Spatial Distribution of Bactrocera dorsalis and Thaumatotibia leucotreta in Smallholder Avocado Orchards along Altitudinal Gradient of Taita Hills and Mount Kilimanjaro

Avocado (Persea americana) fruits are an important source of income and a nutritious food for small-scale growers and other stakeholders involved in farming along the Afrotropical highlands of Taita Hills and Mount Kilimanjaro in Kenya and Tanzania, respectively. Avocado fruits are infested by several insect pests, namely the Asian invasive fruit fly, Bactrocera dorsalis (Hendel) (Diptera: Tephritidae), and the false codling moth, Thaumatotibia leucotreta Meyrick (Lepidoptera: Tortricidae). However, there is inadequate information on the distribution patterns of these pests in small-scale avocado cropping systems in the East African highlands. This study was initiated to generate a spatial distribution map of B. dorsalis and T. leucotreta in avocado orchards at Taita Hills and Mount Kilimanjaro in Kenya and Tanzania, respectively. The two pests were monitored by using their respective parapheromone lures for two years between August 2012 and July 2014. Fruit damage was assessed by computing the proportion of infested fruits for B. dorsalis, whereas the damage score was used for T. leucotreta. Our results indicated that the mean number of B. dorsalis per trap per day differed significantly across elevation, being highest in lowland zone for both Taita Hills (15.90) and Mount Kilimanjaro (24.45). Similarly, the percentage infestation of ground collected fruits by B. dorsalis varied with altitude, being lowest at highlands above 1500 m.a.s.l. (0.66% and 0.83% for Taita Hills and Mount Kilimanjaro, respectively). Conversely, the mean number of T. leucotreta did not vary with altitude in either study area. However, the damage score for T. leucotreta infestation was significantly lower in the highlands of both transects (7.0% and11.1% for Taita Hills and Mount Kilimanjaro, respectively). These findings describe spatial trends that are important in formulating strategies aimed at suppressing the populations of B. dorsalis and T. leucotreta in East African avocado cropping systems.

Bactrocera dorsalis and Thaumatotibia leucotreta cause tremendous yield losses of various fruit crops including avocadoes [10][11][12][16][17][18][19]. Both B. dorsalis and T. leucotreta are considered A1 quarantine pests because of the enormous losses they inflict due to direct infestation of fruits and quarantine restrictions imposed by importing countries [19][20][21]. For example, the invasion of B. dorsalis in Kenya and Tanzania led to a loss of the avocado export market to South Africa in 2007, causing revenue losses of US$1.9 million [20] with cumulative losses amounting to $15.2 million by the end of 2014 [12]. Considering the socioeconomic and nutritional significance of the avocado crop and the magnitude of losses caused by B. dorsalis and T. leucotreta, there is an urgent need for suppression of these pests to enhance quality fruit yield. Such efforts will not only improve the livelihood of the smallholder growers, who account for more than 70% of avocado producers in East Africa [3], but also help the governments of Kenya and Tanzania to regain full access to the lucrative export markets. A prerequisite for such intervention is a sound knowledge of the abundance and distribution patterns of these pests. However, only scanty information is available for commercial or large-scale avocado farming systems in East Africa [11,21]. This study was initiated to establish the spatial distributions of B. dorsalis and T. leucotreta in small-scale avocado cropping systems at Taita Hills in southeastern Kenya and Mount Kilimanjaro in northeastern Tanzania. This study had the following objectives: (i) Assess percentage damage levels in mature avocado fruits by B. dorsalis and T. leucotreta along elevational gradient of Taita Hills and Mount Kilimanjaro; (ii) Describe changes in the population density of B. dorsalis and T. leucotreta along an altitudinal gradient at the two transects and (iii) Determine how environmental variables influence changes in population density of the two pest species in each study area.

Study Areas and Agro-Ecological Conditions
This study was carried out along an altitudinal gradient between 900 and 1800 m.a.s.l. in small-scale avocado orchards at Taita Hills in southeastern Kenya and Mount Kilimanjaro in northeastern Tanzania (Figure 1). The two study transects were selected based on the criteria that they represent a steep elevation gradient with no large commercial farming that would influence the land use patterns and interactions between the lowland and highland environments [22,23]. Furthermore, farming along the two study transects is rain-fed and the smallholder avocado growers do not use any form of pesticide to control insect pests. The Taita (Table 1). Each agro-ecological zone in a study transect was divided into five study blocks. A study transect comprised a total of 15 blocks, with each block consisting of at least 100 avocado trees. Temperature (°C) and relative humidity (%) were recorded daily at each block using iButton™ data loggers (Maxim Integrated Products, San Jose, CA, USA). The data loggers were hung in the lower canopy of avocado trees at a height of 2 m. Rainfall records was obtained from the weather stations set along the study transects in Kenya and Tanzania. The geographical coordinates and elevation of every study site in a block were recorded using a hand-held Garmin GPS eTrex 30 receiver (Garmin International, Olathe, KS, USA).

Insect Monitoring Using Lures
Bactrocera dorsalis and Thaumatotibia leucotreta were monitored along an elevational gradient between August 2012 and July 2014 in each of 15 blocks at each transect, namely Taita Hills and Mount Kilimanjaro. In every block three avocado trees were randomly selected for each target insect pest. For B. dorsalis, one Lynnfield trap baited with methyl eugenol (ME) charged on a cotton wick at a ratio of 4:1 of the ME: othothion as a killing agent was hung on each of the selected trees approximately 2 m above the ground. On the other three trees, yellow delta traps baited with dodecenyl acetate lure were used for monitoring T. leucotreta. The traps were hung at the same height as described for B. dorsalis. During the two years, traps for B. dorsalis and T. leucotreta were randomly rotated every week among 100 mature avocado trees in each block, but they were never concurrently hung on the same tree to avoid cross-contamination.

Fruits Damage Assessment
Five avocado trees were randomly selected from each of the 15 blocks for damage assessment of their fruits. Four fruits, two from the ground and two from the tree, were collected during the peak avocado harvesting season (May to June) and processed according to a protocol described by Ekesi  (Table 1). Each agro-ecological zone in a study transect was divided into five study blocks. A study transect comprised a total of 15 blocks, with each block consisting of at least 100 avocado trees. Temperature ( • C) and relative humidity (%) were recorded daily at each block using iButton™ data loggers (Maxim Integrated Products, San Jose, CA, USA). The data loggers were hung in the lower canopy of avocado trees at a height of 2 m. Rainfall records was obtained from the weather stations set along the study transects in Kenya and Tanzania. The geographical coordinates and elevation of every study site in a block were recorded using a hand-held Garmin GPS eTrex 30 receiver (Garmin International, Olathe, KS, USA).

Insect Monitoring Using Lures
Bactrocera dorsalis and Thaumatotibia leucotreta were monitored along an elevational gradient between August 2012 and July 2014 in each of 15 blocks at each transect, namely Taita Hills and Mount Kilimanjaro. In every block three avocado trees were randomly selected for each target insect pest. For B. dorsalis, one Lynnfield trap baited with methyl eugenol (ME) charged on a cotton wick at a ratio of 4:1 of the ME: othothion as a killing agent was hung on each of the selected trees approximately 2 m above the ground. On the other three trees, yellow delta traps baited with dodecenyl acetate lure were used for monitoring T. leucotreta. The traps were hung at the same height as described for B. dorsalis. During the two years, traps for B. dorsalis and T. leucotreta were randomly rotated every week among 100 mature avocado trees in each block, but they were never concurrently hung on the same tree to avoid cross-contamination.

Fruits Damage Assessment
Five avocado trees were randomly selected from each of the 15 blocks for damage assessment of their fruits. Four fruits, two from the ground and two from the tree, were collected during the peak avocado harvesting season (May to June) and processed according to a protocol described by Ekesi et al. [25]. The collected fruits were transported to a temporary laboratory established in lowland area at each transect. The fruits were kept individually in plastic containers with a layer of heat-sterilized sand. Thereafter, the pupae yield from fruit of each block were collected in a Petri dish and placed in a small cage (30 × 30 × 30 cm) until the emergence of flies. The percentage of fruits infested by B. dorsalis was computed for each block and, thereafter, the emerged fruit flies were killed to reduce possible re-infestation. The percentage of infested fruits by T. leucotreta was based on the damage symptoms rather than the actual larval count, since the larvae leave the avocado fruits before maturation [26]. Therefore, infestation by T. leucotreta was recorded on mature fruits collected from the tree during the harvesting season. Twenty physiologically mature fruits were collected from each block along both transects. In the laboratory, fruits were examined for the typical triangle black scars on the exocarp of avocado fruit caused by the larvae of false codling moth, as described by Du Toit et al. [26].

Statistical Analysis
Data on insect trapping and fruit damage for both of B. dorsalis and T. leucotreta were analysed using Wilcoxon and Kruskal-Wallis Chi-square tests. A Wilcoxon signed rank test was employed to analyse differences between pairs of datasets, whereas a Kruskal-Wallis Chi-square test was used to evaluate differences among three groups of datasets, as described by Crawley and Versani [27][28][29]. Linear mixed effect (LME) models were used to determine the environmental variables that best explained the changes in population density of B. dorsalis and T. leucotreta along an altitudinal gradient in the two study areas [30]. All possible models were constructed based on sets of sampled explanatory (environmental) variables, and model evaluation was done using Akaike Information Criterion (AIC). The environmental variables used in analysis were mean temperature, average relative humidity, mean annual rainfall, elevation and agro-ecological zones of the two study areas. AIC is based on information theory and evaluates models according to model fit and complexity [28][29][30][31][32][33][34][35]. The best model for each transect was selected, and its statistical significance was determined based on classical hypothesis testing [34].

Discussion
Trapping of fruit flies using methyl eugenol indicated that B. dorsalis is more abundant in lower elevations of both transects; this is understandable considering that B. dorsalis is a lowland resident pest [25,[36][37][38]. For example, in Hawaii, along the slope of Maui volcano, Wong et al. [36] found that trap capture of this pest was negatively correlated with elevation. In more recent studies, Geurts et al. [39,40] reported that B. dorsalis is the dominant species at elevation below 1000 m.a.s.l. in the Morogoro region of Tanzania. Moreover, B. dorsalis was more abundant in all agro-ecological zones of Mount Kilimanjaro than their corresponding ones in the Taita Hills. A possible reason for this discrepancy between the two transects in term of B. dorsalis population could be the fact that the former transect is relatively warmer by at least 0.5 • C; though this may seem negligible, B. dorsalis being poikilothermic, this difference in temperature will have a significant influence on the pest population growth. Another explanation for the higher B. dorsalis population in Mount Kilimanjaro could be the availability of more mango trees, the primary host plant of B. dorsalis [10,41,42], in this transect, especially at the lower elevation regions [39,40]. The intertwined nature of avocado with indigenous cloud forests along the Taita Hills transect might have contributed to the reduced population of B. dorsalis in this transect, as the natural, undisturbed vegetation provide refuge for natural enemies of this pest, especially predators such as the African weaver ant (Oecophylla longinoda). Way and Khoo, Van Mele and Cuc, Van Mele and Chien and Duyck et al. [43][44][45][46][47][48] reported that indigenous habitats enhance numbers of natural enemies that enable natural control of pest populations. However, it is worth mentioning that parasitoids are very unlikely to have been a contributing factor in this scenario [49][50][51][52], as this alien pest did not have resident parasitoid species.
The variation of the B. dorsalis population, as shown in the traps' catches, was reflected in fruit infestation levels: highest at lower elevations and quite low at higher elevations of both transects. Our findings concurred with those of Geurts et al. [39,40], who reported that the percentage of sub-tropical fruits infested with fruit flies was significantly lower at elevated zones of the Eastern Arc Mountains in central Tanzania. The avocado orchards at higher elevation (>1600 m.a.s.l.) of Mount Kilimanjaro and Taita Hills can be considered areas of low pest prevalence from which avocado can be exported to lucrative international markets. Additional research is, therefore, required to advance our knowledge of how global warming will influence the envisaged uphill expansion of B. dorsalis in East African avocado cropping systems.
The higher B. dorsalis trap catch at Mount Kilimanjaro did not necessary translate into more fruit infestation compared to that at a similar elevation in the Taita Hills, except for the ground-collected fruits at middle elevations. This finding suggested that B. dorsalis could be utilizing other neighbouring host fruits, such as mango and guava, which are more abundant at Mount Kilimanjaro. It was probably from these host crops that most of the flies were attracted to methyl-eugenol-baited traps, resulting in higher B. dorsalis catches in lowlands. However, the distribution of B. dorsalis along the altitudinal gradient was influenced most by temperature, although other weather variables, such as humidity and rainfall, also played a significant role. In both transects, infestation was higher on fallen avocado fruits than tree-collected ones, which concurs with the results of Ware et al. [21]. This implies that sanitation of avocado orchards can contribute significantly to suppression of B. dorsalis, an approach that is not practiced by small-scale growers, especially in the Taita Hills and Mount Kilimanjaro transects; instead, they consider fallen fruits as manure. Unlike smallholders, commercial avocado farmers embrace efficient pest control methods, including orchard sanitation. Orchard sanitation has been promoted as part of integrated pest management (IPM) for the suppression of B. dorsalis in East Africa by Ekesi et al. [12]. The same approach has been recommended in South Africa for the management of other species related to fruit flies, such as Ceratitis cosyra and Ceratitis rosa [19].
Unlike B. dorsalis, the population density of T. leucotreta was more or less similar across all agro-ecological zones for each transect, implying that the pest has a wider thermal tolerance. For example, Newton, Stotter and Terblanche, Stotter and Boardman et al. [53][54][55][56] reported that T. leucotreta thrives well in the warmer lowlands and colder uplands of South Africa, where it attacks both cultivated and wild plants. Although the abundance of T. leucotreta did not vary with altitude, Mount Kilimanjaro agro-ecological zones had higher trap catches and fruit infestation levels than their corresponding agro-ecological zones in the Taita Hills transect. This could be explained by the fact that in the former transect there is a high diversity and more availability of alternative hosts of T. leucotreta, such as maize, Zea mays (Poaceae), sorghum, Sorghum bicolor (Poaceae), citrus, Citrus sinensis (Rutaceae) and guava, Psidium guajava (Myrtaceae). Moreover, the nature of the avocado plantation in Taita Hills transect, being intertwined with the indigenous forest, may have provided refuge for natural enemies and competitors of T. leucotreta, resulting in a reduced population in this transect. For example, the egg parasitoid (Trichogrammatoidea cryptophlebiae) of T. leucotreta is reported to cause high levels of parasitism in citrus orchards in South Africa [19,57]. However, Trichogrammatoidea cryptophlebiae was not recovered from the samples of avocado fruits collected at the two transects during the current study; this needs to be investigated further in East Africa.

Conclusions
Our findings revealed that B. dorsalis preferred to infest ripened avocado fruits, which were ground-collected in warmer lowlands, suggesting that physiologically healthy unharvested fruits are not the preferred hosts in the Taita Hills or Mount Kilimanjaro. However, damage by T. leucotreta was observed on avocado fruits across all altitudinal zones. Hence, integrated control measures for B. dorsalis in avocado orchards should be concentrated in lowland and adjacent mid-altitudinal zones, whereas the management of T. leucotreta must be tackled in all altitudinal zones. The present findings can be used by policy makers in designing efficient integrated pest management strategies aimed at controlling B. dorsalis and T. leucotreta in order to enhance the livelihood of smallholder avocado farmers in East Africa. Finally, there is a need to establish local farmer-led stakeholder awareness committees to disseminate sustainable management strategies for controlling avocado insect pests in Kenya and Tanzania.