Aerial distribution, flight behaviour and eggload: their inter-relationship during dispersal by the sweetpotato whitefly

Authors


* Present address and correspondence author: Department of Entomology, 203 Pesticide Research Center, Michigan State University, East Lansing, MI 48824, USA.

Abstract

1. The aerial distribution of Bemisia tabaci Gennadius (the sweetpotato whitefly) was studied during the early ascent phase of flight, to test the degree to which dispersal patterns reflect the flight behaviour of individuals.

2. Marked whiteflies were trapped at four heights between 0 and 7·2 m above fallow ground, and at six distances between 0 and 100 m from the insect source. Insects were trapped during a 2–3 h period after the initiation of flight activity during the summers of 1995 and 1996.

3. Analysis of trap catch data revealed a clear negative exponential relationship between height and aerial distribution, and a slightly weaker negative power relationship between distance and aerial distribution. Marked insects were caught in the uppermost traps adjacent to the source, indicating that a portion of the population had a strong capacity for ascent out of the flight boundary layer.

4. Eggload decreased with the height, but not the distance, at which whiteflies were trapped. Mean eggload close to the ground was significantly greater than that for those trapped at 4·8 and 7·2 m, supporting the hypothesis that there is a trade-off between flight and oogenesis in weak-flying insects.

5. Air temperatures during the trapping periods were positively correlated with the proportion of male and female B. tabaci caught in the highest traps, but not in the most distant traps.

6. The significance of these results for accurate prediction of whitefly dispersal is discussed, and the importance of individual's behaviour in determining dispersal patterns of small insects is emphasized.

Introduction

Dispersal has been defined as a change in the spatial distribution of individuals (Odum 1953; Southwood 1978), and is one of the important components of population dynamics. This group process is the product of movements by each individual within the population, and is brought about by specific behaviours. The central role of behaviour in the dispersal of animals has been discussed previously (Baker 1978), often with insects as the organisms on which the theories are based (Taylor & Taylor 1983; Kennedy 1985; Dingle 1996). In the special case of small insects, whose aerial dispersal is highly wind-dependent, the relative importance of climate (Johnson 1953, 1969) and behaviour (Kennedy 1985) in determining their aerial distributions has been debated. This has led to the recognition that although meteorological conditions play a role in the movement of small insects to new habitats, much of their aerial dispersal is an active, rather than a passive, process in which the timing and duration are controlled by the insect (Gatehouse 1997).

Whiteflies are generally less than 2 mm long, and weigh less than 0·1 mg. Despite this small size, Bemisia tabaci Gennadius (the sweetpotato whitefly) has a dispersal strategy that enables the winged adults to colonize ephemeral host plants some distance from the point of take-off (Byrne et al. 1996). The flight behaviour and physiology of this insect have been studied extensively (Byrne & Blackmer 1996; Byrne et al. 1996), but here our focus is on the ascent phase of flight in the field. There are two distinguishable types of flight behaviour (as defined by Kennedy 1985) exhibited by B. tabaci. Adults may make extended migratory flights, sometimes for more than an hour, toward skylight during which they do not respond to vegetative cues. Alternatively, they do not ascend but remain aloft on the wind in foraging flight, characterized by attraction to wavelengths of light reflected by green foliage (Blackmer & Byrne 1993; Blackmer, Byrne & Tu 1995). These behaviours result either in maintenance of altitude close to the ground, or ascent into the higher airflows until the behavioural phase of the insect reverts to foraging flight, when insects orient to host plant cues. The relevance of this to the dispersal of B. tabaci within agricultural systems was recently studied by Byrne et al. (1996). A bimodal distribution was found in the proportion of the total number of whiteflies trapped on the ground up to 2·7 km from their source, indicating that whitefly distributions in the field reflect the flight behaviours seen in the laboratory. Variability in this distribution also indicated that there was a continuum rather than a strict dichotomy in flight behaviour, although the relative importance of this and meteorological variability were not determined. The present understanding of the behaviour associated with flight by B. tabaci provides the opportunity to conduct field studies on dispersal patterns, and to relate these to our knowledge of the flight behaviour of this organism.

Local dispersal in the horizontal plane has been characterized for B. tabaci (Byrne et al. 1996), but the situation is less clear for the vertical distribution after take-off. Although the behaviour of migratory insects is an important component of predictive models of their dispersal (Allen et al. 1996), few studies have been conducted to determine empirical relationships between the aerial distribution of B. tabaci and height above the ground, or horizontal distance from the point of take-off. Some studies have shown that most flight occurs close to the ground (Gerling & Horowitz 1984; Byrne, von Bretzel & Hoffman 1986), while others, using aircraft-mounted traps, have shown that whiteflies can reach high altitudes. Thus, B. tabaci was trapped at 305 m above the ground (Glick & Noble 1961) and Trialeurodes abutilonea Haldeman (the bandedwinged whitefly) has been found at 360 m (G. Kampmeier, personal communication). Live B. tabaci, containing mature eggs, have recently been trapped at 150 m above Arizona (R. Isaacs, unpublished data), providing further evidence that whiteflies can undergo effective long-distance dispersal. The relative ecological and economic importance of these high-flying individuals may be rather small, however, due to mortality factors (Berlinger, Lehmann-Sigura & Taylor 1996), though their capacity to be a founder of a new population undoubtedly provides selection pressures for maintaining this high-risk strategy (Hamilton & May 1977; Davis 1980). Short-range migration and dispersal patterns probably have far greater importance for the more weak-flying insects, and emphasis on those insects flying far above the ground may provide an inflated view of the role of these individuals in the overall population processes of small insects (Loxdale et al. 1993).

Reproductive status is also an important factor in the potential for colonization of new patches by immigrating females. The possibility of a trade-off between the production of eggs or flight apparatus was first proposed by Johnson (1969), as the oogenesis–flight syndrome. Current thinking, based on experimental evidence from many different insect species, is that reproduction and dispersal are not mutually exclusive as originally proposed, but rather exist in a state of balance (Dingle 1996). Here, we examined the variation in eggload of B. tabaci to test the hypothesis that there is a trade-off between investment in egg production and flight.

In the experiments described below, we were interested in studying the aerial distribution of B. tabaci during the first part of dispersal, as the insects take flight from a host plant. We tested the hypothesis that the aerial distribution of this insect would decrease with both distance from the source field and height above the ground, and characterized the shape of these relationships. The eggload of the trapped females was also determined, to examine the relationship between reproduction and flight in B. tabaci. Meteorological variables were recorded to determine their correlation with the proportion of B. tabaci that reached the furthest or highest traps.

Methods

All experiments were carried out at the University of Arizona Agricultural Center near Yuma, in south-west Arizona. Cucumis melo L. (cantaloupe melon, c.v. Top Mark) were planted in a field of approximately 1 ha during May of 1995 & 1996, and maintained until October in both years. This was achieved by planting melons in rotation every three weeks, so that every other double-row strip of melons was always planted with healthy plants. The quality of host plants within the field, a factor known to affect whitefly dispersal (Byrne & Blackmer, unpublished data) was spatially homogeneous throughout this study. Populations of B. tabaci (biotype B) used in these experiments were from natural infestations, and remained at densities of >10 adults per leaf throughout the experiments.

On the evening prior to trapping, a tractor-mounted spray boom was used to apply fluorescent pigment dust (Day-Glo Colour Co., Cleveland, Ohio) to the whitefly source field between 17.00 and 19.00 hours. When experiments were run on adjacent days, two different coloured pigments (AX-14 Fire Orange and AX-18 Signal Green) were applied to provide a means of counting only those insects marked on the evening prior to trapping. A laboratory study had previously shown that when B. tabaci were marked with this dust in the early evening, their flight behaviour on the subsequent morning was not affected (Byrne et al. 1996).

At dawn on the day of trapping, 30 samples of leaves and adults were taken from across the source field to estimate population size and marking efficiency. The samples were collected from positions chosen at random across the whole field, except for a 5-m edge that was not sampled. The fifth true leaf of each plant was removed and sealed in a bag for later assessment and a sample of >10 adults was taken from the same area. These were collected in flight by disturbing the foliage and using a modified hand vacuum cleaner, thus preventing collection of pigment from the leaves. Meteorological variables were recorded during each trapping period from a position adjacent to the source field using a Weather Monitor II portable weather station (Davis Instruments, Hayward, California). Measurements of temperature, relative humidity, wind speed, wind direction and barometric pressure were taken at 5-min intervals during the trapping period. These were later transferred to a computer for analysis.

Insects were trapped for 2–3 h during the morning after pigment application, between 06.00 and 10.00 h, starting when the first flight activity was detected. Insects were sampled from the air using a total of 96 white Versagard sticky traps (33·0 × 18·6 cm) (Ecogen Inc., Langhorne, Pennsylvania) for each trapping period. Each trap was suspended vertically by four small metal binder clips attached to a pair of cord loops that ran between the upper and lower cross bars of wooden frames 7·3 m tall and 2·4 m wide, and there were four pairs of cord loops on each wooden frame, spaced 80 cm apart (Fig. 1). The bottom edges of four traps were placed at either 0, 2·4, 4·8 or 7·2 m above the ground, by attaching each trap to the cords and running it up into its position. A total of six wooden frames was used on each trapping day, arranged in a downwind transect from the edge of the source field, and every 20 m thereafter, up to 100 m from the edge of the field. Each frame was arranged to stand perpendicular to the transect, so that traps faced the prevailing wind. As whitefly aerial dispersal is more of a linear process, in the direction of the wind (Byrne et al. 1996), rather than a diffusion process, the use of the same trap size at different distances was valid for this study. The dimensions of the source fields meant that the mean distance from the point of take-off to the first sticky traps was 70 m. In both years, there were no plants growing in the land between the source field and the most distant frame.

Figure 1.

Apparatus for trapping whiteflies in the field. White sticky traps were raised into position by attaching the corners onto two parallel string loops, and then pulling the traps up. Six of these wooden frames were placed every 20 m along a transect in the direction of wind flow, with the sticky surface facing directly into the wind. Traps were positioned at 0, 2·4, 4·8 and 7·2 m above the ground.

At the end of the sampling period, the traps were removed from the frames and wrapped in one layer of clear plastic film. All traps were then immediately taken to the laboratory and assessed for the number of marked B. tabaci, using an ultraviolet lamp as described by Byrne et al. (1996). For each trap, the total number of marked males and females was recorded.

The glue coating of the trap caused clearing of the normally opaque abdomen of the whiteflies, and this facilitated counting of the number of eggs present in female B. tabaci. To test the hypothesis that eggload of B. tabaci varied with height above ground, or with distance from the point of take-off, the eggload of 10 randomly selected individual females was counted on each sticky trap.

Statistical analysis

Complete sets of sticky traps were collected from the field site on five days during 1995 and six days during 1996 (Fig. 2). A minimum of 200 B. tabaci trapped per day was selected, below which we did not include the data in our analyses. This ensured that days on which no insects were collected at the highest traps were excluded. Thus, we did not include the data from September 23 and 27, and October 10 1995, leaving data from 8 days that were used to determine the distribution patterns of B. tabaci across the two-dimensional transect.

Figure 2.

Number of marked B. tabaci adults trapped during 1995 and 1996. The dotted line indicates the limit, below which the data from these days were not included in the analyses.

All analyses were performed using the GLIM statistical package (Baker & Nelder 1978) to examine separate data sets for catches of male and female B. tabaci. To analyse the proportion of the total daily catch trapped by each sticky trap, the number of B. tabaci trapped in each trap was used as the response variable and the corresponding total daily catch as the denominator. This method corrected for the wide distribution of sample sizes between different days. On initial inspection, the variance distribution was found to be uneven, and so the data were ln (n + 1) transformed prior to analysis. Weighted regressions were then performed on the data using binomial errors and the logit link function (Crawley 1993). The significance of changes in deviance due to removal of terms from the full model was assessed on the basis of χ2 values. Using this method, the significance of the effects of trap height, trap distance, day and the three interaction terms on the proportion of B. tabaci trapped in the sticky traps was determined.

Regression analysis was performed on the mean proportion of B. tabaci trapped on the 8 days, to determine the relationships between the aerial whitefly distribution and the height above the ground and distance from the source field. Regressions of height and distance were performed on transformed data sets. The best-fit equations were determined by examination of r2 values, associated F-values due to fitting the regression equation, and the fit of the predicted values to the data at the extreme proportions near to 0 and 1.

Variation in eggload was analysed using anova to determine the effect of collection day, trap height and trap distance on the mean eggload for each sticky trap sample of 10 females. This analysis used data from the 3 days with greatest abundance of B. tabaci, to ensure a large enough sample size in the highest traps, thus producing a total of 288 samples (3 days, 4 heights, 6 distances, 4 replicates). Mean eggload values were compared between heights and distances with t-tests, after a Bonferroni correction for multiple comparisons (Sokal & Rohlf 1981).

To determine whether the meteorological conditions at the time of trapping could be used as predictors of dispersal by B. tabaci, multiple regression analysis was performed. Mean values of each meteorological variable during each trapping period were the explanatory variables, and response variables were the proportion of the total number of females or males trapped in the highest traps and in the most distant traps from the source field. In addition to the meteorological variables described above, the peak wind (gust) speed and the mean deviation of the wind (in degrees) from the direction of the transect of traps were calculated. Thus, analysis of the effect of meteorological variables on the proportion of male and female B. tabaci trapped was performed for temperature, barometric pressure, relative humidity, wind speed, gust speed and wind direction variation.

Results

Whitefly catches

In total, 38 938 marked adult B. tabaci were caught during the 11 trapping days shown in Fig. 2. Sex determination was not possible or was ambiguous for 1·3% of the total catch, and these insects were excluded from the data set. Selection of the 8 days from all of the samples (see Methods) left a total of 37 685 B. tabaci for the analysis of aerial distribution patterns (Table 1). Marking efficiency decreased from 1·00 at the start of 1995 to 0·58 at the end of 1996, and trapping day was negatively correlated with the total number of whiteflies trapped (r2 = 0·75, F(1,6) = 17·73, P = 0·006), indicating that as the whitefly source field foliage density increased, the probability of an insect being marked decreased.

Table 1.  Catch sizes, sex ratios and proportion of B. tabaci that were marked during the collections in 1995 and 1996. Mean (±SE) sex ratios were calculated from the number of females/number of males for each trap (n = 96 for each collection day
 Source fieldSticky traps
Collection
day
Proportion
marked
Sex
ratio
Total
trapped
Sex
ratio
1995
14 Aug1·002·61 ± 0·43176561·59 ± 0·13
15 Aug1·002·61 ± 0·43129120·78 ± 0·14
1996
28 Jun1·001·41 ± 0·6918452·59 ± 0·26
10 Jul0·831·45 ± 0·309471·82 ± 0·18
11 Jul0·821·45 ± 0·3028970·97 ± 0·07
1 Aug0·471·65 ± 0·242480·74 ± 0·09
7 Aug0·581·31 ± 0·213741·35 ± 0·13
8 Aug0·581·31 ± 0·218061·65 ± 0·17

Whitefly populations were usually female biased (i.e. sex ratio >1) in the samples of adults collected in the source field and on the sticky traps (Table 1). Sex ratios of the insects trapped in 1995 differed markedly between the source field and the aerial catches, even though the two collections were made on consecutive days. The sampling period in 1996 covered a greater time span, and the overall sex ratios of the insects trapped in flight were similar to those found in the source field. There was no significant correlation in sex ratios between these samples (r2 = 0·12, F(1,4) = 0·53, P = 0·51).

Aerial distribution

The distribution of the daily catch of B. tabaci was significantly influenced by the trapping day, the height of the traps above the ground and by their horizontal distance from the source field (Table 2). Of these factors, trap height had the greatest effect, followed by day and distance. Whiteflies started to disperse in the direction of the prevailing wind at dawn, and most of the individuals (70%) were observed flying near to the ground (Fig. 3a,b). The mean proportion of the aerial population trapped decreased exponentially with increasing height above the ground, a consistent pattern as shown by the low variation around the mean values in Fig. 3(a) and (b). Less than 5% of the insects were trapped in the uppermost traps, emphasizing the steep decline in aerial density with height. Some marked whiteflies were, however, caught in traps at this height on the frame placed adjacent to the edge of the field. An interaction between trap height and day was found in the catches of male B. tabaci, in which the vertical distribution was significantly different between the trapping days (Table 2). This relationship was not seen for females.

Table 2.  Results of weighted regression analyses of distribution patterns of male and female B. tabaci. NS = non-significant
  FemalesMales
Termd_f.χ2 Pχ2 P
Height3451·7<0·0001480·9<0·0001
Distance514·1<0·0518·3<0·01
Day7194·5<0·001271·7<0·001
Height × distance1510·5 NS10·0 NS
Height × day2130·5 NS52·4<0·001
Distance × day3535·0 NS26·4 NS
Figure 3.

Mean proportion of the total daily catch of (a) female and (b) male B. tabaci (±SE) trapped at different heights (H) above the ground on 8 different days in 1995 and 1996. Best-fit regression lines for both females and males are shown, with the equations Pfemale = 1·617e−0·994H and Pmale = 1·790e−1·079H, respectively.

Regression analysis of the transformed mean catches of B. tabaci provided equations with which to describe the aerial distribution of this insect in terms of both the height above the ground (H) and the horizontal distance from the point of take-off (D) (Figs 3 and 4, respectively). There was a high degree of correlation between the measured vertical distributions and the descriptive equations for both females (r2 = 0·99, F(1,2) = 425·3, P < 0·002) and males (r2 = 0·99, F(1,2) = 795·3, P < 0·001), and these were best described with negative exponential relationships (Fig. 3). The relationships between distance from the source field and aerial distribution of male and female B. tabaci were of a different mathematical form to those for height. Insects were trapped in greater numbers 20 m from the source field than adjacent to it, although there was no significant difference between the catches at any distance from the source field. This pattern was almost certainly due to obstruction of the lowest traps by the field border topology and foliage adjacent to the first set of traps, i.e. blocking of the lowest traps adjacent to the field. Regression equations were therefore calculated for the data set with the most proximal set of traps removed, to provide a general relationship between distance from the point of alighting and aerial distribution. This revealed that the relationships between trap catch and increasing distance from the source were accurately described by negative power equations for both females (r2 = 0·95, F(1,3) = 79·1, P < 0·003) and males (r2 = 0·99, F(1,3) = 653·4, P < 0·0001) (Fig. 4a,b).

Figure 4.

Mean proportion of the total daily (a) female and (b) male B. tabaci catch (±SE) trapped at different horizontal distances (D) from the whitefly source field on 8 different days in 1995 and 1996. Best-fit regression lines for females and males are shown, with the equations Pfemale = 0·326D−0·546 and Pmale = 0·465D−0·816, respectively.

Eggload

Female B. tabaci were found to contain an average of approximately four eggs, although a small number of individuals were found with as many as 10. Analysis of variance of the mean eggload data showed that there was no significant effect of the day of collection (Table 3). Eggload was, however, significantly influenced by the height at which the insects were trapped (Table 3), due to the decrease in mean eggload with increasing trap height (Fig. 5a). Analysis of the differences between the means revealed that the females trapped at 0 m contained a significantly greater number of eggs than those at 4·8 and 7·2 m above ground level (AGL) (t > 2·90, d_f. = 130, P < 0·0007), but not those at 2·4 m AGL. There was no significant difference between the mean eggloads of the females trapped at 2·4, 4·8 or 7·2 m AGL (t < 1·14, d_f. = 130, for all three comparisons, n.s.). The distance at which the insects were trapped was not a significant factor in the analysis (Fig. 5b).

Table 3.  Analysis of variance on the mean eggload of female B. tabaci trapped in sticky traps at different heights above the ground and distances from the source field
TermSSd_f.msF P
Day6·3323·171·69 NS
Height26·8038·934·79<0.01
Distance21·1754·232·27 NS
Height × distance20·40151·360·73 NS
Error442·182371·87  
Total516·88262   
Figure 5.

Mean eggload (±SE) of female B. tabaci trapped at (a) different heights and (b) different distances from the source field. Means are significantly different between bars with different letters, after Bonferonni correction for multiple comparisons.

Meteorological variables

Weather conditions were variable during this study, due to the range of trapping dates. The mean values of the meteorological variables recorded during the 8 days ± SE were: temperature 31·9 ± 1·5 °C, barometric pressure 752·9 ± 7·2 mmHg, relative humidity 43·6 ± 2·8%, wind speed 4·9 ± 1·2 m s−1, gust speed 7·7 ± 1·4 m s−1, wind distribution 13·7 ± 2·0 °.

The proportion of the total daily catches of B. tabaci caught in the highest traps increased as the mean trapping period temperature increased. This relationship was highly significant for both females (r2 = 72·30, F(1,6) = 15·65, P = 0·007) and males (r2 = 77·20, F(1,6) = 20·32, P = 0·004) (Fig. 6). None of the other meteorological variables explained a significant part of the variation in the proportion of the male or female whiteflies trapped at 7·2 m AGL when added to the regression model containing temperature. The proportion of the daily catches trapped at the most distant traps, 100 m from the source field, was not significantly correlated with any of the meteorological variables measured during the trapping periods. Thus, temperature affected the vertical, but not the horizontal, displacement of B. tabaci.

Figure 6.

Relationship between the percentage of the daily catch of B. tabaci that were trapped 7·2 m above ground level and the mean air temperature (T) during the trapping period. Data are presented separately for females (dark circles and solid line) and males (open circles and broken line). The equations of the regression lines are Pfemale = −5·19 + 0·27T and Pmale = −4·86 + 0·22T.

Discussion

A small proportion of B. tabaci enter a behavioural phase during which their flight is predominantly vertical, toward the short wavelengths of skylight, and when they ignore host plant cues. The rest of the population do not enter this phase of flight, but produce enough lift just to remain airborne, not climbing away from the ground, and respond instead to vegetative cues when they are perceived. These two groups thus exhibit migratory and foraging flight, respectively, as defined by Kennedy (1985). The distribution of these behaviours within the populations of B. tabaci studied here resulted in an aerial distribution, over the 7·2 m vertical measurement in this study, that was strongly biased to flight near to the ground. This is the greatest vertical range yet reported for studies on this insect, and the vertical distribution patterns are in general agreement with those of Gerling & Horowitz (1984) and Byrne et al. (1986), in which a strong negative relationship between aerial distribution and height was reported. Our study was only concerned with the initial phase of dispersal, over the first 100 m after take-off. Such a scope did, however, provide insight into the displacement of this insect and enabled a detailed study of this important phase of dispersal. Although our sample sizes varied by two orders of magnitude between trapping days, there was remarkably low variability in this relationship between samples (Fig. 3), and mathematical descriptions of the general form y = be−ax were generated for the relationships between height (H) and the proportion of males and females trapped at these heights.

The method of trapping was necessarily destructive, thus precluding measurement of the behavioural state or morphometric analysis of the captured insects. We therefore assumed that the majority of whiteflies that flew as high as the traps at 2·4 m above the ground were engaged in active climbing. Airflow at the experimental site is dominated by local mountain/valley wind flows (Brown, Machibya & Russell 1995), and convective air motions were minimal at the time when these experiments were conducted. Some mixing of individuals in these behavioural phases will undoubtedly have occurred, however, and the distinction between migratory and foraging individuals is probably not strictly binary, following more of a continuum (Dingle 1996). The data reported here show that for both male and female B. tabaci 70–75% of the whiteflies in flight were trapped close to the ground, presumably engaging in foraging flight. This percentage is in close agreement with that found in laboratory flight assays with free-flying B. tabaci (Blackmer et al. 1995), in which 76% of the whiteflies were attracted toward a vegetative cue. In that study, 18% were classified as exhibiting ‘intermittent attraction’ and that is approximately the proportion of the males and females that were trapped at 2·4 m. In addition, 6% exhibited a classical migratory behaviour (sensuKennedy 1985) with vertical flight towards white light, and this was the approximate proportion of whiteflies trapped in the traps at 4·8 and 7·2 m above the ground. The close agreement between the distribution of behavioural states found in the laboratory (Blackmer et al. 1995) and the vertical distribution patterns in this study provides some indirect evidence that insects in the uppermost traps were in a migratory behavioural phase. Additionally, the collection of marked whiteflies in the uppermost trap placed adjacent to the source field strengthens the argument that B. tabaci have a strong capacity for directed flight under field conditions, up into the more mobile air layers.

Aerial catches of B. tabaci did not change with distance as much as with height, and, over the 100-m trapping grid, there was only a 40% reduction in trap catch (Fig. 4). Whitefly dispersal is generally linear, in the direction of the wind, so this reduction may be attributed to variation in wind direction, as it was not always in exact alignment with the frame transect. There was, however, no statistical relationship between the proportion of the total catch of either sex at 100 m and wind distribution from the transect line (r2 < 0·01, F(1,6) < 1·0, P > 0·05, for females and males). Removal of insects from the aerial population by the traps may have contributed to this decline, because 5·6% of the area of each frame was covered with sticky traps. Assuming removal of that proportion of insects every 20 m, only 20·6% of the starting insects would have been removed before arrival at the 100 m frame. The increase in trap catch at 20 m from the edge of the source field may have been an artefact, caused by the blockage of the lower traps at the field edges by adjacent crop and field borders. The relationship between distance (D) and aerial density of the general form y = bx−a, provides a method for predicting the proportion of flying insects in flight that will be found at a particular distance from their place of take-off, though noting the cautions above. Overall, the shape of this relationship between distance and trap catch corroborates the field data of Byrne et al. (1996) from the same field site. In that study, the number of marked whiteflies caught in small suction traps on the ground close to the source field was high, dropping off gradually up to 1500 m from the marked field. Their data also showed a peak in whitefly abundance in the traps between 1500 and 2500 m from the source field, possibly associated with an increase in the landing rate of postmigratory individuals. This pattern was not observed in this study, probably due to the short distance of the transect. It is therefore prudent to remember that although insect distributions may be described by mathematical expressions, behaviours can act to modify these general relationships, and to concentrate animals at particular resource sites.

The biology of B. tabaci makes it unlikely that a strict trade-off between egg production and flight would occur as predicted by Johnson's (1969) oogenesis–flight syndrome model, as there is rapid egg maturation in this species. The weight of even a few eggs in such a small insect would, however, provide an energetic cost during flight. A lower eggload would be expected in those individuals that are migratory, to reduce the wing-loading for flight, and yet to provide some eggs for rapid colonization of new habitats. The first test of this hypothesis by Blackmer et al. (1995) demonstrated, however, that egg protein levels were positively correlated with the flight duration of female B. tabaci in laboratory tests. The correlation was weak, however, and protein levels were not corrected for insect weight. In addition, the egg protein could have been in the eggs, fat body or haemolymph and so measurement may not have reflected eggload per se. In the current study, the lowest eggloads were in female B. tabaci that we assume to have been in the migratory phase of flight. This result supports the hypothesis that flight and egg production are competing for the available energy resources, and that eggload is a limiting factor in the ability of B. tabaci to undergo extended migratory flights. In addition, height was the only significant factor in the analysis of variance on eggload, with lower eggloads being found as height increased. It is still unclear whether eggload is the result of the insect being in either a foraging or migratory behavioural state, or whether the behavioural state is dependent on the eggload. Either side of this question will probably be under the control of hormonal factors (Rankin 1989, 1991), but further research is needed to determine the role of oogenesis in whitefly dispersal ecology.

The significant correlation between mean temperature and the proportion of the total whiteflies captured in the uppermost traps indicates that B. tabaci flight activity increases as temperature increases, as seen in other small flying homoptera (Walters & Dixon 1984). This result may also reflect some atmospheric mixing processes that could have become more important at higher temperatures (Drake & Farrow 1988). The larger intercept of the regression equation in Fig. 6 for females, and its slightly steeper slope, shows that a greater proportion of the female catches were trapped in the uppermost traps, compared to males. Thus, females were climbing faster, despite the extra wing-loading caused by their eggs. Initial rates of climb for female and male B. tabaci have been measured in a vertical flight chamber at 28 °C, and were found to be 3·4 and 2·6 cm s−1, respectively, dropping to 1·3 and 0·8 cm s−1 after 40 min continuous flight (Blackmer & Byrne 1993). There are no previous studies on the effects of temperature on B. tabaci rates of climb, but the strong correlation between temperature and the proportion of the total catch found at 7·2 m indicates that whiteflies are temperature sensitive, and may help to explain the formerly reported temporal variation in flight activity of B. tabaci in the field (Byrne & von Bretzel 1987). By substitution, the regression equations of the relationships between the proportion of B. tabaci trapped in the uppermost traps and the mean temperature can also be used to calculate the thermal minimum for significant migratory flight. For females, this was 19·2 °C, while for males it was 22·0 °C. In both cases, these values are greater than the thermal minimum for flight, reported previously for B. tabaci in field studies. Meyerdirk & Moreno (1984) found that whitefly flight activity was restricted at or below 15·9 °C, and although not explicitly stated, the aerial catch data of Bellows et al. (1988) show a thermal minimum of between 6 and 15 °C, depending on the trapping date. Physiological differences in foraging and migratory individuals may explain this discrepancy, and may indicate that in B. tabaci there are different thermal requirements for foraging and migratory flight.

Prediction of population dispersal patterns of mobile insects that exhibit migratory behaviour must be based on a detailed understanding of these behaviours at the individual level (McNeil & Roitberg 1985), and how they in turn interact with physiological variables such as eggload. This is especially important in the case of pest insects in which prediction of the aerial distribution and variation in eggload is of primary concern when modelling the risk of immigration into crops. Our results underscore the relationship between dispersal patterns and the behaviour of migratory insects, and provide clear relationships that can be used in simulations of dispersal by B. tabaci.

Acknowledgements

We are grateful to John All, Andrés Amaya, Joe Bohanick, Chandra Curtin, Erich Draeger, Clayton Mullis, John Palumbo, Fransisco Reyes, Klaas Veenstra and Ryan White for their tireless technical assistance during this study. Thanks also to Davie Brooks of Pasquinalli Brothers Produce for generously providing access to field sites adjacent to Yuma Agricultural Center, and to Spence Behmer and Klaas Veenstra for comments on the manuscript. Financial support was provided, in part, by competitive grant 94–37302–0541 from the United States Department of Agriculture.

Received 3 June 1997;revision received 16 February 1998

Ancillary