ABSTRACT. : We investigated aggregation patterns in three fruit fly species economically important in Brazil, namely Ceratitis capitata, Anastrepha fraterculus, and A. obliqua. The study was carried out in a buffer zone and two neighbourhoods by comparing two-time series associated with the management strategy of fruit flies (systems approach). The abundance of these three species significantly decreased over the years with a negative binomial regression model describing the relationship between abundance and time in the entire area, buffer zone, and their neighbourhoods. In addition, the negative binomial model was also well fitted to the frequency distribution data of fruit flies in all analyzed scenarios. Anastrepha obliqua showed the highest aggregation degree, considering both the entire area and time series. A. fraterculus exhibited the lowest aggregation level, and C. capitata showed an intermediate degree. The buffer zone exhibited the highest aggregation degree for all species, and neighbourhood 2 exhibited the lowest aggregation degree. The aggregation degree was higher in the time series impacted by the systems approach than the series in the first years of its implementation.
Keywords: Ceratitis capitata, Anastrepha fraterculus, Anastrepha obliqua, buffer zone, regression models, frequency distribution.
CROP PROTECTION
Aggregation and spatio-temporal dynamics of fruit flies (Diptera, Tephritidae) in papaya orchards associated with different area delimitations in Brazil
Received: 30 April 2020
Accepted: 15 September 2020
The main tephritid species distributed across the Neotropical Region include two taxonomic groups, Anastrepha, which comprises 328 species native to the Americas (Norrbom et al., 2021) and occurs in all regions of Brazil (Zucchi & Moraes, 2021a), and Ceratitis capitata, a species with worldwide distribution (Vera, Rodriguez, Segura, Cladera, & Sutherst, 2002), and widely distributed in Brazil (Zucchi & Moraes, 2021b). The genus Anastrepha is distributed from a latitude of 35ºN to 35ºS, in different environments, from the sea level to an altitude of 3,000 m or more. Anastrepha spp. are classified in several infrageneric groups (Norrbom, Zucchi & Hernández-Ortiz, 1999), among which the fraterculus group has several economically important species. To understand the distribution of fruit flies, it is essential to associate them with their host plants, emphasizing the fruit-growing regions, making this interpretation a biogeographical criterion (Aluja & Liedo, 1986). A significant diversity of fruits is generally infested by fruit flies, including papaya, a fruit with a high production on a commercial scale (Martins, Uramoto, & Malavasi, 2000).
The papaya-producing areas in Brazil contribute significantly to the international market, owing to the annual production capacity and for currently being in areas considered to have a low prevalence of fruit flies (Martins, Fornazier, Ventura, Malavasi, & Ferreira, 2011). Papaya exports have been possible due to the implementation of an export program in the USA. The program, recognized as systems approach, includes a set of actions during the pre- and post-harvest periods of papaya for the management of fruit flies (Martins & Malavasi, 2003b; 2003a). It started approximately 25 years ago in the state of Espírito Santo, Brazil, following the requirements for quarantine security.
In Brazil, some species of Anastrepha and Ceratitis capitata are the most economically important flies because they attack various commercial orchards and are widely distributed in the country (Zucchi, 2000). The geographical distribution of these species is strongly linked to the distribution of their hosts (Selivon, 2000). However, the degree of association between host plants and fruit flies varies between regions (Malavasi, Morgante, & Zucchi, 1980). Anastrepha fraterculus and A. obliqua exhibit the highest degree of polyphagy (159 and 70 associated hosts, respectively) in Brazil (Zucchi & Moraes, 2021a). Ceratitis capitata, an exotic species, also has a wide range of hosts, being associated with 115 plant species in Brazil (Zucchi & Moraes, 2021b).
Recent studies have been conducted with fruit flies to better understand their spatial distribution in several areas of Brazil (Garcia, Araujo, Uramoto, Walder, & Zucchi, 2017; Frighetto, Machota Junior, Bortoli, Botton, & Guerra, 2019; Santos, Silva, & Miranda, 2020). These studies have taken into account landscape arrangement as one of the factors determining species distribution in different parts of the world (Meyer, Robertson, Peterson, & Mansell, 2008; Kounatidis et al., 2008; Gutierrez & Ponti, 2011; Flores et al., 2016). In general, changes in natural ecosystems, when converting them into agroecosystems, have intensified questions on how to better understand the community structure of pests and their natural enemies as well as investigating the processes determining the abundance of species in different areas (Vandermeer, 1995; Altieri & Nicholls, 2004). Therefore, studies on the spatio-temporal distribution pattern associated with environmental conditions are of high relevance in fruit fly management (Aluja et al., 2003; Aluja & Mangan, 2008; Rull et al., 2018).
Spatio-temporal patterns of distribution in fruit flies have been investigated using different statistical tools such as geostatistics and exponential family models to investigate the frequency distribution of species and populations (Garcia et al., 2017; Nicácio et al., 2019; Oliveira, Uchoa, Pereira, Nicácio, & Faccenda, 2019). These include Poisson and negative binomial functions, usually employed to investigate random and clumped ecological distribution patterns. Insects, in general, particularly fruit flies, tend to exhibit clumped distribution because of the common aggregation of larvae in fruits (Deus et al., 2016). Understanding the aggregation patterns of insects may be useful for pest management in critical areas, such as orchards infested by fruit flies. As for fruit flies in Brazil, this knowledge is still scarce, especially when it involves a time series reflecting real conditions in orchards.
In this study, a long-term time series of C. capitata, A. fraterculus, and A. obliqua was analyzed to investigate the aggregation patterns of fruit fly species, taking into account spatio-temporal distribution in different area delimitations and a time series associated with a management strategy of fruit flies (systems approach). We aim to compare the time series of the three fruit fly species in different spatial delimitations, namely a buffer zone and neighbourhoods 1 and 2, to verify the spatio-temporal distribution patterns in the areas. Therefore, we analyzed the influence of the native forest on the distribution of fruit fly species by comparing the abundance distribution for the three species as a function of time, and under different spatial delimitations in a distance gradient from the Atlantic rainforest remnant. Although previous studies have investigated aspects of population dynamics in fruit flies (Aluja, Ordano, Guillén, & Rull, 2012; Raga, Paula, Souza-Filho, & Castro, 2017; Galli, Michelotto, Bernardes-Soares, Mello-Martins, & Fischer, 2019), a comparative spatio-temporal distribution, taking into account the distance from the native forest as well as pest management, has yet to be performed, especially for the three most important fruit flies (C. capitata, A. fraterculus and A. obliqua).
The data set was obtained from the monitoring of fruit flies in commercial papaya crops in the State of Espírito Santo, Brazil. This monitoring is part of a set of measures adopted for the management of fruit flies in papaya orchards, the systems approach, which aims to maintain crops with a low prevalence of tephritids, i.e., below seven fruit flies per trap per day (FTD), thus enabling the export of fruits to the USA (Martins & Malavasi, 2003b; Martins et al., 2011).
The study region is located in the vicinity of the municipality of Linhares, 33 m of altitude, and it is considered a production center for papaya in the state of Espírito Santo, Brazil, with a mean production of 50 t ha-1 year (Instituto Capixaba de Pesquisa, Assistência Técnica e Extensão Rural [Incaper], 2020). In the region, Ombrophilous Dense Forest, also known as Tableland Forest, is prevalent and the predominant climate is warm tropical, and humid, with a rainy season in summer and dry season in winter, reaching a mean annual temperature of 23ºC (Instituto Chico Mendes de Conservação da Biodiversidade [ICMBio], 2020).
The fruit flies were collected weekly using McPhail traps, with 300 mL of an attractive solution of hydrolyzed corn protein (5%) for collection of Anastrepha specimens, and Jackson traps with the sex pheromone Trimedlure were used to capture males of C. capitata. The traps were distributed at 1 trap per 1 ha of cultivated area, using 50 McPhail traps and 50% Jackson traps, interspersed, at different georeferenced points (18º 02’ 35.7” to 20º 33’ 58.3” S and 39º 13’ 00.7” to 41º 09’ 16.1” W) in commercial papaya production fields in Espírito Santo State (Figure 1). Incaper - Instituto Capixaba de Pesquisa, Assistência Técnica e Extensão Rural - provided us with a time series from 1998 to 2010, containing abundant data for C. capitata, A. fraterculus, and A. obliqua. The nominal A. fraterculus is used here in sensu lato, as it comprises a complex of cryptic species.
The dataset was analyzed in different area delimitations to investigate the relationship between abundance and time. Additionally, the frequency distribution patterns of fruit fly species were analyzed in papaya orchards close and distant from the Reserva Natural Vale, an Atlantic forest remnant that is highly biodiverse with approximately 23,000 ha of protected area (Kierulff, Avelar, Ferreira, Povoa, & Bérnils, 2015). The area delimitations were classified as a buffer zone, neighbourhood 1, and neighbourhood 2 (Figure 1). The buffer zone is a delimitation of the Ministry of Environment of Brazil, an area established around a conservation unit to filter the negative impacts of activities occurring outside it, such as pollution, invasive species, and the advancement of human occupation, especially in units close to intensely occupied areas. Neighbourhood 1 borders the buffer zone and neighbourhood 2 is the pool of all areas further from the buffer zone.
As the present study aimed to investigate temporal space aggregation over a wide area regarding landscape composition, the delimited regions were investigated concerning the distance from the Atlantic forest remnant because it is the main element of the landscape capable of significantly influencing the distribution of fruit flies. Considering the center of the Atlantic Forest Reserve, the distances to the buffer zone, neighbourhood 1, and neighbourhood 2 are 9, 17, and 50 km, respectively.
The papaya orchards are randomly distributed in these regions with sizes between 2 and 275 ha, with a density of 1,100 to 1,500 plants ha-1 (Alves, 2003). A total of 332 orchards were studied in the buffer zone, 251 in neighbourhood 1, and 378 in neighbourhood 2, from 1998 to 2010.

The full-time series (1998-2010) was subdivided into two subseries, 1998 to 2003 and 2004 to 2010, to document and compare the evolution of the systems approach, and their effects on the abundance of C. capitata, A. fraterculus, and A. obliqua. The systems approach is a set of actions during the pre- and post-harvest periods, implemented in 1990s to improve production, harvesting, packaging, and transport of a commodity that meets the requirements for quarantine security (Martins & Malavasi, 2003b; Malavasi & Martins, 2005). For convenience, we named the time series as ‘continuous implementation of the systems approach’ for the series between 1998 and 2003, and ‘impact on abundance’ for the series between 2004 and 2010. Continuous implementation of the systems approach determines sequential, permanent, and uninterrupted management actions since the beginning of its implementation. The impact on abundance describes years after which abundance was visibly influenced by the systems approach.
The Shapiro-Wilk test was used to evaluate data normality, with results indicating that the relationship between abundance and rainfall, temperature, and humidity did not follow a normal distribution. Then, Spearman’s correlation was applied to investigate the correlation between the variables mentioned above. Generalized linear models (GLMs) with Poisson and negative binomial errors were used to analyze the relationship between fruit fly abundance and time, according to the series and subseries mentioned above. The Akaike information criterion (AIC) was used to select the best GLM model to fit the fruit fly dataset.
Based on the results obtained with the GLM models, the frequency distribution of fruit flies was fitted to the negative binomial distribution models to confirm whether the spatio-temporal distribution of C. capitata, A. fraterculus, and A. obliqua followed a clumped pattern, and also to compare the degree of aggregation among areas and time subseries. The number of flies per sampling unit was summarized as a frequency distribution, described as the area delimitations with 0, 1, 2, 3, …n fruit flies. The k parameter in the negative binomial distribution was used to estimate the spatial aggregation degree of flies. The value of k indicates the highest aggregation with k tending to zero. K was calculated by computing the maximum likelihood estimate for aggregation, and the fit of the negative binomial model was tested using the Pearson’s χ2 statistic (Crawley, 2007).
A total of 13,889 adults of C. capitata (FTD = 0.007), 2,999 of A. fraterculus (FTD = 0.006), and 797 of A. obliqua (FTD = 0.006) were collected in papaya orchards in the State of Espírito Santo, during a 13-year monitoring period. The three-time series for the three fruit fly species from 1998 to 2010 described the population dynamics of C. capitata, A. fraterculus, and A. obliqua, indicating a decreasing trend in abundance for the species, particularly visible after 2003 (Figure 2). Regression analyses with Poisson and negative binomial errors were performed for the three datasets showing a significant relationship between time (years) and abundance when data were fitted to the two functions (Table 1). However, the AIC value indicated that the negative binomial model was the best regression function to explain the abundance variation in response to the time series for C. capitata, A. fraterculus, and A. obliqua.


The abundance variation of the three species in response to time was also significant when fly abundances were compared within the entire area from 1998 to 2010 (Table 1). This result was also observed in the delimitated areas, buffer zone (Table 2), neighbourhood 1 (Table 3), and neighbourhood 2 (Table 4).
Spearman’s correlation analysis between the abundance of fruit flies and the variables temperature, humidity, rainfall, and time (1998-2010) indicated a single negative significant correlation (rho = -0.14 p < 0.001), i.e., between abundances of C. capitata and humidity.
The frequency distribution analysis (Table 5) shows significant clumped patterns for the three fruit fly species in all the areas studied. When studying the influence on each area, the three species were more aggregated in the buffer zone. Observing each species separately, it was noticed that the aggregation degree was practically the same for C. capitata in the entire area, neighbourhoods 1, and 2, but the buffer zone exhibited its highest aggregation, reaching k = 0.08 (Table 5). Anastrepha fraterculus also exhibited the maximum aggregation degree in the buffer zone, but with k = 0.13, and in neighbourhood 2, it showed its lowest aggregation degree, with k = 0.25. Anastrepha obliqua exhibited its maximum aggregation in the buffer zone, with k = 0.03, and its lowest aggregation level in neighbourhood 2.
The comparison between the continuous implementation of the systems approach and the impact on abundance also revealed that the two times series exhibited clumped frequency distribution for the three species analyzed (Table 6). The species showed more aggregated patterns during the impact on abundance when the entire area was evaluated. Further evidence of an increase in aggregation can be observed when comparing the C. capitata time series in the buffer zone, with k = 0.13 for the series between 1998 and 2003, k = 0.07 for the series between 2004 and 2010, and A. fraterculus when the k value decreased from 0.2 (1998-2003) to 0.09 (2004-2010; Table 6). Although A. obliqua has exhibited a clear trend to increase its aggregation in the entire area, in the buffer zone, and neighbourhoods 2, its aggregation degree trended to decrease. In both areas, a loss of aggregation was observed, with k values increasing from the 1998-2003 series to the 2004-2010 series. An increase in aggregation was only observed for the entire area and neighbourhood 1.
Among the species of Tephritidae reported in agricultural orchards in Brazil, C. capitata, A. fraterculus, and A. obliqua have been dominant (Querino, Maia, Lopes, Alvarenga, & Zucchi, 2014; Lopes et al., 2015; Duarte, Garcia, & Andaló, 2016; Rosa et al., 2018). These species are widely distributed in the Brazilian territory, are highly polyphagous, and have a wide range of hosts already registered (Zucchi & Moraes, 2021a; 2021b). As observed in surveys of fruit flies in commercial plantations (Aluja et al., 1996), also in the areas of commercial production of papaya in the state of Espírito Santo, Brazil, few species of fruit flies are common.




Ceratitis capitata was more abundant than the other two species in papaya orchards in the State of Espírito Santo. Studies indicate that C. capitata prefers agricultural and/or anthropized environments when compared to Anastrepha species that prefer environments with native vegetation (Alvarenga, Alves, Silva, Lopes, & Lopes, 2010; Querino et al., 2014). The collection methods used in this study may also have influenced the abundance of fruit flies, mainly because C. capitata was captured with a trap using a sexual pheromone as bait, and Anastrepha species were collected using food attractive traps; however, both methods are usually used for fruit fly monitoring (Silva et al., 2011). Additionally, the eventual dominance of C. capitata in papaya orchards in Espírito Santo can also be interpreted from results showing temporal segregation between C. capitata and A. obliqua, as previously documented in loquat orchards (Lopes et al., 2015). There are also results showing high dominance of C. capitata when co-occurring with A. fraterculus in different localities of Argentina (Segura et al., 2006).
The decrease in the abundance of fruit flies, statistically confirmed with Poisson and negative binomial regressions, was potentially due to the impact of the systems approach applied in the studied area. Comparing the abundance values during the years, the decrease in population size is easily noticeable for A. fraterculus and A. obliqua. However, although C. capitata also exhibited a significant decrease in abundance in response to time, it was not as visible as observed in the other species, potentially due to its high variation in abundance values. In 2003, C. capitata exhibited a significantly higher abundance than A. fraterculus and A. obliqua, alongside its abundance in other years. There is no specific reason for this outbreak, although C. capitata outbreaks have commonly been noticed in other parts of the world where outbreaks frequently occur (Montoya, Cancino, Zenil, Gómez, & Villaseñor, 2005; Barnes, Targovska, & Franz, 2006; Carey, 2010).
In the systems approach, papaya fruits are harvested at stages 1 and 2 of ripeness (Martins & Malavasi, 2003b; Brasil, 2008). In these stages, the fruits have high concentrations of benzyl-isothiocyanate (BITC), a natural chemical compound present in fruit latex, and its concentration decreases as the fruit ripen (Seo, Tang, Sanidad, & Takenaka, 1983; Nascimento, Frighetto, Malavasi, & Habibe, 2003). BITC acts as an ovicide, inhibiting oviposition, and killing the initial larvae of fruit flies (Seo & Tang, 1982). There is no risk of infestation by these insects when the papaya is harvested in the early stages of maturation.
Despite having two models to describe significantly the fruit fly dynamics for the three species, the AIC values indicated that the negative binomial model was the best regression function to explain the relationship between time and fly abundance. In addition, the analysis taking into account the frequency distribution of the fruit flies confirmed the suitability of the negative binomial model for all scenarios analyzed (different periods: 1998 to 2003, 2004 to 2010, and different areas: entire area, buffer zone, and neighbourhoods 1 and 2). Insect count data are generally not normally distributed (Sileshi, 2006), and there is evidence showing that the frequency distribution in this taxonomic group is ecologically described by a negative binomial distribution (Deus et al., 2016; Maldonado Jr, Barbosa, Costa, Gonçalves, & Santos, 2016; Nicácio et al., 2019). This function, characterized by variance significantly higher than the mean (Crawley, 2013), was observed in our results.

Landscape heterogeneity in the buffer zone may have contributed to the greater aggregation of species, because it contains a large fragment of preserved natural vegetation exhibiting large quantities of resources available for fruit fly maintenance, which may dissipate the concentration of fruit flies in the same area. Fruit flies in agricultural areas tend to be influenced by the presence of fragments of adjacent native vegetation (Aluja et al., 1996; Uchôa & Nicácio, 2010). Forest fragments have contributed to the presence of Anastrepha species in adjacent guava orchards (Querino et al., 2014). According to Meats (2007), the conformation of the landscape is a relevant factor in the distribution of fruit flies. Natural environments may result in a more aggregated distribution of flies than other types of environments. In all the scenarios analyzed, A. obliqua exhibited abundances significantly lower than those of the other species, and under this condition, any movement of individuals among sampling units may proportionally influence the spatial pattern of distribution. Furthermore, this can result in a decrease in empty units, making the system prone to random occupancy, characterized by k parameter exhibiting values more distant from zero (Reigada & Godoy, 2005).
In neighbourhood 1, the aggregation degree of C. capitata and A. obliqua increased from the first series to the second series, while A. fraterculus maintained the same k value. This result, when compared to the buffer zone, allows us to conclude that the population aggregation behavior may change in response to different local conditions, and this is potentially intrinsic to each species, with responses including intense variation regarding distribution among areas. The aggregation trend from the first to the second series was also maintained in neighbourhood 2, but similarly to the buffer zone, A. obliqua decreased its aggregation from the first to the second series.
At the global scale, it is usual to observe changes in aggregation in insects described by outbreaks over time, mainly in response to climatic changes, with important implications for pest management (Williams & Liebhold, 1995). However, at the local scale, the spectrum of variable values changes, such as temperature and others generally investigated, is much more restricted, making the influence of temperature or other variables less impactful, as observed in our correlation tests. Therefore, at the local scale, there are potentially other factors capable of influencing the spatial distribution of insects. Changes in the spatial distribution patterns of host plants can influence the intensity of ecological interactions between plants and herbivorous insects (Hakes, Halpern, & Underwood, 2018). Papaya orchards in the investigated areas were not distributed significantly far from each other. This condition may be associated with the general aggregation pattern discussed here. However, distance and size are not homogeneous among the orchards, and these factors certainly influenced each species differently, explaining the small differences in aggregation observed in the three fruit fly species, with particular emphasis on A. obliqua.
Comparing the two periods in the entire area, a slight aggregation trend was found for the three species in the series 2004-2010. This result may be due to the management actions implemented in the papaya orchards in Espírito Santo State (Martins & Malavasi, 2003b), mainly because the fruit fly species aggregation is implied in several areas with zero individuals. The decrease in fruit flies in response to the systems approach implementation resulted in more empty areas when compared with the previous condition. This information is relevant, mainly for when sequential sampling plans are implemented to monitor critical population sizes, with direct implications for decisions based on economic threshold and/or prediction of pest outbreaks (Ferreira & Godoy, 2014; Nicácio et al., 2019).
Our results can also be useful as subsidies to optimize monitoring plans and risk mapping in orchards. In this sense, studies should be carried out to discuss the possibility of sampling plans suitable to local realities and critical areas to minimize the costs of installing traps and/or enhancing pest control in locations with a higher risk of infestation.
The aggregation pattern of fruit fly species was significantly influenced by area and time. The buffer zone and the impact on the abundance period provided a maximum aggregation of populations of C. capitata, A. fraterculus, and A. obliqua, indicating a significant association between aggregation degrees and a decrease in the abundance of fruit flies.
We thank the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) for the fellowship for the first author. RAZ and WACG are fellow of the Conselho Nacional de Desenvolvimento Científico e Tecnológico. We also thank Editage (www.editage.com) for English language editing.
*Author for correspondence. E-mail: mayara_araujo@usp.br







