Biology, images, analysis, design...
Clegs etc Ticks Mossies Tsetse
"It has long been an axiom of mine that the little things are infinitely the most important" (Sherlock Holmes)




Why does it matter what tsetse feed on?

Tsetse flies feed on the blood of animals. In the process they transmit various species of trypanosomes which cause the disease trypanosomiasis (also known as trypanosomosis). The disease is commonly known as nagana in livestock (caused by Trypanosoma vivax or Trypanosoma congolense) and sleeping-sickness in humans (caused by Trypanosoma brucei gambiense or Trypanosoma brucei rhodesiense. In most parts of Africa the only threat is from animal trypanosomiasis, and this is the case at Nguruman in southwest Kenya. Under drought conditions, tsetse-borne trypanosomiasis can kill many local cattle, especially those which have been upgraded by cross-breeding.

Trypanosome transmission is dependent on the feeding habits of the tsetse. In order to understand the epidemiology and hence control the disease we need to know what this pattern is and what factors affect it. Tsetse feeding habits can also directly affect which control methods are viable. An increasingly widely-used way to control tsetse is to apply "pour-on" insecticide on the cattle, so the tsetse flies are killed when coming to feed. But this method will only reduce the number of tsetse flies if a large proportion of their population feed on cattle.  

What is the area like?

Nguruman (or Nkuruman) refers to the area lying at the base of the Nguruman escarpment about 150 km south-west of Nairobi in Kenya. As one approaches the escarpment (picture right), the vegetation changes from sparse Cordia bushland to Acacia tortilis woodland and then to dense Acacia-Commiphora thicket on the lower slopes.

Glossina pallidipes (see picture at top biting man) is very common in the denser vegetation. Much smaller numbers of the larger tsetse, Glossina longipennis, fly at dawn and dusk in the more open areas. Both tsetse species will feed on man when available, but they usually prefer other hosts.  

What are the hosts of Glossina pallidipes?

We review three studies on the feeding habits of Glossina pallidipes at Nguruman - Tarimo et al. (1984) Okoth et al. (2007)  and Bett et al. (2010) . They used serological methods  to identify the blood meals of a sample of tsetse flies. The figure below summarizes the results of these studies. We have initially pooled all bovids (cattle, buffalos & antelopes) so we can show their importance in relation to suids (pig species), elephant, giraffe, horses and primates.

These studies indicated that more than half of the blood meals were from bovids and suids. There are no domestic pigs in the area and very few bushpig or giant forest-hog, so the suids are mostly warthogs. Warthog (picture right) are known to be a favoured host of Glossina pallidipes in other parts of Africa.

Torr (1994)  working in Zimbabwe found that tsetse (G. pallidipes & G. morsitans) tended to alight and feed on the head of warthogs especially near the eye. This was thought to be because of a visual response to the darker colour around the eye, The secretion produced around the eyes could also be attractive to tsetse.

We look at which species of bovids are fed on below. The different colours now represent twelve different bovid species, so orange is now buffalos, yellow is cattle and so on. The x-axis is still the percentage of ALL feeds - not just of bovid feeds - so it only goes up to maximum of 52%.

All three studies identified buffalo (first picture below) and cattle (second picture below) as important bovid hosts. Buffalo are frequent in the area - the herd in the picture was seen in early morning dust haze in the open Acacia tortilis woodland.


However there was little agreement between studies on the importance of other bovid hosts. One study found waterbuck, kongoni and wildebeest (picture below right) to be important hosts, another found bushbuck and dikdik as frequent hosts, whilst the third found shoats (sheep and goats) and kudu to be important. We evaluate the likely validity of these blood meal identifications below.

As for hosts other than bovids and suids, elephant (picture right) was sometimes fed on, especially in Bett et al.'s (2008)  study. Another recent study (Muturi et al., 2011 ) found most meals were on elephant or warthog, with only one bovid meal - but this was based on a very small sample size of only 13 flies.

Giraffe (picture below left) was reported as a host in all the main studies, comprising between 5 and 10% of the blood meals. This is also in line with other studies on G. pallidipes. Just one of the studies reported feeding on zebra (picture below right) comprising over 10% of blood meals. This is a very surprising result which we discuss below.


There was one other study (Sasaki, 1995 ) which we have not included above because the blood meals were only tested for a limited range of possible hosts and the tsetse were only sampled in the dry season. Results for 155 blood meals differed sharply from the other studies and had bushbuck as the most frequent blood meal source, followed by ostrich, elephant, buffalo, and warthog.

The higher number of bushbuck may have been because sampling was carried out further south than other studies where bushbuck may be commoner. However, the ostrich result was very surprising since other studies have recorded very few feeds on birds of any species. We now review these results in the light of data given below on the relative abundance of the different hosts - bearing in mind that cannot expect the relative frequency of bloodmeals to be the same as the relative host abundance unless the sample is very large, and each host is equally available, and equally readily fed upon.


How do these results relate to the availability of hosts?

No method of blood meal identification is infallible and Allan (2010)  has rightly cautioned that it is necessary to determine if host blood meal distributions are consistent with the availability of hosts. If there is no evidence that the supposed host occurs in the area, then it reasonable to suspect that the identification is wrong. We can also look at whether the observed distribution of feeds between hosts parallels the level of abundance of the different hosts. A similar distribution suggests (a more or less) uniformly random host selection. A very different distribution suggests selective feeding on a few main hosts.

We have done this below using previously unpublished data on wildlife abundance at Nguruman - namely comprehensive game counts that we kept when carrying out regular tsetse population monitoring from 1983 to 1992 (see Brightwell et al., 1992 ). These involved repeated drives along fixed transects within tsetse habitat on at least five days every month. A total of 48 potential hosts was recorded on these drives. We have also used data from an aerial survey carried out over the area by Kenya Wildlife Service (2005).  The blood meal analyses covered a period of over 25 years, but it seems unlikely that game composition changed much over this period, aside perhaps from elephants.

The ground survey and the aerial survey confirmed the presence in the area of all the species reported in the blood meal analysis. This does not of course prove that the blood meals were identified correctly or that the sample was representative, but it does at least mean they were possible. The top bar in the figure below summarizes the pooled result of the three main studies (671 blood meals). The three studies were relatively homogenous when hosts were grouped in this way, so we can have some confidence in the results. The middle bar shows results from ground survey (1993-1997), and the lower bar results from the aerial survey (2005).

Suids (warthogs) comprised around a third of the blood meals, but made up only 0.5% of the ground counts, and were scarcely present in the aerial survey. The undercounting bias for warthog in aerial surveys is well known (Jachmann, 2002 ). Warthogs may also be under-recorded in ground counts because some are underground in burrows, either reacting to the sound of a vehicle or raising young. Nevertheless, all the evidence suggests that warthog densities are much lower than would be expected from their contribution to blood meals. This indicates Glossina pallidipes strongly prefer feeding on warthog, or they do so more successfully, or both.

Bovids contributed about half of the tsetse blood meals, but comprised about 80% of the medium to large sized warm-blooded animals. To understand this one needs to examine what species comprise the bovids. As before the top bar in the figure below summarizes the summed blood meal results for the three main studies (671 blood meals). This time the three studies were much less homogenous, so we can have rather less confidence in the results. The middle bar shows results from ground survey (1993-1997), and the lower bar results from the aerial survey (2005).

It is immediately apparent that domestic animals (cattle, goats and sheep) contribute far fewer blood meals than would be expected from their numbers.

For cattle this is probably because the livestock owners specifically avoid areas of bush where the challenge is highest. Even if cattle are taken through such an area in the dry season when grazing is limited, they are herded through as fast as possible at times of day when flies are least active.

There is much less avoidance of such areas by people herding goats and sheep (picture right), and given the high densities of small ruminants, we can conclude that tsetse fail to feed upon such hosts even when they are available.

Conversely the contribution of wild bovids especially buffalo to the blood feeds was much higher than would be expected from their numbers, suggesting selective feeding on buffalo.

In the pooled results both bushbuck and eland (picture right) were frequent hosts, both of which have been reported as favoured hosts elsewhere. Two of the studies did not report eland blood meals, but instead reported the (related) lesser kudu. Lesser kudu was either rarely recorded or not recorded at all in the game surveys. It is therefore possible that kudu was a misidentification for eland.

Grants gazelle, wildebeest and impala have been less commonly reported as being fed on by tsetse elsewhere, but all are credible given their abundance.


A questionable bloodmeal identification is waterbuck (picture right), recorded as around 10% of feeds in one study. Waterbuck are present at Nguruman along the flood plain of the Oloiboitoto River. However, they are too infrequent to show up in the graphics showing host abundance, since they comprise less than 0.5% of the total.

Moreover, workers elsewhere have reported waterbuck are seldom fed upon, and there is some evidence that their odour contains repellent constituents. Waterbuck cannot be ruled out as a frequent host, but this should be regarded as uncertain until results are replicated.

Elephant & giraffes were both fed on rather more than would be expected in relation to their abundance, although if one adjusted the figures for biomass that may not be the case. There was some indication that elephant blood meals were more frequent in recent studies. In the 1980s and 1990s poaching was prevalent at Nguruman (picture right), but reduced poaching in recent years may have resulted in an increase in elephant numbers (Mwathe et al., 2006 ).

We should also mention the primate results. The blood meal results mostly identified them as human, whilst all the game survey results for primates refer to baboons and vervet monkeys. Thee is no evidence for any significant feeding of tsetse on these primates, possibly because they are able to kill biting tsetse.


The other questionable identification is zebra, recorded as 10% of feeds in Bett et al.'s (2008)  study. Although zebra are certainly common enough at Nguruman (see picture right of zebra killed by lion), they tend to keep out on the plains and are seldom seen in areas where tsetse are found. Neither Weitz (1981)  nor Moloo (1993)  reported any zebra feeds in a total of over 10000 blood meals of Glossina pallidipes, and Clausen (1998)  only recorded a handful of equid feeds. Waage (1981)  suggested that the reason zebra are seldom fed on because the stripes make them visually unattractive to tsetse.

Lastly all sorts of interesting predators occur at Nguruman, including lion (below left) and the much rarer hunting dog (below right). Tsetse may occasionally feed on lion, but this is uncommon - probably because large predators are always much fewer than their herbivore prey.


How to Assess Feeding habits of Tsetse Flies

There are two main requirements for accurately assessing the feeding habits of tsetse flies:
  1. A representative sample of blood-fed flies
  2. Accurate identification of the blood-meal source.

Obtaining a sample of flies

Flies are sampled either by searching the vegetation for resting flies or by using traps. Searching the vegetation can produce a much larger proportion of recently fed flies, but requires very experienced staff. Traps give far more flies, but only a very small proportion of the flies will be blood fed. In recent years most studies have used traps.

Two trap types have been used for collecting flies at Nguruman. The first is the biconical trap (below left) first developed for sampling riverine tsetse flies. It will catch G. pallidipes but is not very efficient for this species. The second trap type (below right) was one of the NGU series of traps, developed at Nguruman, primarily for community-based tsetse control. They are commonly used baited with cow urine and acetone and are much more efficient than the biconical trap whether baited or unbaited.


Unfortunately neither searching nor traps gives a representative sample of blood meals! Engorged flies tend to settle close to the host (whether amongst vegetation or in a trap to start digesting the blood meal. Hence the samples are mostly composed of flies that have all fed on a few individual hosts. This bias can be counteracted by very large sample sizes. However, samples may still be biased in relation to vegetation types sampled, time of day and so on. The three main studies we considered above all used traps (either biconical or NGU traps) to sample flies over a full year in a variety of vegetation types.

Identifying hosts

Various serological methods have been used to identify blood meals sources, initially precipitation and inhibition tests, and then enzyme-linked immunosorbent assay (ELISA) methods. Serological tests are limited by the availability and specificity of the antisera and by cross-reactions that may occur. More recently DNA based techniques have been used in the identification of bloodmeals. These techniques have the potential to be more accurate and discriminating. However, errors will still occur, so host blood meal identifications must be consistent with the availability of hosts. There remains a problem with unidentified blood meals. Different workers use different criteria for how recent the blood meal should be. The age of the blood meal will determine whether it can be identified or not. Hence with unidentified bloods meals we do not know if it is too old or is a species not included in the range of potential hosts. Other problems are that terminology is sometimes imprecise - for example 'bovine' may be misused used to mean specifically cattle - and we are often not told precisely which hosts are being tested for.



  •  Allan, B.F. (2010). Blood meal analysis to identify reservoir hosts for Amblyomma americanum ticks. Emerging Infectious Diseases, 16, 433-440. Full text

  •  Bett, B. et al. (2008). Estimation of tsetse challenge and its relationship with trypanosomosis incidence in cattle kept under pastoral production systems in Kenya. Veterinary Parasitology 155, 287-298. Abstract 

  •  Brightwell, R. et al. (1992). Factors affecting the dispersal of the tsetse flies Glossina pallidipes and G. longipennis (Diptera: Glossinidae) at Nguruman, south-west Kenya. Bulletin of Entomological Research 82, 167-182. Abstract 

  •  Clausen, P.-H. et al. (1998). Host preferences of tsetse (Diptera: Glossinidae) based on bloodmeal identifications. Medical and Veterinary Entomology 12, 169-180.  Abstract 

  •  Jachmann, H. (2002). Comparison of aerial counts with ground counts for large African herbivores. Journal of Applied Ecology 39, 841-852.  Full text 

  •  Kenya Wildlife Service and Tanzania Wildlife Research Institute (2010). Aerial total count Amboseli - West Kilimanjaro Natron cross border landscape wet season, March 2010. Full text 

  •  Moloo, S.K. (1993).The distribution of Glossina species in Africa and their natural hosts. International Journal of Tropical Insect Science 14,511-527. Abstract 

  •  Muturi C.N. et al. (2011). Tracking the feeding patterns of tsetse flies (Glossina genus) by analysis of bloodmeals using mitochondrial cytochromes genes. PLoS ONE 6(2): e17284.  Full text 

  •  Mwathe, K. et al. (2006). Elephants in Kenya's South Rift: Bridging the Information Gap. Report to the African Elephant Database of the African Elephant Specialist Group. African Conservation Centre.  Full text 

  •  Okoth, S.O. et al. (2007). Glossina pallidipes and host interactions: Implications of host preference on transmission risk of Rhodesian sleeping sickness in Kenya. Trends in Applied Sciences Research 2 (5), 386-394. Full text 

  •  Sasaki, H. et al. (1995). Blood meal sources of Glossina pallidipes and G. longipennis (Diptera: Glossinidae) in Nguruman, southwest Kenya. Journal of Medical Entomology 32, 390-393.  Abstract 

  •  Tarimo, S.A. et al (1984). Preliminary results of blood meal analysis from Glossina pallidipes at Nguruman, Kenya. Unpublished DITH Progress Report, ICIPE, Kenya.

  •  Waage, J.K. (1981). How the zebra got its stripes - biting flies as selective agents in the evolution of zebra coloration. Journal of the Entomological Society of South Africa 44 (2), 351-358.  Abstract 

  •  Weitz, B. (1963). The feeding habits of Glossina. Bulletin of the World Health Organization 28, 711-729. Full text 

Last updated 3 January 2013