Abundance of Coffee berry borer in feral, abandoned and managed coffee on Hawaii Island

Coffee berry borer (Hypothenemus hampei Ferrari), the most damaging insect pest of coffee worldwide, was first detected on Hawaii Island in 2010. Poorly managed, abandoned and feral coffee sites on the island have since been thought to harbour coffee berry borer (CBB) populations, which then negatively impact neighbouring coffee farms. In the present study, we sought to quantify CBB abundance in these sites, which vary in management intensity and vegetation structure and diversity. We collected data on trap catch as a measure of CBB flight activity, fruit production and fruit infestation by CBB in eight well‐managed farms and sites that were either poorly managed, abandoned or feral (wild) coffee. Sites were sampled bi‐weekly over a period of 2 years from 2016 to 2017. We found that CBB flight activity was significantly higher in poorly managed sites relative to abandoned and feral sites, but was not significantly different from well‐managed sites. Coffee production in well‐managed farms was significantly higher than in abandoned and feral sites, but was not significantly different from poorly managed farms. CBB infestation in poorly managed sites was significantly higher than that observed in well‐managed, abandoned and feral sites. We estimated an average load of 11–25 CBB per branch at poorly managed sites, compared to 3–9 per branch at well‐managed sites, 1–16 per branch at abandoned sites and 1–3 per branch at feral sites. Our findings suggest that poorly managed sites should be prioritized for implementation of CBB control measures as part of a landscape‐level integrated pest management (IPM) programme.

in production and yield, pest management in coffee will be essential, and this increasingly must be attained with reduced reliance on chemical control (Watts & Williamson, 2015).
Although coffee has many insect pests (e.g., root-knot nematode Meloidogyne spp., black twig borer Xylosandrus compactus) and diseases (e.g., coffee leaf rust Hemileia vastatrix, coffee berry blotch Cercospora coffeicola), the coffee berry borer Hypothenemus hampei Ferrari (Coleoptera: Curculionidae) is widely considered the most damaging insect to coffee crops worldwide (Damon, 2000;Jaramillo et al., 2006;Vega, 2015). Adult females bore into the coffee fruit, where they build galleries in the seed (bean) and lay eggs. The offspring develop inside the coffee seed where they feed on the endosperm tissue. Male and female siblings mate, and the males remain in the fruit while the mated females leave and search out a new fruit in which to lay their eggs. Coffee berry borer (CBB) affects both the yield and quality of coffee, causing serious economic losses in all coffee-growing regions of the world (Le Pelley, 1968). Additionally, because the CBB completes its entire life cycle inside the coffee fruit, the females are only vulnerable to pesticide sprays when they are out in search of new fruits to infest, making this pest extremely difficult to control.
Hawaii has a relatively small coffee industry (24.6 million pounds of coffee cherry produced in the 2017-2018 season, USDA-NASS, 2018) but commands premium prices on the world specialty market due to its unique origin and high quality (Kinro, 2003). In 2010, the arrival of CBB dramatically changed Hawaii's coffee industry (Burbano, Wright, Bright, & Vega, 2011). The pest rapidly spread across Hawaii Island (~2,400 ha in coffee production; Teuber, 2010) and was later detected on the neighbouring islands of Oahu (2014) and Maui (2016). Both the yield and quality of coffee produced in Hawaii has decreased due to CBB, with significant declines in "Extra Fancy" and "Fancy" grades of coffee (Aristizábal, Bustillo, & Arthurs, 2016;Aristizábal et al., 2017). The CBB invasion has also resulted in increased production costs due to the need to conduct labour intensive control measures such as end-of-season strip picking of all fruits remaining on the trees and regular applications of the entomopathogenic fungus Beauveria bassiana (Aristizábal et al. 2016(Aristizábal et al. , 2017. For the crop years 2011/12 and 2012/13, the estimated economy-wide impact of CBB in Hawaii was a $12.7 M loss in crop value, a $25.7 M loss in sales, a $7.6 M loss in household earnings and a loss of more than 380 jobs (Leung, Kawabata, & Nakamoto, 2014).
Since the arrival of CBB in Hawaii, it has been widely suggested that feral (wild) coffee, abandoned farms and poorly managed farms serve as reservoirs for this pest, and can contribute to infestation in nearby farms (Messing, 2012;Aristizábal et al. 2016;Woodill, Nakamoto, Kawabata, & Leung, 2017). However, no published research exists to our knowledge that quantifies the extent to which these different sites harbour CBB across the coffee-growing landscape in Hawaii. The objective of this study was to examine CBB flight activity, fruit production and fruit infestation by CBB in well-managed coffee farms and compare it with that found in poorly managed, abandoned and feral sites over two growing seasons (2016)(2017)  were added to the study in 2017 ( Figure 1, Table 1). We considered "well-managed" farms as those that conducted active management for coffee production, including regular pruning, fertilizing, weed management, pesticide sprays for CBB (B. bassiana and/or pyrethrinbased pesticides), cherry harvesting and end-of-season strip picking ( Figure 2A). In general, well-managed farms were subject to management interventions multiple times per month. To limit confounding factors due to environmental conditions, we selected a site adjacent to or at a similar elevation to each well-managed coffee farm that was classified as one of the following: a poorly managed farm, an abandoned farm or feral coffee. Farms classified as "poorly managed" were not actively managed aside from mowing the grass and/or cutting vegetation around coffee trees ( Figure 2B); these particular sites were not used for coffee production during our study (i.e. were not harvested). Farms considered "abandoned" received no management for at least 2 years prior to the initiation of the study ( Figure 2C).
Sites designated as "feral" were forested areas that had coffee plants growing wild in the understory ( Figure 2D). In 2017, two of the well-  Table 1). Each study site was approximately 1-2 ha in size.
All well-managed and poorly managed farms grew coffee in full sun ( Figure 2A,B); abandoned farms had few to no trees and were overgrown with tall grass, sometimes head-high ( Figure 2C); feral coffee grew in densely forested areas with generally low levels of sunlight ( Figure 2D). All coffee plants were C. arabica var. typica, with the single exception being farm KC which had primarily C. arabica var. catuai planted. Figure 3 shows a generalized overview of coffee plant phenology on Hawaii Island, along with the recommended timing of CBB management practices (see Kawabata, Nakamoto, & Curtiss, 2015).

| Fruit production and infestation
From March-December, coffee plants were assessed bi-weekly for fruit production and CBB infestation. Tree sampling followed the methods used in Johnson et al. (2018). Briefly, sampling grids were first established to ensure an even distribution of sampling of trees throughout each site. The number of trees sampled depended upon the size of the study area (8-15 trees sampled at small sites, 18-25 trees sampled at large sites). Sampling grids were not established at feral sites due to fewer trees available; at these sites, all coffee trees that could be found were sampled (8-12 trees). To assess fruit production, a single branch was randomly selected at chest height from each tree and the number of infestable fruits (green to ripe fruits that were pea-size and larger) were counted ( Figure 4B). Given that dried coffee fruits (raisins) are known reservoirs of CBB (Johnson, Fortna, Hollingsworth, & Manoukis, 2019), we also counted the number of raisins on each branch. CBB infestation was assessed by examining green fruits for an entrance hole in the central disc ( Figure 4C). Given that ripe and over-ripe fruits were continuously harvested throughout F I G U R E 2 Examples of site types studied: Well-managed farm (a), poorly managed farm (b), abandoned farm (c) and feral coffee (d). Note differences in shade, vegetation structure and diversity, and fruit production [Colour figure can be viewed at wileyonlinelibrary.com] the season on well-managed farms, only green fruits were included in our infestation assessment. When present, 1-3 infested green fruits were collected from each branch and stored at 14°C. Infested fruits were dissected within 24 hr under a dissecting microscope at 30-50× (Leica, Microsystems GmbH). We recorded the number of fruits in which the founding female was in the AB position (i.e., has commenced boring into the fruit but has not entered the endosperm) but was absent due to mortality or disturbance. This was done in order to accurately estimate the number of infested green fruits per branch that had the founding female present (Section 2.4).

| Estimation of CBB load per branch
Given that we did not track the total CBB population per infested green fruit, we calculated a conservative estimate for CBB load per

| Data analysis
We used generalized linear mixed models (GLMM) to determine whether site type (feral, abandoned, poorly managed and well-managed) was a good predictor of: (a) CBB flight activity, (b) fruit produc- were conducted in the package multcomp v. 1.4-13 (Hothorn, Bretz, & Westfall, 2008) in R.

| CBB flight activity
Site type was a significant predictor of CBB flight activity as meas-  Figure 5A). Trap catch was also significantly higher in well-managed compared to feral sites, but there was no difference between abandoned and feral sites (Table 2; Figure 5A). While trap catch remained low year-round in abandoned sites, the main peak in flight activity for well-managed, poorly managed and feral sites was observed from March to May during both years ( Figure 5B).
A second smaller peak was observed in well-managed and poorly managed sites during the end of the season from late October to December for both years ( Figure 5B).

| Fruit production
We found a significant effect of site type on fruit production (X 2 = 11.19, df = 3, p = .011). Tukey contrasts showed that fruit production across all sites/sampling dates was significantly higher in well-managed farms (26.94 ± 2.92 fruits/branch) compared to abandoned (8.66 ± 2.03 fruits/branch) and feral sites (7.78 ± 3.38 fruits/ branch) (Table 2; Figure 6A). Fruit production was not significantly different between poorly managed (19.56 ± 10.50 fruits/branch) and well-managed farms, or between poorly managed, abandoned and feral sites (Table 2; Figure 6A). For all site types, fruit production peaked from June to October ( Figure 6B). Lastly, the mean number of raisins per branch was lowest at feral sites (0.13 raisins/branch) and well-managed farms (0.48 raisins/branch), while abandoned (1.22 raisins/branch) and poorly managed sites (1.13 raisins/branch) had similarly high numbers.
Well-managed and abandoned sites exhibited a similar pattern of infestation over time, peaking at the beginning and end of the season ( Figure 7B). In contrast, infestation in feral sites peaked from August to September and then decreased, while infestation in poorly managed farms increased throughout the season and peaked at an average of 95% infestation in December ( Figure 7B). Across all sampling dates, the average percentage of infested fruits with CBB missing TA B L E 2 Generalized linear mixed model estimates for fruit production, trap catch and fruit infestation as explained by site type (well-managed, poorly managed, abandoned and feral) in the AB position (i.e., boring was initiated but the founding female evacuated the fruit prior to entering the endosperm) was 13.02% in well-managed farms, 12.92% in abandoned sites, 6.59% in poorly managed sites and 6.13% in feral sites.

| D ISCUSS I ON
We compared CBB flight activity, fruit production and fruit infestation in well-managed coffee farms with poorly managed, abandoned and feral coffee sites on Hawaii Island to determine the abundance of CBB in these sites that vary considerably in terms of management intensity. Across all sites and sampling dates, fruit production was highest in well-managed and poorly managed farms, while CBB flight activity and fruit infestation were highest in poorly managed farms. Our results suggest that poorly managed farms are the largest source of CBB on Hawaii Island compared with the other three site types studied, and therefore, these sites should be the focus of control measures to avoid spillover of the pest to neighbouring farms.
The four site types studied here differed in vegetation structure and density, and this likely contributed to differences in CBB flight activity, fruit production and fruit infestation. Feral coffee sites in Hawaii are comprised of a mix of older trees that were planted and abandoned decades ago, and younger trees from volunteer seedlings. Goto and Fukunaga (1986)  The large range in CBB load (1-16/branch) estimated for abandoned sites reflects variation in fruit production and subsequent raisin load at these sites, which will depend on environmental conditions and how long the site has been abandoned. Extreme environmental conditions (high temperatures and low water availability) will limit fruit production in abandoned sites, and the CBB load will continue to diminish the longer a site is out of production and the reservoir of raisins has decreased. We noted that abandoned sites also suffered high coffee plant mortality, likely due to the long periods of drought that characterize the Kona coffee-growing region.
Fruit infestation by CBB in poorly managed farms averaged 63% across all farms/sampling dates. By the end of the harvest season in December, poorly managed farms reached an average of 95% infestation. Such a high level of infestation has generally not been reported in other coffee-growing regions, with infestation typically ranging from >1% up to 35% (Soto-Pinto et al. 2002;Benevides, Bustillo, Cárdenas, Montoya, 2003;Bosselman et al., 2009;Larsen & Philpott, 2010;Jaramillo et al., 2013). However, in Puerto Rico Mariño et al. (2017) reported that of 214 sites surveyed in 2014, the average infestation was 20% and ranged from 1%-95%. These authors suggested that the reasons for such high infestation included little to no management at the majority of sites, high cost and low availability of labour to conduct management and a lack of natural enemies.
The factors mentioned by Mariño et al. (2017) are directly translatable to the situation in Hawaii, where the cost of production and labour is high, labour to conduct management practices such as sanitation and frequent harvesting is difficult to secure, and natural enemies are largely absent (with the exception of flat bark beetles; see Follett et al., 2016). Many of these issues are due to the relatively recent introduction of CBB to the islands. In the present study, we found that for the eight well-managed farms that implemented multiple forms of CBB control, infestation levels averaged 13% (range = 2%-27%) throughout the season. This was significantly lower than that observed in poorly managed farms and highlights the importance of good farm management practices for the control of CBB.
Our findings also suggest that even if a well-managed farm is adjacent to a poorly managed farm with high levels of infestation, it is possible to maintain relatively low CBB populations throughout the season by implementing sanitation, pesticide applications early in the season, and frequent harvesting (see Figure 3; Kawabata et al. 2015). The sites that were neighbouring the poorly managed farms examined here kept infestation levels below 20% despite being adjacent to an area with high pest pressure. Although we did not examine CBB movement between sites, it is likely that females are able to migrate at least short distances between sites as conditions change in suitability. For example, we observed a massive increase in CBB flight activity and infestation at one feral site when the vegetation surrounding coffee trees were removed and fruit production subsequently increased. It is likely that CBB migrated from surrounding coffee farms and feral sites into this area once it became easier to access. This suggests that the use of physical barriers (i.e., exclusion netting) or border crops that are densely planted could inhibit the migration of CBB between sites and therefore minimize costs and labour needed to deal with the continuous influx of CBB from poorly managed sites into well-managed farms. Future research is needed to determine the height and distance that CBB are able to fly, and if their movement patterns change throughout the year depending on coffee plant phenology and differences in land use.  Lastly, if the ultimate goal for a given site is to convert the land to other uses, we recommend removing coffee plants as soon as possible to limit any negative impacts to neighbouring operations.
The present study is the first to our knowledge that has quantified and compared CBB abundance in well-managed, poorly managed, abandoned and feral coffee sites, and thus assessed the potential of these sites to act as population reservoirs for CBB. The information presented here suggests that fruit production is highest in well-managed sites, but that poorly managed sites have the potential for similarly high levels of fruit production. In addition, we found that CBB flight activity and fruit infestation were higher in poorly managed sites relative to well-managed, abandoned and feral sites. Together, these data suggest that poorly managed sites harbour higher CBB loads than well-managed, abandoned and feral coffee sites and thus should be priorities for landscape-level IPM programmes seeking to manage this economically important pest. Bremer for developing the electronic data collection system for our monitoring programme. We are also grateful to the Kona and Ka'u coffee growers that allowed us to conduct this study on their farms and to two anonymous reviewers that provided suggestions for improving the manuscript. This research was funded through the United States Department of Agriculture. Opinions, findings, conclusions or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the USDA.

ACK N OWLED G EM ENTS
The USDA is an equal opportunity provider and employer.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
MAJ and NCM conceived the research, collected data and wrote the manuscript. MAJ analysed data and conducted statistical analyses.
NCM secured funding. All authors read and approved the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data collected in this study are available upon reasonable request to the first author.