Introduction

The increasing number of invasive species is one of the greatest threats to global diversity (Dukes and Mooney 1999) as well as to ecology, evolution (Sakai et al. 2001), and local economies (Sax et al. 2007). Ecologically diverse tropical regions in developing countries are especially vulnerable because of their often-precarious economic systems and security programs, and the limited financial means at their disposal to monitor and control unwanted invasions. A successful invasion depends on transport, colonization, establishment, and spread of the pest (Sakai et al. 2001). Successful completion of each of these stages depends on filters at work on different scales, including climate change (Theoharides and Dukes 2007). Dukes and Mooney (1999) emphasize the significant effect the latter can have on current and future invasions and the need for the incorporation of this element in invasion models. Altitudinal gradients can serve as spatial analogues for climate change (Bale et al. 2002). Like climate change, ecological gradients such as host plants and predators (Vayssières et al. 2008), as well as physical gradients like temperature, rainfall, and humidity (Hodkinson 2005) encountered along an altitudinal transect can have an impact on the density, diversity, and life history of insects and demands for phenotypic flexibility and genotypic adaptability of many species (Bale et al. 2002).

Fruit flies (Diptera: Tephritidae) are a good model taxon for ecological research on invasion success. Many fruit fly species have high reproductive rates and good (passive) dispersive powers amplified by not only the current increase in global trade of fruit products but also by an increasing number of tourists who may carry infested fruits (Liebhold et al. 2006). These characteristics make fruit flies excellent invaders (Ward and Masters 2007) and allow them to become pests in certain places (Ekesi et al. 2007).

Bactrocera cucurbitae (Coquillett), a fruit fly from Asia, became an invasive pest species in Africa after first establishing itself in Eastern Africa and appearing in Western Africa at the beginning of this century (Vayssières et al. 2008; Virgilio et al. 2010). It is considered a major pest of several Cucurbitaceae, while occasionally infesting non-cucurbit fruits (Mwatawala et al. 2010). Another major Asian invasive species, B. invadens Drew, Tsuruta, and White, was first detected in Kenya in February 2003 followed by Tanzania (Mwatawala et al. 2004), and has now been recorded in 27 African countries (Rwomushana et al. 2008b; De Meyer et al. 2010). Currently, its spread appears to be continuing not only in latitude but also in altitude. In Kenya and Tanzania, the highest fruit infestation rates are recorded at lower altitudes (Ekesi et al. 2006; Rwomushana et al. 2008a; Mwatawala et al. 2009a). Recently, however, mango infestation has also been recorded at medium to high altitudes in the Eastern Province of Kenya (Rwomushana et al. 2008a), which was not the case in previous studies (Ekesi et al. 2006). In Tanzania B. invadens occurs most prominently at low and medium altitudes (< 1600 m) (Mwatawala et al. 2006a), but individuals have been detected at higher altitudes (Mwatawala et al. 2006b).

Bacterocera invadens spread and successful colonization of higher altitudes could be limited by climatic conditions (Mwatawala et al. 2006a), host availability and suitability (Rwomushana et al. 2008a), and interspecific competition with cold-tolerant species like Ceratitis rosa Karsch (Mwatawala et al. 2006a). Temperature is usually considered to be the most important factor explaining population dynamics in insect species (Vayssières et al. 2008), as well as humidity to some extent (Duyck et al. 2004). Extreme variations such as drought and inundation are the greatest threats to the most vulnerable stage of fruit fly life history: the pupal stage (Duyck et al. 2004; Duyck et al. 2006b).

In addition to invasive Bactrocera species, several indigenous fruit fly pests occur in Eastern Central Tanzania, especially belonging to the Afrotropical genus Ceratitis (Mwatawala et al. 2006a). The genus Bactrocera is renowned for its strong competitive abilities and capacity to largely displace other (indigenous) fruit fly species (Duyck et al. 2006a, 2006b). Species composition and density along altitudinal gradients change according to the niche width of different species (Hodkinson 2005). In this paper we investigate species composition of fruit flies spatially and temporally in relation to environmental gradients along an altitudinal transect in Eastern Central Tanzania.

Figure 1.

Altitudinal distribution of sampling stations. High quality figures are available online.

Materials and Methods

Statistical analysis

The influence of different environmental factors on species composition and abundance was explored using ordination techniques, correlation analyses, and Kruskal-Wallis tests. Temperature, humidity, and host presence (of mango, peach, and soursop) were the environmental factors used in this study. Ordination was carried out using Canoco 4.5 (ter Braak and Šmilauer 2002). Relative abundance of every species along the transect was used. The species data were square-root transformed and rare species were down weighted. The significance of every variable in the Canonical Correspondence Analysis (CCA) or Redundancy analysis (RDA) was tested using a Monte Carlo permutation (499 permutations). In addition, a series of constrained, detrended CCA (DCCA) were run with the species data and each environmental variable on its own. This was to determine the strength of each variable by its ability to maximize the dispersion of the species scores, expressed as a ratio of the eigenvalue of the first axis to the eigenvalue of the second axis (Zhang et al. 2007; Woodward and Schulmeister 2008). A low value could point out a factor not considered. The unique explanatory power of each significant environmental variable was tested using a series of partial, constrained ordinations (RDA or CCA) with all other significant environmental parameters as covariables (Woodward and Schulmeister 2008). The ordination analyses were an exploratory method to filter out the most important environmental variables. To study their influence in more detail the data were analyzed using correlation techniques. The data did not have a normal distribution, variances were not homogenous, and there was a correlation of the mean and the variance; thus, non-parametric tests needed to be used. Spearman's rank correlation analyses and Kruskal-Wallis tests were performed in R 2.12.2 (R Development Core Team 2008). For Spearman's rank correlations absolute abundances were used, and for Kruskal-Wallis tests proportional abundance of a species at a certain time at a certain site were square-root arcsin transformed. Proportional abundance was used to avoid effects of attractiveness of lures.

Figure 2.

(a) Temperature and (b) RH at different elevations over time. High quality figures are available online.

Results

Temporal and spatial differences

In total, 16 different species were recorded (Appendix 2). Five species were predominant in the trap catches: B. cucurbitae, B. invadens, C. cosyra (Walker), C. rosa, and D. bivittatus (Bigot). The temporal and spatial occurrence for these is shown in Figure 3. The temporal abundance observed throughout the different sites for the dominant species showed two distinct patterns: B. invadens and C. rosa showed a peak period between December (B. invadens) or January and March (C. rosa) (Figure 3a, 3b). Ceratitis cosyra showed an inverse pattern with the highest abundance between August and November (Figure 3c). There was a significant difference in total fruit fly abundances between months (p < 0.01). Significant differences between months in fruit fly abundance were found in the months May to November. Studying every fruit fly species individually showed that for C. rosa and D. bivittatus there was a significant difference in abundances between months (Table 2). Differences for dominant species are summarized in Table 3. For B. invadens there were no significant differences between sites within months. For B. cucurbitae significant differences between sites occurred in March, April, May, June, July, August, and September. For C. rosa significant differences between sites occurred in January, February, March, April, and June. For C. cosyra significant differences between sites within months occurred in January, February, April, June, and July. For D. bivittatus significant differences between sites occurred in June.

Figure 3.

Temporal and altitudinal distribution of five dominant fruit fly species: (a) Bactrocera invadens, (b) Ceratitis rosa, (c) C. cosyra, (d) B. cucurbitae, and (e) Dacus bivittatus. High quality figures are available online.

Looking at spatial patterns for dominant fruit fly species showed that the abundance of some fruit fly species (B. invadens, B. cucurbitae, and C. cosyra) declined with rising altitude, while that of other species (C. rosa and D. bivittatus) increased (Figure 3). There was a significant difference in total fruit fly abundances between sites (p < 0.01). Significant differences between sites were found in Nyandira and Visada. Studying every fruit fly species individually showed that there was a significant difference between sites for each fruit fly species (Table 4). Regarding the dominant species, B. invadens (Table 5), there were significant differences between months in Nyandira. For B. cucurbitae and C. cosyra there were no significant differences between months within sites. For C. rosa there were significant differences between months in Langali, Visada, and Nyandira. For D. bivittatus there were significant differences between months in Kibundi and Visada.

Discussion

Bactrocera invadens, C. rosa, and C. cosyra are reported from a wide variety of hosts throughout Africa (De Meyer et al. 2002; De Meyer et al. 2008), as well as in this particular study area (Mwatawala et al. 2009b). Bactrocera cucurbitae is an oligophagous species attacking a wide variety of wild and commercially grown Cucurbitaceae, while D. bivittatus seems to have a slightly more restricted cucurbit host range in the study area (Mwatawala et al. 2010). Caution should be taken when comparing the trapping results obtained through different attractants, since the relative effectiveness of the different parapheromones used differs widely (see for example Shelly et al. 2010; Vargas et al. 2010). Therefore, proportional abundance was used to correct for any lure attractiveness effect. However, the results obtained allow a comparison for each individual species in time and space since the effectiveness of a particular lure is not affected by these factors.

Figure 4.

Biplot of CCA analysis. October (O), November (N), December (D), January (Ja), February (F), March (M), April (A), May (Ma), June (J), July (Ju), August (Au), September (S), Annona muricata (Am), Psidium guajava (Pg), Prunus persica (Pp), Mangifera indica (Mi), temperature (T), relative humidity (R). High quality figures are available online.

The temporal differences in overall proportional abundances were only significant for C. rosa and D. bivittatus (Table 2). During the wet months (February and March) C. rosa occurred in high abundance in higher altitude areas (Nyandira and Visada) (Figure 3b). This illustrates this fruit fly's preference for periods with a high relative humidity, which can also be deducted from its positive correlation with RH (Table 8). It also only occurs in Kibundi during these months with a high RH (Figure 3b). The distinct peak in February and March in Nyandira, Visada, and Langali corresponds with ideal conditions for this fruit fly: low temperature, high RH, and fruiting of one of its preferred hosts, peach (Figure 3b). This pattern can also be seen on the biplot of the ordination analysis (Figure 4). Dacus bivittatus occurred in periods with a high RH and low temperature. Dacus bivittatus abundance had a positive correlation with RH in high altitude areas (Table 8), illustrated by peaks in abundance in March and April in Visada (Figure 3e), and a negative correlation with temperature in low lying areas (Table 7), illustrated by its peak in abundance in May and June in Kibundi (Figure 3e).

For B. invadens, significant differences between months were found in Nyandira, where it was present in January, March, and April, and absent during the other months (Figure 3a). The abundance of B. invadens was positively correlated with temperature (Table 7) and its presence in Nyandira during these warmer months illustrates this. This can also be explained by a host effect. Earlier host rearing experiments (Mwatawala et al. 2006b, 2009b) have shown that B. invadens has potential hosts available throughout the year, but that one of its preferential hosts, mango, is available from November to February, and that fly populations show a distinct peak coinciding with this fruit, as also seen during this study (Figure 3a). From the ordination analysis it was shown that mango explained a large portion of fruit fly relative abundance variation (Table 6), and that mango was a determining factor for B. invadens (Figure 4). However, mango trees are only present along the altitudinal transect up to 1302 m, and in higher altitude areas alternative hosts such as peach guarantee fruit fly reproduction. Peach becomes available as a food source during February and March, which explains the rise in B. invadens abundance in March and April at higher altitudes. A second peak in March–May at the lower altitudes, coinciding with the guava season, is only marginally present. A large variety of alternative hosts fruiting at different periods in the year clarify the presence of B. invadens throughout the year and the absence of significant differences in abundance between months.

The seasonal abundance observed for B. cucurbitae also largely corresponds with earlier findings in Morogoro (Mwatawala et al. 2010), where a peak period was observed between May and October (Figure 3d), coinciding with the availability of cucurbit hosts (although B. cucurbitae is known to attack non-cucurbits, these infestations are largely occasional and of low significance). These peaks in abundance were also present during environmental conditions that seem to be preferred by this fruit fly species—periods with intermediate to high temperature and intermediate RH.

Ceratitis cosyra showed a positive correlation with temperature (Table 7) and a negative correlation with RH (Table 8), illustrated by its high abundance during dry and relatively warm months in Mlali (November, August, and September 2009) (Figure 3c). These peaks in abundance can also be due to a host effect.

The biplot of the ordination analysis shows a close association of C. cosyra with mango, but an even closer association with soursop (Figure 4a). In addition to mango, the preferential host range for C. cosyra is Annonaceae. A study from 2004–2006 detected a host shift by C. cosyra at the end of the mango season (April/May) towards soursop (A. muricata) (Mwatawala et al. 2009b). In our study, C. cosyra showed a peak in November, corresponding with the early mango season, and a peak in August and September, when soursop is available but mango is not (Figure 3c).

The temporal occurrence for all these species seems to be defined by preferred host availability and environmental conditions. This could also be concluded from the ordination analysis. Temperature and peach explained most of the variation in species composition (Table 6). RH and soursop explained another significant portion of the variation (Table 6). When considering all the other variables as covariables, only temperature still had significant explanatory power (Table 6). The fact that the fruits no longer explained significant portions of the variation is because environmental conditions determine phenology and nutrient quality of hosts (Hodkinson 2005). By separating out the effects it was revealed that underlying the availability of hosts, temperature is the most important factor in explaining fruit fly abundance, which has been concluded from several previous studies (Vargas et al. 1997; Bale et al. 2002; Vayssières 2008). Although RH was no longer a significant factor when considered on its own, it still explained a large part of the variation (Table 6). Rainfall is considered to be the most important factor for growth and quality of host plants (Mwatawala et al. 2006b). At the end of 2008, the short rainy season was delayed, which caused mangoes to ripen later in the season and to be smaller in size.

The spatial patterns can be explained by temperature, RH, and host differences between the sites. Bactrocera invadens and B. cucurbitae showed a positive correlation with temperature (Table 7), and therefore they occur in highest numbers in low-lying areas. Bactrocera invadens spatial occurrence observed here confirms earlier findings (Mwatawala et al. 2006a), showing that its presence declines at higher elevations (Figure 3a). Although suitable hosts such as mango still occur at the higher elevations, the abundance is very low. De Meyer et al. (2010) showed in their climatic modeling that the most suitable areas for establishment of B. invadens corresponded with the Equatorial climate category of the Köppen-Geiger climate classification (Kottek et al. 2006), which is categorized by minimum average temperature at or above 18 ^°C. These conditions are mainly found at the two lowest sites (Mlali and Kibundi). Bactrocera invadens sporadic occurrence at high altitudes (above 1600 m), however, indicates that this fruit fly species has a large invasive potential and may be spreading and/or adapting to higher altitudes with lower temperatures.

Ceratitis cosyra had a positive correlation with temperature (Table 7) and a negative correlation with RH (Table 8), and therefore seems to be even more restricted to lower areas. No climatic model has been developed for C. cosyra, but at our study site, it was only relatively abundant at the lowest sampling site. The abundance of C. cosyra in Benin was observed to be high in dry periods and quickly dropped when the rains started (Vayssieres et al. 2005, 2009), corresponding with our findings. In Kenya, however, C. cosyra was recorded in higher altitudinal areas (Ekesi et al. 2006), which was not found in our study.

Ceratitis rosa showed a positive correlation with RH (Table 8), and for the ordination analysis was associated more with low temperature areas (Fig 4a); therefore, it occurred more often in high altitude areas. Previous studies confirm that C. rosa is a species that can withstand colder temperatures (Duyck et al. 2004, 2006b; De Meyer et al. 2010), although it is has also been reported in warm and wet conditions (Grout and Stoltz 2007). Ceratitis rosa also showed a close association with peach (Figure 4), which is only found at higher altitudes (Visada and Nyandira), explaining its presence there in high abundance.

Dacus bivittatus abundance had a positive correlation with RH in high altitude areas, and a negative correlation with temperature in low-lying areas. Therefore, it can occur in high abundances in areas with intermediate temperatures and RH such as Langali and Visada.

Survival of fruit flies with a narrow temperature and host range is therefore not possible along the entire altitudinal range. For these reasons, spatial shifts from low altitude to the highest altitude seem very unlikely. However, it is possible that the spatial shift of species is in an intermediate phase. Visada should be the ideal location for this phenomenon, as it has hosts from high (e.g., peach) and low (e.g., mango) altitude areas and has an intermediate temperature and humidity range. Therefore, survival of all fruit fly species is possible in this location, and thus host shifts and adaptation to lower and/or higher temperatures could occur here.

In general, the presence of suitable hosts and environmental conditions seem to be decisive for abundance patterns and species composition along the transect. The temporal and spatial patterns observed during this one-year study need to be substantiated by repeated year cycles, but are already supported to some extent through other studies conducted in the study area (Mwatawala et al. 2006a, 2006b, 2009b, 2010). Although B. invadens does not occur permanently at high altitude, future colonization is possible. Every year in January and February, this fruit fly appears to be dispersing towards higher areas (Mwatawala et al. 2006a), especially at times when there is a population surplus. Both the rapid population increase (indicating r-selection) (Duyck et al. 2007) and the broad host range (Mwatawala et al. 2006b) are ideal characteristics to allow for further spread and colonization of new areas. The effect that climate change may have is at this stage hard to predict. However, recent studies indicate warming at higher elevations in the tropics (Eggermont et al. 2010; Chen et al. 2009), which could lead to a more suitable environmental space at high altitudes for this invasive species. In particular, the impact of shifting minimal temperature thresholds on fruit fly distribution because of climate change has been demonstrated in Rhagoletis (Aluja et al. 2011), and a similar pattern might have an impact on the altitudinal move of B. invadens.

Table 1.

Geographical position of trapping stations.

Table 2.

Temporal differences between months for each fruit fly species using Kruskal-Wallis tests.

Table 3.

p-values for Kruskal-Wallis tests for each fruit fly species for differences between sites within months (NA: no individuals caught).

Table 4.

Spatial differences between trapping stations for each fruit fly species using Kruskal-Wallis tests.

Table 5.

p-values for Kruskal-Wallis tests for each fruit fly species for differences between months within sites (NA: no individuals caught).

Table 6.

CCA (only environmental values reported).

Table 7.

Spearman's rank correlation of overall abundance dominant fruit fly species and temperature.

Table 8.

Spearman's rank correlation of overall abundance dominant fruit fly species and RH.