspeed (r2=0.85). In both situations, such meandering virtually ceased in winds of > 1 m s‐1. In woodland, the relationship between the direction of air movement near the surface of bare earth (one potential tsetse landing site) and that c. 0.5 m above ground level (flight height) was often weak (r2=0.2‐0.4), but this problem would be reduced if the fly averaged the ground‐level wind for at least 30 s. Odour (smoke) travelling from a source 15 m ‘upwind’ over open ground arrived at a notional tsetse fly for 80% of the time from a direction within 10d̀ of the true source direction. In typical tsetse woodland, however, the ‘odour’ arrived from all directions (including >90d̀ away from the source), with only a 30% bias towards the true source direction (±10d̀). Evidently, tsetse must navigate up odour plumes by means that get round these difficulties‐simple, moth‐type upwind anemotaxis alone seems unlikely to be adequate.