There are 33 species and subspecies of tsetse flies (Diptera: Glossinidae) in Africa, infesting 36 countries south of the Sahara (Gooding & Krafsur 2005). Tsetse flies transmit trypanosomes that cause African trypanosomiasis (AT) in humans and livestock. The disease has negative effects on the hosts as it causes low productivity, leading to mortality when untreated. In livestock, some of the direct losses caused by the disease include abortion, reduced milk yield and poor calf crop. Given its role in transmitting AT in livestock and humans, the tsetse fly deserves special attention, particularly with regard to its response to different sampling methods in a variety of ecological settings. Studies on tsetse ecology and control have mainly relied on the availability of an efficient sampling method, a number of which have been described. The main sampling methods include the use of fly nets and fly rounds (Pollock 1982), sticky materials (Vreysen, Khamis & Van der Vloedt 1996), and fabric traps (Challier & Laveissiere 1973; Dransfield & Brightwell 1997; Ndegwa & Mihok 1999; Vale 1982; Vale, Flint & Hall 1986). Most of the traps used in the foregoing studies were developed in west and southern Africa for riverine and savannah tsetse species, respectively. Interestingly, the sampling studies carried out thus far have revealed variable efficiency of the traps in capturing tsetse flies. In view of this, the efficiency of traps to capture certain tsetse species has been enhanced through modification of various designs of traps for use against particular target species in relation to the environment (Ndegwa & Mihok 1999; Ndegwa, Mihok & Oyieke 2001).

Traps basically function through visual stimuli. In the field, however, the visual stimuli can be greatly obstructed by vegetation, particularly for forest species. In such cases, attraction of flies to traps is enhanced through the use of odour attractants. Three groups of natural odours have been determined so far from the host animals: those found in urine (e.g. phenols), breath (e.g. acetone) and skin secretions (e.g. octenols (Vale 1993). The response of tsetse flies to particular odours varies amongst species (Green 1986; Kuzoe & Schofield 2004).

Principally, tsetse traps are made up of blue and black textile materials and white netting. The blue colour has been found to be a visual stimulus or attractant to flies. Field observations on tsetse trapping show that all tsetse traps attract tsetse flies to land on them. However, the difference in their relative trapping efficiency is based on their designs, especially the entrance for flies into the retaining cage. This has been the basic reason for modifying conventional traps and designing new ones to suit the target species in different types of habitats. Considerable advances have been made in the development of efficient traps for tsetse flies, as trapping is increasingly being used for population suppression and control of tsetse flies. Currently, the common traps used for sampling and monitoring economically important tsetse species include the biconical trap (Challier et al. 1977), developed for sampling Glossina morsitans; F3 and Epsilon (Flint 1985; Green & Flint 1986) for Glossina pallidipes; NGU (Dransfield & Brightwell 1997) for G. pallidipes; NZI (Mihok 2002) for tsetse flies, horse flies, deer flies and stable flies; S3 (Ndegwa & Mihok 1999) for Glossina swynnertoni and the pyramidal (Goutex & Lancien 1986) for Glossina tachinoides. However, it has been found that the difference in trap efficiency is related to the behavioural differences between the species and varies between different populations of the same species. In some cases, minor modifications are required to improve the efficiency of a trap. For example, the S3 trap underwent a series of modifications before it was perfected (Ndegwa & Mihok 1999). Knowledge of the response of particular tsetse species to specific traps with or without odours is important for enhancing the efficiency of tsetse fly suppression operations and the formation of barrier systems used in tsetse control or eradication campaigns.

In Tanzania, despite the vast area that is infested by tsetse flies, tsetse trapping has mainly depended on traps developed outside the country, targeting different vegetation types and tsetse species, except for the mobile, scoop (Kuzoe & Schofield 2004) and sticky panel (Vreysen et al. 1996). The dominant tsetse species are the savannah tsetse species, which include G. pallidipes, G. swynnertoni and G. morsitans morsitans. Other species that are not widely distributed include Glossina austeni, Glossina brevipalpis, Glossina longipennis, Glossina fuscipes martinii and G. fuscipes fuscipes. Unpublished results (VVBD – Tanga) show that the response of tsetse flies to traps in Tanzania varies from one area to another even within the same species. It is likely that such variation could affect the results obtained from tsetse surveys, particularly data on fly density and distribution. This study was aimed at investigating the efficiency of different traps for different tsetse species so that, if need be, a single trap could be used for sampling different tsetse species if it proved to be efficient against several tsetse species. For example to trap G. swynnertoni, several traps have been used, including pyramidal traps (Malele et al. 2007; Mramba et al. 2013), Epsilon traps (Auty et al. 2012), rectangular cloth targets and small leg panels (Mramba et al. 2013). However, the comparative performance of several traps against the species has not been documented. Variations in response in relation to traps and fly species have a negative impact, especially when planning for tsetse and trypanosomiasis control. Furthermore, such variations could lead to underestimation of the production losses and public health harm caused by tsetse flies and tsetse-borne diseases. On the other hand, if one trap is found to be efficient against several tsetse species, then the cost of making several traps to suit several species present in an area could be avoided. Apart from sampling, efficient tsetse sampling traps have been used elsewhere as cheap control devices against the vector (Madubunyi 1988).

We report the results of studies carried out to determine the relative efficiency of different tsetse traps in trapping various species of tsetse flies in different ecological settings in the Serengeti ecosystem (Mara region), the western ecosystem (Kigoma and Tabora regions) and the southern ecosystem (Selous Game Reserve, which covers Lindi and the south-eastern part of the Morogoro region), and we recommend the most efficient trap for each species according to the ecological zonation in the three ecosystems in Tanzania. The data presented were collected from 2008 to 2012.

Five species of Glossina were recorded in this study. Proportionally, the species ranged in decreasing order from G. swynnertoni (55.9%) to G. pallidipes (31.1%), G. f. martinii (6.9%), G. morsitans (6.0%) and G. brevipalpis (0.2%). G. swynnertoni was the dominant species in the Serengeti ecosystem, whereas G. morsitans was the dominant species in the western ecosystem. Species dominance in specific regions is likely due to the ready and continuous availability of preferred hosts in those regions. Confirmation of preferred hosts is only achievable through blood meal analysis, which was outside the scope of this study. Combining an analysis of tsetse-trapping efficiency with an analysis of the origin of the blood meals for some of the most commonly used tsetse traps would add value and is a subject for further investigation.

On trap performance, the NGU traps caught tsetse flies at only 7 sites compared with the biconical and pyramidal traps, which caught flies in over 25 sites; however, NGU trapped more tsetse flies than the rest of the traps used in the study. Although this demonstrates the superior efficiency of the NGU trap, it is unclear whether the failure to catch flies at most of the sites was a result of inter- and intra-species differences in behaviour and response to NGU.

All the traps used in this study, with the exception of H and NZI, which trapped very few flies, could be used for sampling tsetse flies in the Serengeti ecosystem, although their efficiency varies. In the western zone, all traps except NZI and S3 could be used for sampling or trapping flies. In the southern zone, the mobile, biconical, pyramidal and sticky panel traps could be used for sampling or trapping purposes. Again, G. swynnertoni could be trapped by all the traps used in the study except the H trap.

The performances of the sticky panel and mobile traps showed no significant difference in the proportion of flies caught. The catches were 16% by sticky panel and 15.4% by mobile trap. It was noted, however, that the mobile trap can be used to catch G. swynnertoni, G. brevipalpis and G. morsitans, whereas the sticky panel is more suitable for catching G. swynnertoni. The suitability of the sticky panel for catching G. swynnertoni is consistent with the observations of Mramba et al. (2013), although they used a different type of sticky panel, that is, all blue-legged panels. On the other hand, NGU performed better in catching G. pallidipes, G. morsitans, G. swynnertoni and G. f. martinii. In this study, NGU performed best for G. pallidipes, and this concurs with earlier findings for this trap, which was developed for savannah species (Dransfield et al. 1986).

The pyramidal trap has been the trap of choice in many studies in Tanzania because of its simplicity in deployment; however, its trapping ability was not as superior as NGU, although it still was able to trap about 12.4% of the total flies in this study, second to NGU (amongst stationary traps) and slightly better than the biconical trap, which is usually used as a sampling device (Takken 1984) (Table 6). The suitability of biconical as a sampling trap (unbiased towards any one species) was demonstrated by its being one of the traps that trapped nearly all tsetse species in the study sites (Table 5).

It has been documented that the H trap is good for trapping G. brevipalpis (Kappmeier 2000), which is typically a forest species. However, in this study the performance of the trap against G. brevipaplis was not good; perhaps vegetation cover could have influenced the results in the present study. Trapping was carried out mostly in savannah wooded areas and not in forested areas, where the species is mostly found.

For G. m. morsitans, the mobile trap performed better, followed by NGU (Figure 1). When evaluating the overall performance of traps regardless of species, and by analysing the data considering the density of flies sampled regardless of species, significant statistical difference was found among the traps, with NGU having the highest sampling efficiency (Figure 2). This affirms the superior performance of NGU, as already discussed earlier. Only NGU trapped more flies as a stationary trap than the mobile trap, which sucks in any fly in the vicinity.

From the foregoing observations, it can be concluded that the different tsetse traps being used in Tanzania have varying trapping efficiencies, which in some cases seems to be dependent on the tsetse fly species being sampled and the ecological setting. However, pyramidal and biconical traps can be used for sampling or trapping tsetse flies, hence saving the cost of having different traps developed for each specific Glossina species.