Rapid Extinction of the Moas (Aves: Dinornithiformes): Model, Test, and Implications

See allHide authors and affiliations

Science  24 Mar 2000:
Vol. 287, Issue 5461, pp. 2250-2254
DOI: 10.1126/science.287.5461.2250


A Leslie matrix population model supported by carbon-14 dating of early occupation layers lacking moa remains suggests that human hunting and habitat destruction drove the 11 species of moa to extinction less than 100 years after Polynesian settlement of New Zealand. The rapid extinction contrasts with models that envisage several centuries of exploitation.

All 11 species of the large (20 to 250 kg) flightless birds known as moas (Aves: Dinornithiformes) survived until the arrival of Polynesian colonists in New Zealand (1). Abundant remains of moas in early archaeological sites show that the birds were major items of diet immediately after colonization (2–5). Indeed, the presence of moa remains was formerly used to characterize the earliest or “Archaic” period of human occupation in New Zealand [the “Moa-hunter” period (3,6)]. Polynesian hunting and habitat destruction were responsible for the extinction of all species of moa some time before European contact began in the late 18th century (1–4,7). Sites from the later, “Classic” Maori period lack evidence of moa exploitation. The Classic period is characterized by earthwork fortifications (the Maori term for which is pa) and occupation deposits indicating reliance on fish, shellfish, and plants for food.

Current interpretations of moa extinction implicitly or explicitly require a period of several hundred years of gradual population attrition by hunting and habitat loss: this is the orthodox model (2, 3, 5). The moa-hunting period has been estimated to have lasted some 600 years, peaking 650 to 700 years before the present (yr B.P.) and ending about 400 yr B.P. (2, 3). Anderson (2) estimated the duration of moa hunting from a radiocarbon chronology of moa hunting sites and from moa population parameters based on extant ratites and African bovids. It is difficult to estimate the time of moa extinction from a series of dates on moa bones and from moa hunting sites, because confidence intervals for calibrated ages are greater than those for conventional radiocarbon ages (8), and additional dates could be younger than the presently perceived limit.

Reassessment of some major archaeological sites has suggested that moas were becoming scarce by the end of the 14th century (9, 10). The earliest settlement sites date from the late 13th century (11), not the 10th or 11th century as previously thought (5). A later date for first settlement would imply that moa extinction was much more rapid than suggested by the orthodox model. How might a human population have eliminated complete populations of large birds within a few decades? To explore the possibility of rapid extinction by human predation, we applied a standard Leslie matrix population model (12) to moas using parameters common to long-lived birds and the limited data that can be inferred from fossil evidence (such as clutch size) (13).

The objective was to explore the effects on the total moa population of low levels of exploitation concomitant with a small initial human population, coupled with the habitat loss induced by those people. For the model to be conservative and not to underestimate the time to moa extinction, we used an initial population of 100 people (14). We also selected low to medium human population growth rates, minimal rates of habitat removal in only two areas of the two main islands of New Zealand, and the lowest cropping rates suggested by Anderson (2). We estimated the total population of moas (all species) at the time of human settlement, from a regional analysis of distribution and abundance, to be 158,000 birds, which is about twice the population previously suggested (2). Only consumption of adult moas (>1 year old) was considered. Consumption of moa eggs, known to have been considerable (3), was ignored. At all stages, therefore, we chose population sizes and parameters that would minimize losses to the moa population, to ensure that the model would not underestimate the duration of the extinction process.

The simulations showed that moas, like most long-lived birds, were very vulnerable to any increase in adult mortality. When subjected to even a low level of human predation, moas required a disproportionate (and impossible) increase in recruitment to maintain their numbers (15). Time to extinction under all scenarios simulated was less than 160 years (Fig. 1, A through D). Even without habitat loss, the most conservative simulation suggested that moas were effectively extinct within 160 years of human colonization (Fig. 1A). Hence, the “no habitat loss” curves were useful in defining an upper bound for the duration of the exploitation/extinction process. Substantial areas of habitat were certainly lost early in the settlement period (16), together with their moa populations.

Figure 1

Simulated declines in moa populations (the number of adult females) caused by human predation, with and without habitat loss, in areas II and V (Fig. 2). (A) One female moa per 20 people per week; human population 100, increasing at 1% p.a. (B) One female per 10 people per week; human population 100, increasing at 1% p.a. (C) One female per 20 people per week; human population 100, increasing at 2.2% p.a. (D) One female per 10 people per week; human population 100, increasing at 2.2% p.a. (solid line); one female per 10 people per week; human population 200, increasing at 2.2% p.a. (long-dashed line). (E) Orthodox model of moa extinction versus maximum and minimum times to extinction from the Leslie matrix model. Leslie matrix analyses were done with RAMAS/Age (12), adapted to simulate “inverse density dependence” (increasing crop rates and decreasing prey population). Each simulation started with 78,800 adult females of all species (which is 3% above the population estimate). Survivorship values assumed are as follows (year class, survivorship): 0, 0.25; 1, 0.74; 2, 0.75; 3, 0.83; 4, 0.93; 5 and over, 0.95) based on the tendency of K-selected birds to live longer, have smaller clutches, have lower juvenile survival rates (14), and achieve a stable population given clutch size and assumed longevity. Fecundity rates based on a clutch size of 1 or 2 eggs (22) and an age of first breeding of 5 years, with full breeding potential reached in year 12, were as follows (year, chicks raised): 0 to 4, 0; 5, 0.05; 6, 0.10; 7, 0.15; 8, 0.2; 9, 0.25; 10, 0.50; 11, 0.8; 12 and over, 0.95. No immigration or emigration was assumed. The coefficient of variation (all parameters) is 2%. Predation of eggs and first-year birds was omitted. Starting values resulted in stable to slightly increasing mean populations (minimum, 57,900; maximum, 101,000 females after 50 years). Simulations were run in decadal blocks, and adult survivorship was recalculated after each decade and applied to a starting population of the mean end population (100 simulations for each mean) for the preceding decade. Adult survivorship was reduced from the initial value for the decade by the sum of the percentage loss to cropping (based on the annual crop required by the number of people at the start of each decade) and to habitat loss (23).

Habitat loss in areas II and V (Fig. 2) reduced the time to extinction for the whole fauna by decreasing adult moa survivorship before the human population had increased substantially (Fig. 1). Human population growth rate was more important in determining the time to extinction than was the rate of consumption of moas per individual. Doubling the founding human population to 200 reduced the time to extinction by nearly a third under the same model conditions (Fig. 1D). In combination with habitat loss, this “most likely” scenario resulted in moa extinction 50 years after colonization (Fig. 1D). By then, the human population would have grown to nearly 600, which is much fewer than is presently thought necessary to seriously affect moa populations (2, 3).

Figure 2

Map of New Zealand archaeological sites with radiocarbon dates that are relevant to the duration of moa hunting and of habitat areas used for the estimation of total moa abundance before human colonization. Area I (130,000 km2): tall wet forest dominated by trees of the Podocarpaceae; area II (21,000 km2): drier forest dominated by Podocarpaceae; area III (53,000 km2): high-altitude (>800 m) southern beech (Nothofagus) forest; area IV (12,500 km2): southern South Island mixed forests and shrublands dominated by Podocarpaceae; area V (65,000 km2): drier eastern forest/shrubland/grassland mosaic with abundant Podocarpaceae; area VI (6,200 km2): montane shrubland, herbfield, bare rock, and glaciers. The total moa population was based on current carrying capacities for emus (Dromaius novaehollandiae) (worst, 2.5 kg km−2; medium, 9.1 kg km−2; best, 16.8 kg km−2) (2). Moa species were assumed to occupy habitat independently, and biomass per square kilometer was taken as the sum of that for the dominant species present, using representative species body masses of the following species: Megalapteryx didinus (MEDI), 20 kg; Anomalopteryx didiformis (ANDI), 40 kg; Euryapteryx geranoides (EUGE), 95 kg; E. curtus (EUCU), 20 kg; Emeus crassus (EMCR), 50 kg; Pachyornis elephantopus (PAEL), 145 kg; P. australis (PAAU), 120 kg; P. mappini (PAMA), 20 kg;Dinornis struthoides (DIST), 100 kg; D. novaezealandiae (DINO), 150 kg; and D. giganteus(DIGI), 200 kg. The dominant species in each region were as follows (species, total birds of all species km−2, total birds). Area I: EUCU (North Island only), PAMA (North Island only), DIST, DINO, and MEDI (South Island only); 0.425; 48,025. Area II: ANDI, DINO, DIST, PAMA, and DIGI; 1.54; 31,724. Area III: MEDI (South Island), PAAU (South Island), DINO, DIST, and ANDI; 0.146; 7621. Area IV: DIGI, EUGE, DIST, ANDI, DINO, EMCR, and PAEL; 0.532; 6650. Area V: PAEL, EUGE, EMCR, DIST, and DIGI; 0.88; 57,625. Area VI: PAAU and MEDI; 0.146; 905. Habitat destroyed in areas II and V = 0.9% p.a., with human population increasing at 1.6% p.a., respectively, for the first 100 years of human settlement, with a complete loss of moas occupying the area destroyed each year (that is, no emigration of displaced birds). The actual rate of loss in both areas is about 3% p.a.; a 50% reduction (16) is taken to have occurred within 100 years of colonization.

The effect of the human population being divided into separate groups in different areas was also explored. If 50 of an original 200 people settled in area V (the richest zone of moa distribution), they would have increased to 148 after 50 years at 2.2% per annum (p.a.). In that time, in concert with habitat destruction, they could have eliminated 56,000 adult moas. The time to extinction for the moas would have varied regionally with their abundance, which in turn depended on the terrain and the carrying capacity of the dominant vegetation. Times to extinction in areas with low moa numbers would have been so short that sequential occupation of sites in different regions could have occurred within the range of uncertainty in calibrating radiocarbon dates and hence be difficult to detect. In comparison to the orthodox model, both the longest and the most likely durations simulated here resulted in much steeper declines in the moa population (Fig. 1E). Another important difference is the impossibility of any prolonged initial phase of low-level exploitation or of residual survival of the moa as a resource, which are integral parts of most versions of the orthodox model (2, 3, 5) (Fig. 1E).

To test the conflicting models, it was necessary to determine when moa hunting ceased. Evidence supporting the Leslie matrix model was found in a series of radiocarbon dates establishing a chronology for Monck's Cave (Figs. 2 through 4). The artifacts found in the cave and the materials from which they were made indicate that the site was transitional between moa-hunter and Classic sites (17). In addition, although it is in an area where moa hunting had been intensively practiced (and is close to a major moa-hunter site), Monck's Cave lacked evidence of moa consumption (17). Hence, dating the occupation layers could provide a local terminal date for the moa, assuming that people were likely to eat moa meat if it was available within human foraging range of the site.

Figure 3

Section of excavation at Monck's Cave, South Island, New Zealand, 1998, showing provenience of marine shell samples that were radiocarbon-dated: ovals, shell midden; dark shading, sand; light shading and triangles, laminated cave sediments; cross-hatching, rock. The cave interior (excavation units G13, G14, and L14) comprised two deposits: (i) an in situ remnant that yielded shell and bone (bird, fish, seal, Pacific rat, and one worked fragment of moa bone) in a cleft between two rocks in G14 and in L14; and (ii) two layers interspersed between earlier and later, washed-in, laminated microstrata (black layers in G13, clearly derived from the cultural deposit in G14). Dating samples were submitted from three arbitrary levels in G14 and from both black layers in G13 to test whether more than one occupation was indicated. The results show that the cave was occupied only once, and probably briefly. The cave exterior (excavation units K39 and I45) comprised a single, nonstratified, marine shell midden layer (not necessarily contiguous between these units). K39 is ∼350 mm of shell midden overlying clean sand. K39 was excavated in arbitrary 50-mm levels; the top, middle, and bottom levels were dated. I45 is a fire-scoop feature with marine shell associated, 200 to 400 mm below the surface. A natural shell lens, at 500 to 550 mm, returned an age of 2410 ± 50 years before the present (Wk 6775) (Wk, lab number from the Waikato University Radiocarbon Laboratory). The cave was sealed by a landslide, almost certainly at the time of occupation; later dates for the cave's exterior may reflect the earlier abandonment of the cave itself.

Figure 4

Comparison of radiocarbon ages from early sites with abundant evidence of moa hunting (Wairau Bar, Houhora, Papatowai, and Shag River Mouth), Monck's Cave, and Classic fort (pa) sites. Dates for moa exploitation sites (Orthodox) are expressed as a band encompassing the range of calibrated dates on charcoal, marine shell, and moa bone collagen accepted by Anderson (11). Dates for Monck's Cave are shown as 68% confidence limits (thick line) and 95% confidence limits (thin line) for calibrated dates (24). The band signifying dates for fortified pa summarizes available calibrated dates on various materials (25).

A series of radiocarbon determinations (Table 1) from marine shell (Fig. 3) shows that the site was occupied in the mid-to-late 14th or early 15th centuries inside the cave and perhaps into the 16th century outside (Fig. 4). The dates from Monck's Cave bridged the time between when people unequivocally hunted moas and when they clearly did not, relying instead on other foods, and built fortifiedpa (Fig. 4). The dates therefore showed that moas were not available to people in the area from the late 14th century onward. The evidence for human arrival in the late 13th century (11), and the evidence presented here for the absence of moas from the environment in a major moa-hunting area by the late 14th century, strongly support the rapid extinction model. Gradual extinction models are not supported.

Table 1

Radiocarbon dates from occupation levels at Monck's Cave, Christchurch, New Zealand. All determinations were made on apatite from marine shell (Austrovenus stutchburyi) by the Waikato University Radiocarbon Laboratory (Wk). CRA, conventional radiocarbon age in years before the present (1950), with reference to the Old (Libby) Half-life; the calibrated age is given in calendar years (24).

View this table:

Data from major moa-hunting sites such as Houhora and Tairua in the North Island, and Wairau Bar, Shag River Mouth, and Papatowai in the South Island (Fig. 2) also support the short extinction chronology. Very early sites, such as Wairau Bar, Houhora, and Papatowai show that moa were hunted from the earliest settlement phase. At the Shag River site, 265 km south of Monck's Cave, radiocarbon dates suggest a brief (<50 years) occupation in the middle of the 14th century (9,18). Moas represented about 50% of the food remains in the lower layers but only about 20% by the end of occupation there. We conclude that moas had been severely depleted in the area by this time.

The Leslie matrix model suggests that a protracted period of moa exploitation was not possible, given the life history parameters of the birds, even with the lowest conceivable numbers of Polynesian colonists. Minimal cropping rates (2) initiated a sudden precipitate population collapse, because the birds had little capacity to make good constant losses of adults. Dates for the earliest sites and the end of moa hunting imply that the period of moa-hunting was brief in comparison to the duration of Polynesian occupation of New Zealand. The small founding human population achieved a very high archaeological visibility. Because the hunters moved to new sites shortly after the supply of moas was exhausted, most sites remained intact. The high visibility of massive moa bones on the surface in sites shows that there can have been no “pre-visibility” occupation phase nor a period of slow, archaeologically cryptic, population growth from a small founding human population. Previous suggestions that proxy indicators show anthropogenic environmental change well before 1000 A.D. can be discounted (19). Any occupation of more than 5 years would be visible, both as archaeological remains and as environmental damage (16).

Rapid extinction implies that the number of moas preserved in a site could be directly related to the moa population in that area. Indeed, where the population was small enough to be exterminated in less than 10 to 20 years (as in most parts of the North Island), a site could contain most of the moa population alive in that area when people arrived.

Long-lived birds are very vulnerable to human predation of adults (20). Although habitat destruction by fire (16) increased the effective predation rate on moas, even minimal levels of human hunting pressure caused an irreversible decline in the moa population. Our relatively simple Leslie matrix simulations agreed closely with archaeological data on the beginning and end of moa hunting. The consequences for understanding the pattern and process of the peopling of the last major habitable land mass to be reached by humans are far-reaching. The elimination of the moa by Polynesians was the fastest recorded megafaunal extinction, matched only by the predictions of the “Blitzkrieg” model for North American late Pleistocene extinctions (21).

  • * To whom correspondence should be addressed. E-mail: piopio{at}


View Abstract

Navigate This Article