Published on September 25, 2009 by

Development of baits for tsetse flies: much done, more to do

The diseases of sleeping sickness in humans and nagana in domestic stock severely affect health and development in Sub-Saharan Africa. Resistance to drugs against these diseases is widespread; for sleeping sickness there are no prophylactics, and the curatives are highly toxic. Hence, the safest policy is to attack the insect vectors, i.e. tsetse flies (Glossina spp.). The cheapest and simplest method of tsetse control is the use of insecticide-treated baits. Since male and female tsetse feed exclusively on blood, and need to find hosts every few days, the best baits are host animals, or artificial representations of them. Tsetse visit a variety of hosts, albeit with some selectivity, but most species are strongly responsive to cattle. Hence, where cattle occur it is usually most economical and convenient to treat these. Elsewhere it is necessary to rely on artificial baits that usually consist of cloth screens, about 1m2, called “targets”.


Bait performance has been improved greatly in the last 40 years (1). With artificial baits this has involved attention to the size, colour and shape of the targets, but for the savannah species of tsetse the greatest benefit came from identifying and artificially dispensing the effective components of cattle odour. Overall, the cost-effectiveness of artificial baits for controlling the savannah species has been improved 100 to 1000 fold, so that good control can now be achieved using 2-4 targets/km2. The odours commonly used are: i) 1-octen-3-ol, ii) acetone or butanone, and iii) various phenols. For the riverine species the odours have been less effective, so that targets for these flies are usually deployed without odours and the required number of targets per unit area of habitat is ten times greater than for the savannah tsetse. However, the essential habitat of the riverine flies is often highly localized so that the density in the overall operational area is usually not so high. With insecticide-treated cattle the cost-effectiveness has been improved several-fold by restricting treatment to only the legs and belly, i.e. the body regions where most tsetse feed (2).

Given the great technical advances it might be thought that the job of biological researchers is done, with the focus now on campaign managers to identify the best means of implementation. In part this is true, especially with the restricted application of insecticide to cattle. For example, the costs of this control measure have been reduced to a few US$/km2/year by relying on community implementation (3). For comparison, the costs of government-run campaigns are $120 for whole-body treatment of cattle, $220-385 for targets and $265-$800 for other methods (4). Unfortunately, however, the use of cattle baits to control tsetse in settled areas will be of limited benefit if, as is usual, tsetse can invade from unsettled areas nearby. To prevent such invasion it is necessary to deploy targets for several kilometres into the invasion sources. Since target costs are too high for poor farmers, biologists must reduce the costs of target operations still further.

Unsolved problems

The most important means of reducing the costs of targets might be the discovery of additional odour attractants. For the main savannah species of tsetse, i.e. G. pallidipes and G. morsitans, it is known that when all identified attractants are dispensed at natural dose they are only half as attractive as ox odour, so at least one important attractant remains to be identified. Attempts to identify this attractant in the mid-1990s failed, but the technology for identifications has improved since then so that another try is warranted. With G. tachinoides, a nominally “riverine” species that can also occupy nearby savannah, artificial odours can increase catches by 2.5 times (5), but we have not capitalized on this in control campaigns. For the truly riverine tsetse we need to think again about the response to host odours. In experimental conditions that evoke very strong responses of savannah tsetse to odours, riverine flies respond poorly, if at all. The host-finding strategy of the riverine species must therefore be different, so that distinctive experimental designs may be required to test odour efficacy. In considering how to change the set-ups it would help to theorize on exactly how host-finding strategies might differ between the savannah and riverine flies.

A crucial starting point for such theorizing is an understanding of the strategy adopted by the savannah tsetse. We know much about this, especially for G. pallidipes. The flies can perceive ox odour from a range of about 90m, and can sense the ox visually at about 10m if the ox is stationary, and probably at much greater distance if the ox is mobile. We also know that late in the hunger cycle the flies displace at least 1km per day. Given that such displacement involves a number of hops in seemingly random directions, the actual distance covered by the fly must be much larger than 1km. Hence, in the few days before the fly dies of starvation, it is hardly surprising that hosts at common densities of >10/km2 can be discovered readily, especially since mobile hosts will be found by merely waiting in “ambush” between hops. However, our knowledge is insufficiently detailed for present needs. For example, even with G. pallidipes, we do not know the average hop length and how it is affected by temperature, wind, the density of vegetation and the insect’s degree of starvation. We do not know the relative success of waiting in ambush as against active search, and how this changes with the numbers and types of host present. Hence, we cannot model confidently the full details of the host-finding process of the savannah flies, and so cannot suggest how it might have been modified to best advantage in riverine tsetse.

Consequently, we are reduced merely to offering vague and partially contradictory thoughts. For example, since riverine tsetse often rely on small hosts, such as lizards, that are relatively abundant, the insects might not need to search extensively. Against this, waiting to ambush slow-moving lizards is inherently risky. Dense riverine vegetation makes it difficult to detect hosts visually so that, contrary to available evidence, odours might be expected to be of enhanced importance. Perhaps, odour plumes are difficult to follow far in dense bush, but the odour plumes from small hosts are unlikely to extend far. If the odour plume is short it may be of reduced value in host location, but the range of visual detection of small hosts is also likely to be short. Maybe the flies are reluctant to enter the denser parts of their habitats, but judged by the readiness with which riverine tsetse pass through very narrow entrances into traps, these flies are more willing than savannah tsetse to navigate in restricted spaces. Nevertheless, it might be better to rely on visual searches along well-defined game trails but, since lizards often hide in thicker vegetation, many might be missed. Waiting by, or patrolling along, the edge of a water body, i.e. a place were lizards often go, could be beneficial but many riverine tsetse can be trapped in suitable vegetation away from such edges.


The many apparent contradictions suggest that we have not identified at least one crucial and distinctive factor in the host-finding behaviour of the riverine flies, so that we cannot yet understand the full potential for baits against these insects. The chances of discovering a missing item, or at least assessing where to look for it, could be improved by expanding and combining the present crude models for host-finding behaviour (6) and for movement in bushed landscapes (7), especially if coupled with producing more detailed input data. For example, it would be helpful to repeat with riverine flies the studies that assessed the readiness with which savannah flies travel through leafy obstacles (7). To identify the details of tsetse hops, we need the type of radar tracking used with bees (8). Although such technology is unlikely to track tsetse far in dense bush, knowing how flies move in relatively open situations would be a start.

It would be helpful also to assess the number of insects that discover hosts by active search as against ambush, by comparing the numbers caught at mobile and stationary hosts. This cannot involve handnet catching by persons accompanying the baits since humans repel all but the hungriest ones among the savannah flies, and are very large relative to the small hosts of riverine tsetse. It would be more objective to employ sheets of electrified netting around the hosts, to catch flies efficiently in the absence of men. It is easy to use such devices with a stationary host, but more difficult to get the devices to move with a mobile one. However, a lizard and netting could be placed on a trolley modified from the type of small, radio-controlled vehicle used to deliver equipment onto rugby pitches. A much larger trolley to take an ox and netting, complete with a 2km-long railway for it, was made at Rekomitjie Research Station, Zimbabwe, in the late 1990s but the withdrawal of donor support from the station meant that the device has sat idle.

Additionally, we need a range of speculative studies – trying something strange and seeing what surprises might occur. For example, in a recent study with the riverine tsetse, G. fuscipes fuscipes, Lindh et al. (9) experimented with unusually small targets of only 0.06 m2, i.e. 1/16th of the standard size and a third of the size known to be disastrous for G. pallidipes. Surprisingly, the reduction of target size to 1/16th caused only a 50% reduction in catches of G. f. fuscipes. This has immediate practical implications, suggesting that the cost-effectiveness of target control of G. f. fuscipes could be improved several-fold by using the tiny targets, even if twice as many would be needed. More importantly in the present context, it implies that when testing odours for riverine flies we should dispense them near small visual baits. In further speculation we might ask questions such as whether tsetse respond to trails of residual odours left by a crawling reptile. Do they follow such trails like beagles? Are catches affected if a lizard-sized target has a “head” that nods? How many flies could be caught in plots of about 200m2, in various situations with and without various baits, if we arranged for a tall enclosing wall of netting to be hauled up quickly?

In any case, there is still much to do, especially with riverine tsetse.