Cutmarked bone of drought-tolerant extinct megafauna deposited with traces of fire, human foraging, and introduced animals in SW Madagascar

Each sedimentary sequence from the three excavated ponds (Tampolove [TAMP], Ankatoke [ANKA], and Andranobe [ANDR]) includes a layer of clay (defined as zone 2), which separates the surface soil formation (zone 1) from the underlying fossiliferous muddy sand and bedrock (zone 3, Figs. S4–S7 & S9). Details regarding the composition of this sediment and its microfossils are given in Appendix-Results-Excavation (Figs. S9–S12).

Subfossils and chronology

Coastal survey recovered mostly zebu bones on exposed sandy surfaces, some pygmy hippo and giant tortoise bones on the margins of shallow ponds, and giant tortoise carapace under overhanging limestone outcrops (Appendix-Results-Survey, Fig. S3). A high proportion of surface bone failed 14C analysis (~ 55%, Table S1), yet the successfully analyzed specimens (n = 8) span up to 3390–3220 calibrated years before present (cal BP, PSUAMS 8681, 3150 ± 15 14C BP, a hippo molar). Pond deposits that are relatively deep include bones that cover a relatively long period of time (Figs. S14–S16, Dataset S6). This span ranges from ~ 6000 years at TAMP (~ 120 cm deep) to ~ 2500 years at ANDR (~ 100 cm deep), with the oldest bones present in the fossiliferous sedimentary zone 3 and scarce bones in the overlying clay (zone 2).

Zone 3

Most bones in this layer are relatively intact and include readily identifiable pygmy hippo long bones and cranial fragments (e.g., Fig. S13a,f), giant tortoise carapace and plastron fragments (Fig. S13d), ratite eggshell and long bones (Fig. S13c,m), and crocodile scutes, cranial fragments, and teeth (Fig. S13b). Scarce bones of a duck (genus Anas) were recovered at ANDR. Remains of subfossil lemurs were scarce or absent, but they may be represented by an unknown type of bone fragment identified through protein fingerprinting (ANDR-1-5-55, Dataset S3). The widespread success of collagen extraction from these bones attests to the excellent preservation of organics in this zone. ANKA also includes keratin (mostly in the form of crocodile claws, e.g., Fig. S13i), as well as two rounded agates found associated with ratite eggshell (Fig. S13m).

Remains of a juvenile pygmy hippo were recovered from both TAMP and ANDR (a femur and tibia, respectively, Dataset S3). The epiphyses of some of the pygmy hippo long bones have gnaw marks (Fig. S13f), and none of the bones include chop marks. In association with these bones towards the top of this zone are some large (> 1 cm diameter) charcoal fragments and scarce bones of bushpig (Fig. S13k) and zebu (Fig. S13e). Protein fingerprinting identified a screened fragment of a non-zebu bovid in ANKA zone 3 and confirmed that a tentatively identified bushpig canine fragment (ANKA 1-4-151) belonged to a hippo. This zone at TAMP and ANDR also includes occasional mangrove whelk (Terebralia palustris) shells (Fig. S13g). These whelks currently live at least ~ 500 m distant from these ponds, and whelk shells at ANDR each have an irregular hole above the operculum.

The span of time represented by bones in zone 3 ranges up to ~ 4000 years (~ 6000–2000 cal BP at TAMP, Fig. S14). Confirmed introduced animal bones from zone 3 failed direct 14C analysis. There are multiple examples of directly 14C-dated bone in close stratigraphic association that nonetheless differ in age by > 1000 years, and there are a couple of examples of bones from the same individual that are separated stratigraphically. For example, two giant tortoise carapace and plastron fragments from TAMP that have indistinguishable 14C ages are separated by 22 cm of sediment (PSUAMS 8670 comes from 112 cm depth, and PSUAMS 8668 comes from 90 cm depth).

Although ANKA produced what is thus far the oldest directly 14C dated pygmy hippo bone from a coastal subfossil site (PSUAMS 9383, 4380 ± 25 BP, 5030–4840 cal BP), the mean calibrated age of hippos from the Tampolove excavations (n = 11, x̄ = 2858 cal BP, SD = 972 yr) is significantly less than that of the giant tortoises (n = 9, x̄ = 4582 cal BP, SD = 705 yr, t(18) = − 4.4, p < 0.001). The success rates of directly 14C-dating pygmy hippo versus giant tortoise remains in zone 3 sediment are comparable (10/13 and 9/13, respectively). The pattern of relatively old giant tortoise remains at coastal subfossil sites is conserved through the island-wide review of 14C data (Fig. 3). However, this pattern is inverted at inland sites, and the median calibrated age of hippos from inland sites (n = 57, m = 2595 cal BP) is significantly greater than the median calibrated age of hippos from coastal lowland sites (n = 68, m = 1600 cal BP, Mann–Whitney U = 1459, p = 0.02). The limited timespan of deposition of pygmy hippo bone at low coastal sites cannot be explained by sampling bias as more pygmy hippo bones from low coastal sites have been directly 14C dated (n = 69, as opposed to n = 57 at inland sites), and relatively more low coastal sites have been sampled (n = 10, as opposed to n = 8).

Figure 3

Occurrence of Tampolove fauna on a scale of absolute time relative to local sedimentary zones (Fig. S9), approximate changes in relative sea level on Madagascar, and regional paleoclimate records from points marked in the inset map. Islandwide bone collagen 14C date count densities are given for giant tortoises and hippos and are separated according to whether data come from coastal lowlands or inland sites (Fig. 1). Bayesian change point analysis (BCPA) in the Asafora, Ranobe, and Rodrigues records (red lines) identify intervals that can be approximated reasonably well by a single mean, and the overall means of these records are given by black vertical lines. Intervals with above average values highlighted by BCPA in at least one record (indicative of relatively arid conditions) are highlighted with brown horizontal bars for reference. Note that the plot includes two pygmy hippos and two giant tortoises that are likely duplicate specimens.

Full size image

Many animals that lived near Tampolove during the deposition of zone 3 sediments between ~ 5000 and 2000 cal BP could have experienced rising sea levels30 and a wetting trend marked by the resumption of speleothem formation in multiple caves in SW Madagascar by ~ 3500 cal BP25,26,27. Directly 14C dated remains of both giant tortoises (n = 10) and pygmy hippos (n = 5) attest to the fact that these taxa persisted locally before this wetting trend (Fig. 3). The scarcity of mid-Holocene climate records from SW Madagascar and evidence for asynchronous climate change between the northern and southern parts of the island27 complicate inference of mid-Holocene climate change in the region. However, speleothem records from Rodrigues Island (~ 2100 km distant from our study site yet generally coherent with records from NW Madagascar) suggest that “megadroughts” affected at least parts of the SW Indian Ocean during approximately 4760–4600 cal BP and 3880–3280 cal BP28. Previously published 14C data from bones of pygmy hippos collected from low coastal sites outside of the Tampolove area (n = 41) do not span these intervals, with the exception of a hippo from Nosy-Ve/Nossi-Vey (PSUAMS 5424, 4125 ± 25 14C BP, 4810–4440 cal BP). However, the longest possibly relatively arid interval (~ 600 years, 3880–3280 cal BP) likely encompassed the death of six analyzed individuals: 4 hippos, 1 giant tortoise, and 1 crocodile (with calibrated 95% intervals that span all of the ~ 600-year dry interval).

Zone 2

Bones in this layer are relatively scarce, fragmentary, and chalky, yet readily identifiable fragments of hippo (Fig. S13h), giant tortoise, crocodile, and zebu cattle are present. At TAMP, protein fingerprinting identified a small fragment of a pygmy hippo long bone as shallow as 24 cm depth (TAMP 1-2-48). TAMP zone 2 includes both a chopped distal fragment of a pygmy hippo right femur (Fig. 4) and an associated scatter of marine fish bones (cranial fragments, vertebrae ~ 1 cm in diameter, and spines), one of which (a vertebra) includes a chop mark (Fig. S13j). Both the chopped pygmy hippo fragment and associated fish bones failed 14C analysis due to the exceptionally poor preservation of bone collagen. Charcoal fragments with provenience in zone 2 come from only ANKA and ANDR, and shells of the mangrove whelk are relatively abundant in ANDR zone 2.

The only directly dated collagen from zone 2 (extracted from a fragmentary pygmy hippo molar from ANKA, 55 cm depth, PSUAMS 8733, 3555 ± 20 14C BP, 3880–3700 cal BP, Fig. S15) is > 2000 years older than a closely associated charcoal sample (38 cm depth, PSUAMS 8849, 575 ± 30 14C BP, 630–510 cal BP), which makes this molar comparable in age to bone from zone 3. Consequently, the youngest directly 14C-dated ancient bone from the Tampolove excavations comes from the lowermost zone 3: a pygmy hippo’s vertebra recovered at 90 cm depth at TAMP (PSUAMS 8730, 1865 ± 15 14C BP, 1819–1705 cal BP). Though poorly constrained in time, the deposition of zone 2 sediment came sometime within the past two millennia, which witnessed marine regression and dry intervals recorded in both the δ18O record of a nearby speleothem27 and the salinization of a nearby pan36. Previously directly 14C-dated bone collected around Tampolove attests to the local persistence of at least pygmy hippos and giant tortoises until the start of the last millennium (n = 15), and an atlas from Lamboara/Lamboharana is in fact the most recent confidently dated pygmy hippo bone from the island (PSUAMS 5629, 1100 ± 15 14C BP, 980–930 cal BP).

Figure 4

Cutmarked pygmy hippo femur recovered from Tampolove during recent excavation at ~ 40 cm depth (TAMP-1-2-61, above), and previously-recovered and directly 14C-dated (~ 3500 and 1600 cal BP37) cutmarked pygmy hippo femora from the nearby site of Lamboara/Lamboharana that are currently housed in the National Museum of Natural History in Paris (MAD 1709 & MAD 1710, below). Four views highlight three locations of cutmarks on the broken shaft of TAMP-1-2-61, and the inset frames show 20 × magnification of these areas, with corresponding orientations given by red lines. Note that the false color insets of TAMP-1-2-61 are meant to highlight linear edges and crevices, and the overview photos of all three femur fragments are on the same scale.

Full size image

Zone 1

A fragment of iron (from TAMP, 16 cm depth) and sparse ceramic fragments (from ANKA, 3 & 9 cm depth) are present only in zone 1, and three 14C dates from TAMP and ANKA suggest that these specimens span the past ~ 200 years (Figs. S14–S15).


The directly 14C dated charcoal spans all three stratigraphic zones yet consistently dates to the past millennium (Figs. S14–16). Multiple charcoal samples from different excavated ponds have practically indistinguishable 14C ages (Table S2), and much of the charcoal from Tampolove formed during peaks in the deposition of macrocharcoal at nearby Namonte (17 km distant; Fig. 5A). The onset of directly 14C-dated charcoal deposition approximately coincides with a decrease in Asafora speleothem δ18O values and with multiple directly 14C-dated first and final local occurrences of large animals. While directly 14C dated charcoal is limited to the past millennium, microcharcoal particles were abundant in all TAMP sediment samples (x̄ ± SD = 2.0 × 106 ± 2.8 × 106 particles). Additionally, microcharcoal is relatively abundant near the bottom of TAMP and ANKA, which contains bones that span ~ 6000–2000 cal BP (Fig. 5B).

Figure 5

Records of fire, drought, and faunal turnover from the vicinity of Tampolove within the past 1200 years, with dashed horizontal lines for reference (5A), and macrocharcoal concentrations from the excavated ponds, with depth intervals containing directly 14C-dated charcoal that spans the past millennium marked in red (5B). The past 1200 years includes the entire summed calibrated distribution of the 10 directly dated prebomb charcoal fragments from the Tampolove excavations. The calibrated probability distributions associated with the latest dates from endemic megafauna bone (giant tortoises and pygmy hippos) and earliest dates from introduced animal bone (zebu cattle and bushpigs) are shown as black distributions, and 95% of each distribution is bracketed. Considering directly dated remains within the past 4 ka from hippos (n = 26), giant tortoises (n = 18), and zebu (n = 9) and the assumption that bones were deposited uniformly over time, the grey distributions and bracketed 95% credible intervals give estimates of extirpation and arrival times. As in Fig. 3, the red line on the Asafora record follows from BCPA.

Full size image

Source: Ecology -

Machinery of the state

Extinction magnitude of animals in the near future