# Environmental Entomology

### Editor-in-Chief

E. Alan Cameron

1.295
1.689
33 out of 92

Published on behalf of

## Spatial Autocorrelation Analysis of a Ceratitis capitata (Diptera: Tephritidae) Adult Population in a Mixed Deciduous Fruit Orchard in Northern Greece

Nikos T. Papadopoulos, Byron I. Katsoyannos, David Nestle
319-326 First published online: 1 April 2003

## Abstract

We studied the spatiotemporal dispersion patterns of Mediterranean fruit fly, Ceratitis capitata (Wiedemann), in a mixed, deciduous, fruit orchard in Thessaloniki (northern Greece), using spatial autocorrelation methods to analyze adult trapping data. Each trapping station consisted of a Jackson trap (baited with trimedlure) and a McPhail trap (baited with ammonium acetate, putrescine, and trimethylamine). More males than females were captured throughout the season. Males and females exhibited different spatial dispersion patterns. Females were first detected during the summer (June–July) in apricot and peach trees bearing ripe fruits and significantly aggregated there toward the end of July. In the autumn, females significantly aggregated in apple orchards bearing mature fruits. Early aggregations of males were first detected in August in cherries and plums. In September, males significantly clustered in pears at the edges of the orchard, and by October, after an increase in population density, their spatial dispersion pattern became random (no significant spatial autocorrelation). At the end of the season (November), the dispersion of both sexes became random. Our results show that spatial autocorrelation statistics can provide an important tool in studying the spatial dynamics of this fly even in small orchards. Results also suggest that the incorporation of knowledge on spatial patterns into area-wide control projects may improve monitoring efforts and reduce program costs.

• spatial dispersion patterns
• spatial autocorrelation
• Moran’s I
• mixed fruit orchard
• host phenology

## Introduction

SPATIAL DISPERSION and movement is as important as birth and death rates for the population dynamics of insects, and also one of the most characteristic (fundamental and applied) ecological properties of species (Taylor 1984). Insect populations are spatially heterogeneous in their densities, and this heterogeneity, together with the temporal changes in population loads, constitute an important element in the development of management procedures and in the analysis of insect pest population dynamics (Liebhold et al. 1991, Nestel and Klein 1995). Traditionally, spatial dispersion patterns of insects have been studied by indices that are based on the variance-mean relationship of samples without considering the spatial location of those samples. However, sample observation may be interdependent, and, therefore, other strategies might be more appropriate in analyzing spatial dispersion patterns (Brenner et al. 1998). Spatial autocorrelation is a statistical procedure that determines the degree of dependence of spatially related data. This method provides a more direct measure of spatial dependence than dispersion indices and has been used in the past in the study of insect spatial dispersion patterns (Liebhold et al. 1991, Nestel and Klein 1995).

The Mediterranean fruit fly, Ceratitis capitata (Wiedemann) (Diptera: Tephritidae) is one of the most important fruit pests in the world because it develops in >300 fruit species, most of which are of high commercial value (Liquido et al. 1991). Many aspects of the population dynamics of this fly have been studied during the last few decades. Most of these studies analyze the temporal patterns of adult capture data, with some of them examining the habitat preference of C. capitata mainly on tropical settings (Vargas et al. 1983; Nishida et al. 1985; Harris and Lee 1986, 1987; Harris et al. 1993), and the Mediterranean region (Papadopoulos et al. 1996, Israely et al. 1997, Katsoyannos et al. 1998, Papadopoulos et al. 2001b). These studies suggest a patchy distribution of the Mediterranean fruit fly under low population densities that becomes more random as the population increases and the fly searches for food resources and oviposition sites (Bateman 1972, Prokopy et al. 1996). The importance of the abundance and distribution of host trees, and fruit abundance and availability on the population increase, and on the spatial dispersion patterns of C. capitata has been proposed in the past (Rivnay 1950, Vargas et al. 1983, Eskafi and Kolbe 1990, Katsoyannos et al. 1998, Papadopoulos et al. 2001b). However, there are no studies systematically analyzing the spatiotemporal patterns of this fly either in large areas or at the orchard level. Furthermore, there are no studies that link C. capitata spatial patterns with the environmental heterogeneity of the habitat. Such an analysis might prove to be useful for the development and/or improvement of sampling and management programs for C. capitata and other pest tephritid species.

During the last decades, area-wide pest management projects have been undertaken against C. capitata in many parts of the world (Hendrichs et al. 1995). Such programs rely on the initial and uniform area-wide application of bait sprays, followed, in most cases, by releases of sterile insects. Uniform application of the control measures over large areas has a negative impact both on the cost of the programs and in the environment where habitats may be perturbed by the unnecessary application of insecticides. Identifying key habitats in which the fly develops early in the season under low population densities could help to increase the efficacy of pesticide application, and reduce program costs. Hence, knowledge about the spatial and temporal dispersion patterns of this fly on an areabeing considered for area-wide management may provide an important background for the sound application of management tools and techniques. The study of the spatial patterns of insects, and the effect of environmental heterogeneity upon these patterns, may also provide important information on the ecology of the pest and on the factors that drive their population dynamics. Results derived from such studies could provide the basis for a more coherent application of sampling and control schemes.

This study presents the results of an analysis on the spatial and temporal dispersion patterns of the Mediterranean fruit fly in a farm close to Thessaloniki, northern Greece. We used spatial autocorrelation procedures to analyze data derived from another recent study concerning the early detection and the temporal patterns of C. capitata population in the area using two trapping systems (Papadopoulos et al. 2001c). Special emphasis has been placed on the spatial dispersion of this insect early in the season, toward autumn when the population increases, in the middle of autumn when there is a shift in sex dominance, and toward winter when the population size declines. We also studied differences in spatial dynamics between the two sexes using the best current available adult trapping systems for this species (Heath et al. 1997; Katsoyannos et al. 1999a, 1999b).

## Materials and Methods

### Study Area and Climate

The study was conducted during 1998 in the area of Thessaloniki, northern Greece (40.3° northern latitude, 22.5° longitude), in the farm of the Aristotle University of Thessaloniki, which is located ≈800 m from the sea. A warm and dry summer and a mild and wet winter characterize the climate of the area. In winter, absolute minimum temperature falls often below freezing. Average monthly temperatures range between 18–26°C from May to September and between 5–13°C from October to April (see also Papadopoulos et al. 1996, 1998, 2001c).

### Sampling, Plots, and Procedures

Spatial and temporal patterns of C. capitata were followed in a mixed fruit orchard (≈2 ha), part of the University farm, which was composed of 0.9 ha of stone fruit trees (apricots [PrunusarmeniacaL.],peaches [Prunusper-sica (L.) Batsch], plums [Prunus domestica L.], and cherries [Prunus avium L.]); 0.2 ha of pears (Pyrus communis L.); and 0.9 ha of apples (Malus sylvestris Mill) (mostly of the Granny Smith variety). The fruiting period in the experimental orchard begins in May and ends in late autumn, apricots and peaches maturing first and pears and apples last (Papadopoulos et al. 2001c). Details on fruit availability and the ripening sequence of the different host trees in the experimental farm are given below. The location of the plots of fruit tree within the orchard is schematically represented in Fig. 1. The entire orchard was split into 30 subplots and one sampling station was established in each subplot. Therefore, sampling stations covered the entire orchard and the different fruit species. Each sampling station consisted of one Jackson and one International Pheromone McPhail Trap (IPMT; International Pheromone Systems, South Wirral, England) trap, separated by a distance of ≈10 m. Jackson traps baited with trimedlure (AgriSence, Fresno, CA) are male-specific, and the IPMT baited with three dispensers loaded with the synergistically acting food attractants ammonium acetate, putrescine, and trimethylamine, respectively (Biolure, Consept, Bend, OR), captures predominantly females (Heath et al. 1997, Katsoyannos et al. 1999b). Water (300 ml) plus a 0.01% of the surfactant Triton (Union Carbide, Danbury, CT) was added to the base of IPMT traps as a retention fluid. Sampling stations were kept in the same location throughout the entire study. In the apple plot a total of 15 sampling stations were used, whereas in the smaller apricot, peach, plum, cherry, and pear plots we were limited to three sampling stations per plot. Traps service (sampling) was carried out every 7 d, and all male and female flies trapped in each sampling station were counted. Spatial patterns were inferred from females trapped in IPMT traps, and males trapped in both IPMT and Jackson traps. Both trap types were installed at the beginning of April and serviced until the end of December. The combination of the above two traps is considered to be the current best system available for trapping male and female C. capitata adults (Epsky et al. 1999; Katsoyannos et al. 1999a, 1999b).

Fig. 1.

Schematic representation of the sampling plots (with the center marked) and the relative position of the different host trees in the mixed orchard.

### Ripening Sequence and Fruit Availability

The host that first bore mature fruits in the experimental farm was cherry. Its ripening period began around 15 May, peaked in the first week of June, and ended at approximately 20 June. In the middle of June, apricots started to mature. Their peak of ripeness was observed at the beginning of July, and there were no more fruits left on the trees after 20 July. During this year, peach production was unusually low in the experimental farm. The few peaches of the different varieties began to mature around mid July and there were no more fruits left after the beginning of September. Plums ripening period lasted from mid August until the beginning of September, whereas that of pears from around 10 August until 15 September, with its ripening peak occurring at the beginning of September when pear harvest was conducted. Apples ripening period started in the middle of September: Golden Delicious variety, which constituted approximately 15% of the apple trees in the orchard, peaked first at around 20 September, whereas Granny Smith, the dominant apple variety in the orchard, started its ripening period at the end of September and peaked in the middle of October. By November, there were few apples of either varieties left on the trees. Fruit production was normal in all the tree species in the orchard except for peaches.

### Statistical Analysis

Trap counts were square root transformed to homogenize the variance. The effects of sex and sampling station on C. capitata catches were analyzed with a repeated measures analysis of variance (ANOVA) (Sokal and Rohlf 1995). In this case, the measurement of catches in all the sampling stations over sampling dates provides a restriction on randomization over time. Sampling station and sex effects were tested using the mean square of the interaction between these two terms as error term, whereas the effect of sampling date and the interaction between sampling date and sex were analyzed using the residual error term.

### Spatial Analysis

Spatial autocorrelation of C. capitata adults captured in either IPMT or Jackson traps-grid was investigated using Moran I spatial statistics (Cliff and Ord 1981). Spatial autocorrelation, especially Moran’s I, has been widely used in the study of insect spatial patterns (e.g., Liebhold and Elkinton 1989, Midgarden et al. 1993, Nestel and Klein 1995, Kitron et al. 1996, Efron et al. 2001). Moran I index compares geographic neighbors in terms of their deviation from the mean of all observations, and can be tested for significant deviation from the null hypothesis that there is no spatial autocorrelation. The Moran’s I index for a given lag class (e.g., values included within the boundaries of a distance class, limited by a minimal and a maximal distance size) is calculated as:

$$mathtex$$I = n\sum \sum {w_{ij}}{z_i}{z_j}/{s_0}\sum z_i^2$$mathtex$$

where n is the number of sampling stations,

$$mathtex$${z_i} = ({x_i} + x),{z_j} = ({x_j} + x)$$mathtex$$

xi or xj is the observation at the ith or jth sampling station, wij is a weight denoting the connection between stations i and j (for example, one for neighbors and 0 otherwise), and so is the sum of the weights (Sokal and Oden 1978a). In this study, observations refer to the number of adult C. capitata (males or females) captured in traps located on agiven sampling station and sampling date. Adjacent observations were those related by both orthogonal and diagonal connections. Autocorrelation analyses were performed on up to five lag classes (i.e., separation distances with mutually exclusive intervals). The boundaries of these intervals were chosen in such a way that the largest distance class included a minimum of 30 pair points. Correlograms were tested for spatial autocorrelation significance (under the randomization assumption, “R”) using a Bonferroni approximation and the software “SAAP” (Sokal and Oden 1978b, Wartenberg 1989). Under “R ” the expectation for Moran I in the absence of spatial autocorrelation is:

$$mathtex$${\rm{E}}(I) = - 1/({\rm{n}} - 1)$$mathtex$$

## Results

A total of 9,584 males and 5,542 females were captured throughout the sampling period in all the traps of the sampling grid. The temporal pattern of captures is given in Fig. 2. The first female was detected at the end of June, whereas the first male was detected at the beginning of August. Captures of females were higher than those of males until the beginning of October; however, from the middle of October until the end of the November male captures became dominant. A repeated measures ANOVA revealed a significantly higher male-to-female capture level (F = 9.78; df = 1, 29; P < 0.01), and a significant effect of sampling station on the total number of fly captures (F = 2.36; df = 29, 29; P = 0.01). There was also a significant effect of the sampling date (F = 264.36; df = 25,1450; P < 0.01), because captures for both sexes were significantly higher from mid-September until mid-November than trapping levels before the autumn. The interaction between sampling date and sex was significant (F = 43.07; df = 25, 1450; P < 0.01) corroborating the observed temporal differences in captures between the two sexes (Fig. 2).

Fig. 2.

Temporal patterns of male and female C. capitata in a mixed orchard in northern Greece. Average fly capture per sampling date is derived from a grid of 30 trapping stations, as shown in Fig. 1.

The spatial patterns of trapped males and females on four sampling dates are diagrammatically shown in Fig. 3. The selected dates represent characteristic situations for a given sampling period, such as early detection and clustering, population increase toward the autumn and its effect upon spatial dispersion patterns, shift in sex dominance, and decline on the trapping levels toward winter. Figure 3 also shows areas within the orchard with ripe or ripening fruit (shaded areas). Females were initially detected toward the end of June in apricot and peach trees (Fig. 3), which were at their ripening peak. Significant female spatial autocorrelation was first detected in mid July in the apricot and peach subplots (Table 1 and Fig. 3). This significant aggregation was lost during August. Significant female spatial autocorrelation patterns were again detected in September (Table 1), when trapping of female expanded to the apple and pear plots, which were reaching maturation at that time (Fig. 3). In October, female trapping levels fluctuated, but they continued to significantly aggregate in the apple orchard (Table 1; Fig. 3). By mid November, with the sharp decline in the general trapping levels (Fig. 2), significant female clustering disappeared (Table 1; Fig. 3) and the few female catches were spread at random throughout the orchard. Male captures were first recorded at the beginning of August in the cherry and plum trees, which were beginning to ripen (cherries already passed their ripening period that lasted from mid May until mid June, whereas plums were just starting to ripen) (Fig. 3). Significant male spatial autocorrelation was first detected at the beginning of September (Table 2), when male captures were mainly aggregated in the pear plots, which were at their ripening peak (Fig. 3). This initial aggregation pattern became random during mid October (Table 2), which most probably is related to the dispersion of males to other locations of the orchard, after an enormous increase in their population size at that time (Fig. 2). At the beginning of November, male catches became clustered in the pears and plums-cherries rather than in the apples. At the end of the season, with the decline in trapping levels, male spatial patterns in the mixed orchard became random (Table 2; Fig. 3).

Fig. 3.

Spatial dispersion patterns of male and female captures at four time periods: early fly detection (a), population increase during mid season (b), shift in sex dominance during mid to late season (c), and population decline toward the winter (d). Each graph illustrates the relative density of either male (left graphs) or female (right graphs) captures during a representative sampling date. Rectangles within the graphs stand for trapping levels at specific sampling stations. The level of trapping in a specific location of the orchard is represented by the number of rectangles at that specific location: each rectangle representing a proportion (between 5–30%) of the total number of male or female flies (top right of each graph) captured during that specific date in the entire orchard. The diagrams also show areas (shaded) possessing trees with ripening fruit.

View this table:
View this table:

## Discussion

Our results show that the Mediterranean fruit fly has a dynamicspatial dispersion pattern in which trapped flies, which serve as an indication of fly location and population loads, aggregate in space in response to the changing environment. Moreover, our results also show a different spatial pattern for males and females, which suggests that the two sexes were responding differently to the spatial and temporal environmental variability found in the orchard. In general, female spatial aggregation is closely related to the phenology of the host tree and to the sequential availability of ripe or semiripe fruits in the orchard (Fig. 3). Early in the summer, female captures occurred almost exclusively in the apricot and peach plots, in correspondence with the fruit ripening period for these two crops in the experimental farm. Then, between mid August and early September, female catches concentrated in the pear plots, corresponding with the ripening period for the pear varieties in the orchard. Early in September, at the onset of the apple’s ripening period, clusters of trapped female started to appear in the apple plot. Female clustering on apples continues throughout the apple ripening period (September–October). However, male spatial pattern differed from that of females and did not closely follow the sequential ripening of fruits (Fig. 3). For example, whereas in September females clustered in apple orchard, most males were trapped in pears at the edges of the orchard. Similarly, in October, males had a more random dispersion pattern throughout the entire orchard, whereas females were clustered in the apple plot.

Positive autocorrelation indicate that values at adjacent localities are similar, whereas negative spatial autocorrelation signifies differences between adjacent pairs (Sokal and Oden 1978a). When spatial autocorrelation coefficients on the investigated variable in a geographic space is calculated for several distance classes, the resulting correlograms provide information on spatial interactions of the variable occurring at several geographic scales (Oden 1984). Significant correlograms in which positive autocorrelations are found at short distance lags and negative spatial autocorrelations are found at relatively large distance classes illustrate situations in which values are arranged in space in a gradient or cline way, which is the most common correlogram profile for natural populations (Sokal and Oden 1978b). Most of the significant correlograms in this study conformed to this type of pattern, which suggest that captured flies within the orchard were arranged in space in a gradient pattern in which nearby sampling stations have similar values and farther away sampling stations showed significant negative autocorrelation coefficients. As seen in Fig. 3 and the selected dates that exemplified different periods of the season in terms of spatial arrangements, gradient patterns did not show any stationarity, suggesting that gradient patterns were formed and resulted from an ongoing changing environment. However, the possibility of border effects, as may be suspected from the relative small geographic scale of the study, may be minimal or buffered by the intra-orchard environmental factors that are shaping the spatial patterns of the fly. In the case of this orchard, the surrounding (i.e., border) environment is quite similar in terms of vegetation composition, thus suggesting that the fly population will presumably experience similar environmental conditions that may result in large distance significant positive autocorrelations (i.e., the creation of a circular gradient) (Sokal and Oden 1978b). In general, almost none of the correlograms in this study resulted in such a pattern. In addition, the fact that significant correlograms always had negative spatial autocorrelation coefficients in the largest distance lags (Tables 1 and 2), and that the position of large fly captures differed during the different periods of the study (Fig. 3), strengthens the sense that internal environmental factors inside the orchard had a more stronger effect upon the fly spatial patterns than the effect of border factors.

Even though the few adults captured during the early periods of the season were trapped on neighboring sampling stations (i.e., aggregated), the low captured numbers characteristic of this period usually resulted in nonsignificant spatial patterns. This derives from the inability of the statistics to discern spatial patterns that are based on very low catches that slightly departed from the average trapping level for all the stations. This situation was common for both sexes during the early sampling dates, such as June and July, and extended until August. Significant spatial patterns in both sexes started to appear after September, when population numbers and captures highly increased. Significant spatial patterns extended throughout all the period of high trapping (September–October), except for some specific dates when the correlograms were nonsignificant, suggesting randomness in the spatial arrangement of trapping levels throughout the orchard during these dates. Random patterns for males after the beginning of September (specifically on 22 September and 20 October) coincide with the increases in population trapping levels (see Table 2 and Fig. 2). Similarly, a random spatial pattern in females (especially on 15 September) was also associated with a sharp relative increase in catches during that date (Fig. 2). Thus, seems to be that population increases and the spread of individuals to other trapping stations creates a small temporal situation in which statistically heterogeneous values are randomly positioned in the space of the orchard. The lost of spatial aggregation patterns after November in both sexes were associated with a general decline in captures, and a more widespread distribution of captures throughout the orchard. This situation also suggests that the underlying factors that determined the aggregation of captures in certain areas of the orchard were weakened or disappearing.

The spatiotemporal dispersion patterns of the C. capitata population in the orchard probably reflect the addition of individual adult fly foraging behaviors. Flies require food, mates, egg-laying sites, and refugia as essential resources (Prokopy et al. 1994). In most of the fruit flies, foraging for these resources is a dynamicprocess. Flies are known to adjust their foraging behavior in response to the changes in the spatial, temporal, and seasonal distribution of food and other resources (Hendrichs et al. 1991). Observed differences between male and female spatial dispersion patterns seem to be related to their differential use and need in resources and to the physiological composition of the fly population. Sexually mature females, for example, are expected to forage for egg-laying sites, thus aggregating and getting trapped where host fruit has become ripe and favorable for egg-laying. This may also be the place were females forage for food sources. Males, however, in addition to foraging for food, probably concentrate on areas that provide appropriate shelters and sites to exhibit calling and leking behavior (Prokopy and Hendrichs 1979, Hendrichs and Hendrichs 1990, Hendrichs et al. 1991). Thus, the larger orchard area occupied by males during the autumn could be related to the different physiological and behavioral needs of the male population, which are not only driven by the ripening of fruit. As a result, male flies may search for food resources, and probably other plant substances (Papadopoulos et al. 2001a, Shelly 2001) outside of the female egg-laying areas. Similarly, competition between males for leking sites may act as a dispersing force driving male flies away from very competitive sites resulting in a more ample, random, or uniform occupation of the orchard by males than that observed in females (Fig. 2).

Although this study was conducted in a relatively small geographicarea, our results show that the spatial and temporal patterns of the C. capitata, and probably of other fruit flies, are driven by a multitude of ecological forces. The inclusion of the spatial and geo-graphicdimension to basicstudies of insect ecology, as well as to applied projects of fruit fly area-wide control, could unveil important aspects of the ecology and behavior of the target population. This knowledge can greatly improve the performance of control programs. Not only can we better specify the location and timing of pesticide applications and the application of other control methods, but we may also better understand the gaps that perturb the effective application of control projects. The spatial dimension of the population ecology of a fruit fly could also provide a more holistic view of an area-wide control program, a perspective that will allow suggesting and designing more sound strategies for monitoring and control.

## Acknowledgments

We thank the two anonymous referees for their helpful comments on the manuscript.