A high-precision chronology for the rapid extinction of New Zealand moa (Aves, Dinornithiformes)

Quaternary Science Reviews 105 (2014) 126e135 Contents lists available at ScienceDirect Quaternary Science Reviews journal homepage: www.elsevier.com/locate/quascirev A high-precision chronology for the rapid extinction of New Zealand moa (Aves, Dinornithiformes) George L.W. Perry a, b, *, Andrew B. Wheeler a, Jamie R. Wood c, Janet M. Wilmshurst a, c a School of Environment, The University of Auckland, Private Bag 92019, Auckland 1142, New Zealand b School of Biological Sciences, The University of Auckland, Private Bag 92019, Auckland 1142, New Zealand c Landcare Research, PO Box 69040, Lincoln 7640, New Zealand a r t i c l e i n f o a b s t r a c t Article history: Megafaunal extinction followed the prehistoric human settlement of islands across the globe, but the exact Received 26 June 2014 duration and dynamics of the extinction processes are difficult to determine. The New Zealand moa (Aves, Received in revised form Dinornithiformes) are a prime example, where, despite an extensive fossil and archaeological record, 24 September 2014 debate continues about their extinction chronology and how extinction timings varied among regions and Accepted 26 September 2014 Available online species. We apply probabilistic sightings methods to 111 high-quality radiocarbon dates (from a pool of 653 dates) on moa remains from natural and archaeological sites to provide a high-resolution spatio-temporal chronology of moa extinction. We interpret this alongside an estimated time for the onset of hunting Keywords: Archaeology pressure, obtained by applying the same methods to the most reliable proxies for initial human settlement Holocene of New Zealand: coprolites of and seeds gnawed by the commensal Pacific rat (Rattus exulans). By comparing Human impacts local and national extinction times, we discriminate between the point at which hunting stopped (eco- Island faunas nomic extinction) and the total extinction of moa (ca 150 and 200 years after settlement, respectively). Prehistory Extinction occurred contemporaneously at sites separated by hundreds of kilometres. There was little Radiocarbon difference between the extinction times of the smallest (20e50 kg) and largest (200þ kg) moa species. Our Sightings models results demonstrate how rapidly megafauna were exterminated from even large, topographically- and ecologically-diverse islands such as New Zealand, and highlight the fragility of such ecosystems in the face of human impacts. © 2014 Elsevier Ltd. All rights reserved. 1. Introduction predation from introduced mammals) and large-bodied species (i.e. those targeted by hunters) were most susceptible to extinction The late Holocene settlement of oceanic islands resulted in (Duncan and Blackburn, 2004; Boyer, 2008). abrupt and dramatic ecological transformations (Rick et al., 2014). New Zealand represents the southernmost island group of east Nowhere is this better exemplified than on the recently settled Polynesia and was settled during the final phase of Polynesian islands of east Polynesia (Rolett and Diamond, 2004), where expansion in the late 13th century CE (Wilmshurst et al., 2008, hunting by humans, predation by the introduced commensal Pa- 2011), at a time of relative climatic stability (Wilmshurst et al., cific rat (Rattus exulans), and the destruction of large tracts of forest 2007), and probably with a founding population of 50e100 fe- by anthropogenic fire resulted in severe post-settlement faunal males (based on mtDNA diversity) (Murray-McIntosh et al., 1998). extinctions (Steadman, 1995). Across all east Polynesian islands, an As with other islands across east Polynesia, the settlement of New estimated at least 1000 species of non-passerine land birds suffered Zealand was accompanied by a rapid fire-driven reduction in forest extinction during the prehistoric period (Duncan et al., 2013), cover (McGlone and Wilmshurst, 1999; McWethy et al., 2009) and a representing losses not only of community components, but also of wave of faunal extinctions (Worthy and Holdaway, 2002; Duncan functional diversity (Hansen and Galetti, 2009; Boyer and Jetz, and Blackburn, 2004). Before European settlement (c. 1800 CE), 2014). Small ground-dwelling species (i.e. those vulnerable to three frogs, at least one reptile, and 30 bird species went extinct (Wood, 2013). With a recent human settlement (and the associated faunal extinctions) and extensive radiocarbon dated material from natural and archaeological deposits, New Zealand presents an ideal * Corresponding author. School of Environment, The University of Auckland, Private Bag 92019, Auckland 1142, New Zealand. Tel.: þ64 9 373 7599x84599. situation for constructing a high-precision chronology of the pre- E-mail address: [email protected] (G.L.W. Perry). historic extinction of island megafauna. http://dx.doi.org/10.1016/j.quascirev.2014.09.025 0277-3791/© 2014 Elsevier Ltd. All rights reserved.  G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 127 The flightless ratite moa (Aves, Dinornithiformes) were the guild and provide new insights into offtake rates by early moa largest of New Zealand's extinct avian fauna. At the time of human hunters. settlement, nine species (six genera) of moa (Bunce et al., 2009) occurred across New Zealand. These ecologically diverse herbivores 2. Materials and methods (Wood et al., 2013) ranged in body mass from 20 to 250 kg, with pronounced sexual dimorphism in some taxa (Worthy and 2.1. Radiocarbon dates and quality control Holdaway, 2002; Worthy and Scofield, 2012). Moa had a domi- nantly K-selected demography, characterised by prolonged pre- We compiled a database of 653 radiocarbon dates from moa reproductive periods, long lifespans and a low reproductive remains, with associated metadata including the species, locality, output (Turvey and Holdaway, 2005; Werner and Griebeler, 2012). context, dating method, dated fraction, and citation of data source The total moa population at the time of initial human settlement (Table S1). The remains were categorised (after Worthy and was likely on the order of 50,000e100,000 individuals (SI text), Scofield, 2012) into two adult body mass classes of 20e50 kg with densities possibly higher in the South than the North Island (Megalapteryx didinus, Pachyornis geranoides, Anomalopteryx didi- (Anderson, 1989b) due to greater levels of sympatry (Figs. S1 and formis) and 200þ kg (Pachyornis elephantopus and Dinornis spp.) S2). Moa bones are abundant in prehistoric butchery sites across representing the extreme ends of the body mass spectrum. We also New Zealand (Anderson, 1989b). Both eggs and adults appear to compiled a database of radiocarbon dates on Pacific rat coprolites have been predated, with ancient DNA (aDNA) analyses of midden (n ¼ 5) and rat-gnawed seeds (n ¼ 56) (Table S2). To ensure that our bones providing evidence for hunting targeting incubating males results were not biased by the inclusion of poor-quality or impre- (Oskam et al., 2012). The phylogenetic structure revealed by aDNA cise dates, radiocarbon dates were only included in subsequent from moa bones at individual butchery sites suggest that hunting analyses if they met the following quality criteria: 1) the dates had was relatively localised (Oskam et al., 2012), although dietary re- not previously been categorised as poor-quality by Petchey (1997) constructions from stable isotopes are more ambiguous (Kinaston (relates to the moa dates only); 2) the samples had been analysed et al., 2013). Recent phylogenetic analyses suggest that moa pop- using accelerator mass spectrometry (AMS); 3) the dates were on ulations were stable during the period immediately preceding materials other than carbonate, bulk-collagen or unspecified bone human settlement (Rawlence et al., 2012), and that populations of fractions; and 4) the measurement error was <10% of the radio- some taxa may even have been expanding in size during the late carbon age. A large proportion of the radiocarbon dates that passed Holocene (Allentoft et al., 2014). these criteria was produced since 2000, when the last quantitative While there is little doubt that human activity drove the moa to estimate of moa extinction was published (Holdaway and Jacomb, extinction (Allentoft et al., 2014), the exact mechanisms behind this 2000b), over which period there has been a marked increase in extinction event and its spatio-temporal dynamics remain unre- the quality and precision of radiocarbon dates from moa remains solved and continue to be debated. Population models based on (Fig. S4). All radiocarbon dates were calibrated using the R package best-guess estimates for moa and human demographics, and moa BChron (Parnell, 2013) and the SHCal 13 calibration curve (Hogg, hunting rates, suggest the extinction window (i.e. the period from 2013); all calibrated dates are reported as Cal CE and BP refers to human arrival to moa extinction) may have been as short as 100 years before 1950 AD. years (Holdaway and Jacomb, 2000b). Other estimates, however, suggest a more prolonged extinction window (400e500 years) af- 2.2. Sightings methods ter an initially intensive exploitation phase (Anderson, 1983, 1989b). It is also unclear whether all moa species shared similar Sightings methods (SM) are probabilistic tools for estimating extinction windows (Anderson, 1989c), or if the windows varied extinction or introduction dates based on a series of temporal due to a selective hunting strategy. Extinction seems to have loosely observations. Some SM directly estimate the time of extinction or followed the ‘overkill’ model (Martin, 1967) but whether it occurred introduction (e.g. Roberts and Solow, 2003), while others (e.g. contemporaneously throughout New Zealand or followed a ‘blitz- Bradshaw et al., 2012) estimate the probability of a sighting at kreig’ rolling extinction front is again less clear (Anderson, 1989c). some time in the future or past (requiring a predetermined Sightings methods (Rivadeneira et al., 2009) use a temporal threshold (a) for declaring a species extinct or introduced; in our record of historical observations to estimate the probability of analyses we use a ¼ 0.05). SM provide a statistically robust extinction at some point in time, and have been applied in palae- alternative to summing calibrated radiocarbon dates, a method ontology to estimate extinction times for various fauna (Roberts that is widely used in the archaeological literature but one which and Solow, 2003; Solow et al., 2006; Bradshaw et al., 2012; Lima- has been questioned on statistical grounds (Williams, 2012). Ribeiro and Diniz-Filho, 2014). Here, we critically examine the When applying SM to a fossil or archaeological record, quanti- date of moa extinction by applying sightings methods to a geore- fiable uncertainty exists around the timing of each ‘sighting’ (i.e. ferenced set of high-quality (based on criteria outlined in the radiocarbon error). We dealt with this uncertainty by resampling Methods section) radiocarbon dates on moa remains from natural from the 68% (1s) and 95% (2s) percentiles of the calibrated age and archaeological deposits. We apply the same methods to high- distributions of each date under consideration to generate a large quality radiocarbon dates on Pacific Rat coprolites and gnawed number (n ¼ 2500) of synthetic sighting records. This approach seeds (Wilmshurst et al., 2008, 2011) to estimate the time of initial accounts for uncertainties both in the age of each individual human settlement for New Zealand. When combined these two sample and the set of samples used; note that the set of cali- estimates enable us to determine the time taken for moa to have brated ‘sightings’ used in each of the 2,500 records will vary, as been driven to extinction. The application of these methods allows will the date attached to each. We used two sampling ranges us to evaluate moa extinction patterns in space and time (syn- (68th and 95th percentiles of calibrated age ranges) to provide chronicity across New Zealand) and across taxa (with respect to both narrow and conservative estimates of extinction and different body masses). By estimating the timing of initial settle- introduction times. We generated our synthetic sightings records ment and final extinction, and taking into consideration radio- using dates where the minimum boundary of the 95% calibrated carbon dating errors via calibrated calendar distributions of age range was <1000 years BP (i.e. significantly earlier than radiocarbon dates, we provide the highest-precision chronology human settlement of NZ in the 13th century). A total of 111 yet for the prehistoric extinction of an entire island megafaunal radiocarbon dates met both this temporal cut-off point, and the 128 G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 Fig. 1. Geographic distribution of radiocarbon dates on moa remains with a temporal range younger than 1250 ybp (conventional radiocarbon age [CRA], i.e. age relative to 1950 AD); the oldest date used in any of the analyses had a CRA of 1248 14C ybp (see Table S1). Red circles and labels denote specific sites referred to in the text. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) quality criteria outlined above (Fig. 1; Table S1). These dates were 2.3. Population decline models widely distributed across the South Island, with only four from the North Island. The oldest sample used in any analysis was a P. Holdaway and Jacomb (2000b) used a Leslie matrix model to elephantopus bone (NZA 30114), which dated to 1248 ± 25 BP estimate the time taken for moa to become extinct after human (cal. 766e894, 940e948 CE; 95%). By using a consistent timespan settlement. They estimated moa population size at the time of we mitigated issues around the sensitivity of SM to the length of human settlement of NZ as being 78,800 individuals. In their time that they cover (Solow, 2005). However, this approach led model, and in the absence of human hunting pressure, the moa to each sighting record for a given analysis comprising of a population grows to a maximum size of 101,000 individuals after slightly different suite and number of samples (Table S3). We 50 years (p. 2251), equating to a moa population growth rate used the 15 and 25 most recent dates on rat-gnawed seeds and (l ¼ (Nt/N0)(1/t)) of 1.005. We use this (favourable) moa population rat coprolites to estimate the time of human arrival in NZ and the growth rate as the basis for a discrete-time logistic model of pop- South Island, respectively (Table S2). Finally, we applied a suite of ulation growth with fixed-effort harvesting (i.e. a constant number, SM to each of the synthetic sighting records, generating a dis- rather than proportion, of the population is removed) to estimate tribution of estimated extinction or introduction times. In each offtake (harvest) rates that would result in population extinction analysis we randomly and with replacement selected dates (see description in Haddon, 2011): (n ¼ 10,000) from the probability distributions for the timing of Nt   human arrival and moa extinction, and subtracted the latter from Ntþ1 ¼ Nt þ rt Nt 1 H (1) the former to derive a probability distribution of the extinction K window (i.e. time taken to drive moa to extinction). We did not perform statistical tests on the differences or simi- where: Nt ¼ population size at time t, K ¼ the carrying capacity, rt is larities between extinction times for different regions and body the intrinsic rate of increase at time t (lt ¼ ln(rt)) and H is the offtake masses, because the number of resamplings (n ¼ 2500) was high rate (birds per year). enough that a statistically significant result (i.e. low p-value) would The model developed by Holdaway and Jacomb (2000b) had have almost been guaranteed, irrespective of the actual effect size multiple stage-classes but ours considers only one, so is more (White et al., 2014). All radiocarbon calibration and sightings ana- aggregated. However, our simplified representation has the lyses were conducted in R version 3.0.2 (R Development Core Team, advantage that we can avoid the many uncertainties associated 2014). with estimating demographic rates for different age-classes across  G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 129 the nine moa taxa. We also, by default, assume that all age-classes bone from Bulmer Cave in north-west Nelson (Rawlence and were predated, whereas Holdaway and Jacomb (2000b) only Cooper, 2012) (Fig. 1). consider predation on adult birds; aDNA analyses suggest that in- dividuals of all ages, as well as eggs, were killed (Oskam et al., 2012) 3.2. Extinction chronology and numerous middens contain eggshell (e.g., Wairau Bar, Higham et al., 1999). The maximum number of individuals that can be Sightings models estimate that moa were extinct in New Zea- sustainably harvested from any population is the number of in- land by 1418-(1440)-1473 CE or 1427-(1445)-1461 CE (Fig. 2; red dividuals being added to the population when it is growing most numbers). The two sightings methods provide estimated dates for quickly. In a continuous-time logistic growth model, the population human settlement of 1096-(1231)-1291 CE or 1162-(1235)-1281 CE grows most quickly when N ¼ K/2, at which point rK/4 individuals (Fig. 2). On this basis, we estimate that moa became extinct are added at each time period (see Pastor, 2008; Haddon, 2011 for throughout NZ within windows of 141-(211)-346 years and 158- details). A revised estimate of the population size of moa across all (209)-286 years (Fig. 3). Therefore, even under the most conser- species (see SI text and Fig. SM 2) at the time of human settlement vative estimates (that is, using 95% calibrated distributions), moa of NZ, yields a median moa population size of 57,681. Thus, erring extinction occurred at least two centuries prior to European set- on the side of conservatism, we set our baseline (N0) above this at tlement of New Zealand in the early-19th century. N0 ¼ 6.0  104. Our estimate of moa population size is in line with For the South Island (where most radiocarbon dated moa re- an independent recent estimate derived using population genetic mains are from e Fig. 1), we estimate that moa were extinct by methods (Allentoft et al., 2014). We use a value double that of the 1417-(1439)-1471 CE or 1426-(1446)-1462 CE. Our models predict estimated moa population at human arrival as the carrying capacity human settlement of the South Island in 1181-(1266)-1331 CE or (K ¼ 1.2  105); it is at N0 ¼ K/2 that a logistically growing popu- 1188-(1241)-1303 CE. Accordingly, we estimate that moa were lation is increasing most rapidly. We explored the speed of driven to extinction across the South Island within 106-(174)-260 extinction under a range of offtake (H) and human population or 142-(204)-260 years, again well before European settlement. growth rates, with the former based on values in Anderson (1989c) and Holdaway and Jacomb (2000b). Unless otherwise stated we 3.3. Local extinctions assume an initial human population size of 100 individuals. We also evaluated the model with the initial moa population size doubled Two sites e one natural, the other archaeological e contained and the rate of moa population growth doubled to represent pop- sufficient radiocarbon dated moa remains for an evaluation of local ulation dynamics under estimates extremely favourable for moa extinction chronologies (Figs. 2 and 3). Wairau Bar in the north- population persistence. Finally we evaluate the impact of a eastern South Island is one of the earliest known M aori sites, and doubling of the initial human population. Our model is determin- contains abundant evidence of moa butchery associated with an istic and does not include habitat loss. These two assumptions are occupation period of possibly just a few decades (Higham et al., highly conservative. The addition of demographic and/or environ- 1999; Jacomb et al., 2014). Local extinction of moa was rapid at mental stochasticity and/or habitat loss would reduce the offtake this site and occurred ca 1383-(1402)-1433 CE or 1395-(1414)-1432 rates required to effect any given extinction timeline. (Fig. 2), within ca 150 years of human settlement. At Daley's Flat in the Dart River Valley, southern South Island (ca 550 km from 3. Results Wairau Bar), which is a naturally deposited bone and coprolite assemblage less than 30 km up-river from an early Ma ori campsite We tested five different SMs and found their estimated time containing abundant moa remains (Anderson and Ritchie, 1986), distributions to be relatively consistent, with their medians all very the local extinction of moa was also rapid, occurring at ca 1366- similar (Fig. S3). Therefore we report just the results obtained using (1391)-1424 CE or 1402-(1427)-1447 CE (Fig. 2). Although the two two methods: one described by Roberts and Solow (2003), which SMs provide slightly different estimates of the time of extinction, directly estimates the time of extinction, and the other by McInerny for each method the estimates are within a decade of each other et al. (2006), which estimates the probability of extinction at a between sites. given time. These two methods produced extinction timing esti- mates close to the other SMs used, and unlike some others (e.g. the 3.4. Extinction of different body mass classes GRIWM method of Bradshaw et al., 2012) did not yield long low probability tails of early arrival or prolonged persistence (Fig. S3). In There was little difference in the estimated time of extinction for the following text, settlement (rat) and extinction (moa) timings, the largest (body masses > 200 kg) as opposed to the smallest and length of extinction windows, are shown as the median in (body mass < 50 kg) moa taxa, although one of the sightings parentheses, with 2.5 and 97.5 percentiles, for the Solow and methods shows a slight lag between the two (likely as the result of Roberts (2003) method, and the McInerny et al. (2006) method, the relatively few [n ¼ 16] and patchily distributed [in time] dates respectively, based on sampling from 95% of the calibrated date for the smaller size class). The estimated extinction windows were distributions. Thus, each arrival or extinction estimate comprises 1371-(1400)-1457 CE or 1398-(1423)-1455 CE and 1366-(1399)- two ranges: one for each sighting method. 1443 CE or 1425-(1461)-1488 CE) for large and small moa, respectively (Fig. 4). 3.1. Youngest radiocarbon ages for moa remains 3.5. Hunting dynamics The youngest radiocarbon date to pass our quality criteria was 560 ± 45 BP (Wk-2417: cal. 1320e1455 CE; 95% confidence), for an Based on our new chronology for moa extinction, and an eggshell fragment from the Shag River archaeological site in south- updated population estimate for moa of ca 58,000 (95% quantile eastern Otago (Fig. 1). Seven of the eight youngest high-quality range, 8834e166,794) birds at the time of human settlement (SI radiocarbon dates were from archaeological sites in the South Is- text), we can also revisit the harvest rates that would have been land (Shag River, Warrington, Pleasant River and Redcliffs Flat) required to bring about the extinction pattern shown in Fig. 2. If we (Fig. 1). The youngest radiocarbon date from a natural deposit was assume the total moa population followed logistic growth from a 564 ± 26 BP (OXA 20287: cal. 1396e1442 CE; 95% confidence), on a population size (N0) at the time of settlement of 6.0  104 with a 130 G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 Fig. 2. Estimates (a e via method of (Solow and Roberts, 2003) and b e via method of (McInerny et al., 2006)) for the timing of human (Polynesian) settlement in NZ and timings of subsequent moa extinction. Histograms show estimates based on sampling under 68% (1s; blue) and 95% (2s; red) of the calibrated age curves (based on n ¼ 2500 replicates for each), with purple showing the area of overlap. Grey histograms are the distributions based on the 1s calibrated curves for human arrival. Boxplots show the calibrated 95% distribution for the terminal (oldest for rats, youngest for moa) radiocarbon dates; boxplot hinges cover the first and third quartile, the whiskers extend to 1.5 the inter-quartile range and the points are outliers beyond that. Triangles and text above them show the median time of arrival or extinction (colours as for probability distributions). (For inter- pretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) carrying capacity (K) of 1.2  105 individuals then, theoretically, the offtake assumes, however, either a declining offtake per person maximum sustainable harvest rate was ca 150 birds/year (i.e. rK/4, over time as the human population grows or no human population (Pastor, 2008)). Under constant harvest at this rate, extinction growth. Assuming that the offtake increased over time at a rate would have taken ca 630 years e much longer than our analyses directly proportional to the human population growth rate (i.e. the predict. Holdaway and Jacomb (2000b) evaluated offtake rates per person offtake is constant) further reduces the time to extinc- equivalent to 1 bird/10e20 people/week (or 2.6e5.2 birds/person/ tion (Fig. 5a). Our sightings models predict the time for total year). Assuming an initial human population size of ca 100 in- extinction was between 140 and 346 years (the 95% quantile range dividuals (Murray-McIntosh et al., 1998), these rates are equivalent of time to extinction; Fig. 3) and extinction within this time horizon to a total harvest of 260e520 individuals/year e an offtake rate requires only moderate levels of human population growth and comfortably in excess of the maximum sustainable rate, and one offtake. that would have grown over time as the human population Clearly, demographic estimates for both moa and human pop- increased. A constant total offtake of 500 birds/year would have ulations at the time of settlement are uncertain. However, even if resulted in total moa extinction after 135 years (with the initial moa population size is doubled or the rate of moa pop- N0 ¼ 6.0  104 moa, l ¼ 1.005 and K ¼ 1.2  105 moa). A constant ulation growth rate is doubled then a human population growing at  G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 131 Fig. 3. Duration of the extinction process across all taxa of moa at Wairau Bar, Daley's Flat and all of NZ (a e via method of (Solow and Roberts, 2003) and b e via method of (McInerny et al., 2006)). Colours and symbology as per Fig. 2. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) the low rate of 1% per year would have driven moa to extinction investigated and dismissed claims of late moa survival). Previous under plausible offtake rates (ca 5 birds/person/year) within ca 190 debate about the timing of moa extinction has focussed on 14 years (Fig. 5b, c). If the initial human population size is doubled to relatively recent (post 1500 CE) calibrated radiocarbon ages on moa 200 this further reduces time to moa extinction (Fig. 5d). The total bones from archaeological sites (Anderson, 2000; Holdaway and number of birds killed over the extinction event falls as human Jacomb, 2000a,b). However, all of these dates failed to meet our population growth and offtake rates increase, or if the initial human quality-control criteria as they either did not pass the protocol of population size is doubled. Petchey (1997) or their measurement error was too large to be precise (i.e., >10% of the radiocarbon age). Furthermore, we ob- 4. Discussion tained two additional radiocarbon dates (lab numbers Wk-33994 and Wk-33996) on moa bone fragments from Tumbledown Bay, Our analyses of high-precision radiocarbon dates show that an archaeological site in sand dunes on the east cost of the South extinction of moa in New Zealand was rapid, occurring in little Island, which had yielded one of these recent ages. The original more than 200 years of initial human settlement. The extinction bone that gave the recent age (NZA338: 307 ± 85 BP) could not be event was contemporaneous at sites that were geographically relocated for re-dating and may have been completely consumed in distant from each other and there did not appear to be preferential the original analysis. Our new dates (Wk-33994: 1837 ± 28 BP [cal. hunting of the nine species of moa on the basis of their size. Our 128-339 CE; 95% confidence] and Wk-33996: 839 ± 26 BP [1201- estimated extinction windows are broadly consistent with, if a few 1280 CE; 95% confidence]) failed to replicate the young age, but decades longer than, population model estimates based on mod- indicated that older, naturally deposited moa bones were incor- erate hunting pressure (Holdaway and Jacomb, 2000b), but are porated into the lower archaeological layers at this site. inconsistent with prolonged survival of moa into the post- Owens and Bennett (2000) describe two main pathways to European era (supporting the view of Anderson (1989a) who extinction: (i) direct demographic pressure (e.g. increased mortality 132 G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 Fig. 4. Probability density curves (a e via method of (Solow and Roberts, 2003) and b e via method of (McInerny et al., 2006)) for extinction timing of the smallest and largest moa body size classes (<50 kg and >200 kg). Colours and symbology as per Fig. 2. Grey histograms are the 68% curves for human arrival. Boxplots show the calibrated 95% distribution for the youngest radiocarbon dates on moa. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Fig. 5. Extinction timelines as a function of offtake rate and human population growth rate assuming logistic moa population growth and fixed effort hunting; unless otherwise stated the moa growth rate and carrying capacity were l ¼ 1.005 and Kmoa ¼ 1.2  105, and the initial human population was 100: (a) Nmoa ¼ 6.0  104; (b) Nmoa ¼ Kmoa ¼ 1.2  105, (c) Nmoa ¼ 6.0  104, lmoa ¼ 1.01 and (d) N ¼ 6.0  104, Nhuman ¼ 200. The heavy contours show the 68% and the lighter the 95% estimated intervals for time to extinction (137e279 y and 140e346 y, respectively). Greyed out areas show parameter combinations where extinction did not occur after 1000 simulated years. Grey arrows show the offtake rates (used by 4) of one moa per 10 and 20 people per week, respectively.  G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 133 via hunting, reduced recruitment through egg harvesting) and (ii) drive moa to extinction within the timeframes we estimate. fragmentation and loss of habitat. A taxon's susceptibility to either of Furthermore, Allentoft et al. (2010) suggest that the sex ratios in these pathways is influenced by a variety of factors including their some moa species were highly female-biased, perhaps by up to a life-history strategy and the environments they inhabit (McKinney, 3:1 ratio. Together with the fact that males incubating nests may 1997; Lee and Jetz, 2010). Taxa with muted demographic rates are have been more vulnerable and targeted by hunters (Oskam particularly vulnerable to demographic pressures, whereas ecolog- et al., 2012), such biased sex-ratios would have dramatically ical specialists are especially at risk from habitat disruption and loss lowered the offtake rates required to drive moa to extinction. (Owens and Bennett, 2000). Because a large amount of suitable Moa may also have been concentrated into specific parts of the habitat was still present across New Zealand at the time of European landscape, which would have increased their vulnerability to settlement (Perry et al., 2012) habitat loss is unlikely to be have been intense hunting activity (McGlone et al., 1994). Accordingly, we the primary driver of the extinction of the moa (see also Turvey and suggest hunting pressure was likely to have been the main factor Holdaway, 2005). Current evidence suggests that moa were ecolog- contributing to the rapid demise of moa. A lack of genetic bot- ical generalists (Wood et al., 2013). Few moa species were restricted tlenecking in the moa population immediately prior to extinction to forest habitats, which experienced widespread destruction supports the argument that extinction was abrupt and rapid, as immediately following Ma ori settlement (McGlone and Wilmshurst, expected under overkill, rather than being prolonged (Allentoft 1999; McWethy et al., 2009), and some actually preferred non-forest et al., 2014). habitats (Worthy and Holdaway, 2002; Wood et al., 2013). However, Assuming hunting was the key factor that drove moa to a rapid it is possible that habitat fragmentation and associated isolation of extinction, the results from our extinction models provide insights sub-populations may have played some role in the extinction process into the movements and strategies of prehistoric Polynesian (e.g. Gibson et al., 2013). hunters in New Zealand. It has been proposed that early Polynesian Low recruitment rates and extended juvenile phases made settlers spread rapidly through the New Zealand archipelago megafauna such as moa (Turvey and Holdaway, 2005) particu- following a ‘chaotic colonisation’ pattern (Anderson and larly vulnerable to overharvesting, and our modelling indicates McGovern-Wilson, 1990, p. 41), rather than a wave-like expansion that even moderate offtake rates would have been sufficient to front. Several lines of evidence, including the presence of early Fig. 6. Illustration of the effect of dating error on estimations of human arrival and moa extinction. While they both pass our discard protocol, the oldest two rat bones have dates with relatively large error (WK-8548: 780 ± 70 and WK-8991: 754 ± 65 BP, respectively). Artificially changing the error to 25 y results in much tighter estimates of human arrival (a and c from the methods of Solow and Roberts (2003) and McInerny et al. (2006), respectively) and hence the time before moa became extinct (b and d). These shifts are due to the typically smaller gaps between the last two dates in the resampled sightings records (e and f). Analyses based on all rat bones for NZ and all moa material across NZ where the minimum boundary of the 95% calibrated age range was <1000 years BP. Blue represents estimates as presented in Fig. 2 and red represents reanalyses with radiocarbon laboratory error reduced to 25 y. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 134 G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 archaeological sites throughout the country, prehistoric human 5. Conclusions diets reconstructed from stable isotope analyses of bones from Wairau Bar (Kinaston et al., 2013) and patterns of lithic exchange The post-settlement extinction of New Zealand's large birds is ori were highly mobile and (Walter et al., 2010), all suggest early Ma certainly not an isolated instance of a rapid loss of megafauna, but spread rapidly through NZ after settlement. The rapid dispersal of because it happened so recently and is so well-dated it now pro- people throughout New Zealand is evident in our results, reflected vides the best resolved case. In Tonga, which was settled ca 2980 in the near-synchronous local moa extinction windows in years ago (Burley et al., 2012), the extinction of large fauna (giant geographically distant areas of hunting activity (Wairau Bar and iguana [Brachylophus gibbonsi] and megapodes [Megapodius spp.]) Dart River Valley) by the early 15th century. Our results show that seems to have been so rapid that the time between human settle- moa population collapse was rapid even in areas that were not ment (using chicken bones as a proxy) and extinction is obscured deforested, such as the Dart River Valley, likely because people by the error margin of radiocarbon dates (Steadman et al., 2002). were active in such landscapes collecting other resources (McGlone Conversely, there are also instances where humans appear to have et al., 1994). Local extinction in both the Wairu Bar and Dart River coexisted for long periods alongside megafauna (e.g. Madagascar; Valley sites is ca 30e50 years earlier than that estimated for the Dewar et al., 2013). The approach we have adopted has yielded entire South Island and New Zealand landmasses. We interpret this what is probably the most precise examination of the extinction of lag time as reflecting the period between economic and final an entire guild of animals yet and it could be applied to continental extinction (where economic extinction is defined as the point at situations by using human artefacts alongside megafaunal remains which an organism is sufficiently rare that the reward gained from (e.g. as in Boulanger and Lyman, 2014) and to islands (such as the hunting is less than the associated costs; see Courchamp et al. Canary Islands; Rando et al. (2014)) where human arrival and (2006)). Once populations of moa in areas surrounding human faunal extinction are both precisely dated. habitation or butchery sites were quickly exterminated, the depo- sition of moa remains in such sites became minor relative to other Data accessibility dietary items. The chance of finding and dating the youngest moa remains in any site is low, but by using many sites as we do here it is Data used in the analyses are available as supplementary ma- possible to detect the ‘tail’ of persistence. This tail reflects small terial accompanying this article. populations that would have persisted in more remote and inac- cessible areas, although our results suggest that these only survived Author contributions for a few decades before they too were exterminated. Hunting of these last remaining birds is evident in the archaeological record, Author contributions: G.L.W.P. J.R.W and J.M.W designed the where taphonomic changes between early and late moa hunting study; G.L.W.P. and A.W. developed and conducted quantitative strata indicate a shift towards butchering at sites distant to where analyses; all authors contributed to database compilation; all au- remains were consumed and deposited (Nagaoka, 2005). We found thors contributed to the writing of the manuscript. little evidence for sorting by body mass in the extinction chronol- ogy, which implies hunting was not size-selective, although one of Acknowledgements the sightings methods suggests that larger body sizes may have gone extinct slightly earlier than smaller ones (Fig. 4). We thank radiocarbon dating laboratories, and in particular Where extinction is rapid, establishing robust chronologies is Fiona Petchey (Waikato), Dawn Chambers (GNS), Diane Baker challenging, and necessitates the use of high-precision radiometric (Oxford), John Southon (UCI) and Mitzi DeMartino (Arizona), for dates. In our case the largest uncertainties are those associated with providing details of sample analyses and pretreatments. Nic Raw- the timing of human settlement (Fig. 2) and stem from the rela- lence provided pre-publication access to some Pachyornis radio- tively high errors associated with the two oldest rat gnawed seeds carbon dates. The radiocarbon database of moa material that (WK-8548: 780 ± 70 and WK-8991: 754 ± 65 BP, respectively). The underpins this work builds on earlier efforts by Trevor Worthy, median dates for human arrival (Fig. 2) suggest invisibility win- which we gratefully acknowledge. Matt McGlone provided useful dows of 40e50 years when compared to the oldest dates reported comments on an earlier version of this manuscript. Constructive by Wilmshurst et al. (2008). The longer tail of earlier arrival sug- comments from two anonymous referees helped to improve the gested by the 95th percentile (Fig. 2) is ecologically implausible manuscript. This research was supported by Core funding for given what is known about the rapid spread of Norway rats (Rattus Crown Research Institutes from the New Zealand Ministry of norvegicus) across NZ following European arrival (Innes, 2005; Business, Innovation and Employment's Science and Innovation Wilmshurst et al., 2008). While Wk-8548 and Wk-8991 dates Group. AW was supported by a summer scholarship from the pass our discard protocols, their inclusion means that our estimates University of Auckland. for the time horizon of the extinction event is conservative and they increase its uncertainty. To illustrate how significant radiocarbon Appendix A. Supplementary data error in the ‘terminal’ dates can be, we repeated the estimations of human settlement with the same CRA determinations, but with the Supplementary data related to this article can be found at http:// error precision reduced to 25 years (Fig. 6). Reducing the error on dx.doi.org/10.1016/j.quascirev.2014.09.025. these dates has the effect of reducing the average gap (the ‘terminal gap’) between the oldest and second oldest dates in each sighting References record. This reduction acts both to make the estimated timing of arrival slightly later (from 1231 to 1262 and 1235 to 1256 to AD, for Allentoft, M.E., Bunce, M., Scofield, R.P., Hale, M.L., Holdaway, R.N., 2010. Highly skewed sex ratios and biased fossil deposition of moa: ancient DNA provides the methods of Solow and Roberts (2003) and McInerny et al. new insight on New Zealand's extinct megafauna. Quat. Sci. Rev. 29, 753e762. (2006), respectively; Fig. 6a and c) and, more importantly, to Allentoft, M.E., Heller, R., Oskam, C.L., Lorenzen, E.D., Hale, M.L., Gilbert, M.T.P., considerably reduce the uncertainty around the estimated duration Jacomb, C., Holdaway, R.N., Bunce, M., 2014. Extinct New Zealand megafauna of the extinction event (by 69 (33.6%) and 77 (42.0%) years for the were not in decline before human colonization. Proc. Natl. Acad. Sci. U. S. A. 111, 4922e4927. two methods; Fig. 6b and d). Anderson, A., 1989a. On evidence for the survival of moa in European Fiordland. N. Z. J. Ecol. 12, 39e44.  G.L.W. Perry et al. / Quaternary Science Reviews 105 (2014) 126e135 135 Anderson, A., 2000. Less is moa. Science 289, 1472e1473. Murray-McIntosh, R.P., Scrimshaw, B.J., Hatfield, P.J., Penny, D., 1998. Testing Anderson, A., McGovern-Wilson, R., 1990. The pattern of prehistoric Polynesian migration patterns and estimating founding population size in Polynesia by colonization in New Zealand. J. Roy. Soc. N. Z. 20, 41e63. using human mtDNA sequences. Proc. Natl. Acad. Sci. U. S. A. 95, 9047e9052. Anderson, A., Ritchie, N., 1986. Pavements, pounamu and ti: Dart Bridge site in Nagaoka, L., 2005. Declining foraging efficiency and moa carcass exploitation in western Otago, New Zealand. N. Z. J. Archaeol. 8, 115e141. southern New Zealand. J. Archaeol. Sci. 32, 1328e1338. Anderson, A.J., 1983. Faunal depletion and subsistence change in the early prehis- Oskam, C.L., Allentoft, M.E., Walter, R., Scofield, R.P., Haile, J., Holdaway, R.N., tory of southern New Zealand. Archaeol. Ocean 18, 1e10. Bunce, M., Jacomb, C., 2012. Ancient DNA analyses of early archaeological sites Anderson, A.J., 1989b. Mechanics of overkill in the extinction of New Zealand moas. in New Zealand reveal extreme exploitation of moa (Aves: Dinornithiformes) at J. Archaeol. Sci. 16, 137e151. all life stages. Quat. Sci. Rev. 52, 41e48. Anderson, A.J., 1989c. Prodigious Birds: Moas and Moa-hunting in Prehistoric New Owens, I.P.F., Bennett, P.M., 2000. Ecological basis of extinction risk in birds: habitat Zealand. Cambridge University Press, Cambridge. loss versus human persecution and introduced predators. Proc. Natl. Acad. Sci. Boulanger, M.T., Lyman, R.L., 2014. Northeastern North American Pleistocene U. S. A. 97, 12144e12148. megafauna chronologically overlapped minimally with Paleoindians. Quat. Sci. Parnell, A., 2013. Bayesian Radiocarbon Chronologies and Relative Sea Level Anal- Rev. 85, 35e46. ysis. R Package Version 3.2. Boyer, A.G., 2008. Extinction patterns in the avifauna of the Hawaiian islands. Pastor, J., 2008. Mathematical Ecology of Populations and Ecosystems. Wiley- Divers. Distrib. 14, 509e517. Blackwell, Oxford; Hoboken, NJ. Boyer, A.G., Jetz, W., 2014. Extinctions and the loss of ecological function in island Perry, G.L.W., Wilmshurst, J.M., McGlone, M.S., Napier, A., 2012. Reconstructing bird communities. Glob. Ecol. Biogeogr. 23, 679e688. spatial vulnerability to forest loss by fire in pre-historic New Zealand. Glob. Ecol. Bradshaw, C.J.A., Cooper, A., Turney, C.S.M., Brook, B.W., 2012. Robust estimates of Biogeogr. 21, 1029e1041. extinction time in the geological record. Quat. Sci. Rev. 33, 14e19. Petchey, F.J., 1997. New Zealand bone dating revisited: a radiocarbon discard pro- Bunce, M., Worthy, T.H., Phillips, M.J., Holdaway, R.N., Willerslev, E., Haile, J., tocol for bone. N. Z. J. Archaeol. 19, 81e124. Shapiro, B., Scofield, R.P., Drummond, A., Kamp, P.J.J., Cooper, A., 2009. The R Development Core Team, 2014. R: a Language and Environment for Statistical evolutionary history of the extinct ratite moa and New Zealand Neogene Computing. R Foundation for Statistical Computing, Vienna, Austria. paleogeography. Proc. Natl. Acad. Sci. U. S. A. 106, 20646e20651. n, B., Navarro, J.F., 2014. Reappraisal of the extinction Rando, J.C., Alcover, J.A., Galva Burley, D., Weisler, M.I., Zhao, J.-x., 2012. High precision U/Th dating of first Poly- of Canariomys bravoi, the giant rat from Tenerife (Canary Islands). Quat. Sci. Rev. nesian settlement. PLoS One 7, e48769. 94, 22e27. Courchamp, F., Angulo, E., Rivalan, P., Hall, R.J., Signoret, L., Bull, L., Meinard, Y., 2006. Rawlence, N.J., Cooper, A., 2012. Youngest reported radiocarbon age of a moa (Aves: Rarity value and species extinction: the anthropogenic Allee effect. PLoS Biol. 4, Dinornithiformes) dated from a natural site in New Zealand. J. R. Soc. N. Z. 1e8. 3415. Rawlence, N.J., Metcalf, J.L., Wood, J.R., Worthy, T.H., Austin, J.J., Cooper, A., 2012. The Dewar, R.E., Radimilahy, C., Wright, H.T., Jacobs, Z., Kelly, G.O., Berna, F., 2013. Stone effect of climate and environmental change on the megafaunal moa of New tools and foraging in northern Madagascar challenge Holocene extinction Zealand in the absence of humans. Quat. Sci. Rev. 50, 141e153. models. Proc. Natl. Acad. Sci. U. S. A. 110, 12583e12588. Rick, T.C., Kirch, P.V., Erlandson, J.M., Fitzpatrick, S.M., 2014. Archeology, deep his- Duncan, R.P., Blackburn, T.M., 2004. Extinction and endemism in the New Zealand tory, and the human transformation of island ecosystems. Anthropocene 4, avifauna. Glob. Ecol. Biogeogr. 13, 509e517. 33e45. Duncan, R.P., Boyer, A.G., Blackburn, T.M., 2013. Magnitude and variation of pre- Rivadeneira, M.M., Hunt, G., Roy, K., 2009. The use of sighting records to infer historic bird extinctions in the Pacific. Proc. Natl. Acad. Sci. U. S. A. 110, species extinctions: an evaluation of different methods. Ecology 90, 1291e1300. 6436e6441. Roberts, D.L., Solow, A.R., 2003. Flightless birds e when did the dodo become Gibson, L., Lynam, A.J., Bradshaw, C.J.A., He, F.L., Bickford, D.P., Woodruff, D.S., extinct? Nature 426, 245. Bumrungsri, S., Laurance, W.F., 2013. Near-complete extinction of native Rolett, B., Diamond, J., 2004. Environmental predictors of pre-European defores- small mammal fauna 25 years after forest fragmentation. Science 341, tation on Pacific islands. Nature 431, 443e446. 1508e1510. Solow, A.R., 2005. Inferring extinction from a sighting record. Math. Biosci. 195, Haddon, M., 2011. Modelling and Quantitative Methods in Fisheries, second ed. CRC 47e55. Press, Boca Raton. Solow, A.R., Roberts, D.L., 2003. A nonparametric test for extinction based on a Hansen, D.M., Galetti, M., 2009. The forgotten megafauna. Science 324, 42e43. sighting record. Ecology 84, 1329e1332. Higham, T., Anderson, A., Jacomb, C., 1999. Dating the first new Zealanders: the Solow, A.R., Roberts, D.L., Robbirt, K.M., 2006. On the Pleistocene extinctions of chronology of Wairau Bar. Antiquity 73, 420e427. Alaskan mammoths and horses. Proc. Natl. Acad. Sci. U. S. A. 103, 7351e7353. Hogg, A., 2013. SHCal13 Southern Hemisphere calibration, 0e50,000 Years cal BP. Steadman, D.W., 1995. Prehistoric extinction of Pacific Island birds e biodiversity Radiocarbon 55, 1889e1903. meets zooarchaeology. Science 268, 625. Holdaway, R.N., Jacomb, C., 2000a. Less is moa - Response. Science 289, 1473e1474. Steadman, D.W., Pregill, G.K., Burley, D.V., 2002. Rapid prehistoric extinction of Holdaway, R.N., Jacomb, C., 2000b. Rapid extinction of the moas (Aves: Dinorni- iguanas and birds in Polynesia. Proc. Natl. Acad. Sci. U. S. A. 99, 3673e3677. thiformes): model, test, and implications. Science 287, 2250e2254. Turvey, S.T., Holdaway, R.N., 2005. Postnatal ontogeny, population structure, and Innes, J.G., 2005. Norway rat. In: King, C.M. (Ed.), The Handbook of New Zealand extinction of the giant moa Dinornis. J. Morphol. 265, 70e86. Mammals, second ed. Oxford University Press, Melbourne, pp. 174e187. Walter, R., Jacomb, C., Bowron-Muth, S., 2010. Colonisation, mobility and exchange Jacomb, C., Holdaway, R.N., Allentoft, M.E., Bunce, M., Oskam, C.L., Walter, R., in New Zealand prehistory. Antiquity 84, 497e513. Brooks, E., 2014. High-precision dating and ancient DNA profiling of moa (Aves: Werner, J., Griebeler, E., 2012. Reproductive investment in moa: a K-selected life- Dinornithiformes) eggshell documents a complex feature at Wairau Bar and history strategy? Evol. Ecol. 26, 1391e1419. refines the chronology of New Zealand settlement by Polynesians. J. Archaeol. White, J.W., Rassweiler, A., Samhouri, J.F., Stier, A.C., White, C., 2014. Ecologists Sci. 50, 24e30. should not use statistical significance tests to interpret simulation model re- Kinaston, R.L., Walter, R.K., Jacomb, C., Brooks, E., Tayles, N., Halcrow, S.E., Stirling, C., sults. Oikos 123, 385e388. Reid, M., Gray, A.R., Spinks, J., Shaw, B., Fyfe, R., Buckley, H.R., 2013. The first Williams, A.N., 2012. The use of summed radiocarbon probability distributions in New Zealanders: patterns of diet and mobility revealed through isotope anal- archaeology: a review of methods. J. Archaeol. Sci. 39, 578e589. ysis. PLoS One 8, e64580. Wilmshurst, J.M., Anderson, A.J., Higham, T.F.G., Worthy, T.H., 2008. Dating the late Lee, T.M., Jetz, W., 2010. Unravelling the structure of species extinction risk for prehistoric dispersal of polynesians to New Zealand using the commensal Pa- predictive conservation science. Proc. R. Soc. B Biol. Sci. 278, 1329e1338. cific rat. Proc. Natl. Acad. Sci. U. S. A. 105, 7676e7680. Lima-Ribeiro, M.S., Diniz-Filho, J.A.F., 2014. Obstinate overkill in Tasmania? The Wilmshurst, J.M., Hunt, T.L., Lipo, C.P., Anderson, A.J., 2011. High-precision radio- closest gaps do not probabilistically support human involvement in megafaunal carbon dating shows recent and rapid initial human colonization of East Pol- extinctions. Earth Sci. Rev. 135, 59e64. ynesia. Proc. Natl. Acad. Sci. U. S. A. 108, 1815e1820. Martin, P.S., 1967. Prehistoric overkill. In: Martin, P.S., Wright, H.E. (Eds.), Pleistocene Wilmshurst, J.M., McGlone, M.S., Leathwick, J.R., Newnham, R.M., 2007. A pre-defor- Extinctions: the Search for Cause. Yale University Press, pp. 75e120. estation pollen-climate calibration model for New Zealand and quantitative McGlone, M.S., Anderson, A.J., Holdaway, R.N., 1994. An Ecological Approach to the temperature reconstructions for the past 18, 000 years BP. J. Quat. Sci. 22, 535e547. Polynesian Settlement of New Zealand, the Origins of the First New Zealanders. Wood, J.R., 2013. New Zealand, 500 years ago. In: MacLeod, N. (Ed.), Grzimek's Auckland University Press, Auckland, New Zealand, pp. 136e163. Animal Life Encyclopedia: Extinction. Gale, Farmington Hills, USA, pp. 595e604. McGlone, M.S., Wilmshurst, J.M., 1999. Dating initial Maori environmental impact in Wood, J.R., Wilmshurst, J.M., Richardson, S.J., Rawlence, N.J., Wagstaff, S.J., New Zealand. Quat. Int. 59, 5e16. Worthy, T.H., Cooper, A., 2013. Resolving lost herbivore community structure McInerny, G.J., Roberts, D.L., Davy, A.J., Cribb, P.J., 2006. Significance of sighting rate using coprolites of four sympatric moa species (Aves: Dinornithiformes). Proc. in inferring extinction and threat. Conserv. Biol. 20, 562e567. Natl. Acad. Sci. U. S. A. 110, 16910e16915. McKinney, M.L., 1997. Extinction vulnerability and selectivity: combining ecological Worthy, T.H., Holdaway, R.N., 2002. The Lost World of the Moa: Prehistoric Life of and paleontological views. Annu. Rev. Ecol. Syst. 28, 495e516. New Zealand. Indiana University Press, Bloomington. McWethy, D.B., Whitlock, C., Wilmshurst, J.M., McGlone, M.S., Li, X., 2009. Rapid Worthy, T.H., Scofield, R.P., 2012. Twenty-first century advances in knowledge of the deforestation of South Islands, New Zealands, by early Polynesian fires. Holo- biology of moa (Aves: Dinornithiformes): a new morphological analysis and cene 19, 883e897. moa diagnoses revised. N. Z. J. Zool. 39, 87e153.