Changes in a West Indian bird community since the late Pleistocene

To establish a chronology for late Quaternary avian extinction, extirpation and persistence in the Bahamas, thereby testing the relative roles of climate change and human impact as causes of extinction.


INTRODUCTION
During the late Quaternary, the most dramatic biotic phenomenon on the American continents was the extinction of most species of large terrestrial mammals (megafauna), North and South America, the causes and consequences of continental megafaunal collapse continue to be debated (Barnosky et al., 2004;Gill et al., 2009).
In the West Indies, on the other hand, insular megafauna, such as sloths, monkeys and large rodents, survived until the mid-Holocene or even later, which corresponds to the first human arrival in the Greater Antilles, but was characterized by changes in climate, sea level and island area that were much less drastic than those during the PHT (Steadman et al., 2005;MacPhee, 2008). On Great Abaco Island (Abaco), located in the Bahamas north of the Greater Antilles, the largest terrestrial herbivore and carnivore at first human contact were both reptiles, namely the tortoise Chelonoidis alburyorum and the Cuban crocodile, Crocodylus rhombifer (Steadman et al., 2007;Franz & Franz, 2009;Morgan & Albury, 2013). Both of these species perished on Abaco shortly after human colonization c. 1000 years ago (Hastings et al., 2014;Steadman et al., 2014).
Birds are the most diverse group of terrestrial vertebrates in the Bahamas and across the West Indies. As with mammals and reptiles, West Indian resident birds have undergone considerable late Quaternary losses of populations (extirpation) and species (extinction). Many of these losses have been documented through bones identified in cultural (archaeological) sites (e.g. Steadman et al., 1984a;Carlson & Steadman, 2009), thereby demonstrating that the population or species survived into the period of human occupation (late Holocene). Other avian losses are based on bones identified from non-cultural (palaeontological) sites, the majority of which have not been radiocarbon-dated and their ages are therefore unknown beyond being generally 'late Quaternary' (e.g. Olson & Steadman, 1977;Su arez & Olson, 2001).
This study is based on remarkably well-preserved vertebrate fossils from Abaco in the northern Bahamas. To examine the relative rates of natural and anthropogenic faunal change, we establish a chronology for avian extinction, extirpation and persistence on Abaco, an endeavour not previously pursued for any West Indian island except Hispaniola (Steadman & Takano, 2013). The chronology allows us to test the role of climate change (and associated changes in sea level and island area) by analysing the extent to which species of Bahamian birds survived the PHT. We compare PHT patterns of extinction, extirpation and persistence with independently established estimates of climate and sea level for glacial (late Pleistocene) and interglacial (Holocene) times. We then examine the fates of the species that survived the substantial changes in climate, habitat, sea level and island area during the PHT, but were subjected to human activities (habitat loss, hunting, etc.) around 1 ka.

Site setting
The Bahamian Archipelago consists of islands that lie just off the south-eastern coast of Florida and the northern coasts of Cuba and Hispaniola (Fig. 1). The archipelago comprises 23 major islands (> 50 km 2 ) and many smaller ones, lying on shallow carbonate banks separated by deep water. All the exposed bedrock in this tectonically stable archipelago is Quaternary aeolianite and shallow marine limestone, with the major build-up of the islands taking place as late as Marine Isotope Stage 5e, 125 ka (Carew & Mylroie, 1995;Hearty et al., 1998;Hearty & Kaufman, 2000). The archipelago stretches 980 km from c. 27°N, 79°W in the northwest to c. 21°N, 71°W in the south-east. The island group comprises the independent Commonwealth of the Bahamas and the Turks and Caicos Islands, a British protectorate. Our study focuses on Great Abaco Island (hereafter, 'Abaco'), the third-largest island in the group today (1214 km 2 ). All Bahamian islands are low-lying, with none exceeding 70 m elevation today, and most of the land is below 10 m elevation. Although lying on the North American plate, no Bahamian island was ever connected to mainland North America or any other continent.
The fossils we studied are from three types of site on Abaco ( Fig. 1)flooded caves known as blue holes (Sawmill Sink and Ralph's Cave), inorganic sediments in dry caves (Holein-the-Wall Cave) and intertidal peat deposits (Gilpin Point). Of these sites (described in detail in Appendix S1 in the Supporting Information), Sawmill Sink and Hole-in-the-Wall Cave yielded the greatest number of bird fossils. Sawmill Sink (26.21767°N, 77.21052°W, WGS84 datum) is an inland blue hole with Holocene vertebrate fossils from a peat deposit, as well as late Pleistocene vertebrate fossils from inorganic sediments (Steadman et al., 2007;Franz & Franz, 2009;Morgan & Albury, 2013). Ralph's Cave (26.24997°N, 77.19004°W) is an inland blue hole where vertebrate fossils occur in a peat deposit. Hole-in-the-Wall Cave (25.86021°N, 77.18369°W) yielded fossils from dry, unstratified, inorganic sediments excavated by N. A. Albury and G. S. Morgan in January 1989. The Gilpin Point site (26.10457°N, 77.17767°W) is a bonerich intertidal peat deposit on the windward (Atlantic Ocean) coast of southern Abaco .
Following Cohen et al. (2013), we set the age of the Pleistocene-Holocene boundary at 11.7 ka. This age is near the mid-point of the PHT, when sea level rose from c. À100 m to c. À25 m (Clark & Mix, 2000;Bard et al., 2010). Based on these reference points, sea level was ≥ 35 m lower than present at ≥ 11 ka. This is the minimum age for the Pleistocene 'owl roost' bone deposits in Sawmill Sink, which occur on ledges 27-35 m below modern sea level. These deposits represent prey remains of barn owls that roosted in this currently flooded sinkhole when it was above sea level and dry. We used a global topographic database that includes seafloor topography (Smith & Sandwell, 1997) to estimate that, at that time, Abaco and Grand Bahama (on the Little Bahama Bank) were joined as a single 'superisland' with an area of c. 14,250 km 2 (13,880-14,550 km 2 at À35 to À45 m sea level), which is around six times the current land area of all the islands on this bank combined (2320 km 2 ). This latest Pleistocene land area was only slightly smaller than the largest estimated extent of land on the Little Bahama Bank (16,492-17,004 km 2 ), which occurred 10 kyr earlier, during the Last Glacial Maximum (LGM: 19-23 ka; Fig. 2, Appendix S2).

Comparative osteology
Many of the prehistoric Bahamian bones retain every detail of the birds' original anatomy, and much of their original chemical composition. There is no clear-cut criterion as to when a late Quaternary bone should be called a fossil, given that the post-mortem period extends from decades into centuries, millennia and so on, with highly variable rates of mineralization, organic preservation and physical degradation.
We therefore use 'fossil' and 'bone' interchangeably because any distinction between these two words, whether based on chronology or organic preservation, would be arbitrary.
The avian fossils from Abaco are housed in Marsh Harbour, Abaco, at the scientific collections of the National Museum of the Bahamas (NMB), and in the Florida Museum of Natural History, University of Florida (UF; Gainesville, FL, USA). Identifications were based on direct comparisons with modern skeletal specimens in the UF Ornithology Collection, which includes every species of bird that currently lives in the Bahamas. Additional comparisons were made with avian fossils from various sites in the Bahamas and Hispaniola in the UF Vertebrate Paleontology Collection.  For some species, the taxonomic resolution of our identifications is conservative because of limited fossil material and/ or limited modern skeletal specimens from Cuba and Hispaniola. For this reason, we classify Glaucidium sp., Colaptes sp., Myadestes sp., Xenoligea sp. and Loxia sp. as extirpated rather than extinct species. In each case, an endemic congener lives today on Cuba or Hispaniola, but we cannot yet resolve whether the Bahamian fossils are conspecific or represent an extinct species that may have been endemic to the Bahamas. As fossil and modern collections grow, we hope to resolve these issues.

Radiocarbon dating
Until now, the radiocarbon ( 14 C) chronology for late Quaternary vertebrates from Abaco had been based on accelerator mass spectrometer (AMS) 14 C dating of ultrapurified collagen from individual fossils of an extinct tortoise, Chelonoidis alburyorum, and the extirpated Cuban crocodile, Crocodylus rhombifer (Steadman et al., 2007. AMS 14 C dating of bird fossils is challenging because only minute amounts of purified collagen can be extracted even from well-preserved bird bones, which are typically much smaller than those of tortoises and crocodiles. To build a chronology for the prehistoric bird communities, we compiled all AMS 14 C age determinations performed thus far on fossils from Abaco, including reptiles, birds and mammals. These datings were carried out at Beta Analytic (Miami, FL, USA; for laboratory and calibration methods, see http://www.radiocarbon.com/). The 14 C dates are presented as calendar years before present (cal. yr bp). In the text, we use 'ka' (thousand years ago) for general chronological values, whether or not based on 14 C dating (for example, 4.5 ka = 4500 cal. yr bp).

Species climate envelopes
For six extirpated species of bird identified from Pleistocene but not Holocene deposits, and that are widely distributed today in the Western Hemisphere (not restricted to the Greater Antilles), we examined the range of climate conditions found in their current distribution (their 'climate envelope'; Hijmans & Graham, 2006) as well as their current habitat associations. We compared those ranges with the climate conditions found today and those modelled for the LGM in the Bahamas to learn whether these species are associated with different climates from those now found on Bahamian islands, but perhaps similar to those found there during the late Pleistocene. Species locality data came from the North American Breeding Bird Survey 1966-2010 annual surveys (BBS; http:// www.pwrc.usgs.gov/bbs; accessed on 2 September 2013) (Sauer et al., 2011) and, for species whose ranges extend well into Mexico and regions further south, we included data from the Global Biodiversity Information Facility (GBIF, http:// www.gbif.org/; accessed 10 October 2013). BBS survey routes are c. 40 km long. GBIF occurrence records are compiled from a variety of sources, primarily museum collections.
LGM climate was reconstructed using global climate model outputs from the Paleoclimate Modelling Intercomparison Project Phase II (http://pmip2.lsce.ipsl.fr/) and statistically down-scaled using the WorldClim current climate data. To account for uncertainty in the modelled palaeoclimate, we used both available models of LGM climate, the Community Climate System Model (CCSM) and the Model for Interdisciplinary Research on Climate (MIROC). We examined the bioclimatic ('BIOCLIM') variables available from WorldClim, because these seasonal summaries of precipitation and temperature means and extremes are frequently used to analyse species' ranges (Beaumont et al., 2005;Franklin, 2010;Booth et al., 2014). Climate data interpolated at this scale are strongly spatially autocorrelated (Franklin et al., 2001), which helps to obviate the effect of the coarse resolution of the BBS data and the locational uncertainty of GBIF data in our analyses, the aim of which is to broadly circumscribe the climatic conditions within each species' range.
Values for all BIOCLIM variables were extracted for each location of the species' current distributions. To characterize the climate of the northern and central Bahamas, we extracted climate data at each grid point for the BIOCLIM grid cells intersecting the land area of the Little Bahama Bank and Great Bahama Bank. The frequency distributions of climatic variables were examined to compare the conditions where these species occur today with the conditions in the Bahamas today and in the Pleistocene (LGM). We expected average minimum temperature and precipitation to be the climatic factors that most limited the distributions of habitats preferred by these species. We tested for multivariate differences among climate envelopes of each species and the climate of current and LGM Bahamas using the multiresponse permutation procedure (MRPP) applied to eight largely uncorrelated (r < 0.7) BIOCLIM variables (annual mean temperature, mean diurnal temperature range, temperature seasonality, maximum temperature of the warmest month, minimum temperature of the coldest month, annual mean precipitation, precipitation of the driest month and precipitation of the coldest quarter).
The northern Bahamian annual mean and average maximum temperatures were estimated from the WorldClim datasets to be c. 1-2°C lower at LGM than present (Table 1), whereas the average minimum temperatures were c. 2-3°C lower, and annual precipitation was c. 20-25% lower. Although these changes are substantial, they still reflect a subtropical climate that was likely to be frost-free. These modelled reconstructions of palaeoclimate portray the LGM Bahamas as cooler and drier than now, but did not show temperature declines as large as the 3-4°C estimated for the southern Caribbean (Lin et al., 1997;Lea et al., 2003) or the 5-6.4°C estimated for the continental tropics (Porter, 2001). Other studies comparing these same climate reconstructions to distributions of fossil vertebrate taxa have noted that inconsistencies suggest potential inaccuracies in the modelled palaeoclimatic variables (McGuire & Davis, 2013).

RESULTS
We submitted 11 individual avian fossils from Abaco for AMS radiocarbon ( 14 C) dating, as well as 20 samples of associated, more readily dated, plant material (wood), reptile bones and mammal bones (Table 2). Of the seven avian fossils submitted from the owl roost deposits in Sawmill Sink (representing the extinct Small Abaco flightless rail, the extirpated burrowing owl and the extirpated eastern meadowlark; Table 3), none retained enough collagen to yield a 14 C date. Based on independently derived sea-level data, the owl roost deposits must be older than 11 ka (see Appendix S2). From the peat deposit at Sawmill Sink, the extirpated white ibis and the extinct Creighton's caracara (Fig. 3) both have direct late Holocene AMS 14 C dates, whereas three extirpated species (Cooper's/Gundlach's hawk, scaly-naped pigeon and burrowing owl) are associated with late Holocene AMS 14 C dates on reptiles. From Hole-inthe-Wall Cave, a direct late Holocene AMS 14 C date was determined on the extirpated burrowing owl, whereas a sandhill crane bone did not have enough collagen. Seven other extirpated species also occur in at Hole-in-the-Wall Cave (Audubon's shearwater, Swainson's hawk, purple gallinule, doublestriped thick-knee, pygmy-owl, pearly-eyed thrasher and grasshopper sparrow). From the Gilpin Point site, the two extirpated species of birds (Bermuda petrel, Cuban crow) are associated with five late Holocene AMS 14 C dates (on wood or reptile bones) that cluster around 920 to 900 cal. yr bp. In total, nine extinct or extirpated species from Abaco are associated with cultural deposits (Table 3), thus demonstrating their presence into the time of human occupation.
From Pleistocene contexts (the owl roost in Sawmill Sink), 25 of 45 species (21 extirpated, 4 extinct; total 56%) no longer reside on Abaco (Table 3). The 37 species of birds recorded as Holocene fossils feature 15 (14 extirpated, 1 extinct; total 41%) that are gone from Abaco. Nine of the total 30 extirpated or extinct species have been recovered in both Holocene and Pleistocene deposits, i.e. they survived the PHT only to be lost in the late Holocene. Incomplete and uneven sampling undoubtedly accounts for some of this small amount of overlap (Nowak et al., 2000); our Holocene sample (280 identified fossils) is much smaller than that from the Pleistocene (1803 identified fossils). Nevertheless, the evidence suggests two major times of avian extirpation and extinctionthe Pleistocene-Holocene transition (15-9 ka) and the late Holocene (< 4 ka). From the Pleistocene sample, 96% of the individual fossils (1726 of 1803) represent extinct or extirpated species; this value for Holocene fossils is 54% (150 of 280).
Our data also document the persistence (Pleistocene fossil + Holocene fossil + extant on Abaco) of an ecologically diverse set of 12 species (Table 3) including aquatic/estuarine predators (night-herons, osprey), columbid granivores and frugivores (pigeons, doves), crepuscular/nocturnal predators (barn owl, nighthawk), and insectivorous or frugivorous songbirds (kingbird, yellowthroat, spindalis). This list will grow as more fossils are discovered and identified. Further evidence for incompleteness in the fossil data is that 12 species extant on Abaco were recorded as Pleistocene but not Holocene fossils.
For the 15 extinct or extirpated species of birds recorded only from Pleistocene contexts, the preferred habitats today for them or their nearest living relatives are pine woodlands, grasslands, or unknown (Table 3). In this group are five extirpated species, mainly continental today (brown-headed nuthatch, eastern bluebird, cliff swallow, short-eared owl and eastern meadowlark), that now occupy areas with lower average winter minimum temperatures and/or drier conditions or more open, grassy habitats than are found in the Bahamas now (Table 4, Fig. 4, Appendix S3). Their ranges may have shifted out of the Bahamas as the climate warmed and sea level rose after the last ice age, although this seems unlikely for the brown-headed nuthatch, which is still found on nearby Grand Bahama. Furthermore, the hindcast ranges of LGM minimum temperatures for the Bahamas were cooler but still more similar to those of the modern Bahamas than to those for any of these species (Fig. 4, Appendix S3). We note, however, that one of these species is a long-distance migrant today (the cliff swallow), so the minimum temperatures on its breeding grounds are irrelevant. The cave swallow is currently found in climatic conditions most similar to those found in the Bahamas today and during the LGM (Table 4, Fig. 4).

Biotic limitations of the Bahamas
For perspective, we note that the Bahamian islands sustain many fewer resident species of birds (and other vertebrates)  (Hearty & Kaufman, 2000). We note, nevertheless, that small oceanic islands can have considerable potential to sustain substantial bird communities, including islands two to three orders of magnitude smaller than Abaco (Franklin & Steadman, 2008). Perhaps a factor in the current faunal impoverishment of small Bahamian islands is their greater vulnerability to human-caused losses (see 4, below). 3. The lower elevations and minimal topographic relief of Bahamian islands, even during glacial intervals, means that these islands have had much less diversity in climate and habitats than large, high islands such as Cuba or Hispaniola. 4. The Bahamian islands may have undergone more drastic human impact in the late Holocene than the larger, higher and more rugged Greater Antillean islands because of the relative ease of access to all parts of the islands by people, a  trend seen on oceanic islands worldwide (Steadman, 2006, pp. 466-469).

Chronology of extinction
Without chronological data, we have no basis for understanding when species became extinct or extirpated. Across the West Indies, we are still in the early stages of developing precise chronologies for fossil sites, most of which remain undated. Even at the Banana Hole site (New Providence Island), a dry cave site that had yielded the largest sample of Bahamian bird fossils until our work on Abaco, scientists assumed that the fossils dated to the Pleistocene (Brodkorb, 1959;Olson & Hilgartner, 1982) although no radiocarbon dates had been determined. Having a chronology for the sites on Abaco, we can interpret the fossil assemblages in terms of independent estimates of climate, sea level and island area for glacial (late Pleistocene) versus interglacial (Holocene) times. Within the Holocene, we can also evaluate the fossil assemblages by whether they pre-date or post-date human arrival. Most of the extinct or extirpated species recorded on Abaco only from Pleistocene contexts are characteristic of open habitats, such as pine woodlands or grasslands. Furthermore, some of the extirpated species are found today where winters are cooler than in the modern or Pleistocene Bahamas. By contrast, the extinct or extirpated species recorded from Holocene contexts, regardless of their presence or absence as a Pleistocene fossil, are characteristic of a number of habitats, with no clear trend toward any one type. Because we have a much larger sample of fossils from Pleistocene than Holocene deposits on Abaco (1803 vs. 280 identified specimens as of 21 December 2013), a future priority will be to excavate more Holocene sites. We suspect that some of the Holocene absences will prove to be sampling artefacts; further work is likely to increase the number of documented Holocene losses.
None of the Bahamian or Cuban sites with fossils of Creighton's caracara (Brodkorb, 1959;Su arez & Olson, 2001 had been 14 C dated until now. The dated Sawmill Sink specimen shows that this extinct carrion-feeder lived until the late Holocene, and is likely to have died out when the Bahamas and Cuba lost their large reptiles and mammals. This is comparable to the losses of avian scavengers in North America when most large mammals went extinct at the end of the Pleistocene (Steadman & Martin, 1984). With large species, whether reptiles, birds or mammals, it is intuitively easier to propose that direct human impact accounts for their extinction. Smaller species tend to have larger populations, to be less conspicuous, and to provide less meat and other raw materials, making it less likely that direct human exploitation caused their extirpation or extinction. Our results show, however, that even small species on West Indian islands can be vulnerable to extirpation and extinction following human arrival, regardless of the precise cause. Among the many examples of this phenomenon are lizards from the Lesser Antilles (Pregill et al., 1994) and bats on Abaco (J. A. Soto-Centeno, American Museum of Natural History, New York, unpublished data).
That a number of extirpated or extinct species of birds survived on Abaco into the late Holocene is supported by the remains of such birds in Amerindian middens elsewhere in the archipelago, such as sites MC-6 (Jones O'Day, 2002) and MC-37 (Indian Cave; D.W.S., unpublished data) on Middle Caicos, and site GT-2 (Coralie) on Grand Turk (Carlson & Keegan, 2004). Late Holocene bones of extirpated or extinct birds also occur in prehistoric Amerindian middens elsewhere in the West Indies, such as Puerto Rico and the Virgin Islands (Wetmore, 1918;Carlson & Steadman, 2009), Antigua (Steadman et al., 1984b), and Montserrat (Steadman et al., 1984a). These species form a taxonomically diverse group of water birds and land birds. Large and sudden increases in rates of sedimentation, charcoal concentration, and pine pollen (pine woodland is a pyrogenic vegetation community; O'Brien et al., 2008) took place 1.2-0.8 ka in the Bahamas (Kjellmark, 1996;Slayton, 2010). This is when people arrived in the northern Bahamas (Steadman et al., 2007) and strongly suggests an increase in fire from human burning of the landscape. Similarly pronounced increases in charcoal have also been reported for earlier Holocene time intervals corresponding to human arrival in the Greater Antilles (Burney et al., 1994;Horn et al., 2000).
Other late Holocene changes in pollen spectra, charcoal influx or other sedimentary features from wetland cores in the Caribbean have been attributed to fluctuating wetter or drier conditions (Hodell et al., 1991;Higuera-Gundy et al., 1999;Lane et al., 2014). Although these changes are not nearly as dramatic as the increase in charcoal when people arrived, non-anthropogenic late Holocene changes in wetlands could account for the loss on Abaco of at least one species, the white ibis, which is known to have nomadic breeding colonies that are affected by weather patterns and wetland conditions (Frederick et al., 1996;Kushlan, 2010).

The long view
When sea level was À40 m, c. 10-12 ka, the two superislands of the Little Bahama Bank and Great Bahama Bank were separated by less than 34 km of ocean (Appendix S2). (The closest major islands on these separate banks are now c. 90 km apart.) In addition, the Great Bahama Bank superisland was only 20 km from Cuba at that time (now c. 180 km) and had a land area of 102,231-103,670 km 2 , nearly the same size as the largest West Indian island today  (Cuba, 111,463 km 2 ) and exceeding that of the next largest island (Hispaniola, 76,193 km 2 ). Thus, we must entertain the fact that the Bahamian Archipelago, although never higher than 200 m elevation even at the lowest late-Pleistocene sea levels, was essentially Greater Antillean in its land area, location and proximity to other islands during glacial times. Climate fluctuations on Hispaniola were small in the Holocene compared to those of the PHT (Hodell et al., 1991). Given the 20 or more glacial-interglacial cycles of the past 2 million years, perhaps it should not be surprising that most species of West Indian birds persisted on high islands during these drastic shifts in climate, sea level and land area. Nevertheless, extinctions and extirpations could have occurred on Bahamian banks, because the PHT's rising sea levels caused the superislands to be replaced by smaller and more isolated islands, or as a result of the associated change in climate and habitat (becoming somewhat warmer and wetter; Pregill & Olson, 1981). For any of the 16 species of bird that lack evidence of persistence into the Holocene on Abaco, we must entertain this possibility, although we have no real understanding of the lapse time to extinction from island shrinkage during the glacial-to-interglacial sea-level rise. Our analysis of climates and habitats associated with five of these species supports the notion that they would have found climatic conditions in the Bahamas somewhat more suitable in the LGM Pleistocene than currently, and/or that they prefer open habitats that probably were more widespread. The climate-based models do not consider plasticity in habitat preferences, or species interactions, which can be important in shaping distributions independent of climate (Cole, 1995;Stafford et al., 1999;Semken et al., 2010). Based on the evidence presented here, we can demonstrate that a taxonomically and ecologically diverse set of species did persist through the PHT, but did not survive the past millennium of human activities.

ACKNOWLEDGEMENTS
For help in the field or laboratory, we thank Nancy Albury, Bray B eltran, Richard Franz, Richard Hulbert, Brian Kakuk, Annual precipitation (mm) Figure 4 Box-and-whisker plots showing the distributions (median heavy line, 25%/ 75% quartiles edges of box, 5%/95% whiskers, and outliers as points) of average minimum temperature of the coldest period, average maximum temperature of the warmest period and annual mean precipitation based on modern climate for modern localities [from the North American Breeding Bird Survey 1966-2010 annual surveys (BBS) and the Global Biodiversity Information Facility (GBIF); see text] for six extant bird species identified on Great Abaco Island (Bahamas) from Pleistocene but not Holocene deposits and that no longer occur in the Bahamas (