Automated tephritid fruit fly semiochemical mass feeding structure: design, construction and testing


Lin T. Tan (corresponding author), Department of Civil, Environmental & Geomatic Engineering, University College London, London WC1E 6BT, UK. E-mail:


Research has shown that exposure to semiochemicals significantly improves the mating performance of males from some of the most economically important agricultural tephritid fruit fly pests species. There is consequently a growing awareness of the need to include a pre-release semiochemical treatment in sterile insect technique pest control programmes, to ensure optimal sexual performance of sterile male flies vis-à-vis wild males and to ensure their survival in the wild. This is, however, not a matter of simply spraying flies with the semiochemical as several species (including Bactrocera dorsalis and Bactrocera carambolae) need to actively ingest and metabolize the chemical for it to have the desired effect. Currently, no economical or efficient method exists for large-scale administering/feeding of semiochemicals to such species. This may be due to the complicated requirements of (i) administering the semiochemical only to flies that are close to sexual maturity, (ii) assuring fixed-dose feeding to avoid damaging over exposure to flies and (iii) segregating semiochemically satiated from non-satiated flies prior to release. Using an engineering approach, while leveraging biological behaviour patterns (such as attraction to the semiochemicals and light), a mass semiochemical feeding structure was designed that fulfils all three requirements. The aim of the design was furthermore to maximize survival of flies and to minimize semiochemical wastage and hence cost. To validate the proposed design and to ascertain the behaviour of insects interacting within such a structure, a prototype was constructed and tested for B. dorsalis (with methyl eugenol as the attractant) and Ceratitis capitata (with citrus oil as the attractant). Our results show that the used semiochemicals and drivers, especially the novel conveyor belt system, are effective in encouraging the flies through the facility with minimum damage and without backward motion of flies.


One of the most environmentally friendly insect pest control methods ever developed, the sterile insect technique (SIT), involves the release of large numbers of sterile male insects in targeted areas to mate with wild females that will then yield non-viable offspring (Enkerlin 2005; Hendrichs and Robinson 2009). Numerous studies have shown that exposure to semiochemicals significantly improves the mating performance of males of several tephritid (Diptera) species, including Bactrocera dorsalis (Hendel), Bactrocera carambolae Drew and Handcock and Ceratitis capitata (Wiedemann) (Shelly and Dewire 1994; Tan and Nishida 1996, 1998; Katsoyannos et al. 1997; Shelly 2000; Papadopoulos et al. 2001; Shelly et al. 2004; Wee et al. 2007). These species are also some of the world’s most economically important agricultural pests. It is therefore important to expose sterile males to the relevant semiochemicals before releases, in order for them to compete more effectively with wild males. However, this is not a simple matter of spraying the insects with the semiochemicals or including the chemicals in the insect diet. In the case of methyl eugenol (ME), the semiochemical attractant for B. dorsalis and B. carambolae, direct bodily contact with this oil-based chemical is harmful as it can block the spiracles and tracheal system of the flies because of its low surface tension. In terms of a dietary supplement, experiments with B. dorsalis have shown that adding ME to the larval diet is ineffective (Shelly and Nishida 2004). Ideally, the semiochemical should be administered through voluntary feeding of males close to sexual maturity (Tan and Nishida 1996, 1998) and release (Tan et al. 1987). Thus far, only small-scale experiments have been carried out, with the ME being administered through netting (Wee et al. 2002), on small pieces of blotting paper (Khrimian et al. 2006) or on cotton wicks (D.O. McInnis personal communication).

Given that larger SIT programmes produce over 80 million sterile males per week (USDA 2009), a large-scale semiochemical feeding structure must therefore be capable of processing more than 0.5 million adult males per hour. Feeding of semiochemicals on such a massive scale in a fully automated, timed and fixed-dose manner has never been performed. In our approach, we combined the biological attraction of fruit flies to semiochemicals and light with the design of one-way funnel-shaped valves, conveyor belts, guidance flaps and airflow, to lure flies through the semiochemical feeding system with minimal to no damage. While we used ME with B. dorsalis and citrus oil with C. capitata as the respective semiochemicals in our model, and subsequently successfully validated it with these species, the structure is designed to be used with several different species of fruit flies. This paper is presented in three main sections: the first detailing the design concepts and the challenges faced, the second describing the construction of the prototype and third the results of testing the small-scale structure with live insects and their respective semiochemicals.

Design Concepts of the Semiochemical Feeding Structure

The feeding structure was designed to have systems that are variable and adjustable to ascertain the optimal conditions. Fruit fly attraction to semiochemicals and light together with directional airflow was used as tools to lure flies through the system, while devices such as funnel holes/valves, brushes and air jets were used to ensure one-way progress of flies throughout the structure.

The structure comprises four main sections (fig. 1), three of which (A, B and C) contain insects at some point during the system operation and the last section (D) containing fans for venting and a light source. Adult flies emerge in an emergence box which is inserted when flies are close to sexual maturity into section A. Attracted by the scent coming from the semiochemically soaked conveyor belt, maturing males crawl through the funnel valves into the semiochemical feeding area (section B). Once males start leaving section A, the moving belt is turned on. It moves constantly and is timed to carry flies from the start of section B into section C in 7 min, the maximum time period defined for semiochemical feeding by male flies (Hee and Tan 1998; Wee et al. 2002).

Figure 1.

 Model fruit fly semiochemical feeding structure: top – 3D schematic representation. The numbers 1, 2 and 3 represent the different routes flies can take to exit section B.; bottom-sizing of prototype with side view.

For a full-scale automated SIT semiochemical feeding structure, adult fly cages with maturing males of the relevant age can be automatically positioned at section A, while cages of semiochemically satiated male flies can be removed from section C, held for the required time to metabolize the semiochemical and then readied for aerial release. This automated process will enable the continuous production of large numbers of semiochemically satiated sterile males.

The four key elements of the design for the semiochemical feeding structure are (i) fly separation; (ii) male feeding; (iii) subsequent metabolism by males of consumed semiochemical and (iv) knockdown for collection in case of aerial release.

It should be emphasized from the start that as the prototype structure is totally dependent on electrical, lighting, airflow, conveyor belt rotational movement, to avoid flies escaping during a power outage, the semiochemical feeding structure requires a backup power supply.

Fly separation

Separation of flies is required at three different stages (see fig. 1), i.e. between:

  •  male and female adults – occurs between sections A and B;
  •  sexually mature and immature male adults – occurs between sections A and B;
  •  semiochemically satiated and non-satiated males – occurs between sections B and C.

Although the majority of C. capitata SIT rearing facilities use sexing strains to obtain only male flies for release, no Bactrocera rearing facility currently uses male-only strains. Thus, separation of Bactrocera males and females between sections A and B will be essential.

Using Tan’s (1985) principle for clear/transparent fly traps, separation at these stages is achieved by the use of separation walls with cylindrical holes for the flies to crawl through (fig. 2), i.e. flies not responding to the semiochemical do not cross over through the valves because there is even lighting in A owing to the clear/transparent wall.

Figure 2.

 Design of funnel separator walls AB and BC of the semiochemical feeding structure.

Separation between male and female flies and between sexually mature and immature males takes place at the separator wall between sections A and B as only the sexually mature males will be attracted to the semiochemical and hence move into section B, where they can feed on the semiochemical. Experiments have shown that with the added use of different light intensities, the flies are discouraged from moving back into the preceding section. This ‘one-way’ system is necessary to prevent inefficiency and waste of sterile males.

Separation between semiochemically satiated and non-satiated males occurs at the separator wall between sections B and C as the light and airflow aimed towards section C will direct males into section C.

The structure is designed so that subsequent sections are progressively brighter, and hence, once flies have passed through a separator wall, the differences in light intensity will make those walls appear to be very dark, thus preventing flies from backtracking.

To further encourage the flies through the system, a constant airflow (at low rate) is imposed in the direction of A to C. With suitable design, this airflow can be accelerated as it reaches the holes and hence push the flies through the holes faster. Furthermore, the windward side of the separator sheet is opaque black to enhance the light appearing through the holes and hence to attract the flies specifically to these holes.

As the structure will have airflow provided by extraction fans, a curved profile of funnel-shaped holes provides a smoother airflow around the hole gradually increasing the air speed as the fly approaches it. This is predominantly attributed to laminar flow characteristics with an attached boundary layer along the surface. Sharp edges can cause detachment of the boundary layer causing the onset of turbulent flow, which could potentially disrupt the flight path of the insects.

Male feeding

In designing the semiochemical feeding stage, the following points were considered:

  •  only males close to or at sexual maturation will feed on the semiochemical;
  •  all male flies feed until satiated (up to 7 min for B. dorsalis on ME);
  •  the use of the semiochemical should be kept to a minimum; therefore, feeding should be limited but still satiate the flies;
  •  direct spraying or excessive bodily contact with semiochemicals can be fatal for flies (at least for ME).

For B. dorsalis, ME ingestion takes place through the male proboscis (approximately 1 mm long); the fly requires between 45 s and 7 min to reach satiation, ingesting approximately 0.1 μl (Hee and Tan 1998; Khoo et al. 2000; Wee et al. 2002). Precaution needs to be taken as certain semiochemicals are not soluble in water, and high doses or physical contact can be lethal to the flies. Hence, a system that will transport the flies from the emergence box placed into section A to the holding box placed into section C while facilitating the timed feeding on the semiochemical is required. We propose a conveyor belt design (fig. 1) as being the most efficient method to continually administer the semiochemical, whereby the belt gradually transports the feeding flies from the entrance (AB) to the exit (BC) of the feeding chamber B, and the speed of the belt can be timed according to the maximum permissible feeding duration. The fully assembled prototype (without conveyor belt motor, emergence box inserted in section A and holding box C ready for insertion in section C) is shown in fig. 4 prior to painting. To provide some physical separation between the flies and the semiochemically saturated conveyor belt, it is covered by fine mesh netting with holes <1 mm2, allowing flies to settle on the mesh and extend their proboscises through the mesh to feed on the semiochemical on the belt. The belt is semiochemically replenished throughout the cycle by passing through a bath. In our current design, only the floor of section B is fitted with a conveyor belt; however, to further maximize the surface area for feeding, also the two sides of the chamber may also be fitted with conveyor belt systems, the ceiling being left clear for inspection purposes.

Transporting of males to holding area and removal from the conveyor belt

Once satiated, the males should leave the belt and head towards the light coming through the BC funnel valves/separator wall as section C is considerably brighter than section B. In addition, they can be transported to section C by the conveyor, as the conveyor belt not just administers the semiochemical to the males but also provides a mode of transport to take the males into the holding area C. However, initial experiments showed that voluntary movement of flies from section B to section C is slow, with the majority of satiated flies settling on the belt or the stationary walls of the chamber. Hence, rather than relying solely on insect behaviour, we investigated and implemented two further concepts to move the flies more efficiently off the conveyor belt and safely into section C. These involved a startle brush and a guiding air jet.

The startle brush is a soft brush mounted at the end of the conveyor belt (fig. 3) to startle the males, once in section C, off the belt; it also acts as a seal preventing flies from following the belt around and escaping. A similar seal is mounted at the beginning of the conveyor belt under wall AB to prevent flies escaping back also at that end.

Figure 3.

 Design detail below separator wall BC of the semiochemical feeding structure, showing guidance flap and startle brush.

The air jet concept uses a controllable jet of air to gently blow any remaining flies off the belt. Compared to a solid brush and provided it is adjusted correctly, the controllable air jet can minimize impact and damage to the flies. Although the idea is simple in theory, in practice it is difficult to adjust the direction and the velocity of the airflow so that flies are shifted in the right direction without air disturbances startling the flies too soon, and forcing them back into section B, or the air jet being too weak and allowing the flies to fly against the airflow and hence escape from the system.

To prevent males already in section C from moving back into section B by flying through the clearance area above the conveyor belt, a transparent air guidance flap is connected to the last funnel separator sheet (separator BC) that separates the holding area C from the ME feeding area B (figs 1, 3 and 4). The flap is adjusted to a height above the belt slightly larger than the height of a fly. This allows flies settling on the conveyor belt netting to pass below it into section C, while leaving too little space for insects in section C to fly in the opposite direction and re-enter section B. To prevent flies from being startled by the proximity of the guidance flap, it is manufactured from clear plastic so as to appear invisible to the flies.

Figure 4.

 Assembled prototype feeding structure (without belt motor) prior to painting, and with emergence box A inserted and holding box C ready for insertion.

As the males on the belt pass under this first air guidance flap, they will encounter the startle brush at the end of the belt (fig. 3), which is used to startle them into flight. As they fly off, they will head upwards initially. However, as the guidance flap only leaves a very small clearance and the directional airflow is towards section C, the flies will have to take a forward flight path into the holding box C rather than dropping down low to get out from the second adjustable flap (fig. 1, top).

Thus, the model semiochemical feeding structure described here combines light, funnel valves, airflow and conveyor belt motion to lure flies from feeding area B into holding area C. It also provides three routes that a fly may take to ultimately end up in the holding area (see fig. 1, top):

  •  Route 1: the semiochemically satiated fly leaves the conveyor netting to go through the funnel valves of the wall separating sections B and C (guided by airflow and light);
  •  Route 2: it remains on the conveyor netting and continues under the adjustable air guidance flap, and then, the extra air jet encourages it to leave the conveyor belt and travel with the airflow into the holding section C;
  •  Route 3: it persists to stay on the conveyor belt netting, until it comes into contact with the startle brush that triggers it to fly off the belt and travel with the airflow into the holding section.

This contingency approach is aimed to ensure that the flies move in a controlled and predictable fashion from sections A through B to C, and arrive, with no human intervention, in holding area C as semiochemically satiated flies.

To further enhance the effectiveness of the model and to encourage flies to land on the belt instead of on the walls, all surfaces in section B, except for the belt and the leeward side of the funnel separator walls, were covered with a thin layer of silicon oil; all wooden surfaces were covered with clear adhesive plastic prior to being silicon-coated.

Metabolism and knockdown

The holding box that is inserted into section C is designed to provide the maximum surface area for the flies to rest after feeding in order to be able to metabolize the semiochemical before release. The vertical side of the box facing the conveyor belt (adjacent to separator wall B–C) is a removable plastic shutter, while the side facing the fans (adjacent to section D) holds vertical frames that extend into the holding box (fig. 1). All the other sides of the box are covered with netting to allow maximum air flow. Depending on the number of frames included, the surface area can be doubled, tripled or quadrupled. The removable frames may be coated with a food substrate to sustain the flies. Once full, the holding box is removed from section C and replaced with an empty holding box. The semiochemical satiated flies are then left to metabolize the semiochemical.

When fly knockdown is required for aerial release, release boxes are placed below the holding box, the base of the holding box is removed, and cold air or carbon dioxide is applied from the top of the holding cage to immobilize or anaesthetize the flies. As the holding box has several sides of netting (except for the removable sides), temporary covers are applied to all sides to prevent leakage of the anaesthetic. When the anaesthetic is administered, the flies in the holding cages become drowsy and drop down into the release boxes.

Construction of Prototype

For ease of testing in the University College London (UCL) insect laboratory, the prototype feeding structure had to be no longer than 1200 mm in length. The conveyor belt length was set to be half the total length (600 mm). The width of the structure was set at 170 mm, with 170 mm square areas at each end, respectively, to insert an emergence box in section A and a holding box in section C. Together with the spacing for the extraction fans and the electrical control console, this gave an initial size for the structure of 1200 mm in length and 170 mm in width; the height of the structure was 400 mm (fig. 1, bottom). The structure is mounted onto a wooden backing board to secure the unit for easy transport. The electrical controls for the fans, lighting, drive motor, video equipment and plug sockets for power are all mounted on the backing board.

The design for the funnel holes has to satisfy three different criteria:

  •  it has to provide an attractive light source to utilize the flies’ phototactic behaviour,
  •  the holes have to be long and narrow enough so that the flies cannot turn around and crawl back out once they have entered the holes – the ‘no return’ valve concept by Tan (1985),
  •  they also need to incorporate the curved funnel profile for smooth laminar airflow through the holes.

The thickness of the plastic used for walls AB and BC is 23 mm (fig. 2). The cylindrical length of the funnel holes is 10 mm in diameter and 16 mm long, leaving 7 mm for the funnel curvature on the windward side of the funnel to be formed. The funnel starts at a diameter of 50 mm narrowing to the diameter of the cylindrical section (10 mm). The funnel holes are spaces a distance of 60 mm apart. In wall AB, the curvatures face area A, while in BC they face area B (fig. 1).

The transparent guidance flap hinged to the lower end of wall BC (fig. 3) allows the angle of the airflow to be adjusted for optimum performance. The maximum clearance is 30 mm and the minimum approaches zero. The startle brush is mounted under the base of area C at the end of the conveyor belt. A similar seal is mounted also at the beginning of the conveyor belt under wall AB.

Semiochemical feeding conveyor belt

The speed of the conveyor belt with 600-mm-long feeding area is determined by the maximum feeding time of 7 min. For a conveyor belt roller radius of, e.g., 20 mm, the required roller speed is 0.68 revolutions/min; hence, the roller completes 4.8 revolutions in 7 min. As the netting which rests on semiochemical saturated belt depends on the belt to stay in alignment, it is of paramount importance that the belt has the correct mechanical properties to perform well while wet. Four different materials (three cottons – poplin, denim and twill and one linen) were tested to decide which had the most suitable mechanical properties. Samples measuring 30 cm2 were soaked in water for 2 h, from which the liquid retention (absorption) and the resistance to stretching were tested. From these tests, twill was found to have both excellent absorption and in-plane resistance to shear. The twill sample was also tested in a warm machine wash at 40°C and showed no shrinkage. The edges of the conveyor belt are coated in resin to prevent fraying.

The simplicity in the conveyor belt design lies in the fact that the netting/mesh layer relies on the conveyor belt to move and thus is always synchronized with this belt. To keep the mesh layer sitting securely on the soaked belt while this is moving and also to separate the mesh layer from the semiochemical bath, a weighted roller is added just underneath the semiochemical bath (fig. 1). The belt is 1370 mm in length and 170 mm wide. To keep the belt aligned, guidance ribs or cords are sewn into the long edges of the belt which then move in grooves cut into the rollers.

Emergence and holding boxes

The emergence (A) and holding (C) boxes are designed to be able to slide into place in sections A and C of the feeding structure and are made of wooden frames and gauze mesh on up to four faces. A plastic sliding shutter (using grooves in the wooden frames) closes the open side that faces the separator walls, sealing the container completely. The holding box C is similar to the emergence box A, except that in box C the base of the wooden frame sits behind the plastic shutter to allow flies more space to enter the box. When filled with flies, the holding box C can be sealed and removed from the feeding system. To remove the flies, anaesthetic (cold or carbon dioxide) can be applied from the top of box C; the removal of a second shutter at the bottom of box C then allows flies to drop down into separate release containers.


The fans pull air through the model of the feeding structure (from the emergence box section A → semiochemical feeding area B → holding box section C) (fig. 1). The airflow is created by four personal computer fans run on a standard 12-volt direct current supply with independent controls, so that the lower fans can operate independently and generate a different airflow compared to the upper fans. The fans are mounted in 80-mm holes on 10-mm plywood at the end of section D and connected to a four-channel variable resistor for individual airspeed control (fig. 1).

Lighting and viewing

The prototype is completely blacked out to give maximum control of lighting levels within the model. Light is provided by a 150-watt incandescent light bulb positioned below area D (figs 1 and 4) and connected to a dimmer switch (Lutron Electronics Co. Coopersburg, PA).

The light will pass from the light bulb to area D→C→B→A, losing intensity through the different sections. The effect is an increasing light gradient for the flies as they progress through the different sections (A→B→C), encouraging them to move only one way through the system, i.e. from A to B and then to C.

Viewing windows are designed at prominent locations in order to be able to monitor fly movement. To prevent external light from disrupting the designed light gradients in the system, windows are covered with red cellophane and removable black cardboard flaps.

The model is designed so that it can be calibrated for different light intensities to test the response of the flies to light. The dimmer switch is graded from 1 to 7. Table 1 shows, for each grade, the light intensity (lux) in each of sections A, B and C measured using a Lutron Lx-101 light meter (Lutron Electronics).

Table 1. Light intensity levels (lux) measured at different dimmer settings in sections A, B and C, using a 150-watt standard incandescent light source
Dimmer settingSection
1 (min)013
7 (max)05105

Video camera

To monitor the flies during semiochemical feeding, a video camera is mounted on a bracket on the backing board of area B close to the top of wall AB and connected to an external monitor. Small openings are cut into the backing board to fit the wires and maintain a good seal. For filming purposes, an external light source can be placed in an angled hole cut into the backing board.


The airflow was calibrated using an anemometer; measurements were taken in section C only. For these measurements, the fans were calibrated in two ways: (i) fan controllers were set at low, medium and high settings with a constant 12 volt supply, and (ii) fan controllers were set at the high setting and input voltage was varied between 12 and 3 volts. The second method of varying the input voltage provides a more accurate adjustment of the air speeds generated by the fans (table 2).

Table 2. Variation of air velocity depending on fan setting and input voltage
Input voltage (V)Fan settingAnemometer reading (m/s)

Owing to airflows through the funnel holes at AB and BC, the flow through these funnel holes could not be measured because of the comparatively large size of the probe. Furthermore, the holes slow the airflow down significantly, and as a result, there is relatively little air movement in section B. However, because the main purpose of the airflow is to pull air away from section A and to have some flow through the funnel holes BC to encourage flies through to section C, the design was not altered.


The prototype feeding structure was the one described above (figs 1–4). The light intensities in the relevant tests were 0 lux for section A (emergence box), 5 lux for section B (feeding area) and 105 lux for section C (holding box). The maximum fan speed used produces an airflow of 0.91 m/s. At the end of each test, flies in each section were anaesthetized with CO2 applied through the mesh of sections A and C and through a vent in section B.

All tests with insects were carried out to ascertain the efficacy of the system for different insect species, using C. capitata (tests 1–3) with citrus oil as the semiochemical (Katsoyannos et al. 1997; Papadopoulos et al. 2001; Shelly et al. 2004) and B. dorsalis (tests 4–6) with ME (Shelly and Dewire 1994; Shelly 2000; Tan and Nishida 1996, 1998; Tan et al. 2002, 2006; Wee et al. 2007).

All tests were carried out with adult flies reared at 26°C, 60% relative humidity and 12-h light/dark cycles at the UCL insect laboratory. The flies were fed with sugar and yeast hydrolysate (3 : 1). The number of insects in each section was recorded by video monitoring of the feeding area (tests 1–4) and visually (tests 5–6) by counting the files through viewing panels located on the sides of each section. Owing to the low light intensity in the feeding area (<5 lux), a liquid crystal display (LCD) light tube was inserted and lighted during video monitoring. To minimize disturbance to the flies, the viewing panels were covered with red cellophane to filter out shorter wavelength light from the external environment and by a black cardboard that can be lifted when viewing is required.

Test 1 was to monitor in general how the model structure performed with live flies. Four drops of citrus oil were diluted with 5 ml of absolute alcohol and spread evenly on the belt. Light and airflow were both at their highest settings. Adult C. capitata (150 specimens) male flies (20–35 days old) were anaesthetized with CO2 and immediately placed into the emergence box. At this point, the experiment timer was started. The belt was initially stationary so that the response to just the light and airflow could be observed, and was activated 30 min after the start of the test.

In test 2, 150 flies were given 30 min to recover from knockdown before being placed in section A. All other conditions were identical to those in test 1, except that the LCD light was switched off after 18 min and the belt was activated after 60 min.

Test 3 was identical to test 2, but this time without internal light and airflow. In all three tests, the height of the lower end of the guidance flap was set at 7 mm above the surface of the belt.

Tests 4–6 were performed with adult B. dorsalis (2–3 weeks old) and ME as the semiochemical, and with the guidance flap set at 10 mm above the surface of the belt to allow for the larger size of B. dorsalis. Test 4 had both light gradient and airflow. The belt was initially stationary but switched on after 30 min. CO2 was applied to a rearing box of 7-day-old B. dorsalis, and from, this 147 males were separated out and placed in the emergence box. Flies were given an hour to recover. About 17 min after the start of the experiment, the LCD light was switched off for 12 min to create a better light gradient between sections B and C.

Test 5 was conducted with 100 15-day-old B. dorsalis males, which were given 24 h to recover from knockdown. The belt was switched on from the start, and counting of flies was performed through the viewing panels to eliminate the LCD light. The fan speed was changed at 15-min intervals from maximum power (100%) through 75% and 50% to 0% during the test.

Test 6 was performed without internal airflow, LCD light and external video monitoring, i.e. only with the designed light gradient. The test was again conducted with 100 15-day-old males given a day to recover from knockdown.

Results and Discussion

Observations from all tests showed that no flies escaped from the structure during the course of the tests, nor were any injured while moving through the system. Despite the strong LCD video light in the feeding area (section B), the directional airflow and the scent of citrus oil, only 39 (26%) C. capitata males in test 1 crossed the AB separator wall to enter section B, clearly indicating that the duration of the test (38 min) was insufficient for the majority of flies (74%) to recover from the immediately preceding CO2 sedation and to enter section B (table 3). Also the rate at which flies crossed AB increased only slightly during the course of the test. Of the 25 flies that had crossed into feeding area prior to switching on the conveyor belt, only two crossed over to holding box (section C) voluntarily. The reason for this is most likely that the LCD video light is very bright and highly directional, so that the gradient between sections B and C during video filming is very low. However, as the video was the only means of counting and observing fly behaviour, it had to be switched on in this initial experiment. When the belt was started, the flies were observed to remain on the belt and to appear unconcerned that it was moving; equally no startled behaviour or sudden reaction was observed as they were transported along with the belt and under the clear guidance flap into the holding box. These are crucial observations as one of the main concerns when designing the belt was that flies might be startled by the movement and as a result either not settle on the belt or fly off when the belt started moving. The other major concern was that the flies might be startled by the sudden narrowing of space as they move under the guidance flap and hence fly away from the flap and back into feeding area. During the 8 min that the conveyor belt was moving, a further 24 of the 39 males that had crossed from section A to section B entered into section C, confirming the observation that this is the main point of entry for the flies.

Table 3. Test 1 results after 38 min of testing for Ceratitis capitata (20–35 days old) with citrus oil internal light and airflow both at their highest settings. Flies were anaesthetized with CO2 and immediately placed into the emergence box
Time (mins)Number of flies in section
30Belt switched on

The half-hour pretest recovery time given to flies in test 2 resulted in a much higher number of flies moving through the system than in test 1, with 84 (table 4) vs. 25 males crossing AB within the first 38 min of the tests; hence, sufficient recovery time is required following sedation for flies to become active again. Most settled on the conveyor belt and very few flies were resting on the walls of section B (except in some corners that may have been too thinly coated), thus demonstrating the effectiveness of silicon oil in deterring flies from resting on coated surfaces.

Table 4. Test 2 results after 77 min of testing for Ceratitis capitata (20–35 days old) with citrus oil and internal light and airflow on (*indicates when the video light was switched off). Flies were given 30 min to recover from knockdown before being placed into section A
Time (mins)Number of flies in section
2143 7 07
60Belt switched on

On activation of the conveyor belt at 60 min, flies again remained on it and passed under the clear guidance flap into holding cage C. Flies also continued to land on the moving belt. After 7 min of belt operation, all flies on the belt had been transported to section C. In total, 134 flies (82%) were transported through sections A→B→C, again confirming the conveyor belt as the main point of entry into section C.

In test 3 without lights and airflow, the effects of the absence of these two parameters were assessed (table 5). During the first 18 min of the test, when the LCD light was on, only 48 flies crossed AB (compared to 62 flies in test 2), suggesting that the airflow in test 2 increased the rate of passage by 29%. The internal light gradient increased the rate of passage through AB by a further 24% above that of the airflow alone, although the belt was again the main point of entry into section C. The total number of flies transported through the system is almost a third less when comparing both tests (57% in test 3, compared with 82% in test 2). The two main sources of encouraging flies to move from section A to section B are the light gradient and scent, but as in test 2 both light and airflow were switched off, there was only the scent of the semiochemical to lure the flies into section B and the belt to transport them into section C. Furthermore unlike B. dorsalis, C. capitata does not need to feed on the semiochemical being in the vicinity of the fragrance suffices, and hence, they are not as attracted to move to section B. However, it needs to be noted that apart from the citrus oil attractant, the video light source was still on for part of the test as the video is required to monitor the numbers of flies in section B, and hence, the system was not in complete darkness.

Table 5. Test 3 results after 77 min of testing for Ceratitis Capitata (20–35 days old). Conditions were identical to test 2 but without internal light and without airflow (*indicates when the video light was switched off)
Time (mins)Number of flies in section
60Belt switched on

We tested the model semiochemical feeding structure also with B. dorsalis males given 1 h to recover from sedation, in the presence or absence of directional airflow, internal lighting, the scent of ME and the LCD video light. Also with this species, the conveyor belt proved to be the main point of entry into section C. In fact, the five flies that crossed separator wall BC voluntarily in test 4 (table 6) did so only after the LCD was switched off, and none crossed during the remaining 21 min when the LCD was back on. This demonstrates the distortion of the video light source, and in later experiments, the video was no longer used and the flies in section B were monitored manually through viewing flaps.

Table 6. Test 4 results after 50 min of testing for Bactrocera dorsalis (2–3 weeks old) with methyl eugenol and internal light and airflow (* and indicate, respectively, switching off and on of video light). Flies were given an hour to recover
Time (mins)Number of flies in section
45Belt switched on

It was apparent that the B. dorsalis flies were attracted by the ME and hence moved very quickly into section B. Even after the video light was switched off, there was a steady flow of flies into section B. Once in section B, the majority of flies settled on the areas of the belt nearest the AB separator sheet and began to feed on the ME. After 45 min, the belt was switched on and the feeding flies were transported under the guidance flap into section C.

In test 5 (table 7), we set out to determine the effect of fan speed on the movement of B. dorsalis in the model structure. The fan appeared to play no major role in the movement of flies from A to B, possibly due to its channelling the ME fragrance in section B away from section A. Also, when comparing the movement rate to that observed in test 4, it becomes apparent that without the external LCD video light, the number of flies entering into section B drops significantly, the internal light gradient being perhaps too low. Nonetheless, once flies have entered section B, they quickly settle on the conveyor belt and are transported at a steady rate into section C.

Table 7. Test 5 results after seven time 5 min of testing for Bactrocera dorsalis (2–3 weeks old) with methyl eugenol and belt running constantly, internal light gradient, no video light and with varying fan settings. Flies were given 24 h to recover from knockdown. The belt was switched on from the start and the fan speed was changed at 15-min intervals from maximum power (100%) through 75% and 50% to 0% during the test
Time (mins)Number of flies in section
0 (100% fan)100000
15 (75% fan)96314
30 (50% fan)902810
45 (0% fan)8621214

A direct comparison of the tests with C. capitata and B. dorsalis cannot be made because of the different test designs. Nevertheless, while the model structure was capable of moving 82% of C. capitata through the system in 50 min, the 41% observed for B. dorsalis still leaves room for improvement. One reason for this may be that the fans exhaust the volatile ME molecules into the air surrounding the structure, hence after some time in operation the scent of the ME emanates throughout the entire laboratory — this may confuse the flies in section A as to the actual source of the ME, and hence they do not move into section B. This could potentially be the reason why lower numbers of B. dorsalis moved through the system as this species has a very much higher sensitivity to its semiochemical, ca 1 nanogram ME attract native B. dorsalis in the field (Tan and Nishida 2000), compared to C. capitata, and the concentrations of ME were probably higher than those of the citrus oil to bring out the equivalent behaviour of flies. Furthermore, the model was designed for the B. dorsalis size of fly and C. capitata are less than half the size; hence, voluntary movement through the funnel valves is much easier for C. capitata.

We therefore conducted test 6 (table 8) several days after test 5 in order for the residual ME scent to have completely dissipated. The test was performed without internal airflow, and we found that (table 6) when determined at 15-min intervals, 18%, 20%, 6%, 4% and 5% of the flies, respectively, passed through AB; in addition, flies remaining in section A were now down from the 74% observed in test 5–47% in test 6 within the same 75-min period.

Table 8. Test 6 results after 80 min of testing for 15-day-old Bactrocera dorsalis males with internal light gradient but without airflow and video light. Flies were given 24 h to recover from knockdown
Time (mins)Number of flies in section


To obtain the benefits of the inclusion of a pre-release semiochemical treatment of sterile males in SIT programmes requires for Bactrocera spp. a fully automated, timed, and fixed dosage semiochemical feeding structure, while for C. capitata where there is no need for direct feeding, the exposure of sterile males to ginger root oil based on aromatherapy has been shown to be effective at all scales tested (Shelly et al. 2010) and is already being used in fly emergence and release centres (see references this issue).

Sterile flies of non-sexing strains (without separation or elimination of sterile females) have to be released in operational SIT programmes before reaching sexual maturity. The reason is that one cannot wait to release them when fully mature, because sterile males and females start mating among each other in the holding containers. Therefore for the feeding system to be viable, the response to and feeding on ME has to occur at an earlier age, and this fortunately is the case for Bactrocera spp. males (Obra and Resilva 2011), thuconfirming ts he potential of routinely applying ME in emergence and release facilities to enhance sterile male mating competitiveness. An additional benefit when using non-sexing strain of Bactrocera spp. is that some degree of separation between male and female files can occur between sections A and B of the developed feeding system, as females respond less aggressively to the semiochemical.

We have presented a model design for the first large-scale semiochemical feeding structure of its kind. This feeding structure not only minimizes human intervention but also maximizes the quality of flies while minimizing wastage of the semiochemical. Separation of flies (male/female, sexually mature/immature, semiochemically satiated/non-satiated) is achieved by different light intensities, airflow and separator walls with valves. The administration of the semiochemical and the safe transport of the flies are obtained through the use of a carefully designed conveyor belt system that combines the semiochemically saturated belt with a netting layer. Flies are also discouraged from moving back through the structure and prevented from escaping by a system of guidance flaps, funnel valves and brush seals. The design of the feeding structure allows flies to take different routes from feeding area (B) to holding box (C). Should the need arise, mesh sides in sections A and C together with air vents in section B allow flies to be knocked down effectively and independently of other sections. Sliding panels in the emergence and holding boxes allow flies to be securely loaded into and removed from the system. A prototype, with fully calibrated light and airflow systems, was constructed and has been shown to be effective in encouraging C. capitata (with citrus oil as the semiochemical) and B. dorsalis (with ME) through the structure.

The design and construction of the model attained its primary objectives. The flies were anaesthetized, loaded, tested and removed from the model securely and with ease. The results on the scaled prototype presented here demonstrate that the proposed design concepts (conveyor belt, guidance flap, funnel separators, startle device, light gradients and airflow) used to feed and encourage flies through the semiochemical feeding structure has potential. However, further testing and fine-tuning and, if necessary, design modifications may be required to be able to recover most males entering the feeding system. In terms of a large-scale semiochemical feeding structure within an SIT mass-rearing factory, the prototype shows how sterile male flies can be automatically and effectively moved from one area, fed on a semiochemical in another and subsequently transferred into a third area for collection and release.


This work would not have been possible without the efforts of Helen Blacker, Helen Collie, Chitru Ratnayake and Yvette Park. Thanks also go to Jorge Hendrichs and Eric Jang for their help and correspondence and to Tracy Chapman for rearing advice and the use of the insectary. This project was financially supported through FAO/IAEA Research contract No 13962.