FAO Home Page French version Spanish version Waicent Free Text Search

Use of attractive devices for tsetse survey and control

Chapter 4: Evaluating attractant devices

General principles

The effective use of attractive odours, traps and targets depends very much on the responses of the flies to these devices. Different species of tsetse respond very differently to particular odours, traps and targets and there is some evidence that even separate populations of the same species may not respond in the same manner.

In addition, the materials that traps and targets are made from can greatly affect the catch. A control programme may start by using certain materials which may later have to be substituted by others according to availability. These may perform worse, or better, and there may be a choice of materials to use. Hence before commencing full scale deployment, simple but accurate comparative experiments must be carried out to test the devices and their components.


Experimental designs for comparative experiments

The Latin square design
The first design to consider, which is also the most important, is the latin square design. If there are several different trap/target configurations available, the tsetse worker will wish to determine which is best for the target species.

A sample of each design to be tested can be put in different places and the catches compared; the main problem with this approach is that there are always big differences in catches between different sites, even if the same design is used in each. It will, therefore, not be possible to tell whether the differences in catch result from site differences or differences in the attractiveness of the devices.

Alternatively just one site could be used and a different device put there each day. The same problem arises, this time because catches vary from day to day, even in the same site and using the same device. Hence there are two unwanted sources of variation in catches, between sites and between days, as well as the main factor of interest which is the variation in attraction of the trap or target design.

The latin square experiment deals with just this sort of problem. If there are four different devices (A,B, C,D), four sites are used and the experiment is operated over four days. Each day the devices are moved round so that each is operated once in each site. This design, a 4 x 4 latin square, is shown diagrammatically below.

See Table

Note that a latin 'square' does not mean the traps have to be arranged in a square, but refers to the statistical design.


The design should be randomised as follows to prevent systematic errors. Firstly the order of the rows should be randomised. This can be done by writing the numbers 1-4 on four pieces of paper, putting them in a box and then taking them out one at a time. Say the numbers are taken out in the order 2,1,4,3. This means the square is rewritten, with the second row first and so on:

See Table

Then the columns are randomised in the same way, say in the order 4,1,3,2, so again the square is rewritten to give the final design for use in the field:

See Table

A latin square can be used to compare any number of combinations, three or more, providing there are the same number of sites and the experiment is done over the same number of days. In practice, large latin square experiments, more than 5 designs, are more difficult to operate and do have some theoretical disadvantages.

When an experiment like this is carried out, it is assumed that, although catches may vary between sites, the relative difference stays the same over the period of the experiment. In other words if site one is twice as good for flies as site two at the beginning of the experiment, it remains that way for all the days of the experiment. The longer the experiment lasts the less likely this becomes.


The other disadvantage of large latin squares, is that they still need to replicated, minimum of three replicates. This is partly for statistical reasons, but there is a more important reason if trap designs are being compared. It is very difficult to precisely standardise a trap design, and quite small differences in construction can make a big difference to the catch. If there is only one of each design, then it is particular traps that are being compared, rather than different designs with their inherent variability.

At least three of each trap design should, therefore, be made, and these used for three replicates at the same time. This is really only feasible for latin squares up to about 5 x 5, i.e. 3 replicates of a 5 x 5 square.

The latin square design is suitable if there are more than two treatments, trap designs, odours etc. But what if there are only two comparisons to make? This may happen if the effectiveness of odours with targets is being compared with electric screens. There may not be enough electric screens available to make more than two comparisons.

In this situation positions are simply alternated every day. This is known as a cross-over design. Each pair of two days constitutes one replicate, and at least five replicates should be carried out, i.e. total duration of experiment would be 10 days.


The use of Control
When any experiment is carried out, a control must be included as one of the treatments. For an experiment to compare trap designs, the royal blue and white biconical trap is often used as the control as it has been the most widely used trap type. Alternatively, if you are interested in comparing several types of locally made cloth as alternatives to an imported cloth that has been used for some time, the traps should be made from this imported cloth and used as the control.

If odours are being compared, a trap without odour makes a good control to start with. Once an effective combination of chemicals has been found and thoroughly tested, that can be used as the control, against which to check the addition, or substitution of other odours.


Selection of Trap/target sites
Care must be taken when designing experiments that treatments do not interfere with each other; in other words one trap or odour should not affect the catch of another. There are two types of interference that may occur:

(a) Direct interference
This occurs if the traps are close enough such that they are visible from each other to a tsetse fly, or if odours are being compared, and the wind is carrying the odour over another trap and hence affecting the catch.

This problem is solved by having trap sites far enough from each other. Tsetse cannot see a trap from more than about 20 m, so if odours are not being used, putting the traps about 100 m apart should prevent any direct interference. If odours are being used, positions should be at least 200 m apart, and preferably more in open country.

(b) Indirect interference
This occurs when there is a marked 'trapping-out' effect around a particular trap position. This will occur if a very efficient trap/odour is being used against a tsetse species that does not move very much, e.g. some of the palpalis group species in particular habitat. The best treatment may reduce numbers in one site and thus affect the catch by another treatment in that site the following day.

The problem can be solved by only trapping one day a week, so that flies have time to move back into an area. This has the disadvantage that experiments can take a long time to complete, for a 4 x 4 latin square it will take four weeks instead of four days, but it may be the only way to get good results for such species.

The assumptions of the latin square design also have to be taken into account when siting the devices. As has been said previously, the design assumes that the relative catch between sites stays the same over the duration of the experiment. Therefore traps should not be put near game or cattle trails, because catches may be temporarily increased by animals moving close to a trap on one day and not on another.

For the same reasons, all trap sites within each replicate should be put in the same sort of vegetation. If some sites are in the open, and some in dense vegetation, on a hot day flies may move into the dense vegetation, whilst on a cool day they may move out into the open. This will make it more difficult to demonstrate differences between the treatments.


Duration of experiment
It has been assumed up till now that the treatments are moved daily, but very few flies may be caught in only one day. If this is the case, and especially if some of the treatments catch no flies in one day, the traps should be kept in position for a longer period, e.g. only moved to a new position or rotated every three days. If however, a low catch is due to exceptional circumstances, e.g. heavy rain, that day could be immediately repeated. This is also a good way to deal with 'missing' data points, e.g. where a device has been damaged.

It has to be decided what time of day to remove the cage and rotate the treatments. If possible this should be done at a time of day when the flies are inactive, and always at the same time of day for each trap site. It is generally best to have one day as the minimum time before rotation, e.g. rather than just the afternoon, because the flies may respond differently to the treatments at different times of day and one is interested in the change in catch over the full activity period of the fly.


Comparison of odours

Types of comparisons
There are three types of comparative experiments that can be carried out on odours:

- comparison of release rates of a single odour;
- comparison of different odours;
- comparison of different combinations of odours.

When odours are being tested in a new area or on a new species, the above sequence should be followed, starting with a comparison of release rates. This is because for most odours there is an optimum release rate and if an odour at too high or too low a release rate is used, it may not increase the catch at all. Octenol for example decreases the catch at high dose rates.

Once the best release rates for various chemicals have been determined, they can then be compared directly in the same experiment. Various combinations of chemicals should then be compared. This is very important because some chemicals have little or no effect on their own, but may increase the catch when used in combination with other chemicals. This is true for octenol/phenols for G. tachinoides and acetone/cow urine for G. longipennis. In contrast some odours have little or no additional effect when used in combination e.g. two ketones together. Some odours such as octenol and the phenols can be mixed in the same dispenser provided their optimum release rates have already been determined, however, other odour baits such as cow urine and acetone are ineffective when mixed, and separate dispensers must be used.

Possible experimental designs for such combination trials are as follows:
1. unbaited
2. acetone
3. cow urine
4. acetone + cow urine

This would be done in a 4 x 4 latin square.

1. unbaited
2. acetone
3. octenol
4. phenols
5. acetone + octenol
6. acetone + phenols
7. phenols + octenol
8. acetone + phenols + octenol

This would be done in an 8 x 8 latin square, but remember the problems of using large latin squares.


Comparison of traps with targets
Either traps or targets can be used to compare the odour baits. If the odours are to be used for sampling, traps will be more appropriate. If they are also to be used for control, both traps and targets may need to be tested.

Odours may not necessarily work in the same way for traps and targets. This is because the odour may increase trap catch by improving the entry response to the trap, but have little effect on the number of flies landing on or colliding with a target. Hence the comparison must be appropriate for the way the odours are going to be used.

If traps are used, the comparison is straightforward. At least 10-12 traps of the same design will be needed, and the odours must be rotated each day between the different sites. If targets are used, electric screens will be needed to capture the flies and so make the comparison.


Determination of release rate
In any experiment with odours, the optimum release rate of the chemical or residue being tested should be determined.

For a single chemical, the simplest method is to record the volume in the dispenser before and after a set number of days. To get the release rate in mg/hour, this volume should then be multiplied by the specific gravity of the chemical, and divided by the number of hours over which the chemical was being released.

For example; 12 ml of acetone, used over 4 days
Volume used = 12/4 = 3 ml per day = 0.125 ml/hour
Weight used = 0.125 x S.G.of acetone (0.8) g/hour
= 0.1 g/hour = 100 mg/hour

If urine is being used, the specific gravity can be assumed to be approximately 1, and again the volume can be measured.

For mixes of chemicals, it is better to weigh the odour container at the start and end of the experiment, since the specific gravity will be unknown. However, a very sensitive balance may be needed to measure the difference, and the weight of each chemical that has been released will not be known. This can only be determined by knowing the initial weight of each chemical and then sending the odour dispenser to a specialised laboratory after the experiment.

If the optimal dose rate is fairly high, e.g. acetone at 100-1000 mg/hour, there should be a measurable change in volume or weight over a few days. If the normal release rate is very low, e.g. octenol at 0.5 mg/hour, the dispenser will have to be left in the field for several weeks before reweighing it to get the release rate. If sachets with mixtures of chemicals made up by other workers in the field are being used, they should be asked what the expected release rates are for a given temperature. Dispensers should be protected from rain and direct sunlight to avoid dilution, undue heating or damage by sunlight.


Precautions for odour experiments
- odour dispensers must be completely clean before being filled. The contents of each dispenser should be labelled clearly;

- if rain/sun covers are used on the odour containers, they should be moved round with the container as they may be contaminated with the chemical;

- if polythene sachet dispensers are being used they should be made a few days in advance and left in the open for 24 hours before the start of the experiment. This will give a more even release of chemical;

- a delay of about one hour should be left between moving an odour to a new position and putting on the trap cage in the new site. This will give time for the previous odours and hopefully any flies attracted to them to fully disperse and the new 'odour plume' to become established.


Comparison of materials and designs

Types of comparisons

There are three comparisons that can be made:
- comparison of one design using different materials;
- comparison of small modifications of the same basic design, e.g. the collecting device;
- comparison of different designs.

All too often comparisons of traps have been done by mixing all three factors, i.e. comparing trap designs made with different materials and with different collecting devices. The problem with this approach is that, if there are differences, the reason for them is unclear.

A suitable design for comparing materials would be as follows:
1. trap with all imported materials
2. blue material local, others imported
3. black material local, others imported
4. netting local, others imported
5. trap with all local materials

For this experiment the same collecting device and the same design should be used for all the traps.

For comparing different trap designs, all designs should be made from the same material, and if possible from the same batch of material. This is because the colour may vary between batches of the same cloth. As far as possible, the same collecting device should be used for each trap design.


Precautions with trap design experiments
- all the traps should be set in the same vegetation type and in similar sites, i.e. all shaded or all sunny;
- if the traps only have one entrance, e.g. NGU or epsilon traps, they should all be faced in the same direction. North or south is probably best so that the sun does not shine into the trap in early morning or late evening;
- all traps should be set at the same height above ground level unless an experiment has previously been made to determine the optimum height for each trap;
-the distance of odours from the trap should be standardised;
-all traps should be handled in exactly the same way; old cloth may give different results from new, as may recently washed compared to unwashed cloth.


Recording and analysis of data

Data recording
When treatments are being set and rotated, a book should be maintained in the field to record all actions taken so that mistakes are not made in moving the treatments to their new sites. Start and finish times for setting the trap and odour at each site should be recorded. It is easy to make mistakes when rotating treatments and labelling cages, so these must be closely supervised. There should be a section in the record book for comments to be made during collections. Helpful information should include notes on any game animals or livestock near the traps, ants in the cages or similar problems. Avoid transcribing field notes to a book at some later date.

A label should be put inside the cage of each trap with both the trap site number, the treatment letter and the collection date. When the flies are being counted, they should be separated by species, sorted into males and females, and tenerals/non tenerals if required, and then arranged in groups of twenty or fifty on the counting tray. The number of males and females of each species should be written on the back of the label, and then the count should be checked by someone else and the result written in the book. Careful checking is essential, because even one mistake could ruin the experiment.


Calculation of indices of increase
Once the experiment is completed, it may seem obvious that the first thing to do is to add up all the totals for each treatment and compare them to see which trap or odours caught most. This may give an initial indication of which is best but it may also be very misleading.

This is because the latin square design is a multiplicative model, not an additive model, as demonstrated by the following hypothetical example.

If there are four trap types (A-D), four trap positions and the experiment is being done over four days in a latin square:

See Table

If there were no differences between positions, between days and between trap types all the catches might be expected to be the same:

See Table

This is never the case because some positions are always better than others. It may be that position 2 is twice as good as position 1, position 3 catches the same as position 1, whilst position 4 catches half as many as position 1, note the difference is multiplicative i.e. half as good or twice as good; position 2 does not always catch 20 more flies or 40 more flies than position 1. The following catches would then be expected:

See Table

Catches also vary from day to day. Thus on day 2 twice as many flies may be caught as on day 1, on day 3, three times as many flies, and on day 4, half as many flies:

See Table

Lastly, the effects of the different trap types can be included so that type B is half as good as type A, type C is the same as type A and type D is three times as good as type A:

See Table

This then is the hypothetical result. If the numbers for each treatment are just added up, the following totals and arithmetic means will result. Each can then be divided by the mean number caught in A (60) to give the index of increase (I) relative to A:

See Table

As was said earlier, the arithmetic totals do give a rough indication of the result (type D is best, Type B is worst) but some of the indices of increase are quite wrong. The correct indices from our initial assumptions for C and D are 1.0 and 3.0 respectively, type C is the same as type A and type D is three times as good as type A.

To get the right answer, logarithms of each number in our table must first be taken. This can easily be done using a calculator (use the 'log' key) or a table of logarithms.

See Table

Then the totals are added up, each of them is divided by four and the antilog (often 10x key on a calculator) taken to get the detransformed or geometric mean. The indices of increase are then worked out using the geometric means.

See Table

This procedure now gives the right answer. Many people think the logarithmic transformation used here is just for the later statistical analysis; in fact it is the only way to get the correct results from the experiment. If there are any zeros in the catches, 1 should be added to all the catches before taking logarithms.


Statistical Analysis of data
Many biologists are reluctant to use statistics, and the data could be taken to a statistician to analyze. However, it is much better for those doing the fieldwork to learn how to interpret the results themselves because they are the ones who have gathered the data, and will know the problems associated with the experiment. A statistician may even do the wrong analysis if the design is not properly explained to him.

A detailed account of the statistical analysis is beyond the scope of this manual, although a summary is given in Appendix 1.

The reason a statistical analysis of the data is required is that even if there were no position, day or treatment effect exactly the same number of flies (in our example 40) would not be caught in each trap. Numbers would instead vary from trap to trap around an average number.

This is known as the inherent variability and occurs with all biological data. Hence even if there were no real differences between the trap types, some would still apparently be better than others. The aim of a statistical analysis is to determine whether differences are significant or not; in other words is it likely that the differences are real or could they just have arisen by chance.

One way this can be tested is to carry out an analysis of variance on the data to see if the variability between treatments is significantly greater than the inherent or residual variability. If it is, then the next step is to test the traps which seem significantly better Only when this has been done, can it be reasonably certain that a particular trap or odour is the best in a particular situation.


Electric nets

Uses of electric nets
Electric nets can be used for the following:

- comparison of the efficacy of different target designs which do not in themselves incorporate a catching device. This is the only way this can be done, other than using the highly expensive method of doing several full scale trials;

- to determine the attractive efficiency of different trap designs;
- to surround odour sources, live hosts or chemicals, to compare number of flies attracted;
- as a sampling device usually fixed to the back of a moving vehicle or person.

In this section the first three of these uses will be covered; the last is covered in Chapter 5. It should be remembered that electric nets are not easy to use, and it will take some time to get worthwhile results. If possible advice should be sought from someone who has experience in this technique.


Comparison of target designs
Target designs can only be compared using electric nets. Some attempts have been made to use a sticky substance on the targets to trap flies on contact, but this is not very efficient and invariably changes the appearance of the target. Its use is not, therefore, recommended.

The target should be stretched between the two wire grids, and wires should be added or removed as necessary to only cover the exact area of the target. Specially designed electric nets, the 'sandwich' or 'cassette' design, can be obtained which make it much easier to insert the target under test.

Different targets can be compared in this way using either a cross-over design, two target types, or a replicated latin square design, more than two target types. The target catch on the electric nets could also be compared with a trap catch, without electric nets.

These experiments are not without problems. The electric nets may change the appearance of the target and thus affect the catch. The electric net may also vary in efficiency depending on whether flies collide with the wires, as in the case of any netting component of the target, or attempt to alight on the material, e.g. black cloth.

However, despite these disadvantages electric screens are the only practical way to compare target designs, other than to undertake several large scale control trials and see how well the different target designs perform.


Estimation of trap efficiency
A clear definition of the term trap efficiency is essential. The most commonly used definition is to take the percentage of flies approaching the trap that are actually caught. Thus if we say a trap is 50% efficient, we mean that of every 100 flies approaching the trap, only 50 are actually caught.

Most traps are not very efficient. Some tsetse may fly around the trap or land on it, but not actually enter and be trapped. Many factors can affect the efficiency of a trap including not only its design but also environmental factors, time of day, temperature, etc. and the categories of fly approaching the trap, age, hunger stage, sex. Clearly, for control, traps should be as efficient as possible.

Trap efficiency can be measured using electric nets. There are two basic methods:

(a) Incomplete ring of electric nets
Electric nets are placed in an incomplete ring around the trap to be tested (see Fig. 22). Nets should be placed at a radius of at least 4 m from the trap, which will give a circumference of about 25 m. At least 20% of the incoming flies should be intercepted, which would require 5 electric nets each of 1 m width. To avoid catching tsetse flying around the trap before entering, the system can be improved by increasing the radius and using more, or wider, electric nets.

Fig. 22 Incomplete ring of electric nets used to estimate trap efficiency.

In this case, 20% of attracted flies will be caught on the outside of the net. Therefore 80% will get through to the trap. Of these some will be caught by the trap and the remainder will leave, of these again 20% will be caught on the inside of the nets.

If the trap catch is divided by four times the total catch on the outside of the screens, this will give an estimate of trap efficiency (method 1). Alternatively the trap catch can be added to four times the total catch from the inside of the nets, and then the trap catch divided by this total (method 2). These two methods should give the same answer. Since the screens are not 100% efficient, and some flies may fly over the top of the nets, the formulae to use are as follows:

Method 1.

Trap efficiency (E) = 100.y/(y + (x/e.p))

where y is the number of flies caught in the trap cage, x is the total number of flies caught on the inside of the electric nets, p is the proportion of the perimeter covered by the nets, and e is efficiency of the electric nets expressed as a proportion.

Method 2.

Trap efficiency (E) = 100.y/((z/e.p)-z))

where z is the total number of flies caught on the outside of the nets, and other symbols are the same as above.

These methods make several assumptions. The main one is that flies approach the trap only once, and that once they have entered the ring of nets, they either enter the traps or leave directly. In fact some flies may fly around the trap before entering, and these may get caught on either side of the nets. This will tend to give an underestimate of trap efficiency.

Another problem, especially if odours are used, is that flies will not be approaching the trap randomly from all directions but will tend to fly upwind to the trap. Provided they leave the trap randomly, method 2 will still give the correct answer. Lastly, the ring of electric nets around the trap may itself affect the number of flies approaching.


(b) Complete ring of electric nets
The first two of the above problems may be avoided by using a complete ring of electric nets and comparing the catch on the nets, divided by the electric net efficiency, with that in another trap not surrounded by electric nets (Fig. 23). With this design two trap positions must be used and the electric net arrangement moved each day between sites, a cross-over design; two days will constitute one replicate and at least three replicates should be carried out. Alternatively two trap designs, each with and without electric screens, could be compared using a 4 x 4 latin square design.

Fig. 23 Complete ring of electric nets used in a cross-over design to estimate trap efficiency.

Various factors can affect the efficiency of a given trap design for a particular species, and these should be taken into account when designing an experiment. The presence of an odour source may affect efficiency, so efficiency could be compared with and without an odour. Males and females may respond differently to the trap, as may different age or hunger stage categories of flies.

Trap efficiency can also be affected by a range of environmental factors, including the time of day, temperature, wind strength etc. There is little detailed information available on the effect of these factors, so any experiment to investigate them would be very useful.


Natural hosts and their odours
Electric screens have been used extensively for research purposes to look at the number of tsetse attracted by natural hosts, both without visual attraction when only the natural host odours are present, by concealing the animal in a pit, and when the host is also visible.

Ways have already been discussed to compare host residues and chemicals known to be effective for some species of tsetse. But for some other species, none of the odours so far identified have an effect on the catch, e.g. G. fuscipes. In this situation, it is worthwhile to test odours from the preferred natural hosts of the species concerned.

This can be done by constructing a pit within which to put the animal, so that it is invisible to the tsetse (Fig. 24). The pit has a ventilation system powered by a small fan which blows air from the pit through an electric net, or a trap, at ground level about 4 m from the pit. Alternating with and without odours will indicate whether the odours are attractive.

Since carbon dioxide will be present in exhaled breath from the animal, the next step is to use activated charcoal filters to remove all the carbon dioxide, expert advice should be sought on this as conventional filters are often not fully effective. If the odour is still attractive to tsetse, and none of the known attractants are effective, there may be a new attractant present in the odour.


Some of this odour then has to be 'captured', e.g. in special resins, the chemicals have to be identified and screened for activity. This is very specialised and can best be done in a well-equipped and specialised laboratory. The discovery of new odour attractants would be a major contribution to tsetse control.

It may also be of interest to know the number of flies attracted when the host is visible, e.g. if one is trying to quantify tsetse 'challenge'. Either a complete or incomplete ring of electric nets can be put around an animal, often a cow in order to compare the number of flies attracted to a baited trap compared with the number attracted to the cow. Although the number landing on the cow is difficult to determine, the number feeding can be estimated by looking at the number of fed flies captured on the inside of an incomplete ring of nets.

For this experiment the animal has to be restrained in some way to prevent damage to the electric nets. It can be tethered, but tethered animals tend to lie down which changes the visual stimulus. Alternatively, a small individual cattle crush can be constructed; this will hold the animal safely, but may also change the appearance of the animal to tsetse.

Fig. 24 Cross section through a ventilated pit used to investigate effects of host odours. The electric net could be replaced by a trap if appropriate.


Practical hints for use of electric nets
- the net must always be stored dry and protected from damage to the wires;

- they should be handled carefully during transportation and they are best carried vertically in strong boxes. This will reduce the chance of adjacent wires wrapping round each other;

- all nets should be checked carefully before the start of an experiment. Wires that are wrapped around each other must be separated, and any broken wires, old springs that have lost their tension, and broken ties holding the terylene netting in place, must be replaced. Shiny wires should be blackened with a felt pen, to make them less visible. Spare copper wire, springs and nylon loops should always be available, as these sometimes burn through;

- batteries must be fully charged at the start of each experimental session; pulsed units operated from poorly charged, small or defective batteries rapidly loose efficiency. The electrical system must be well insulated and all wires pegged to the ground to prevent people tripping over them when setting up or approaching the net;

- the high-tension units should be shaded from direct sunlight, allowing for good ventilation, or they may overheat and stop working. The most common results of overheating are that the ballast resistors or the storage capacitors become unsoldered. In some units the thyrister tends to 'latch on' if hot; this causes the ballast resistor to overheat, and could also damage the oscillator;

- pulsed units are much more dangerous than continuous ones, especially when used with water trays. The battery negative and screen base should be connected to the ground using metal pegs and the screen should never be operated upside down. People nearby should be warned when the nets are turned on, and made aware of the danger of touching wires when the screen is operating. If an insect caught on the wires needs to be removed while the screen is on, it can best be done safely with a dry twig;

- nets should be checked frequently to ensure they are working properly. Since the smell and sight of man or a moving motor vehicle may affect catches, this should be minimised by the person checking the nets going no closer than 20 m to ensure that the nets are sparking correctly;

- plastic covers should be readily available so that in the event of rain the nets can be quickly turned off and covered;

- water trays can be cleared very quickly using small rectangular sieves, but care should be taken not to scratch the paint on the bottom, nor to splash water on the net.


Table of Contents

Other Chapters


Comments: AGA-Webmaster