Chapter 5: Attractant devices for surveys

The objective of a survey must be clearly defined before the sampling technique to be used is decided upon. We can broadly divide the objectives into four categories:

(a) Extensive survey
The objective here is primarily to establish the presence or absence of flies over a wide area, usually to determine distribution limits rather than the level of abundance. Since it may often involve more than one species, it usually requires several sampling methods which can be readily and practically implemented over such large areas.

(b) Intensive monitoring
This is carried out in a defined area before, during and after a trial or control operation to assess the effectiveness of the technique, or to monitor the progress and success of field operations. This may again necessitate the use of several sampling methods since each will have its own bias. The methods should have high sensitivity so that low density populations can be detected, and may be combined with other techniques, e.g. ovarian ageing or mark-release-recapture, to improve our understanding of the effects of control on the fly population.

(c) Minimal monitoring
This is used to monitor the effectiveness of an established control technique. Where large areas are involved cost may become a major constraining factor and thus a method is required which minimises the cost relative to the actual control measures, yet gives sufficient data to assess whether the control is effective or not. Where traps are used both for control and monitoring, they can indicate areas of reinvasion pressure and those with insufficient trapping densities to achieve the desired result.

(d) Population dynamics/behavioural studies
These are undertaken to increase our knowledge of tsetse ecology and behaviour. Here any number of sampling methods may be used to maximise the information obtained, cost is not a major consideration since such studies are usually carried out over a relatively small area. In population dynamics studies, the sampling method should only kill a very small percentage of the individuals so as not to affect the overall population structure.

It is important at this stage to distinguish between collecting tsetse for laboratory work or estimation of trypanosome infection rate, and sampling to assess the distribution and/or abundance of tsetse in an area. In the former case, the traps are put where they catch the most flies. In the latter case a methodical sampling programme must be designed.

Back

General Principles
Because not all the tsetse flies in a population can be counted, aged, sexed etc. reliance has to be placed on the collection of representative samples. Ideally the absolute population density, the number of tsetse per unit area, could then be estimated. In practice this is often not possible, so instead the relative population density is determined, that is the number per section of fly round or the number per trap per day. Such measures are then used to compare apparent densities in different localities, or at different times of year in the same locality, to determine seasonal population fluctuations.

Ideally the sampling method would take a fixed proportion of the population, regardless of the vegetation type or the time of year, i.e. the catch would give a consistent indication of the actual number of flies present. However, when the method depends on the tsetse being attracted to a man or to a trap, the numbers depend not only the density of flies but also on their degree of activity. The activity of flies depends on environmental factors on age and physiological state, so the proportion of the population sampled may not be constant.

The sample may also not be representative of the population composition for the same reason. Certain categories of the population, age groups, hunger stages, pregnancy stages, are more active than others, and respond differently to the sampling methods. Hence samples tend to be biased in favour of certain sections of the population.

Back

Pattern of spatial sampling
Many species of tsetse change in their seasonal distribution. Generally, they may occupy dense vegetation in the dry season and extend out into more open areas during the rains. Such seasonal changes in distribution take place not only with savanna species, such as G. pallidipes, but also with riverine species, for example, G. fuscipes and G. tachinoides.

In this situation, sampling must be carried out over the full range of vegetation types that the flies occupy at different times of year. If sampling is only carried out in dense vegetation, there may be a marked decline in apparent density in the rains mainly because the population becomes spread out over a larger area. For minimal monitoring, it may not be feasible to keep traps where they catch little or nothing for much of the year, so changes in fly distribution must be allowed for when interpreting such data.

Sampling with traps can be carried out in two ways in this situation:
- stratified random sampling; to do this vegetation types are categorised and a number of traps located randomly within each;
- systematic sampling. This is the most commonly used pattern of tsetse sampling. Transects are cut to pass through all vegetation types and traps positioned at regular intervals along the transects.

To intensively monitor a tsetse control trial sampling should be done both within and beyond the control zone. In this way the percentage reduction can be estimated in relation to the natural numbers where there is no control. For savanna species such sampling should extend at least 5 km outside the control zone, since the control will affect populations well outside the area. For less mobile species such as G. fuscipes, an extension of two or three kilometres should be sufficient.

Back

Pattern of sampling in time
The next questions to be addressed are how often should sampling be conducted, and for how many days at a time. If the purpose is to determine the presence or absence of flies over a large area, objective (a), sampling should be carried out at least once during the dry season and again during or immediately after the rains. Sampling should continue for at least a month, but the longer the sampling is continued, the greater the chance of detecting low fly densities.It has been recorded that in extreme situations continuous surveying for over one year have been required to detect low density populations. In such situations the monitoring of livestock for trypanosomiasis should also be considered.

If the sampling aims are to detect changes in the level of abundance over time, objectives b-d, then there are two factors that must be taken into account:

- tsetse have a relatively slow rate of reproduction, so very rapid changes in the density of flies are unlikely, unless some unusual intervention occurs such as a bush fire, the introduction of insecticide or sudden invasions of flies;

- catches may vary considerably from day to day depending on fly activity in relation to climatic factors, their physiological state or the availability of hosts.

Recent work suggests that sampling once a month for a 7-10 day period is adequate to detect the major changes in density, and should be regarded as the sampling intensity required for minimal monitoring of a control programme. When testing new control techniques, it may be desirable to sample more frequently, e.g. every two weeks.

If intensive monitoring of a control trial is being carried out, there should also be some pre-trial monitoring. Ideally this should cover a full year to give information on the natural seasonal fluctuations in numbers, although this may not always be possible.

The techniques available for sampling adult tsetse flies depend primarily on exploiting their responses to colour, shape, movement and odour. In this manual, emphasis will be put on the latest developments in sampling techniques, especially the use of odours with fly rounds, traps, electric screens and artificial refuges.

Back

Hand nets can be used to capture tsetse flies that have been attracted to a man or a bait animal, or to an odour baited target. One of the earliest ways of sampling tsetse flies was for two or more men to walk along a marked path through the bush stopping at set intervals to collect tsetse flies using hand nets. This is known as a fly round. Details of this technique can be found in Vol 1, section 7.8.

The addition of odours makes this method more sensitive for certain species. A bottle of acetone with a release rate of about 500 mg/h carried at one end of a black cloth screen will increase numbers of G.m. morsitans. Traps are not very sensitive for this species.

The effectiveness of the man fly round may be considerably increased for many species through the provision of a domesticated animal, usually an halter led cow. This it is commonly known as an ox fly round.

Back

Recommended traps and odours
The recent developments in the use of odours and traps have been covered in detail in Chapters 2 and 3, and are summarised here in tabular form.

Table 5.1 Odours for various Tsetse species

It should be borne in mind that carbon dioxide exhibits good attractive properties to several species but has not been included in the above table due to its inconvenience for field use.

Table 5.2 The efficiency of traps for various Tsetse species

Back

Practical hints on the use of traps
- traps should be stored in a clean, dry place. Before they are set, they should be checked carefully to make sure there are no holes or tears in the material, particularly in the net cone or in the cages. One of the commonest problems with trap samples is that old cages with holes are often used and the data gathered in this way is worthless. If a hole is found in a cage it should be recorded as a missing data point;

- the siting of a trap in a survey or ecological study is different from siting them for control. For surveys, the aim is to find out where tsetse are present or absent and which vegetation types have the greatest numbers of flies at different times of year; for control it may already be known where catches are greatest and traps may, therefore, be concentrated in a predetermined area;

- vegetation should be cleared within a set radius around the trap. This helps to standardise visibility of the trap and minimises the possible damage by fire. Each trap position should be marked and numbered;

- once a monitoring trap has been sited in a given position, the position should not be changed. Moving a trap only a few metres can greatly affect the catch. This is known as the site effect;

- all traps should be set at the same height above ground level;

- the top of the collecting cage should not be directly above the exit hole from the trap into the cage; if it is, and catches are high, many flies will remain in the cone, because it is too dark above the entry hole. New cage designs avoid this problem, but Geigy cages should be set at an angle, (Fig. 3B);

- the caught flies must be adequately protected from ants, even though signs of ant damage, severed wings, legs, etc, are not visible, it does not mean that there is no problem. Some species of ants may remove the whole fly from the cage. The best protection is to coat all trap supports with car grease or, if safari ants are present, Stickem. Provided the flies are not required alive the cage can be treated with insecticide or a collecting bottle with a preservative can be used. Such chemicals, may, however, be repellent to tsetse and thus reduce the catch; grease has been shown not to affect the catch;

- if possible flies should be collected from the traps at the same time each day, preferably at a time when tsetse are inactive, e.g. very early morning. Any damage to the trap or odour dispensers should be recorded, and a needle and thread, or stapler, should always be carried to repair any minor damage. Cages must be checked for holes every day before putting them on the trap, and odours replenished when necessary;

- the base of the lower cone of a biconical trap, and the shelf of an F3 trap, should be checked for any dead flies that may have dropped down. These should be put into the cage for counting. If there are live flies in the cone that have not yet entered the cage, they can be caught and included in that days catch or left to enter the cage to be counted the next day;

- a label giving trap position should be put inside the cage, and the cages kept in a box, preferably covered with a wet black cloth to keep the flies quiet, especially if they are to be dissected later;

- lastly, in field work there are many factors affecting the number of flies caught over which we have no control, e.g. climate, animals etc. This means more care has to taken, not less, to minimise the sources of variability, such as holes in cages, over which one does have control.

Back

Relationship between trap catches and absolute density
In many situations it would be very useful to know how daily catches compare with the real or absolute population density. If traps are sited in all the vegetation types occupied by the flies, and the trapping intensity is related to the area covered by each vegetation type, i.e. if most of the area is covered by open woodland, most of the traps are in this vegetation type, then there is a good chance that the changes in catches are mainly related to changes in density.

Several studies on G. pallidipes in Kenya, G. pallidipes and G. morsitans in Zimbabwe, and on G. palpalis and G. pallicera in Ivory Coast have shown quite good correlations between trap catches and estimates of population size by mark-release-recapture, (see Vol 1.7.7.). Such relationships may not, however, be linear. It would be unsafe to use these relationships to predict absolute densities in other areas, because they will depend on many factors, especially on the distribution of traps in vegetation types and particularly if the fly distribution is uneven.

Back

Sampling bias with traps
All methods of sampling tsetse are biased and traps are no exception. Bias can result from:

- flies of certain categories being less active, and therefore less likely to encounter the trap;

- flies of certain categories responding differently to the trap once they have encountered it.

For a given species, the sample may be biased with respect to sex ratio, age category, pregnancy stage and hunger stage. It is useful to understand and measure the biases of different sampling methods, e.g. low catches in cold weather may be due to the flies being inactive or low trap efficiency rather than a small population.

(a) Sex ratio
Most traps catch a higher proportion of females than other sampling methods, especially fly rounds. This does not mean that traps are necessarily unbiased in this respect, only that they are less biased than other sampling methods. In fact electric screen experiments have shown that most trap designs are more efficient for males than they are for females.

Mark-release-recapture studies in Cote d'Ivoire showed that whilst biconical trap catches give an unbiased estimate of the percentage females of G. palpalis in villages, they underestimate the percentage females of G. palpalis and G. pallicera in plantations.

Different trap designs vary in their sex-ratio bias. For G. pallidipes, NGU and F3 traps catch a higher proportion of females than do biconical traps. The same applies for G. longipennis; in biconical trap samples females make up only about 25% of the total compared to 50% in NG2B traps.

Back

(b) Age composition
It is generally accepted that, for some species, very young flies are under-represented in the sample. However, this may be because they are less active than older ones. Electric screen experiments have shown that the biconical and NGU trap efficiency for tenurial G. pallidipes is the same as for other age categories. For 4-9 day old flies, the efficiency is actually slightly higher than for older flies.

It has recently been suggested that not only very young flies are less active. With savanna species, activity may increase gradually for the first 40 days of adult life, resulting in undersampling of all the younger age groups. If this is the case, estimates of mortality rates derived from age distributions would be underestimated.

Again, different trap designs may give quite different age compositions. Biconical trap samples of G. longipennis usually have a high proportion of 0-9 day old flies, at least 20%, but NG2B trap samples have less than 10% of this category.

Back

(c) Pregnancy stage
The later pregnancy stages are usually under represented in trap samples, probably because they are less active and/or are responding to larviposition site stimuli rather than to hosts.

d) Hunger stage
Once flies have fed they are inactive for a period of time, and therefore unlikely to be trapped. After this period they become more active and hence more vulnerable to capture; whether this is a sudden or gradual change in the fly behaviour is still not clear. Once they encounter a trap, their entry response may also depend on their hunger stage.

The result of these biases is that samples are composed predominantly of flies in the latter part of the hunger cycle. Much can be learnt about fly populations from looking at the hunger stage using fat-haematein analysis, but there are differing opinions on the interpretation of such data.

Back

Advantages and disadvantages of traps
The advantages of using traps for survey are:

- they provide a standardized system of sampling that is not dependent on the varying ability of people to catch flies with hand nets;

- they can be used for species that are somewhat repelled by the odour of man, G. pallidipes and G. morsitans, without catches being affected.

- compared to fly rounds, they catch a more representative sample of the two sexes, with a higher proportion of females;

- they can operate over the full activity period of the fly, rather than being restricted to only the morning or the afternoon. If the activity period changes, depending on climatic conditions, methods that only sample for part of the day can give misleading results;

- they provide a relatively cheap method of sampling that can be managed by only a few staff;

The disadvantages of traps are:

- traps have to be well constructed and maintained to provide a standardised sample;

- like all sampling methods, trap samples are biased and sufficient information is not yet available to correct the data accordingly;

- stationary traps are very dependent for efficiency, on the site chosen.

- effective traps are still not available for some species, e.g. G. austeni, and they are not sensitive enough to readily detect low density populations of other species, e.g. G. morsitans.

Back

Stationary electric nets
Stationary electric nets supplied with an odour can be used for sampling species which are reluctant to enter traps. They are especially useful for looking at the daily activity patterns of tsetse. Traps can give a misleading picture of activity patterns because:

- trap efficiency varies through the day;
- some species are crepuscular, active at dawn and dusk, and may not be able to see the trap when light intensities are low.

To improve the efficiency of electric nets in such situations it may be advisable to erect a 1 m2 black or blue target to enhance the visual stimulus, together with an odour, if appropriate.

Mobile electric nets
Two main devices have been used:

(a) Electric back packs (+ continuous HT unit).
A portable electric screen (Fig. 25) is strapped to the back of a man who then walks along a fly round; a modified back pack has been designed, the COPR electric trap, that can be held close to the ground to catch flies approaching below waist level.

Fig. 25 Backpack mounted electric screen for monitoring. A. three dimensional view; B. side view.

Electric back packs have been used to sample G.m. submorsitans, G. tachinoides, G. palpalis, and G.m. centralis. They have also been used for behaviour studies on G.m. morsitans and G. pallidipes. The back pack can be worn all the time, or it can be taken off and held close to the ground at regular stops along the fly round, mainly for G. palpalis.

Electric back packs are more efficient for catching flies attracted to man when densities are high, but may be less so at low densities. The female percentage is similar to hand net catches for G. palpalis but higher for G.m. submorsitans.

(b) Vehicle electric nets ( + pulsed HT unit).
An electric net is fixed at the back of a motor bike or in the back of an open backed vehicle which is then driven slowly along a sampling transect (Fig. 26).

Fig. 26 Vehicle mounted electric screen for monitoring G. morsitans

Advantages and disadvantages of electric nets

The advantages of electric nets for survey are:

- they provide the most sensitive method of sampling G. morsitans when put on a motor bike or a larger vehicle;

- the efficiency of catching flies by electric nets is less dependent on the number of flies attracted than is catching them by hand nets.

The disadvantages are:

- they require a much higher level of maintenance than other sampling methods, and tend to break down frequently under field conditions;

- they are more expensive than other sampling techniques;

- although their efficiency was formerly thought to be close to 100%, it now seems that the figure is closer to 50% and some flies may actively avoid the nets;

- electric nets can only be used when the weather is dry;

- vehicle mounted nets can only be operated where access is suitable.

Back

Artificial refuges have only been used for savanna species, namely G. pallidipes and G. morsitans and recently for G. longipennis. There is evidence that they give a less biased sample of the age distribution, and can also be used to collect recently fed flies for blood meal analysis.

They cannot be used to give a measure of apparent density because the number of flies entering is directly dependent on the temperature, flies only seek refuges above about 32^oC, and the higher the temperature, the more flies will enter.

Back

Counting of flies
The same points apply here as were made in the section on methods for comparison of odours, traps and targets (Section 4.5.1). The flies should be separated by sex, and by tenurial/non-tenurial if required, and put in groups of 20 or 50 in the counting tray. The result should be written on the back of the label, and the count should always be verified by a second person before being recorded.

Recording and analysis of data
The number of flies caught for each trap position, each day should be recorded. Data for male and female captures should be analyzed separately.  The total of males and females should be added up for each day, either over all trap positions, or in each of the vegetation types orareas. Arithmetic mean catches per day per trap are then calculated for the full sampling period.  The data can be displayed as a graph of numbers, on the vertical axis, against time, on the horizontal axis.

If a control operation is being monitored, numbers will often be declining at a constant rate.  For example if the population declines at a constant 50% per month, and sampling is done each month, the trap catches might drop from 2000 to 1000 to 500 and so on.

If the rate of decline is constant and the numbers are plotted on an arithmetic scale on the y axis, the decline will not be linear(Fig. 27A).  Instead numbers will go down in a series of steps that get progressively smaller with time. If, however, a logarithmic scale is used, the plot will be a straight line (Fig. 27B); deviations from this straight line indicate that numbers are not dropping as they should, possibly because of reinvasion pressure.

There are two commonly used types of logarithms, logarithms to base 10 (log₁₀) and natural logarithms (log_e -  e is a constant equal to 2.71828).  For the plot of numbers against time, natural logarithms should be used.

The slope of the line (b) then gives the instantaneous daily rate of reduction of the tsetse population.  The best way to calculate the slope is to do a regression analysis. This is a statistical analysis which gives the equation of the best-fit straight line to the points. A worked example is given in Appendix 1.

Once the slope of the line (b) has been determined, the percentage the population falls each day, the finite daily rate of decline, can be estimated from the following formula:

Finite daily rate of decline = 100(1-e^b)%

To calculate how long it will take at this rate to get a given percentage reduction in numbers (p), the formula is:

Number of days = (log_e (1-p/100))/b

So for a 99% reduction, with a value of b of 0.04, a 3.9% decline per day, it would take ((log_e (0.01))/0.04) or 115 days.

Fig. 27 Two ways of expressing reduction in tsetse numbers A. arithmetic scale; B. logarithmic scale.

Back