Insects

Wild male Z. cucurbitae were captured from a papaya field in Keaau, HI (19.6276, -155.0406). Small pieces of dental wick treated with cuelure (4-(3-oxobutyl)phenyl acetate) were placed in plastic containers (8 cm x 12 cm, cylindrical) which were set on the ground between papaya trees for 10–15 mins. Captured flies were transferred to screen cages (40 cm x 40 cm x 40 cm) and held under controlled conditions at the USDA-ARS-PBARC facility in Hilo, HI. Holding conditions were 24–27°C, 65–70% RH, and a photoperiod of 12:12 h (L:D). Flies were provided with water, honey (spread on the cage screen), and a solid diet of three-parts sucrose, one-part protein yeast hydrolysate (Enzymatic, United States Biochemical Corporation, Cleveland, OH, USA), and 0.5-part torula yeast (Lake States Division, Rhinelander Paper Co., Rhinelander, WI, USA).

Harmonic radar

Dipole harmonic radar tags were fabricated from a Schottky diode (RECCO AB, Lidingö, Sweden) and nitinol wire. Two 4 cm lengths of wire were attached to the diode with UV activated adhesive (Bondic, Niagara Falls, NY, USA) so that each wire touched one of the diode contacts while avoiding the opposite diode contact and the other wire. Electrical connections between the wires and the diode contacts were secured using conductive silver paint (GC Electronics, Rockford, IL, USA). Nitinol wire from two manufacturers were used to make tags. Initial tags, used in experiment 2, were made with 0.0254 mm diameter superelastic nitinol wire purchased from Kellogg’s Research Labs (Salem, NH, USA). This wire had a slight curvature which resulted in an overall slight S-shape to the tags. Tags for all other experiments were fabricated with straight annealed 0.0254 mm diameter superelastic nitinol wire purchased from Fort Wayne Metals (Fort Wayne, IN, USA). Early attempts at tag fabrication utilized tungsten (0.071 mm diameter), copper, and platinum (Wollaston process, 0.00254 mm diameter, Leico Industries, Inc., Lyndhurst, NJ, USA) wires which were not found to be suitable.

Tags were attached to the fly thorax in a longitudinal orientation (Fig 1) using a UV activated adhesive (Bondic, Niagara Falls, NY, USA). This was accomplished by chilling wild male melon flies for 20 mins before being held by their legs between the fingers. Tags were then dipped in the adhesive and positioned on the thorax prior to being cured with light from a UV LED. Care was taken not to glue the wings or the head during tag attachment. Initial flight testing was also conducted with the tags attached to the underside of the abdomen in a lateral orientation but this attachment position was observed to impede normal flight behaviors.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. A wild male melon fly, Zeugodacus cucurbitae, with a harmonic radar tag attached.

The tag shown weighed approximately 0.8 mg and was fabricated from a diode and superelastic nitinol wire.

https://doi.org/10.1371/journal.pone.0276987.g001

Harmonic radar transceiver units (R9) were purchased from RECCO AB (Lidingö, Sweden). Locating flies with HR was accomplished either by searching an area to which a fly was visually noted to have flown to or by moving around either cages or fields in a regular pattern. During searching, the RECCO unit was rotated in the hand and moved from side to side to maximize signal detection by aligning the transceiver with the tag attached to the fly. Under optimal conditions, alignment of the RECCO unit with the tag, without vegetation interference, had a maximum detection range of approximately 10 m. For determining search patterns, a 5 m detection range was assumed.

Flight ability testing (20 May 2021)

Experiment 1 tested the flight ability of tagged flies versus untagged flies. The experimental design was based on quality control measures outlined for tephritids by FAO/IAEA/USDA [44]. Twenty HR tagged and 20 untagged wild male melon flies were placed into chilled storage (4°C) for 5 mins in separate containers. Tags used in this experiment were fabricated from straight annealed superelastic nitinol wire with an approximate mass of 0.8 mg. Each fly was taken individually to the flight-testing room, which had a temperature of 26.5°C and 60% RH and dropped out of the container into the bottom of a PVC flight tube (20 cm x 14.5 cm x 15.5 cm) inside a cage (30.5 cm x 30.5 cm x 76 cm). The time for the fly to become responsive and exit the tube was recorded. Two flights were recorded for each fly and the times were averaged. Flies were considered a non-flier after 15 mins.

Field cage experiments

Experiment 2 (29 December 2020–5 January 2021) investigated the movement of tagged flies in outdoor field cages. In this initial experiment, tags made with non-annealed nitinol wire were used. Flies were released from the center of one of two outdoor field cages (15 m x 6 m x 2.5 m) located at USDA-ARS-PBARC in Hilo, HI. Ten flights were recorded for each fly. After each flight the landing position was recorded and the fly was recaptured and rereleased from the center of the cage. Step-distances (the length of a single flight from take-off to landing point), flight directionality (the angle from take-off to landing point), and pseudo-turning angles (the difference in angle between successive flight directions) were calculated from recorded fly positions. This initial experiment used slightly heavier tags (approximately 1.1 mg, only used in this experiment), wind information was not recorded, and three flies were tracked twice (N (tracking episodes) = 11).

Experiment 3 (12–25 May 2021) was conducted in a manner similar to experiment 2 except that slightly lighter tags (approximately 0.8 mg) made with annealed nitinol wire were used, wind direction and speed were recorded in the near vicinity of moving flies using a Kestrel 4500 Weather Meter (Kestrel Instruments, Boothwyn, PA, USA), and flies were run only once (N = 25).

Papaya field experiment (28 May 2021–11 June 2021)

Experiment 4 investigated the movement of tagged flies in a papaya field in Keaau, HI (19.6280, -155.0407). The field was approximately 11,000 m² in size with papaya trees ranging in height from 2–6 m. The field was bordered on three sides by dense tropical forest. The fourth field border consisted of tall grasses and sparse trees. No permit was required for access to this field site. Flies were tracked one at a time after being released into the field. Papaya trees were used as release sites with new trees used for successive releases. No more than two tagged flies were tracked at the same time (separate crews followed each fly). After release, tagged flies were tracked through the field with landing locations (specific tree) recorded after each flight. At most landing locations the fly was visually located, however, sometimes the canopy height allowed the landing location to be identified only by signal detection. Flies were allowed to rest for 5 mins following each flight. If a fly had not flown again after 5 mins, it was gently poked or the surrounding foliage was disturbed to induce flight to another tree. Flies with at least five recorded flights were used in the analysis. Up to 10 steps were recorded for each tagged fly. When possible, flies were recaptured and removed from the field after the 10^th recorded flight. Wind speeds and directions were recorded in the near vicinity of moving flies. Step-distances (flight distances), flight directionality (angle from take-off to landing), and turning angles (angle between successive flight directions) were calculated from recorded fly positions.

Open field experiment (7 January 2021 and 8 June 2021)

Experiment 5 investigated the movement of tagged flies in an open field in a public park (Lokahi Park, approximately 25,000 m² in size, 19.6979, -155.0734) in Hilo, HI. No permit was required for access to this field site. Tagged flies were released from two release points (varying by date) at the center of an open grassy (mowed) field (approximately 10,000 m²). The open field was bordered by a mixture of forest, tall grass, and basketball and tennis courts. Flies were tracked after release with landing locations recorded after each flight. When possible, flies were recaptured and released from the same central point. Wind speeds and directions were recorded in the near vicinity of moving flies. Step-distances and flight directionality were calculated from recorded fly positions.

Statistical analysis

For experiment 1, the proportion of fliers to non-fliers was tested using Fisher’s exact test. The time to exit the tube was compared by t-test calculated using Microsoft Excel (version 2108, Microsoft Corporation, Redmond WA).

For experiments 2–4, the Watson-Williams test for homogeneity of means was used to determine if the flight directions varied between flies. Subsequently the Rayleigh test and the Hermans–Rasson test [45] were used to determine if flight directions were random for each set of flights for an individual fly. Additionally, the V-test was used to assess the effect of wind direction on flight direction in experiment 3–5 [46]. All circular statistical analyses were performed using R packages CircStats, circular, and CircMLE [47]. Chi-squared for turning angles were performed in R [47]. For experiment 4, individual flights were grouped into quadrants and mean step distances were compared using ANOVA followed by Tukey’s HSD (α = 0.05) while the proportion of steps in each quadrant was analyzed using a contingency table approach followed by the Marascuilo procedure, both performed in R [47]. Equations for step frequencies vs. step-distances were calculated using Microsoft Excel.

Harmonic radar tags

Tag mass and size [48], tag entanglement [49], and antenna deformation [48] have all presented obstacles to HR tracking of insects. HR tags are required to be extremely small and light for this application, with tag mass less than 5% of the insect mass, although the empirical basis for this limit is weak [32]. Additionally, tag antennas need to be flexible enough so as not to impede movement while simultaneously avoiding permanent deformation when bent. Copper [42], copper-cladded steel [41], silver-plated copper [50], steel [50], and aluminum [33, 43] wires have been used as antennas for HR tags in previous studies with occasional reports of wire deformation [41, 51] and entanglement and/or encumbrance [43, 50].

The successful tracking of melon flies in this study was largely due to the use of superelastic nitinol wire as the antennas for the HR tags. Nitinol is light, relatively inexpensive, and possesses the conductive properties needed to operate as an HR antenna. The superelasticity of nitinol allows this nickel-titanium alloy to undergo substantial deformations (up to 30 times greater than other metals) before returning to its original undeformed shape [52]. The advantages of nitinol were clearly seen when compared with the other two metal wires used in this study. Wollaston process wire, a platinum wire clad in silver, was light and flexible but would quickly become deformed by bending and tangle, thus critically reducing the detection distance of the tags. In contrast, Tungsten wire is one of the hardest and most dense metals. While antennas made with tungsten did not deform or tangle, they restricted the movement of the flies due to the inflexibility of the forward-facing wire, which was observed to obstruct forward movements of the fly. Also, given the density of tungsten, tags made with this wire were heavier than both those made with nitinol or Wollaston process wire.

There are a limited number of previous studies using HR to track dipteran species [32], likely due to their small size and lack of adaptation to carrying loads in flight. Roland et al. [33] were the first to demonstrate that it was possible to track flies using the tachinid fly Patelloa pachypyga (Aldrich & Webber) and the sarcophagid Arachnidomyia aldrichi (Parker).

Larger Coleoptera (many of which were tracked as they walked) and Hymenoptera have been more commonly tagged [32]. Interestingly, the other dipteran species that has been studied in the field with any depth is a tephritid fruit fly, the Chinese citrus fruit fly, B. minax (Enderlein) [35]. This species is among the largest tephritids (approximate male weight 44 mg) and is roughly three times the mass of the melon fly (approximate male weight 15 mg). The work with B. minax utilized a HR tag weighing 3.4 mg, which is more than four times the mass of the tags in the current study (0.8 mg). However, the tag to fly mass ratios are similar in both studies with each being around 5% of the total insect body mass. Tags used on B. minax were roughly 8% of the fly body mass with repeated experiments finding little effect on the flight characteristics of the fly [34, 36, 37]. Although the flight testing in the current study was not as extensive as that carried out with B. minax, we similarly found minimal negative effect on Z. cucurbitae flight (experiment 1).

Flight directionality

Melon flies tracked in this study showed both individual level flight directional biases (Figs 2, 5, and 7) and, in some experiments, collective directional biases, both turning angles (Figs 3 and 8) and combined absolute flight directions (Figs 6 and S1). At the individual level in field cage experiments (experiments 2–3), 30 of the 36 flies observed (83%) showed directionally biased flights while similar biases were observed in roughly half the flies tracked in a papaya field (experiment 4). Additionally, mean flight directions varied between flies (Watson-Williams test) in experiments 2–4 showing strong inter-individual differences in directional orientation. These individual-level flight directional biases may be an example of biased behaviors in insects, such as turning biases, leg preferences, and antennal asymmetries. These are linked to lateralization in the central nervous system [53, 54]. Insect individual-level bias include turning bias in the human body louse [55], a staphylinid [56], bumblebees [57], 7-spot ladybird beetles [58, 59], drosophila [60], and honeybees [61], as well as forelimb lateralization in desert locusts [62] and the biased use of laterally paired penises in earwigs [63].

There are examples of insect population-level turning biases which include giant water bugs [64], several ant spp. [65–67], cockroaches [68], honeybees [69], and bumblebees [70]. Population-level biases have also been found in a number of diptera including turning in drosophila [60], and mating behaviors in Culex pipiens L. [71] as well as two tephritids, B. oleae (Rossi) [72] and C. capitata [73]. These left-right asymmetries, either at the individual or population level, demonstrate innate behavioral biases which in many instances relate to locomotion and more specifically turning. However, while these left-right movement associated biases share may similarities with the movement biases observed with male melon flies in this study, the complexity of 360° flight directional orientation is likely higher than a simpler binary left-right choice.

Similar observations of individual insects displaying biased directional flight can be seen in the migratory butterfly Pieris brassicae L. Individual butterflies of subsequent generations have been shown to fly in opposite directions, the spring generation generally flies north while the autumn generation flies south, both in field observation [74–76] and cage bioassays [77–81]. Larranaga et al. [81] found that P. brassicae butterflies show both between and within individual variations in flight directional biases strikingly similar to those observed in the field cage experiments in the present study. Larranaga et al. [81] propose that inter-individual flight direction variation indicates that directional orientation is a heritable trait and that the migratory habit and directional orientation may be rapidly lost without selection, as was seen for butterflies bred for multiple generations in captivity.

While the P. brassicae cage experiment results [81] are similar to those observed in this study melon flies are not migratory and it is unclear what mechanism might be driving the observed directional flights. Perhaps the behavior leads to a greater dispersal of the entire fly population as flies having different directional biases would be more likely to move away from each other, minimizing competition.

Strongly biased directional melon fly movement can also be seen in the observed turning angles (Fig 8) and pseudo-turning angles (Fig 3). The term ‘pseudo-turning angle’ is adopted here to highlight that these flies were physically moved (returned to the release point) between successive flights. The observation that flies in the field cages successively choose similar flight directions is even more striking when considering that flies faced a random direction when returned to the release point (no attempt was made to orient them in the same direction as they were captured on landing). Flies making successive flights (steps) in the papaya field might show a “forward bias” simply because they continue is the same direction after landing (Fig 8). Flies in the field cage experiments had to reorient after being replaced at the release point and yet, as a group, flies chose subsequent flight directions within 90° left or right of the directly previous flight (Fig 3). Turning angle biases observed throughout this study may be examples what is known as ‘persistence’ or ‘forward persistence’ [82], the tendency observed in many animals towards forward movement [83–85]. Correlations between successive step orientations led to the development of the correlated random walk model (CRW) [82] and successive models such as biased random walk (BRW, consistent bias in a preferred direction or towards a target), and biased and correlated random walk (BCRW) [85]. These models have been applied to tephritid movement previously [30, 86] and the biased turning angle data observed for melon fly in this study support the use of modeling techniques that incorporate persistence and/or bias when modeling tephritid movement in the future.

Another possibility is that the observed flight directional biases are the result of flights being in an ‘aroused state’ from being handled and that the ballistic flight paths taken by flies are a mechanism to escape a dangerous situation. This hypothesis is supported by the fact that flies were subjected to handling or disturbances to induce movement throughout the experiment in which directionally biased movement was observed. In this case, future tracking studies that are less agitating to the flies should show greater randomness in the directions of flights, thus flight paths should look less like escape movement.

Collectively, when all flight (cardinal) directions are combined and analyzed for each of the experiments 2–5 (Figs 5 and S1). Flight directions are random for experiment 2 (first cage experiment), suggesting that flies, as a group, are not orienting towards one strong directional cue (e.g. visual, light, wind cues), but they do show directionality for experiment 3 (second cage experiment), experiment 4 (papaya field), and experiment 5 (open field) so directional cues may be at play in these experiments. Interestingly, a similar collective biased directional dispersal behavior was observed for B. minax [35] with the biased B. minax movement attributed to flies moving out of an orchard and into an adjoining forest. The effect of environmental cues on insect movement are notoriously difficult to document and analyze [87]. Visual cues in the cage experiments were likely less intense and varied when compared to the papaya and open field experiments. The cage walls were made of white screen which obscured the view outside the cage. The cages were also erected in a gravel parking lot which further reduced the complexity of short-range cues. Light cues were largely similar for experiments 2–5 as all were conducted from late-morning through mid-afternoon. Variations in shade were greatest in the papaya field experiment.

No attempt was made to quantify either visual or light cues during this study. In contrast, wind speed and direction were recorded for experiments 3–5. Wind speeds and directions were recorded in the near vicinity of moving flies but may not reflect the actual microconditions experienced by the flies. Additionally, mean wind values were used for V-tests and while the variation in wind speed and direction were relatively low, this likely further weakens the ability to correlate wind parameters with fly movements. Even with these caveats, the mean wind direction was correlated with mean flight directions for both some individual flies and for combined flights from experiments 3–5. Where there was a correlation downwind movement was generally observed (negative anemotaxis). The observed mix of relationships between flight directions and wind, both flies with mean flight directions that correlate with the wind direction (19/42) and those that don’t correlate (23/42), suggest that wind direction is one factor influencing fly flight direction but not the only factor.

Unsurprisingly, wind affects flight directionality for many tephritids but the magnitude of this effect and the interaction with other contributing factors is still largely unknown. Mean flight direction has been found to correlate with mean wind direction in studies of olive fruit fly, B. oleae [88], medfly [89], and Mexican fruit fly, Anastrepha ludens (Loew) [89]. However, a study conducted by Plant and Cunningham [90] observed the opposite–that medfly dispersal and wind were not related. Additional studies with Queensland fruit flies, B. tryoni (Froggatt) and A. ludens suggest that the relationship between flight direction and wind is complicated by environmental factors [91, 92] and wind speed [93].

Flight distances

Step-distances (meters per flight) for male melon flies differed between the field cage, papaya field, and open field experiments with the longest mean flight distances observed in the open field (Fig 3). Step-distances were truncated in the two field cage experiments as flies were limited to maximum flight lengths of 3–9 m dependent of the direction of flight (flies would encounter the screen wall at these distances). The most common step-distances in experiment 2 and 3 (field cages) fell between 3–4 m and the mean step-distances were 3.6 ± 0.2 m and 4.1 ± 0.1 m respectively, which largely reflects instances of flies landing on the nearest cage wall. Given the limited possible maximum flight distances in the cages, it is difficult to draw conclusions about mean step-distances from these experiments.

In contrast, step-distances recorded in experiment 4 (papaya field) are not truncated. The most common step-distances in the papaya field were from 1–2 m with a mean step distance of 6.0 ± 0.5 m. The relationship between step frequency vs step-distance (categorized into 1 m intervals) was well described by a power function (S3 Fig) showing that flies generally make short flights between nearby trees with occasional longer flights. This result differed from what was seen in experiment 5 (open field) where there was no relationship between step frequency and step-distance (S3 Fig). Step-distances in the open field had the highest mean, 18 ± 3 m, with flights ending either in the first tall grass or trees encountered or on the ground if the flight did not reach dense foliage. Flies in the open field were harder to track given the longer flights and tagged flies were lost at a higher rate when compared with experiments 2–4. Similar higher dispersal rates of melon flies in less favorable environments (fewer roosting sites and host plants) have been previously noted [94].

For B. minax the mean step-distances ranged from 2.3 ± 0.4 m to 6 ± 5 m depending on the experiment [35]. This is similar to the 6.0 ± 0.5 m step-distance found for melon fly in a papaya field (experiment 4). In the current study, flies were disrupted to keep them moving, which differs from the B. minax experiments and likely accounts for the longer total distance observed with melon fly. Bactrocera minax were never observed to move more than 25 m from the release point while this was routinely observed in melon fly (Fig 6).

While melon flies have been recorded to move up to 200 km [95], when host plants are plentiful, overall fly movements are thought to be nondispersive [24, 96, 97]. Melon flies are generally thought to show two types of movement, long-distance dispersal movement after emergence and more localized daily movement patterns in and around crop fields [96, 97]. Given that many detection and control efforts are based on understanding how melon fly, and other pest tephritids, move within the landscape, numerous previous studies have sought to quantify fly dispersal parameters [24, 94, 95, 98–101]. Additional studies have reported similar movement behavior for other pest tephritids such as B. dorsalis [99, 102–104], B. tryoni [93, 105–107], B. latifrons (Hendel) [108], B. oleae [88], C. capitata [89, 90], Rhagoletis mendax Curran [20, 86], and Anastrepha spp. [89, 92, 109]. While flight direction and distance were assessed in the current study, the timing to fly movements were artificially accelerated (disturbance to induce flight) making direct comparisons of flight parameters with previous MMR results complicated. Future tracking studies in which tagged flies are allowed to move more naturally (without artificial disturbance) are needed.

The data obtained in this study will be very valuable to melon fly detection and control. This is primarily because they enable application of more realistic and accurate models of melon fly movement such as the CRW [110, 111]. This model includes rare events and is an individual-level model of movement, compared with the more commonly used diffusion models which produce the distributional outcome of the insect’s movement process but don’t model the movement itself [112, 113]. Delimitation arrays can be optimized for improved response following an incursion, and quarantine and eradication boundaries can be set more effectively. In the event a melon fly is caught in a surveillance trap, the potential origin of that fly can now be better set to a given radius, enhancing the likelihood of an effective response.