We collected Dacini flies using handmade bottle traps. A 3 cm-diameter hole was cut 15 cm from the base of a 500 ml plastic water bottle. Male attractant lures: methyl eugenol 10 g cones (Scentry Biologicals Inc., Billings MT, USA), cue lure 2 g cones (Scentry Biologicals Inc., Billings MT, USA), and zingerone (Sigma-Aldrich, St. Louis MO, USA) were individually suspended by a string inside the bottle, 5 cm from the top. A 100 ml water solution of Fisherbrand Sparkleen detergent (Fisher Scientific, Pittsburgh, PA, USA) poured at the bottom of each bottle trap was used as a killing agent. The traps were then hung from a cacao tree (Theobroma cacao L.) branch at 1.5 m high. Traps were checked every 2–5 days, and trapped flies were transferred to 95% ethanol. Trapping was mainly conducted in Wotu, Kabupaten Luwu Timur, South Sulawesi at sites named “Insitu” [WGS84 N 2.5587 E 120.7935], “MCRC Tarengge” [WGS84 N 2.5547 E 120.8047] and “Arni field” [WGS84 N 2.5587 E 120.7935] (Fig. 1).

Planting at Insitu was composed primarily of cacao clones PBC123 and BR25 that were planted 3.5 m apart within a row, with 3.5 m spacing between rows, irregularly shaded by a diversity of fruit trees. This site was the most diversified among the trapping sites, including more than 100 banana (Musa sp.), four large durian tree (Durio sp.), eight rambutan (Nephelium lappaceum L.), six coconut trees (Cocos nucifera L.), as well as some ginger (Alpinia sp.), Luffa (Luffa acutangula L.), papaya (Carica papaya L.), chili peppers (Capsicum sp.) and corns. The cacao trees were not regularly pruned but were treated with an unknown pesticide, and were not artificially irrigated. This farm was surrounded by neighboring cacao farms with a similar diversified composition. In addition, some jackfruit (Artocarpus heterophyllus Lam.), mango (Mangifera indica L.), guava (Psidium guavaja L.), rose apple (Syzygium sp.), as well as breadfruit (Artocarpus altilis (Parkinson) Fosberg) were present around the neighboring farms. The site at MCRC Tarengge represents a 1 ha of cacao trees of clone M01 with a 1.5 × 3 m density, without any other fruit trees within the block. However, several langsat trees (Lansium parasiticum (Osbeck) Sahni & Bennet), banana and a couple of durian trees were present in the neighborhood farms 100 m away from the trapping sites, as well as 20 papaya, 10 rambutan trees (Nephelium lappaceum L.) within 200 m, and several mango trees, jackfruit, guava, and rose apple trees within 400 m radius from the trapping site. No pesticide was applied during our field collection, but both surrounding blocks were regularly treated with pesticides. The site ‘Arni field’ was also mainly composed of cacao trees at lower density (3 × 3 m). Various fruit trees disseminated around the farm, including some banana, rambutans, jackfruit, mangos, guava, and Ambarella (Spondias dulcis L.), with rows of corn (Zea mays L.) and several durian trees within 50 m, as well as jambu putik (Syzygium sp.), rose apple, and breadfruit within 300 m.

In total, the trapping effort at Arni field was approximately four months, five months at MCRC Tarengge, and six and half months at Insitu, spread over different periods during 2016–2019 (Suppl. material 1: Table S1). The site “Manado” [WGS84 1.3973N, 124.6488E], near the city of Manado in North Sulawesi, had three trapping days. At all sites combined, we collected 4,517 Dacini flies and identified all specimens to species level, initially based on external morphology. In cases where the morphology was inconclusive, we used DNA sequences of Cytochrome C Oxidase I and/or Elongation Factor 1-alpha for a total evidence identification approach. In 2016, some additional collecting was done with torula yeast dissolved in water, which attracts females. A full overview of all traps and localities can be found in Suppl. material 1: Table S1. All voucher material is stored at the University of Hawaii Insect Museum (UHIM). Photographs of adult specimens were taken using a Zeiss Discovery.V8 stereomicroscope with an attached Sony alpha-6300 camera. Photographs from multiple focal plains were combined into a single stacked image using Affinity Photo 1.7.3 and optimized for publication. Plates with multiple images were assembled in Affinity Designer 1.7.3. Wings of selected specimens were removed and mounted in euparal on glass-slides and photographed in a similar manner as the adults.

Methods for DNA extraction, PCR primers and conditions, and Sanger sequencing follow those of San Jose et al. (2018a). For the present study, we sequenced a Cytochrome C Oxidase I (COI) 809 base pair 3P’ fragment and an Elongation Factor 1-alpha (EF1-alpha) 762 bp gene fragment for Bactrocera niogreta. We compared the sequences to our (partially unpublished) sequence database and here release sequences of the most closely related species to establish the diagnostic discrimination of COI and EF1-a sequences. We also sequenced COI and EF1-a for several specimens of Bactrocera melastomatos Drew & Hancock, 1994 and Dacus longicornis (Wiedemann, 1830), to confirm if the different morphological forms were mirrored in mitochondrial and/or nuclear genetic variation. Finally, we sequenced EF1-alpha for the two specimens of B. carambolae Drew & Hancock, 1994, which is diagnostic at five positions, to confirm its identity (see also Leblanc et al. 2019). All specimen collecting details and DNA sequences are available through BOLD dataset DOI: http://dx.doi.org/10.5883/DS-DACSU, and GenBank accessions: MT456325–MT456363 [COI] and MT456286–MT456324 [EF1-alpha]). We performed maximum likelihood analyses for each subset of sequences using IQTree 1.6.10 (Nguyen et al. 2015). We allowed IQTree to determine the substitution model via its integrated modeltest and ran maximum likelihood analyses with 5,000 ultrafast bootstraps and 5,000 Sh-aLRT bootstraps. We consider branches with support values > 95 % for ultrafast bootstraps and > 80 % for Sh-aLRT bootstraps as well supported. Resulting trees were optimized for publication using FigTree 1.4.3 and Affinity Designer 1.7.3.

Sulawesi is a dispersal crossroads for the biotas of Southeast Asia, Australia, and Oceania. The updated species checklist we present here shows that Sulawesi is unique, with many endemic species, but that there are also strong connections with Southeast Asia, at least for the taxa under study. This finding does not support the earlier working hypotheses that posited a closer connection to the Sahul fauna, including Papua (Hardy 1982). Based on this recent data, it seems likely that Wallacea has been a “stepping-stone” for Dacini to reach Australasia and Oceania; 30 Sulawesi species are also found in (former) Sunda, while only one is shared with Sahul. This can provide crucial insight into the timing of the diversification of the group in the latter areas. Wallacea, and Sulawesi with it, was separated in discontinuous landmasses with fluctuating sea-levels until the mid-Miocene, 10–15 Ma (Hall 2009). Before this connection, Sunda and Sahul were separated by vast oceanic distances that were unlikely to be crossed by fruit flies. There are currently hundreds of Dacini species known from Australia and Oceania (Drew 1989, Drew and Romig 2001), which may have resulted from rapid radiation after reaching these new areas of ecological opportunity. Similarly, the timing of the formation of Wallacea suggests a relatively recent origin for the 34 Sulawesi endemic Dacini species. However, it should be noted that the Dacini fauna of Papua is understudied (White and Evenhuis 1999) and further surveys in this area may reveal shared geographic ranges with some of the species now presumed to be Wallacean endemics.

It has been advocated by some that the categorization of biogeographic regions should follow more quantitative measures (Kreft and Jetz 2010), as opposed to the qualitative assessments from the early explorers. Using a non-metric multidimensional scaling approach across a wide range of taxa, Kreft and Jetz (2010) suggested that Lydekker’s line was most appropriate for separating the Oriental region from Australasia. This agrees with our findings for Dacini, although, depending on the scale of the patterns in question, maintaining Wallacea as a separate biogeographic region can equally well be argued. Other authors have further included a phylogenetic component in the delimitation of biogeographic areas, which resulted in a suggested split of Wallacea: Sulawesi and the Lesser Sunda Islands were grouped with Southeast Asia into the Oriental region, whereas the Moluccas were grouped with Papua in the Oceanian region (Holt et al. 2013). However, none of these broad-scale assessments of biogeographic categorization include or consider invertebrate taxa. Moreover, as phytophagous insects, it might be expected that the biogeographic pattern of Dacini more closely tracks phytogeographic regions. A broad study that included 7,340 plant species across Southeast Asia suggested ‘central Wallacea’ [defined to encompass the Philippines, Sulawesi, lesser Sunda islands, Moluccas, and Java] as a separate region (van Welzen et al. 2011). This categorization is further corroborated by the climatic conditions; central Wallacea has a yearly dry season and monsoon, whereas both neighboring regions lack a prolonged dry season. For Dacini, we find few connections between the Philippines and Wallacea, but there are some, e.g., B. commensurata. Future surveys of the Papuan Dacini fauna, and placement of the Wallacean taxa in a phylogenetic framework, can further inform which biogeographic delimitation fits best with this group of fruit flies, and it is clear that more studies on invertebrate groups will be important to fully understand the biogeographic affinities of the islands that connect Asia with Australia and Oceania.

As a tropical island, Sulawesi has a rich diversity of fruiting plants and, consequently, insects that utilize them. Our surveys were performed in cacao plantations; the only Dacini that is known to feed on cacao is the polyphagous Bactrocera dorsalis (Allwood et al. 1999). However, we have never observed flies attempting to oviposit on cacao nor have we found maggots inside the pods. Cacao is likely a very rare host for B. dorsalis, if at all, and potentially only fallen and dehiscent fruit where the tough skin is cracked is susceptible. We commonly found B. dorsalis, but also another pest species; B. albistrigata at all sites. Together they made up 70.6 % of all individuals collected. The host records of B. albistrigata include jackfruit (Artocarpus heterophyllus), jambu putik (Syzygium sp.), mango (Mangifera indica L.), guava (Psidium guajava L.) and rose apple (Syzygium sp.) some of which were planted near the cacao. It is interesting to note that although we encountered the cucurbit pest Zeugodacus cucurbitae, it was surprisingly rare. Possibly, this is due to the limited availability of melon hosts in the area (none were observed), although it is also known to feed on papaya, which were recurrent in the cacao orchards sampled, and non-commercial cucurbits that commonly occur as weeds in gardens and plantations. We further encountered small numbers of B. umbrosa, a pest of breadfruit (Artocarpus altilis) that has an extraordinarily wide distribution; it is the only species in the checklist that is known to be naturally dispersed across Southeast Asia, Wallacea, Australia and Oceania (Krosch et al. 2018). We suggest that, if desired, the population densities of pests in our survey areas can likely be decreased significantly with sanitation measures, most importantly the removal of fallen fruit and pruning of damaged fruit unfit for consumption.