Studies in Avian Biology No. 22:51-60, 2001.

MOLECULAR SYSTEMATICS AND BIOGEOGRAPHY OF THE HAWAIIAN AVIFAUNA

ROBERT C. FLEISCHER AND CARL E. MCINTOSH

Abstract. The Hawaiian avifauna is exceptional for its high proportion of endemic taxa, its spectac- ular adaptive radiations, and its level of human induced . Little has been known about the phylogenetic relationships, geographical origins, and timing of colonization of individual avian lin- eages until recently. Here we review the results of molecular studies that address these topics. Mo- lecular data (mostly mitochondrial DNA sequences) are available for 14 of the 21 or more lineages of Hawaiian . We briefly review results of phylogenetic analyses of these data for lineages that have experienced major and minor radiations, and for single differentiated and probable recent colonists. When possible, we determine the mainland species that are genetically most closely related. We find evidence that roughly half of the >21 lineages colonized from North America; not even a quarter appear to have come from South Pacific Islands. Our data also provide little evidence that Hawaiian lineages predate the formation of the current set of main islands (i.e., >5 Ma), as has been found for Hawaiian Drosophila and lobeliads. Key Words: adaptive radiation; biogeography; Hawaiian avifauna; mitochondrial DNA; molecular systematics.

In 1943 Ernst Mayr published a short paper in tially mislead on issues of common ancestry via The Condor summarizing his hypotheses about homoplasy. DNA sequences, on the other hand, the geographic origins and closest living rela- while obviously not evolving in a perfect clock- tives of each known lineage in the Hawaiian avi- like fashion (see below), do change over time, fauna. Mayr (1943) concluded that half of 14 and evolve more continuously than morphology. hypothesized colonizations were of American Also, with the exception of a relatively few non- origin and only two lineages arose from Poly- synonymous changes within protein sequences, nesia. Therefore, although ‘ is considered they generally evolve via mutation and drift (Nei part of the “Polynesian Region” because most 1987, Avise 1994), and are not as subject to ho- of its biota and its human inhabitants had Pol- moplasy via convergence or stasis as are mor- ynesian ancestors, in terms of its birds Hawaii‘ phological or other characters. Thus major adap- is in the Nearctic Region. Since Mayrs’ paper, tive shifts in, for example, the bills of Hawaiian other authors have posited similar systematic hy- honeycreepers, may occur within some lineages potheses and biogeographic scenarios based on (e.g., to thin and decurved in the nectarivorous morphological, ecological, and distributional I‘ iwi,‘ Vestiaria coccinea), while not in others data (e.g., Amadon 1950, Pratt 1979, Berger (e.g., conical and finchlike in the Laysan Finch, 1981). Paleontology has offered only minor res- Telespiza cantans), in spite of an identical olution of the relationships of ancestral lineages amount of time since evolving from their puta- or the timing of speciation events; although tively “finch-billed” common ancestor. There there is an excellent Holocene record in are methods for detecting symplesiomorphic Hawaii‘ (Olson and James 1982a, 1991; James versus synapomorphic characters in phylogenet- and Olson 1991), the pre-Holocene record is ex- ic analysis, but the higher variance in rates of tremely limited (though one excellent fauna change of morphological characters remains a dates to >0.12 Ma ago; James 1987). problem for phylogenetic reconstruction (Hillis In recent years, molecular methods have prov- et al. 1996). en extremely useful for inferring evolutionary While there have been significant molecular relationships among taxa and the relative time investigations of particular Hawaiian plant and frames during which taxa evolved (Avise 1994, invertebrate taxa (especially Drosophila; e.g., Hillis et al. 1996). Inference from molecular data Hunt and Carson 1983, DeSalle and Hunt 1987, may be the best available way to reconstruct DeSalle 1992), few molecular studies detailing phylogenetic relationships and determine geo- evolutionary histories of the Hawaiian avifauna graphical origins and evolutionary time frames have been made until recently (e.g., Tar-r and for Hawaiian taxa. In part this is because mor- Fleischer 1993, 1995; Feldman 1994, Cooper et phological or behavioral changes are often adap- al. 1996; Fleischer et al. 1998, 2000, this vol- tive responses subject to natural or sexual selec- ume; Paxinos 1998, Sorenson et al. 1999, tion (i.e., as part of the process of adaptive ra- Fleischer et al. in press, Rhymer this volume; C. diation), and they do not usually show constancy Tarr, E. Paxinos, B. Slikas, H. James, S. Olson, in their rates of change. Thus they can poten- A. Cooper, and R. Fleischer, unpubl. data).

51 52 STUDIES IN AVIAN BIOLOGY NO. 22

TABLE 1. THE ELEMENTS OF THE HAWAIIAN AVIFALINA

TaXOn Family No. of species” Geographic ori& Comments‘ Non-passeriformes: Plataleidae 22 N.A. minor radiation, flight- less, Apteribist Night Heron Ardeidae 1 N.A. recent colonist, Nyctico- rax nycticorax Moa-nalos Anatidae 24 W. Hemisphere minor radiation?, 3 flightless duck gen- eral Trne Geese Anatidae 23 N.A. minor radiation, Bran- ta,te Modem Ducks Anatidae 2 N.A. and Asia 1 Z differentiated, 1 re- cent colonist, Anas e Porzana Rails Rallidae 212 Pacific/unknown major radiation?, ~2 colonizationsl Large rallids Rallidae 2 N.A.? recent colonists?, coot and moorhen Black-necked Stilt Recurvirostridae 1 N.A. recent colonist, Himan- topus knudseni e Acciptridae 1 Asia recent colonist, Haliaee- tus leucophryst Acciptridae 1 N.A. differentiated, Buteo so- litarius e Harrier Acciptridae Unknown differentiated, Circus dossenust Long-legged Owls Strigidae Unknown minor radiation, Grullis- trix spp. 4t Short-eared Owl Strigidae Unknown recent colonist, Asio flammeus sandwichen- sis Passeriformes: crows Corvidae 24 Unknown minor radiation?, Corvus SPP., 3t, 1 e Millerbird Sylviidae 1 South Pacific differentiated, Acroce- phalus familiaris e Elepaio‘ Myiagridae 21 Australasia differentiated, Chasiem- pis sandwichensis Thrushes Muscicapidae 5 W. Hemisphere minor radiation, Myad- estes spp., 37, 1 e Meliphagidae ~6 South Pacific minor radiation, spp., Chaetoptila, alli_ Honeycreepers Fringillidae 250 Asia or N.A.? major radiation, drepani- dines, most? or e >21 lineages 13 families 2 IO2 species

a Number of species within each lmeagelfamily, based on James and Olson (1991). Olson and James (1991), and H. James (pers.comm.). h N.A. = North America; W = West ’ ’ t denotes at least some members extinct; e denotes at least some members endangered.

Components of the Hawaiian avifauna vary flightless waterfowl (moa-nalos) show extreme greatly in the degrees to which they have spe- morphological modification in their apparent ciated and become modified morphologically shift into a ratite/grazing /tortoise niche and ecologically (Table 1). For example, the Ha- (Olson and James 1991; Sorenson et al. 1999). waiian drepanidines (Hawaiian finches or hon- Other avian lineages have not speciated and eycreepers) have evolved incredible morpholog- have changed morphologically little or not at all ical, ecological, and behavioral diversity across from putative mainland relatives (e.g., Black- more than 50 species and are one of the most crowned Night Heron, Nycticorax nycticorax often cited cases of adaptive radiation (Roths- hoactli; Short-eared Owl or Pueo, Asioflammeus child 1893-1900, Perkins 1903, Amadon 1950, sandwichensis). Is this variance in levels of spe- Raikow 1977, Freed et al. 1987a, James and Ol- ciation and phenotypic differentiation related son 1991, Tarr and Fleischer 1995, Fleischer et merely to the lengths of time that lineages have al. 1998). Several species of extinct, large, been evolving in the islands (Simon 1987, Car- HAWAIIAN BIRD MOLECULAR SYSTEMATICS--Fleischer aizd Mclntosh 53

FIGURE 1. Map of the main Hawaiian Islands (plus inset map of main and leeward Hawaiian Islands). Ages of the oldest rocks from the main islands based on K-Ar dating are noted. -Nui is composed of the islands of Maui, Lgna‘i, Kaho‘olawe, and Moloka‘i, all of which were connected until about 0.3-0.4 Ma ago and again during more recentperiods of low sea level. son and Clague 1995)? Or are there other factors 1993, Fleischer et al. 1998). We then consider that have promoted stasis in some lineages and the origins and phylogenetic histories of each change in others, regardless of length of time in lineage within the avifauna, addressing exten- the islands? As noted above, the fossil record sive and minor radiations, well-differentiated provides little resolution of this question. Thus, single species, and undifferentiated (and likely estimates of the age of separation from ancestors recent) colonists. We also apply a molecular outside of the Hawaiian Archipelago, or the age clock approach to obtain rough estimates of the of a radiation within the islands, can only be maximum period of time that a lineage could inferred from molecular data. have existed in the Hawaiian Islands. The Hawaiian Islands and its avifauna are ex- tremely isolated from continental and other Pa- GEOLOGICAL HISTORY AND THE cific island avifaunas. This is likely the primary CALIBRATION OF MOLECULAR reason for the relatively low number of indepen- EVOLUTIONARY RATES dent taxonomic avian lineages that occur in the The Hawaiian Islands have an unusual geo- islands (Mayr 1943, Pratt 1979). While the total logical history (Clague and Dalrymple 1987, number of such lineages has been increased (and Walker 1990, Carson and Clague 1995; Fig. 1). continues to increase) from recent fossil findings They form as the Pacific Plate drifts northwest (Olson and James 1982a, 1991; James and Olson over a “hot spot” where magma extrudes from 1991), the islands still appear to have far fewer the earths’ mantle through the crust to build independent avian lineages than one might ex- huge shield volcanos (often to >4 km above sea pect for a tropical archipelago of this size and level). The extreme weight of a new island, topographic diversity, and there may be addi- combined with the cooling of the crust as it tional factors involved that limit the primary di- moves away from the hot spot, causes a rela- versity of the avifauna. tively rapid subsidence in island elevation and Here we summarize molecular and other data area. Subsidence continues slowly beyond this relevant to systematics and biogeography of the point, as does erosion, and islands shrink to be- Hawaiian aviafauna. We first provide a brief come small coral and sand atolls and ultimately overview of the geological history of the Ha- undersea mounts (Fig. 1). waiian Archipelago and its utility for calibrating The Hawaiian Islands are ordered by age in a rates of molecular evolution (Tarr and Fleischer linear pattern, with the oldest main island in the 54 STUDIES IN AVIAN BIOLOGY NO. 22 northwest (’ at 5.1 Ma) and the youngest obtain a rate of about 5.O%/Ma); RFLP variation in the southeast (Hawaii‘ at 0.43 Ma; Fig. 1). in New World quail at 2.O%/Ma (reported in This volcanic conveyor belt provides an excep- Klicka and Zink 1997); woodpecker cyt b at tional system for evolutionary studies, as it sets 2.O%/Ma (Moore et al. in press); cyt b in cranes up a temporal framework that can be used to at 0.7%/Ma for Balearicines versus Gruines (old estimate the timing of evolutionary events and split) and up to 1.7%/Ma for comparisons within rates of evolution. The age of an island is the the Gruines (Krajewski and King 1996); and cyt maximum age for a population inhabiting the is- b in albatross at 0.65%/Ma (Nunn et al. 1996, land. These ages can be used to calibrate rates recalculated for total sequence change in Klicka of molecular change if phylogenies reveal that and Zink 1997). In the crane and albatross stud- the pattern of cladogenesis parallels the timing ies the slower rates could be caused by the lon- of island formation, and if populations colonize ger generation times in these species, or perhaps near to the time of island emergence (Bishop by reduced metabolic rates in these larger-bod- and Hunt 1988, Tarr and Fleischer 1993, Givnish ied taxa (Martin and Palumbi 1993, Rand 1994, et al. 1995, Fleischer et al. 1998). Bromham et al. 1996, Nunn and Stanley 1998). We used this rationale to calibrate part of the Alternatively, the difference may relate to the mitochondrial cytochrome b (cyt b) gene in Ha- fossil dates used for calibration: for both studies waiian drepanidines (Fleischer et al. 1998). The these dates are older than 10 Ma, whereas for overall rate of cyt b divergence, corrected for all but the partridge/Callus comparison (Randi minor saturation, transition bias, rate variation 1996) the dates are before 5 Ma. Both studies among sites, and potential lineage sorting is attempt to correct for saturation (Krajewski and 1.6% sequence divergence/Ma. This value is King 1996, Nunn et al. 1996), but may severely similar to a rate we estimated for overall restric- underestimate divergence (Arbogast and Slow- tion site divergence in mitochondrial DNA inski 1998). This could be considered an inverse (mtDNA) in drepanidines (-2%/Ma; Tarr and prediction of the findings of Moore et al. (in Fleischer 1993). Note that rates calibrated using press): using dates older than 5 Ma to calibrate this approach are based on a time period of di- may result in an underestimate of the rate. Sup- vergence up to only about 4 Ma. Recently, porting this is a negative correlation between di- Moore et al. (in press) showed through simula- vergence times and divergence rates (Spearman tion modeling that cyt b sequence divergence is rho = -0.51, P = 0.042) from Table 2 of Martin accurate as a predictor of time of divergence and Palumbi (1993). Avian rates are similar to only to about 5 Ma (i.e., about 10% overall se- most mtDNA/cyt b rates calculated for mammal quence divergence). Predictions of dates older taxa (e.g., -2%/Ma; Brown et al. 1979, Irwin et than 5 Ma are generally underestimated. Nonlin- al. 1991, Stanley et al. 1994, Janacek et al. earity of sequence divergence due to saturation 1996). and rate variation among sites appears to be- In general, then, calibrated rates of mtDNA come problematic above about 10% overall se- protein coding sequence divergence in birds and quence divergence for birds (Krajewski and do not appear to vary greatly from King 1996, Randi 1996, Moore and DeFilippis about 2%/Ma. Most rate variation appears to be 1997). Thus the drepanidine or other cyt b rates correlated with variation in body size and its are not likely to be applicable to events that hap- correlates (i.e., metabolic rate, generation time; pened appreciably earlier than 5 Ma, and caution Martin and Palumbi 1993, Rand 1994), although must be exercised when making predictions or some of the variation may be due to differing calibrations from cyt b sequence divergences selective constraints on proteins in different lin- over 10%. eages or to fluctuations in population size (Ohta Our drepanidine rates (Tarr and Fleischer 1976). In summary, with the exception of the 1993, Fleischer et al. 1998) are within the range very rapidly evolving control region (which in of estimates for avian and mammalian taxa some sections may be evolving an order of mag- based on calibrations derived from relatively re- nitude faster than the average for mtDNA; e.g., cent fossil evidence of cladogenesis. This is true Quinn 1992) most avian and mammalian rate for both restriction fragment length polymor- calibrations based on corrected mtDNA diver- phisms (RFLPs) in total mtDNA and sequence gence and dates before 5 Ma ago reveal rates at divergence in the cyt b gene. Examples of avian about, or above, 2% divergence/Ma. Based on rates include RFLP variation in geese at -2%/ the rather detailed rationale described above we Ma (Shields and Wilson 1987); cyt b sequences feel that mtDNA (RFLP or cyt b) sequence di- in partridges versus Gallus at 2.O%lMa (Randi vergence between a Hawaiian taxon and its clos- 1996; however, Arbogast and Slowinski [ 19981, est non-Hawaiian relatives that is below about corrected the divergences using an HKY [Has- 10% would indicate an origin near the time of egawa et al. 198.51 model with a I-correction to or after the formation of the island of Kauai.‘ HAWAIIAN BIRD MOLECULAR SYSTEMATICS-Fleischer and McIntosh 55

ORIGINS AND EVOLUTION OF THE led to the suggestion that the drepanidines are HAWAIIAN AVIFAUNA not monophyletic (Pratt 1992a,b). Molecular data may prove especially useful for assessing There were more than 102 species of native breeding land- or waterbirds (i.e., non-seabirds) evolutionary relationships in this group, and in the Hawaiian Islands (Table 1; constructed they do support a cardueline ancestry and, thus from James and Olson 1991, Olson and James far, monophyly of the drepanidines (Fleischer et 1991; and H. James, pers. comm.). These 102 al. 1998; Fig. 2~). species sort into six families (Passeri- Molecular data may also be effective in esti- formes) and seven non-songbird families (Table mating a time frame for the drepanidine radia- 1). Some families have a relatively large number tion. The radiation of the drepanidines would of species (i.e., >4) and, in some cases, it is seem quite deep based on their relative degree fairly clear that each group of species in a family of phenotypic diversity. Molecular evolutionary represents an in situ radiation from a single col- rate estimates based on DNA-DNA hybridiza- onization (e.g., drepanidines, thrushes). It is tion data (Sibley and Ahlquist 1982) are in sup- clear that in some families (e.g., anatids, rallids) port of this prediction with an estimated split of there has been more than a single colonization drepanidines from a cardueline outgroup of event, while for others (e.g., corvids, meliphag- about 15-20 Ma. Molecular rate estimates from ids) it is difficult to determine how many inde- both allozyme (Johnson et al. 1989, Fleischer et pendent colonization events have occurred. al. 1998) and mtDNA data (Tarr and Fleischer Avian biologists working in the islands have 1993, 1995; Fleischer et al. 1998), however, been fortunate to have an excellent Holocene strongly contradict the results of Sibley and fossil record (Olson and James 1982a, 1991; Ahlquist (1982) and suggest a basal split that James and Olson 1991). Without this record, we began about 4 Ma ago and a separation from a would be missing a tremendous amount of in- mainland cardueline ancestor (not necessarily formation about distributions, phylogeny, bio- the closest outgroup; Fig. 2c) of <5-6 Ma ago. geography, and ecology of these birds. Even so, These mtDNA results are based on several in- additional fossil taxa continue to be discovered ternal rate calibrations estimated as outlined and, thus, our knowledge remains incomplete. above for cyt b. Sibley and Ahlquists’ (1982) The advent of genetic studies employing the results may be biased by their use of continental polymerase chain reaction (PCR) has opened a biogeographic points in their calibration (Quinn new and exciting avenue for study of these fos- et al. 1991) or by use of too distant outgroups sils. Our laboratory has had considerable success for comparison. amplifying mtDNA sequences from these sub- No other avian radiation in Hawaii‘ is so di- fossil remains. Here we summarize what has been learned about the evolution of Hawaiian verse in morphology or number of lineages as birds from phylogenetic analyses of mtDNA se- the drepanidines. Extinct flightless rails, classi- quences from a number of extinct and extant fied as Porzana (Olson and James 1991), in- taxa. cluded perhaps more than 12 species, with as many as three species on each major island. Un- EXTENSIVE RADIATIONS til recently it has not been clear whether these species comprise a single highly radiated clade, The drepanidines (Hawaiian finches or hon- or represent a number of independent coloniza- eycreepers) are by far the most speciose group tions from mainland or other Pacific island in Hawaii,‘ with 33 species known from histor- sources. Molecular phylogenetic analyses (B. ical collections and more than 17 known from subfossil remains (totaling over 50 species; Slikas, S. Olson, R. Fleischer, unpubl. data) in- James and Olson 1991; H. James, pers. comm.). dicate that each of the two historically collected The drepanidine radiation is remarkable for its Porzana species resulted from independent col- extreme morphological, ecological, and behav- onizations. For Porzana palmeri the Kimura 2- ioral diversity (Rothschild 1893-1900, Perkins parameter corrected distance (Kimurd 1980; dis- 1903, Amadon 1950, Baldwin 1953, Raikow tance and SE calculated in MEGA, Kumar et al. 1977, Pratt 1979, Freed et al. 1987a, James and 1993) for 197 base pairs (bp) of ATPase8 was Olson 1991). However, major adaptive shifts ap- 2.1 -+ 1.1% distant from its closest non-Hawai- pear to have modified many characters tradition- ian Porzana relative. For P. sandwichensis the ally used for phylogenetic reconstruction, while ATPase8 Kimura 2-parameter corrected distance others less subject to selection have been con- was 5.9 + 1.8% to its closest non-Hawaiian Por- served and provide little or no phylogenetic in- zana relative. Molecular analyses of Porzuna formation. The somewhat chimeric associations taxa known only from subfossil remains are un- of morphological traits in the group have even derway. 56 STUDIES IN AVIAN BIOLOGY NO. 22

a. Hawaiian Hawk, 13.solitaries Old World Ibises (9 spp.)

Swainson’sHawk, B swamsoni (Pegad’ s)’ Short-talled Hawk, 6 brachyurus

Wh,te ,b,s (Eudoc;;mus) Red-talled Hawk, 6. /ama!censis Lanai Flightless

New World lblses (5 spp ) “Woodland” Buteo

L Other ButeoiParabuteo L Bufeogallus?Inbmga _ Drepanidines (18 spp.) I Y. .. d. ~____--____--______--__~ White-browed‘ koseffnch (Carpodacus Ihum) Townsends’ Solltaire : c - _ _.__.. _ _ __.._.-._ _._..____-.-House Finch._... (C_.-._ ~nex~canusj .,_.._.._._..-.._._-._._ Carduelini _._ Green Honeycreeper (Chlorophanes spm) RufLthroated Solltalre : Scarlet-rumped Tanager Summer Tanager (~iranga rubra) (Ram~hoce’“s passerrr”~ : Swa~nsons’ Thrush Prothonotary Warbler (Protoi?oinnacilrea) Emberizinae Carharus : _ _._ _._._._._._Brown-headed _ _ _CowbIrd _._ _._._._._ (Mololhrusaler) _._._ _ _ _._._._._._._._._ Buff-barred Warbler (Phyiloscopus pum~i)

FIGURE 2. Abbreviated phylogenetic reconstructions for six Hawaiian taxa. a. Summarized maximum parsi- mony tree based on 407 nucleotide sites of 12s ribosomal RNA (A. Cooper, S. Olson, H. James, R. Fleischer, unpubl. data). b. Summarized parsimony phylogram based on preliminary analysis of over 1500 bp of mtDNA sequence (ATPaseB, ND2, cyt b, and COI) in Buteo and related taxa (R. Fleischer, I? Cordero, C. McIntosh, I. Jones, and A. Helbig, unpublished). c. Summary of relationships of outgroups and drepanidines based on par- simony analysis of 675 bp of cyt h sequence. d. Parsimony phylogram constructed from 700 bp of cyt b sequence from two Myuclesfesand three Catharustaxa with bma‘ o‘ and Turdus outgroup. e. Parsimony tree of two moa- nalo genera and a wide sampling of other waterfowl taxa showing two moa-nalo genera to be sister taxa and related to dabbling ducks. Tree is summarized from Sorenson et al. (1999), and based on over 1200 bp of mtDNA sequence. f. Parsimony phylogram showing summary of jay relationships to Corvus and a sampling of Corvus taxa based on 1008 bp of cyt b. The Alaki‘ is most closely related to the Common Raven.

MINOR RADIATIONS Most of the morphological and other evidence (e.g., Kepler and Kepler 1983) clearly favors Seven other Hawaiian avian groups have un- placement of thrushes in Myadestes (Pratt 1982). dergone what appear to be minor radiations, We analyzed variation in about 700 bp of the each with fewer than six species (Table 1). cyt b gene of mtDNA (C. McIntosh and R. These include thrushes ( Myadestes), hon- Fleischer, unpubl. data), for the Hawaii‘ Thrush eyeaters (genera Moho and Chaetoptilu), a lin- (or Oma‘ o,‘ M. obscurus), three Cutharus, two eage of owls (genus Grallistrix), several crows American Myadestes and a Turdus species, (genus Corvus), flightless ibises (genus Apteri- along with outgroup taxa. The resulting trees bis), and two waterfowl (Anatidae) lineages: clearly place the Oma‘ o‘ within the Myadestes true geese (genus Brunta) and the highly modi- clade, regardless of the tree building algorithm fied dabbling duck relatives called “moa-nalos” (i.e., maximum parsimony, Fig. 2d; maximum (genera Chelychelynechen, Ptaiochen, and likelihood or minimum evolution). We could not Thambetochen). resolve with certainty using this data set whether The five species of thrushes were placed orig- the Gma‘ o‘ is more closely related to M. geni- inally in their own genus, Phaeornis, but were barbis, a Caribbean solitaire, or M. townsendi of considered aligned with solitaires (Myadestes; western North America. The Kimura_2-parame- Stejneger 1887, Amadon 1950), robins (Turdus) ter corrected distance between the Oma‘ o‘ and or nightingale-thrushes (Cutharus; Ripley 1962). the solitaires is 6.7% for the 700 bp. HAWAIIAN BIRD MOLECULAR SYSTEMATICS-Fleischer and McIntosh 57

The meliphagid genera Chaetoptila (Kioea; 2 DNA analyses have yet been made on this spp.) and Moho (the 0‘ 6s;‘ 4 spp.) may repre- group, it appears likely that they represent the sent independent colonizations from south Pa- results of a single colonization and subsequent cific meliphagids (Perkins 1903), although Mayr minor radiation. (1943) considers both genera derived from a sin- At least four lineages of waterfowl have col- gle colonist. One species of Moho occurs on onized the Hawaiian Islands. Of these, only two, each of Kauai,‘ ,‘ Maui Nui (Maui, Lanai,‘ the moa-nalos (Olson and James 199 1, Sorenson ,‘ and Kahoolawe),‘ and Hawaii,‘ and et al. 1999) and the modern geese (Branta; Ol- this well-differentiated lineage (Pratt 1979) may son and James 1991, Paxinos 1998; E. Paxinos provide an opportunity to estimate a rate cali- et al. unpubl. data), have speciated beyond a sin- bration. The closest sister groups for the Ha- gle endemic species. All of the moa-nalos waiian meliphagids are unknown, with some au- evolved to very large size, flightlessness, and thors suggesting Gymrzomyza of Fiji and Samoa highly modified cranial morphology. They have (e.g., Mayr 1943) and others favoring Foulehaio become convergent in morphology to ratites in of Samoa or the New Zealand tuis’ (Prosthe- terms of postcranial morphology, and one spe- madera; e.g., Munro 1944, Pratt 1979). Molec- cies in particular has converged to tortoise-like ular studies are underway to address the origin cranial morphology. Like the moas of New Zea- and monophyly of the Hawaiian forms and the land (Darwin 1859), the moa-nalos occupied a possibility of a rate calibration from the four grazing mammal or tortoise niche (Olson and Moho species. A calibration could be used to James 1991). One genus and species (Chelyche- estimate the date of separation from the most lynechen quassus, the Turtlejawed Goose) is re- recent common ancestor. This date is important stricted to Kauai‘ and one (Ptaiochen) to Maui, because we estimate from our drepanidine cali- but Thambetochen is found on both Maui Nui brations that nectarivorous drepanidines evolved and Oahu,‘ suggesting the genus may have orig- only 2-3 Ma ago, while Givnish et al. (1995) inated on Oahu‘ and later walked across the Pen- used a calibration of chloroplast DNA restriction guin Bank land bridge (Fig. 1) to Molokai.‘ No fragment variation to estimate that bird-pollinat- moa-nalo is known from the young island of Ha- ed flowering lobeliads (genus Cyanea) evolved waii‘ (but see below). 8-17 Ma ago. Thus it is highly unlikely that dre- Olson and James (1991) suggested that the panidines “coevolved” with these plants in the moa-nalos were related to either dabbling ducks islands (as was suggested by Givnish et al. or shelducks (tadornines) on the basis of skeletal 1995). The meliphagids are the only other characters, primarily the presence and shape of known native, obligate nectarivores in the is- their syringeal bullae. Livezey (1996) tentatively lands and, if they are older, could be the coe- concluded from a cladistic analysis of morphol- volved taxon. ogy that the moa-nalos were sister to a “true” At least four crows (Corvus) occurred in the geese and swan clade, and not to anatids. Mi- islands (James and Olson 1991; H. James, pers. tochrondrial DNA analyses for two of the three comm.). Three of these are known only from genera (Thambetochen and Ptaiochen; Sorenson subfossils; two of which have been described et al. 1999) have provided a phylogenetic hy- and the fourth is the highly endangered Hawai- pothesis and estimates of minimum genetic di- ian Crow (Corvus hawaiiensis), hereafter re- vergence from anatid outgroups. The two genera ferred to as Alala.‘ It is unclear at present wheth- form a well-supported clade that is itself sister er these represent a single colonization and sub- to the “dabbling” ducks, although perhaps sequent radiation, or multiple colonizations by somewhat more similar to several South Amer- the same or different ancestral taxa (James and ican Anas or Anas relatives than to North Amer- Olson 1991). Preliminary phylogenetic analyses ican dabblers (Fig. 2e). Molecular data do not of the Alala‘ and seven other Corvus taxa indi- support a close relationship with either tadorni- cate that it is more closely related to the Com- nes or true geese. The distance between the mon Raven (Cowus corax) than to more typical moa-nalos and their closest anatid outgroup, crows, including two South Pacific island crows based on 1,009 mtDNA sites, is 6.9 -t 0.5%. (R. Fleischer and C. McIntosh, unpubl. data; The N&e or Hawaiian Goose (B. sandvicen- Fig. 2f). The Kimura 2-parameter corrected se- sis) is the only extant representative of what ap- quence divergence for 1,008 bp of cyt b between pears to be a minor radiation of Branta in the Alala‘ and North American Common Raven is islands (Olson and James 1991, Paxinos 1998; about 8.4 + 1.0%. E. Paxinos et al., unpubl data.). N&e are clearly Subfossil bones and owl pellets are all that derived from Canada Geese (B. canadensis; remain of four species of long-legged owls Quinn et al. 1991), and distances based on (Grallistrix) that apparently were morphologi- mtDNA restriction fragment and cyt b sequence cally adapted to feeding on birds. While no data suggest that the two taxa shared a common 58 STUDIES IN AVIAN BIOLOGY NO. 22 ancestor sometime within the past 1 Ma (Quinn Swainsons’ Hawk (Buteo swainsoni; as suggest- et al. 1991). At least two, and probably more ed by Mayr 1943), and the endemic Galapagos than three additional Branta species existed in Hawk (Buteo galapagoensis). The 10‘ does not the islands (Olson and James 1991, Paxinos have a close relationship with any Old World 1998; E. Paxinos et al., unpubl. data). One of Buteo we assessed. The Kimura 2-parameter these, the “very large Hawaii‘ goose” is the (Kimura 1980) corrected sequence divergence largest land vertebrate known from Hawaii‘ and from Buteo brachyurus is only 1.4 ? 0.8% for is restricted in distribution to the island of Ha- part of cyt b. We have no molecular data for the waii‘ (Giffin 1993). The species is highly mod- extinct and highly modified Circus. ified morphologically with a massive body, The Laysan Duck (Anas Zaysanensis) is a rel- short, stout wings (it was flightless, but may atively differentiated, small duck whose very have used its wings for fighting; S. Olson, pers. small and vulnerable wild population inhabits comm.); and cranially quite similar to the moa- only the tiny leeward island of Laysan. It has nalos. In fact, it appears to be a superb example been consistently classified as either a subspe- of convergent evolution to the moa-nalos. Mi- cies of the Hawaiian Duck (Anas wyvilliana), tochrondrial DNA sequence analyses (Paxinos hereafter referred to as Koloa, or of the Mallard 1998) strongly support placement of the very (Anas platyrhynchos) on the basis of morphol- large Hawaiian goose Branta and also indicate ogy and allozyme data (see Amadon 1950, Liv- a sister taxon relationship with the Nene and its ezey 1991, Browne et al. 1993). Recent DNA close, larger relative, B. hylobadistes. analyses (Cooper et al. 1996; J. Rhymer, unpubl. Two species of ibis (Apteribis) have been de- data), however, have strongly countered the scribed from subfossil material (Olson and Wet- above scenarios, indicating instead that the Lay- more 1976, Olson and James 1991). Apteribis san Duck is differentiated from the Koloa and had stouter legs and shorter wings than other Mallard and may be more closely aligned with ibises and were flightless. The two or more spe- the South Pacific Black Duck (Anas supercilio- cies were limited to Maui Nui, and the discon- sa) clade. The Koloa, on the other hand, does nection of Maui, Lanai,‘ and Molokai‘ 0.3-0.4 cluster closely with the North American Mallard Ma ago may have initiated the speciation or Mottled Duck (Anas fulvigula) clades. Anal- event(s). Analyses of mitochondrial 12s ribo- yses of mitochondrial control region sequences somal DNA sequences of Apteribis and 21 other of subfossil bones (Cooper et al. 1996) have also ibis species (Fig. 2a; A. Cooper, S. Olson, H. revealed that the Laysan Duck occurred in the James and R. Fleischer, unpubl. data) indicate main Hawaiian Islands well into the period of that the closest sister taxon to Apteribis is the Polynesian settlement, and in forested habitats New World White Ibis ( albus). The and higher elevations (> 1,500 m) not considered Kimura 2-parameter pairwise distance between typical for a dabbling duck. The level of mito- the two taxa for 407 bp of 12s rRNA sequence chondrial control region sequence divergence is 3.2 +- 1 .O%. between the Laysan Duck and its closest out- group taxon is about 10%; overall mtDNA di- SINGLE DIFFERENTIATED SPECIES vergence is lower than this (J. Rhymer, unpubl. Two raptors, a duck, and two rep- data). resent single differentiated species. These taxa The fourth “nonradiating” species, the Ele-‘ apparently colonized the islands and differenti- paio (Chasiempis sandwichensis), is polytypic at ated considerably from their ancestors but did the subspecies level and occurs on the islands of not undergo subsequent speciation. The two rap- Kauai,‘ Oahu,‘ and Hawaii‘ (enigmatically, no tors are the endangered Hawaiian Hawk or 10‘ have been found of this species on Maui (Buteo solitarius) and an extinct accipiter-like Nui; James and Olson 1991). The Elepaio‘ is harrier (Circus dossenus). The IO‘ is currently likely related to Polynesian flycatchers in the ge- restricted to the island of Hawaii‘ but has been nus Monarcha (Mayr 1943, Amadon 1950) and found in fossil form on other islands (Olson and is one of the few species for which differentiated James 1991; S. Olson, pers. comm.). Like many subspecies have been identified on a single small other species of Buteo, the 10‘ exhibits a light island (Hawaii;‘ Pratt 1980). Molecular analyses and a dark color morph. Preliminary phyloge- of each island subspecies may, however, reveal netic analyses of more than 1,500 bp of mtDNA differentiation sufficient to elevate them to spe- sequence in 18 species of Buteo (R. Fleischer, P cies level. Cordero, C. McIntosh, I. Jones, and A. Helbig, unpubl. data) provides weak support for a clade PROBABLE RECENT COLONIZATIONS containing the 10,‘ the North American Short- Several taxa show little phenotypic diver- tailed Hawk (Buteo brachyurus; to which it is gence from mainland outgroups, suggestive of a least divergent; Fig. 2b), the North American very recent colonization (Table 1). These in- HAWAIIAN BIRD MOLECULAR SYSTEMATICS--Fleischeu and Mclntnsh 59 elude the Black-necked Stilt (Hirnantopus mex- divergence between the Hawaiian taxa and their icanus knudseni), Hawaiian Coot (Fulica alai), closest non-Hawaiian (mostly mainland) rela- Common Moorhen (Gallinula chloropus sand- tives (i.e., from zero to 10.3% sequence diver- vicensis), Koloa, Black-crowned Night Heron gence for 14 lineages). Based on these results, (Nycticorax nycticorax hoactli), an eagle (Hal- none of these Hawaiian lineages split from iaeetus), and the Short-eared Owl. Of these, only mainland ancestors earlier than about 6.4 Ma. In the Black-crowned Night Heron is not currently fact, most of our estimates, although rough and considered to be distinct from mainland forms lacking meaningful standard errors, fall well at the subspecies or species levels, but the Short- within the period of formation of the current set eared Owl, in spite of its subspecific designation, of main islands (i.e., Kaua‘i at 5.1 Ma and later, is thought to be a post-Polynesian colonist (Ol- Fig. 1). Only the drepanidines (10.3%) the son and James 1991). corvids (8.4%), and perhaps the moa-nalos The Common Moorhen, Hawaiian Coot, (6.9%) and the thrushes (6.7%) have Kimura 2- Black-crowned Night Heron, and Short-eared parameter sequence divergences from mainland Owl are extremely similar morphologically to relatives that suggest colonization prior to even outgroup relatives (Amadon 1950), but no DNA the formation of O‘ahu (3.7 Ma), and in each of data currently exist with which to assess the age these cases we may not have obtained sequence of their splits. As noted above, the Koloa is a for the closest mainland outgroup (which we very close relative of the Mottled Duck and may not have sampled or it might be extinct). Mallard (<3% mitochondrial control region di- The overall picture suggests that while native vergence; Cooper et al. 1996). The endemic sub- Hawaiian Drosophila (Beverley and Wilson species of the Black-necked Stilt differs from 1985, Thomas and Hunt 1991, DeSalle 1992, North American Black-necked Stilts (H. m. mex- Russo et al. 1995) and lobeliads (Givnish et al. icanus) by only about 1.5 + 0.6% sequence di- 1995) may have colonized the archipelago well vergence in 447 bp of mtDNA control region (R. before the formation of Kaua‘i, thus far we have Fleischer et al., unpubl. data). The North Amer- little evidence that any bird lineages have done ican Black-necked Stilts are considered to be the so. closest mainland relatives on the basis of mor- These findings lead us to consider factors be- phology. Cyt b and 12s rRNA sequences from yond simple isolation by distance and the an- a subfossil bone of the extinct eagle (Haliaeetus thropogenically induced Holocene extinction sp.; Fleischer et al. 2000) are not different from that may help to explain Hawai‘i’s low primary the Old World White-tailed Eagle (H. alhicilla), avian diversity. First, the unique geology of the and the two species differ by 1.5% for the ATP- islands (Carson and Clague 1995) results in a ase8 gene. Skeletal characteristics could not dif- situation in which individual islands have a lim- ferentiate the Hawaiian eagle bones from either ited “lifespan” (-5-7 Ma) as a high island. Lin- White-tailed Eagle or (H. leucoce- eages that have colonized older islands, but for phalus; Olson and James 1991). Thus, for at some reason cannot succeed onto younger is- least three of these seven taxa the supposition of lands, will be ultimately lost as their island dis- a recent split from a mainland ancestor and re- appears into the sea (this may be especially true cent arrival in the islands is supported by the for forms that have evolved to be flightless). molecular data. There may be reduced chance for taxonomic di- SUMMARY: GEOGRAPHIC ORIGINS AND versity to build up over long evolutionary peri- TEMPORAL FRAMEWORK ods relative to archipelagos with longer surviv- Above we summarize recent molecular sys- ing islands. Secondarily, what secondary enrich- tematic studies of the Hawaiian avifauna. We ment of avifaunal lineages by speciation that use these data to infer, if possible, the closest does occur in the islands may allow “niches” to living relatives and the geographic origins of the be filled (perhaps by now locally adapted taxa) Hawaiian taxa we sampled. Our biogeographic such that they are no longer available for occu- analyses indicate (Table I) that at least 9 or 10 pation by new (and not locally adapted) colo- of the 2 21 independent lineages appear to be nists from elsewhere. Thus, primary diversity of North American or at least Western Hemi- could be reduced by competitive exclusion. sphere origin, 4 appear to be of South Pacific or Continued paleontological research in the is- Australasian origin, 2 or 3 are of Asian origin, lands combined with studies of DNA sequence and 5 are of currently unknown geographic or- variation should help us to address these hy- igin. Thus Mayr’s (1943) conclusion that about potheses. We hope these new fossils and se- half the Hawaiian avifauna is of American origin quences will continue to shed light on the sys- is still supported by our molecular data. tematics, biogeography, and timescale of avian We found a relatively low level of molecular evolution on the Hawaiian conveyor belt. 60 STUDIES IN AVIAN BIOLOGY NO. 22

ACKNOWLEDGMENTS (Academy of Natural Sciences-Philadelphia), P Bruner (Brigham Young University-Hawai‘i), E. Bermingham We thank C. Tan; S. Olson, H. James, E. Paxinos, (Smithsonian Tropical Research Institute), S. Olson, l? A. Cooper, B. Slikas, J. Rhymer, B. Arbogast, S. Co- Angle, and M. Braun (U.S. National Museum), R. nant, M. Sorenson, T. Quinn, A. Helbig, I. Jones, A. Cann (University of Hawai‘i), S. Rowher and S. Ed- Driskell, and T. Pratt for information concerning and wards (Burke Museum), and C. Cicero (Museum of discussion of many of the topics covered in this paper, Vertebrate Zoology-Berkeley). We greatly appreciate and C. Tan; B. Slikas, J. M. Scott, reviewer #l, and permission from Dave Swofford to use PAUP* (a gem especially H. James for comments on an earlier draft of a program). Funding for many of the results pre- of the manuscript. Samples for many of our analyses sented above was provided by the Smithsonian Insti- were provided by museum or field collections of tis- tution Scholarly Studies Program, Friends of the Na- sues and we gratefully acknowledge the cooperation of tional Zoo, U.S. National Science Foundation, U.S. C. Kishinami and A. Allison (B. P Bishop Museum), Fish and Wildlife Service, the National Geographic S. Conant (University of Hawai‘i), E Sheldon (Loui- Society, and the Biological Resources Division of the siana State University), M. Robbins and B. Slikas U.S. Geological Survey.