Patterns of Nestedness in Remote Polynesian Ant Faunas (Hymenoptera: Formicidae)1 Lloyd W. Morrison2 Abstract: The entire ant faunas of remote Polynesian islands consist of intro- duced species. An important question concerning the assembly of Pacific island ant faunas is whether these species are a random assortment of the available spe- cies pool, or whether they exhibit highly ordered occurrence patterns (i.e., nested subsets of species). I evaluated nestedness for the ant faunas of two island groups in remote Polynesia: (1) the Hawaiian Islands, and (2) French Polynesia and the Cook Islands. Wilcoxon two-sample tests were used to analyze nested- ness patterns for individual species and islands; the degree of nestedness for species assemblages and archipelagos was determined by combining tail proba- bilities of individual species and islands. Both island groups revealed highly sig- nificant nestedness at the level of the assemblage (a per-species approach) as well as the archipelago (a per-island approach). Considered individually, most species (73–95%) and most islands (89–100%) demonstrated significant nested- ness. Instances of nonsignificant nestedness were frequently associated with low statistical power. These results reveal a strong deterministic element in the as- semblage of remote Polynesian ant faunas. Dispersal opportunities along with presence of appropriate habitat type are likely the most important mechanisms underlying the observed patterns. The entire ant faunas of the more distribution patterns (Wilson and Taylor remote Polynesian islands (east of Rotuma, 1967b); this remained the conventional wis- Samoa, and Tonga) consist of introduced dom for many years (Ho¨lldobler and Wilson species. Although most introductions have 1990). A more comprehensive reevaluation occurred within the last 200 yr, some intro- based on more thorough collections revealed ductions may date to the ancient Polynesians more species were present than previously (Wilson and Taylor 1967a). These islands thought and found no evidence of mutually represent rare opportunities to study patterns exclusive distribution patterns among islands of community organization among species (Morrison 1996a). Ant communities exam- that share no common coevolutionary history ined on the island of Moorea revealed pat- in the region. terns of structure and function similar to A collection of early survey records sug- those found in more speciose, coevolved gested that the ant faunas of these islands mainland communities (Morrison 1996b). were depauperate, and that strong interspe- Here I examine whether the ant species cific aggression resulted in complementary occurring on remote Polynesian islands are a random assortment of the available species pool, or whether they exhibit highly ordered occurrence patterns (i.e., nestedness or nested 1 Manuscript accepted 20 February 2007. 2 Department of Biology, Missouri State University, subsets of species). Nestedness occurs when 901 South National Avenue, Springfield, Missouri the species composition of each island repre- 65897 (phone: 417-836-3119; fax: 417-836-8886; e-mail: sents a subset of the species assemblage [email protected]). on more speciose islands. A high degree of nestedness implies that the geographic dis- Pacific Science (2008), vol. 62, no. 1:117–127 tribution of species is a deterministic (and : 2008 by University of Hawai‘i Press predictable) function of physical, biological, All rights reserved or anthropogenic processes, and this has 117 118 PACIFIC SCIENCE . January 2008 important implications for biogeography, Bora (one unnamed island) (Morrison 1995). evolutionary ecology, and conservation (e.g., Morrison (1997) surveyed 10 islands in the Patterson and Atmar 1986, Patterson 1987, Marquesas, Tuamotus, Gambiers, and Cook Simberloff and Martin 1991, Cook 1995, Islands. Collections from the main Society Cook and Quinn 1995, Kadmon 1995, Hec- group of Tahiti, Moorea, Huahine, and Bora nar and M’Closkey 1997, Wright et al. 1998, Bora were relatively thorough (Morrison Hecnar et al. 2002). 1996a), as were records from the five small The following questions are posed: (1) Are islands on the barrier reefs of those main the introduced ant faunas of remote Polyne- islands (Morrison 1995). Collections from sian islands significantly nested at the level the other islands in this group were not as of species assemblages and archipelagos? (2) thorough (Morrison 1997). Which individual species or islands contrib- In the records from French Polynesia and ute to this nestedness, and which do not? the Cook Islands, a few forms were not iden- and (3) What are the potential mechanisms tified to species. Considering these records as of the observed nestedness? distinct species could artificially increase the species richness in the island group and de- crease the degree of nestedness. Thus these materials and methods forms (five ponerines and one Monomorium sp.) were not included in these analyses. A Ant Species Records Brachymyrmex sp. from Tahiti was retained, Nestedness was evaluated for the ant faunas because no other Brachymyrmex were re- of two of the more thoroughly collected is- corded from French Polynesia or the Cook land groups in remote Polynesia: (1) the Ha- Islands. waiian Islands and (2) French Polynesia and Seifert (2003) revised the Cardiocondyla the Cook Islands. All of the ant species pres- species groups, changing the identities of ent in both of these island groups are intro- Cardiocondyla species known from Polynesia. duced. The main Hawaiian islands of O‘ahu, Seifert did not examine all Cardiocondyla Hawai‘i, Kaua‘i, Maui, La¯na‘i, and Moloka‘i material from these island groups, however. are arguably the best-collected islands for Thus it is not possible in all cases to deter- ants in Polynesia. Collections of ants have mine with certainty which species now known also been made from a number of the smaller and described are present on which islands. Northwestern Hawaiian Islands. For my anal- Furthermore, distributional records for Ha- ysis, I used the list of established species in wai‘i (Nishida 2002) and French Polynesia Krushelnycky et al. (2005) (except for some are all based on species concepts before Sei- Cardiocondyla species, see later in this section). fert’s revisions. For the analyses presented Geographical distribution records were ob- here, I used the species concepts in place be- tained from Nishida (2002), with the fol- fore Seifert’s revision. lowing exceptions for recently established species: Pheidole moerens and Solenopsis sp. Quantification of Nestedness (Gruner 2003), Tetramorium insolens (Kuma- shiro et al. 2001), and Wasmannia auropunc- Wilcoxon two-sample tests (also known as tata (Krushelnycky et al. 2005). Mann-Whitney U tests) were used to analyze Records of ants from French Polynesia nestedness patterns. This approach was first and the Cook Islands were compiled from used by Schoener and Schoener (1983) to several sources. Perrault (1988) recorded the measure the ‘‘occurrence sequences’’ of spe- ant fauna of Tahiti and that of the atoll Fan- cies on islands, by ranking islands according gataufa (Perrault 1993). The Society Islands to the magnitude of a particular variable (e.g., of Moorea, Bora Bora, and Huahine were area) and then determining whether the asso- surveyed by Morrison (1996a), as well as five ciated presence/absence sequence of the spe- small islands on the barrier reefs of Moorea cies of interest was significantly nonrandom. (Tiahura, Fareone, Ahi, and Irioa) and Bora This procedure has since been used to ana- Nestedness in Polynesian Ants . Morrison 119 lyze nestedness patterns for individual species (individuals and assemblages), this approach (Patterson 1984, Simberloff and Levin 1985, could also be used to evaluate nestedness as a Simberloff and Martin 1991, Kadmon 1995, function of islands (both individual islands Nores 1995, Hecnar and M’Closkey 1997, and archipelagos). Once ranked as described Hecnar et al. 2002). Although nestedness is, here, the Wilcoxon test is applied to the se- strictly speaking, a property of a species as- quence of presences and absences of species semblage rather than an individual species, on each island; combining tail probabilities referring to individual species that have non- of individual islands yields the degree of random occurrence sequences as being ‘‘sig- nestedness for the entire archipelago. Here I nificantly nested’’ is usually done as a matter use this technique to evaluate nestedness for of convenience (e.g., Hecnar et al. 2002). individual species and entire species assemb- A nested matrix is constructed with species lages, and also for individual islands and rank-ordered in terms of decreasing numbers entire island archipelagoes. Analyses were of occupied islands, and islands rank-ordered conducted for: (1) 14 Hawaiian islands, (2) in terms of decreasing numbers of species. If 20 islands in French Polynesia and the Cook perfectly nested, all cells in the upper left- Islands, and (3) a subset of the nine most hand corner of the matrix would consist of thoroughly collected islands in French Poly- ‘‘presences’’ and all cells in the lower right- nesia, all of which are in the Society group. hand corner would consist of ‘‘absences.’’ SPSS 11.0.4 for Mac OSX (SPSS 2005) was The greater the departure from this pattern used for all analyses. in the ordered matrix, the greater the depar- ture from nestedness. results For each species, the Wilcoxon test (one- tailed) determines the degree to which