Ants of Ambon Island – diversity survey and checklist

Abstract The present checklist of ants (Hymenoptera: Formicidae) of Ambon is the first comprehensive overview of ant species recorded on the island during the last 150 years. The species list is based on literature and museum collections’ records combined with data from our field survey in 2010. In total, there are 74 ant species and subspecies representing 34 genera and six subfamilies known from Ambon. Five of the species found in undisturbed forest were exotic and indicate the overall habitat degradation on the island. The largest proportion of Ambon ant fauna are species with affinities to the Oriental region and species of Oriental-Austro-Melanesian origin. At least 20% of the species are regional endemics. In comparison to other islands in the region, the Ambon fauna seems more diverse and better sampled; however it is clear that a large part of it still remains to be described.


Introduction
The Indonesian region of Moluccas (Maluku) is one of the most biogeographically and biologically complex areas on Earth. Many of its local invertebrate groups remain undescribed, including ant diversity, which is mostly unknown from this region. Despite being a frequent subject of ecological and taxonomic studies in surrounding regions, there are a limited number of ant species records from Moluccas (e.g. Chapman and Capco 1951, Trainor and Andersen 2010, Andersen et al. 2013. Ambon is a smaller (775 km 2 ), densely populated island (estimate for 2014 is 369 000 inhabitants; BPS-Statistics, Indonesia 2014) at the north of Banda islands arc, adjacent to the much larger Seram to the north. Ambon consists of two parts, Latimor and Hitu, which are connected by a short isthmus. Both parts are characterized by cordierite-rich dacites, granites, ultramafic rocks, and uplifted Quaternary coral reefs (Pownall et al. 2013). There are also two inactive volcanoes present on Hitu.
Ambon and Seram are located in the eastern Indonesia region, which is a geologically complex zone, influenced by the activity of three major plate regions: the Eurasian, Indian-Australian and Pacific-Philippine Sea plates. Boundaries among them form deformation zones that contain many small plates, crustal fragments or terranes and are characterized by the presence of subduction zones (Hall and Wilson 2000). The Banda Arc consists of an inner volcanic arc and an outer non-volcanic arc of islands formed by sedimentary, metamorphic and igneous rocks of Permian and Quaternary age (Hall 2000). Both Ambon and Seram lay within a belt of subducted Australian crust that is south of a deep subduction trench encircling the Banda Sea. Similar to most of the Banda Sea islands, Ambon and Seram are true oceanic islands, having never been connected to the mainland, although the distances between them and the Asian and Australian continents decreased significantly during periods of glaciation (Voris 2000). It is estimated that Ambon and Seram emerged above the sea about six to five million years ago (Fortuin andDe Smet 1991, de Jong 1998).
Due to its position between the Oriental and Australian regions, the Moluccas host an interesting mixture of fauna and flora from both biogeographic origins, in addition to an important proportion of local endemics. Which faunal elements prevail is often taxon and island specific and has been a subject of multiple debates (Gressit 1958, Keast andMiller 1996). The distance between New Guinea and Ambon (and Seram) is shorter (140 km) then between Sulawesi (550 km). However, there are several islands between Ambon and Sulawesi which could serve as stepping stones. We can also expect that the exchange of biotas was mostly unidirectional as Ambon, Seram and other islands are younger than New Guinea and Sulawesi (de Jong 1998).
Based on the distribution of butterfly species, Vane-Wright and Peggie (1994) suggest that fauna of northern Moluccan islands (Halmahera, Ternate, Morotai and others) is biologically related to New Guinea, while the Buru, Ambon, Seram arc is related to North-West Australia and the Sula islands in west Moluccas are part of the 'Sulawesi region'. A similar pattern was confirmed for the butterfly genus, Hypolycaena by Cassidy and Rawlins (2011), although the authors also found links among Ambon, Seram and Sulawesi as well as Mindanao in the Oriental zone. Close relationship between Moluccan and Sulawesi species was also suggested by phylogenetic reconstruction of the grasshopper genus Chitaura (Acrididae: Oxyinae) (Butlin et al. 1998).
A distributional analysis of hawk moths (Lepidoptera: Sphingidae) revealed that Ambon species are nested within Seram fauna and both are close to Halmahera and Sulawesi. They all are however less similar to south Philippines fauna than to New Guinean fauna (Beck and Kitching 1998). Finally, a comprehensive comparison based on the distribution of over 200 Lepidopteran species (de Jong 1998) supports greater similarity of Moluccan islands to New Guinea than to Sulawesi, suggesting a much higher diversity of New Guinean fauna and shorter distance as the main contributing factors. This pattern is also repeated at the generic level (de Jong 1998). These studies suggest a complex pattern of relationships among Mollucan Islands and the surrounding regions, with the most prominent links to Melanesia or Oriental regions (often through Sulawesi).
Unfortunately, no comparable data exist for Moluccan ant fauna. Although a number of species were described early on from specimens collected by A. R. Wallace (Smith 1859(Smith , 1860(Smith , 1871 and the regional list was expanded in the following 150 years, a detailed overview of the regional fauna is missing. In this respect, the most relevant studies are by Karavaiev (1925Karavaiev ( , 1927Karavaiev ( , 1929Karavaiev ( , 1930Karavaiev ( , 1935 and by Chapman and Capco (1951) which provide summaries of ant species from the Oriental and Austro-Melanesian regions.
In recent years, studies focusing on the ants of Timor Island were conducted, contributing important knowledge of ants of the region (Trainor at al. 2010;Andersen et al. 2013). In Timor, the largest proportion of ant species (76%) had South-East Asian (Oriental) affinities, while Australian taxa accounted for 18% of the fauna and the proportion was even smaller for other neighboring islands. In addition, a distributional analysis of ants of genus Polyrhachis confirms that at least some other Moluccan islands have high levels of endemism, even higher than plants and birds (Andersen et al. 2013). However, it is unclear to what extent these patterns are representative of northern Moluccan islands, as they are positioned closer to New Guinea which contributes Austro-Melanesian taxa, as well as to Sulawesi, which could serve as a source of Oriental fauna. Understanding the distribution patterns and origin of the ant fauna of Moluccas is desirable, since ants can provide another line of evidence for reconstructing the biogeographic history of the region. Due to their diverse biological roles, ants are ideal for assessing the effects of life strategy on dispersal and diversification patterns across oceanic islands.
Here, we assemble all available information on ants occurring on Ambon Island and combine it with new data from a recent field survey. The main objectives of our study are 1) to create a checklist which can be used as a reference for follow up studies; 2) to identify new species records collected during our survey of Sirimau forest in Ambon; and 3) to compare habitat preferences of collected ants, analyze biogeographical affinities, and asses the levels of endemism of ant fauna currently known from Ambon.

Survey of Sirimau area
We surveyed ants in forested areas of Sirimau, one of two protected areas of Ambon Island. Sirimau forest covers an area of approximately 3.5 Ha and is located on the eastern side of Ambon Island between -3.6707°; -3.6918° and 128.2394°; 128.2859°, at elevations between 100 to 200 meters above sea level. Although the study area has been referred to as 'primary' forest, it is more appropriate to call it 'undisturbed', as many parts of Ambon underwent habitat transformation in the last two centuries and a large proportion of the island's surface was altered at some point. Today, urban areas cover more than 40% of the island (Miller 1999). The undisturbed forest at the collecting site was dominated by the following plant taxa: Eucalyptus sp., Acacia mangium (Willdenow, 1806), Tectona grandis (Linnaeus, 1782), Nephelium lappaceum (Linnaeus), Anacardium occidentale (Linnaeus), and Shorea sp.
Ants were collected in 500 m long transects each of which included 25 pitfall traps, 25 bait traps on the ground and 25 bait traps on vegetation. A mixture of sardines and honey was used for bait. The pitfall traps were exposed for two days and bait traps for one hour. Six such transects, located at least 200 m apart, were surveyed in undisturbed forest and six transects in secondary forest. In addition to collections in transects the area was repeatedly surveyed with general hand collecting and species not recorded in transects were added to the checklist. Samples were collected between July and September 2010 and preserved in 95% ethanol.

Species identification and compilation of species list
Collected specimens were initially sorted to morphospecies, mounted and then identified by the authors using several reference collections: Basic Entomology Laboratory, Agriculture Faculty Gajah Mada University Jogyakarta, Indonesia; Ant Collection at Museum of Comparative Zoology, Harvard University, USA (MCZ); Melanesian Ants Collection at Czech Academy of Sciences (CAS); and taxonomic literature for relevant genera. For selected species which were not possible to match with museum, literature or online records, we sequenced ~ 650 bp fragment of mitochondrial gene cytochrome c oxidase (COI), using primers LCO1490 and HCO2198 (Folmer et al. 1994). The COI sequences were used to confirm species identities among samples and to improve the identification of unknown species. The sequences were compared with records available at the online databases, Barcodes of Life -BOLD (http://barcodinglife.org) and GenBank (www.ncbi.org), and used to identify the most closely related described species or species group and to confirm morphology-based identification. When we were not able to reliably match the specimen with a described species we used a morphospecies number and attempted to identify it to at least the closest described species or species group. The specimen duplicates were deposited in CAS, Czech Republic and the MCZ, USA. The DNA sequences of COI were deposited in the Barcode of Life database (http://barcodinglife.org), as part of the Ants of New Guinea project (ASPNA) and photos of selected specimens are available online at Ants of New Guinea database (www.newguineants.org, search query 'Ambon'). Sequences were deposited in BOLD under accession numbers: ASPNA3025-14 -ASPNA3033-14.
The species checklist was compiled from collected specimens, literature records and database records. We reviewed the names of all taxa that were originally described from Ambon and included specimen records found on Antwiki (www. antwiki.org) and Antweb (www.antweb.org). Other resources including Formis 2013 (Porter and Wojcik 2013), Google Scholar (www.scholar.google.com) and Web of Science (www.webofknowledge.com) were also searched for relevant records (Ambon, Amboina etc.). We also searched for ant records associated with Seram Island and checked if species were recorded on Ambon as well. Searching the Indo-Australian collections deposited at the MCZ did not result in additional species records. We used the taxonomic nomenclature based on the online version of An Online Catalogue of the Ants of the World (Bolton 2014), as published at online Ant Catalogue (www.antcat.org). We eliminated names where the records could not be confirmed due to possible misidentification or erroneous and insufficient labeling. We retained trinomial names for all cases where status of the names were not disputed and considered valid.

Data analysis
The specimen records from all traps and transects within the same habitat type were combined and the occurrence frequency of each species was used as input data. Diversity statistics and accumulation curves were calculated in EstimateS (Colwell 2013), using 100 iterations. To compare species diversity, evenness and species occurrence frequency between habitats we used the Shannon diversity index, Shannon diversity t-test and Mann-Whitney test in the program PAST 2.14 (Hammer et al. 2001).
For each described species in the checklist, we assembled data about its distribution range using online resources (Antwiki.org, Antweb.org) and the literature. The distribution categories reflected main areas of occurrence of the given taxa and were based on recognized biogeographic regions (Holt et al. 2013) with Oceania more finely divided into regions of Melanesia, Micronesia and Polynesia to provide more sensitivity for distributional comparisons. A species was considered widespread if it occurred in more than five major biogeographic regions. A species was considered introduced if it was known as non-native in the area or unlikely originating in Moluccas. A species was considered endemic it if was recorded only from Ambon or other Moluccan islands. This latter category was broader as the ant fauna of the whole region is highly undersampled and unknown. It is thus possible that species described from a particular island are in fact distributed across several neighboring islands. For these reasons our endemic species category included regional endemics.

Results
Based on our literature survey, 48 species have been reported to occur on Ambon. In our field survey, we collected 31 ant species, 16 of which could be identified, while 15 did not match any described taxa on the basis of morphological or molecular characters. Of the 16 described species, only 5 species were previously reported. Our results suggest there are at least 74 ant species from 34 genera and six subfamilies on Ambon Island. The combined checklist of described ant species currently known from the island is 59.
Habitat types differed in species richness: 30 ant species were found in undisturbed forest, while 19 ant species were found in secondary forest. All species occurring in secondary forest were also recorded in undisturbed forest; the overlap as expressed by Jaccard Index was 0.63. Shannon index of diversity for undisturbed forest was 2.96, and 2.6 for secondary forest. The habitats differed significantly in species occurrences, calculated as the proportion of visited traps (Shannon diversity t-test p < 0.01, t = 10.9, d.f. = 1010.5). The accumulation curves suggest that both habitats were sufficiently sampled, as both curves and the statistical indicators (Mao Tau, ICE) reached an asymptote (Figure 1).
The comparisons of biogeographic associations of described species suggest that the ant fauna of Ambon is dominated by species with Oriental affinities (13 spp., 22%)  and by species distributed equally across Oriental and Austro-Melanesian regions (13 spp., Figure 2). The proportion of species of solely Melanesian origin is slightly smaller (10 spp.) and there are 12 species (20%) which can be considered as local or regional endemics (restricted to Ambon and/or Moluccas). In addition, we detected five species which are most likely introduced on the island. A smaller proportion of species are either widespread (occurring in five or more biogeographic regions) or have more restricted distribution throughout Moluccas (Figure 2).

Discussion
Our study provides the first formal checklist for ant species occurring at Ambon Island. In addition to nearly 50 species reported previously from the literature and databases, our field survey contributed new distributional records for 11 species (marked in Table 1). The additional 15 species which did not match with any currently described taxa are likely to expand the checklist once incorporated into taxonomic revisions of corresponding groups. Our survey also confirms the presence of several species recorded more than 100 years ago and provides evidence of several introduced species in the area.
Despite the addition of 26 species to the checklist, the total number of ant species recorded from Ambon (74) is most likely underestimated. Our survey focused on forested area of only one of the two Ambon islands and in the current checklist some of the genera (e.g. Pheidole, Tetramorium, Camponotus, Nylanderia) which are generally diverse in this region are represented by only few species or completely missing. Our study also did not include sampling of more cryptic leaf litter fauna which would probably contribute substantially to the local species list. For example, we anticipate that species diversity of the genera Pheidole, Nylanderia, Hypoponera and Strumigenys will be higher than reported here. On the other hand, we report significantly higher species richness than the number of species recorded from the southern Moluccan islands of Wetar (3600 km 2 ), Pantar (728 km 2 ) and Lembata (Trainor and Andersen 2010) or the Solomon Islands (e. g. Lembata -70 ant spp., Sarnat et al. 2013) which are larger than or of similar size to Ambon. We anticipate that the ant species richness of Ambon might be even higher than that of other comparably sized islands, as it is adjacent to the much larger Seram, which could potentially serve as a source area for ant species that would otherwise go extinct on small and disturbed islands. The accumulation curves indicate that both our sites were sampled exhaustively, within the limits of the methods used. This suggests that our survey provides a representative view of the above ground foraging ant fauna of the area. General collecting in the area outside the transects yielded only one additional species (Anochetus graeffei Mayr, 1870). The fact that all the species from secondary forest were also found in undisturbed forest is somewhat surprising and probably indicates a limited regional species pool and relative similarity of both types of habitats. A thorough sampling of other microhabitats (leaf litter, canopy) could contribute additional species.
Both forests were also inhabited by several invasive species not native to the area, including: Anoplolepis gracilipes, Paratrechina longicornis (Latreille, 1802), Solenopsis geminata (Fabricius, 1804) and Monomorium pharaonis. Although, these species mostly did not reach high frequencies in either type of forest (only A. gracilipes reached a high frequency, being the fourth most frequent species in disturbed forest), their presence is indicative of medium to strong habitat disturbance (Invasive Species Specialist Group, www.issg.org, New Zealand biosecurity system, http://www.biosecurity. govt.nz). The most frequently collected taxa in both habitats were several species of Pheidole and Polyrhachis, as well as Oecophylla smaragdina and Dolichoderus thoracicus.
Although the undisturbed forest site was inhabited by 11 more species than the disturbed site, the high abundances of O. smaragdina, D. thoracicus and moderate abundances of M. pharaonis and Tapinoma melanocephalum (Fabricius, 1793) indicate that this area has likely been disturbed in the past, as these species are often associated with secondary or disturbed forests (Invasive Species Specialist Group database, http:// www.issg.org, Way andKhoo 1991, Klimes et al. 2012).

Biogeographic affinities
Out of the 59 valid species currently known from Ambon, the largest proportion (13 spp.) is of Oriental (South East Asian) and Oriental-Austro-Melanesian origin (13 spp.). The latter category represents species which are equally distributed across Oriental and Austro-Melanesian regions, mostly spread across islands of South East Asia, New Guinea and/or Australia. The third largest group of Ambon myrmecofauna is of local origin (12 spp.). Here, they represent a smaller proportion of fauna compared to data from the ant genus Polyrhachis on the much larger Timor (~70% endemics, Andersen et al. 2013). Our data suggest that the Oriental faunal elements are somewhat more prominent in Ambon than Austro-Melanesian elements. The most important part of the fauna are however regional endemics and species distributed equally across parts of the Oriental and Austro-Melanesian regions -an area which corresponds with a largely extended Wallacea. This group includes taxa that either originated within this region and/or lineages which expanded into the region from either Oriental or Australian continental shelves. However, without detailed phylogenetic reconstruction for each of these genera it is not possible to reveal their exact area of origin. A detection of five introduced species in a relatively preserved habitat should be of conservational concern and suggests that these exotic species are well established in Ambon. This is indeed a case in other parts of the island, especially in urban and suburban areas (FL personal observation).

Conclusions
Our field survey and literature search revealed that there are over 70 ant species and subspecies found on Ambon Island, of which at least 59 have been previously described. The total number is likely to increase once data from other parts of the island and from understudied microhabitats such as tree canopy or forest leaf litter become available. The occurrence of several non-native ant species recorded in old preserved forest is of conservational concern and suggests that the introduced species are well established on the island. The overall structure of the ant community recorded in the forested area is indicative of a strong effect of habitat fragmentation and large habitat conversion on the island, where urban habitat is prominent. From a biogeographic perspective, the largest portions of Ambon ant fauna are species with Oriental affinities and species native to the Oriental-Austro-Melanesian region. Interestingly, the portion of Austro-Melanesian species is smaller which contrasts to patterns observed in Moluccan Lepidoptera. Our study provides the first basis for future research on ants of Ambon Island and surrounding areas. As habitat conversion in this region continues, the unique and endemic fauna of Moluccas is particularly endangered and in urgent need of proper conservational efforts. Ants, along with moths and butterflies can serve as useful indicators of habitat quality and uniqueness and for identification of conservational priorities.