Hominin adaptations in the Lesser Sunda Islands: exploring the vertebrate record to investigate fauna diversity before, during and after the Last Glacial Maximum

ABSTRACT

This paper reviews the available vertebrate record from the Lesser Sunda Islands to explore the effect the Last Glacial Maximum had on human subsistence strategies. By focusing on vertebrate assemblages from Laili and Matja Kuru 2 in Timor Leste, Tron Bon Lei in Alor Island, and Here Sorot Entapa in Kisar, this paper investigates biodiversity and resource availability in these nearby islands through the application of standardising indices and statistical testing. Results indicate that vertebrate biodiversity remained fairly stable through and after the Last Glacial Maximum, suggesting that in terms of available mammals, birds and reptiles, this period did not led to severe resource depletion. Hence, potential variations in human subsistence practices or occupation dynamics might not be due to changes in vertebrate diversity. As such, this analysis contributes to investigating anatomically modern humans’ subsistence adaptation in the Lesser Sunda Islands pre- and post-Last Glacial Maximum.

KEYWORDS:

• Lesser Sunda Islands
• Last Glacial Maximum
• zooarchaeology
• anatomically modern humans
• subsistence
• biodiversity

1. Introduction

Archaeological research in Wallacea, the biogeographic region separating the Asian and Australian continental shelves, is igniting exciting research questions. These include the antiquity of human occupations and artistic manifestations (e.g. Brumm et al. 2017, 2021) migratory routes and the development of navigation techniques (O’Connor and Veth 2000; O’Connor 2007; Kealy, Louys, and O’Connor 2016, 2018; O’Connor et al. 2017; Norman et al. 2018), insights into genetic diversity from the mid-Holocene onwards (McColl et al. 2018; Purnomo et al. 2021), diversity of cultural practices (Langley, Clarkson, and Ulm 2019; Samper-Carro et al. 2022), as well as the effect of human interactions with endemic biota (e.g. Louys et al. 2021).

Research in Wallacea has traditionally focused on the large islands of Sulawesi, Flores and Timor. Since the beginning of new excavations in Timor-Leste in 2000, a number of sites dating back from the Pleistocene, before and through the Last Glacial Maximu (LGM) have been documented (O’Connor et al., 2002; O’Connor and Veth 2005; O’Connor 2007; O’Connor and Aplin 2007; O’Connor et al. 2010; O’Connor, Ono, and Clarkson 2011; Hawkins et al. 2017). To date, the archaeological deposits from Timor-Leste yielded some of the earliest dates for anatomically modern human (AMH) occupations in Wallacea, with Laili (44.6 ka cal BP), Asitau Kuru (former Jerimalai; ca.42 ka cal BP) and Lene Hara (ca.42 ka cal BP) providing the oldest dates (Figure 1). Sites excavated in Timor-Leste have evidence of episodic occupation, although when the occupations documented in every site are combined, they suggest a continuous occupation of the coastal regions from ca. 44 ka, throughout the LGM, across the Pleistocene-Holocene boundary and throughout the Holocene (O’Connor and Aplin 2007).

Figure 1. Map showing the location of the archaeological sites mentioned in the text.

The modern native terrestrial fauna in Timor is poor, limited to a small murid and a tiny shrew, and introduced species such as the New Guinean cuscus (Phalanger orientalis), and Asian domesticates and commensals, translocated in the Late Holocene (Glover 1986; O’Connor 2015). Fossils of two pygmy stegodons, a giant land tortoise, a large lizard and four to seven genera of giant rat have been reported in Timor (O’Connor and Aplin 2007; Aplin and Helgen 2010; Louys, Price, and O’Connor 2016; Louys et al. 2018).

However, not only these larger Wallacean islands have yielded evidence of human occupations. Smaller island are also suitable to sustain human populations, especially when considering how factors such as distance to another landmasses, or the degree of maritime skills, can dramatically affect insular habitability (Keegan et al. 2008; Fitzpatrick et al. 2016; O’Connor et al. 2018). In recent years, a series of archaeological excavations and surveys in the Lesser Sunda islands (i.e. the archipelago from Bali to the West to Wetar to the East; Figure 1) are yielding interesting data from these previously poorly explored insular environments (e.g. Louys et al. 2017; O’Connor et al. 2017). Since 2014, excavations in the islands of Alor, Kisar, Lembata and Pantar are demonstrating how these small insular environments were suitable for human survival from, at least, the late Pleistocene onwards (Samper Carro et al. 2016, 2018, 2019, 2021, 2022; O’Connor et al. 2018; Hawkins et al. 2020).

The land-based fauna of the Lesser Sunda Islands is impoverished, consisting only of reptiles (lizards and snakes), several rodent species, shrews, a few species of bats and several birds species (Trainor 2005, 2007; Trainor et al. 2006, 2008). However, the diversity of marine ichthyofauna is immense, with thousands of species identified. The introduction of domesticates and agriculture has transformed human subsistence practices in most of these islands.

The Last Glacial Maximum (LGM) – dated from 26,500 to 19,000 years ago based on relative sea level data – entailed the most recent interval in Earth history when ice sheets reached their maximum volume (Mix, Bard, and Schneider 2001; Clark et al. 2009). In the Northern Hemisphere, especially in northern Europe and America, the limit of the ice sheets during the LGM and their subsequent retreat are well documented, with plentiful records for local mean sea levels (e.g. Yokoyama et al. 2000; Lambeck and Chappell 2001; Lambeck et al. 2014). In the Southern Hemisphere, changes in rainfall and monsoon regimes led to a southward shift of the Intertropical Convergence Zone (ITCZ) (Denniston et al. 2013, 2017). These shifts affected the Indo-Pacific Warm Pool (IPWP), the ‘global heat engine’, affected its size and other processes, such as deep atmospheric convection and sea surface temperatures (SST) (DiNezio and Tierney 2013; De Deckker 2016). Additionally, marine sediment samples from offshore Borneo yielded no evidence of a pronounced dry season. For the Lesser Sunda Islands, palynological analysis indicate a reduced overall precipitation, with great interannual variability in some areas (Hope 2001, 2005; O’Connor and Aplin 2007). Sediment samples from cores extracted offshore Sumba, dated to the LGM, indicate enhanced aridity and water stress during the dry season, with dominance of grassland -C⁴ vegetation- (Dubois et al. 2014).

The large number of vertebrate and invertebrate remains recovered in recent excavations in the Lesser Sunda Islands are suitable proxies to investigate the effect that the LGM had on vertebrate diversity. Through the standardisation and statistical testing of the vertebrate assemblages from two coastal sites – Tron Bon Lei (Alor) and Heret Sorot Entapa (Kisar)- and two inland sites – Laili and Matja Kuru 2 (Timor-Leste) – this paper explores the current evidence for vertebrate fauna accumulations dating prior, during and post LGM in these islands. By combining palaeoecological observations based on taxonomic composition with statistical testing of raw and standardised values, new data on human settlement dynamics and fauna diversity in these insular environments from the late Pleistocene to the mid-Holocene are presented.

2. Material and methods

2.1. Archaeological context

2.1.1. Laili (Laleia, Timor-Leste)

Laili is a partially collapsed cave located ca. 86 m asl, about 350 m east to the closest water source, the Laleia River, and 4.3 km from the coast. Three occupational phases are identified in Laili: an earliest period, dated to the Late Pleistocene (44.6–27 ka cal BP); a second period coinciding with the LGM (22.5–18.5 ka cal BP) and a younger period dated to the terminal Pleistocene (17.5–11.2 ka cal BP) (Hawkins et al. 2017). A large number of stone artefacts were recovered, with the highest concentration documented in stratigraphic units postdating the LGM (Hawkins et al. 2017; Table 1).

Table 1. Total number of remains (NR) for each of the sites analysed in this paper.

		Laili			Matja Kuru 2		Tron Bon Lei		Here Sorot Entapa
		pre-LGM	LGM	post-LGM	pre-LGM	post-LGM	LGM	post-LGM	post-LGM
Mammalia		33	14	97			43	96	9
	Small muridae	2272	1456	797	6678	9653	100	39	166
	Large muridae	*	*	*	377	1439	8	1	12
	Giant muridae	74	76	64	234	590	5	0	2
	Chiroptera	13	29	49	73	446	0	1	43
Reptilia				3	0	0
	Lacertilia	66	53	23	87	200	7	8	160
	Testudines	1	1	0	520	898	8	12	97
	Serpentes	27	25	57	344	800	199	15	2265
Amphibia					0	0
	Anura	60	31	13	89	310		2
Aves
	Aves	274	294	162	117	241	15	10	197
Unid. Tetrapoda		1647	1001	957	8571	24,197	24	25	602
Total Tetrapoda		4467	2980	2222	17,090	38,774	409	209	3553
Total Icthyes		1	0	1	84	283	3770	11,571	93,672
Total NR		4468		2223	17,174	39,057	4179	11,780	97,225

Vertebrate remains were recorded throughout the archaeological sequence, being most abundant in the earlier periods of occupation and declining over time (Hawkins et al. 2017). The faunal assemblage is dominated by mammals, with sixteen mammal taxa documented, including five small and four extinct giant rodents, bats, birds and translocated domesticates (Hawkins et al. 2017). The zooarchaeological material contains evidence of the exploitation of aquatic resources, with abundant molluscs (NISP = 6,686) and few remains of freshwater (i.e. eel) and marine bony fishes (parrotfish), mostly found in the earlier and terminal Pleistocene period, with an abrupt decline during the LGM (Hawkins et al. 2017).

2.1.2. Matja Kuru 2 (Poros, Timor-Leste)

Matja Kuru 2 (MK2) is a large cave in an uplifted limestone (ca. 370 m asl), located ca. 8 km inland, north of the Lake Ira Lalaro, the largest freshwater source in Timor. Three occupational phases, based on radiocarbon dates on charcoal and shell, were recorded at MK2: a Late Pleistocene phase (40–31 ka cal BP), an early Holocene phase (11–9.5 ka cal BP) and a late Holocene phase (3.5–2 ka cal BP) (Veth, Spriggs, and O’Connor 2005; O’Connor, Robertson, and Aplin 2014; Roberts et al. 2020).

The faunal assemblage from Matja Kuru 2 is dominated by terrestrial resources, with abundant remains of murids (including giant murids), reptiles, birds and bats, including marine shellfish and fish, as well as a freshwater chelid turtle (O’Connor and Aplin 2007; O’Connor, Robertson, and Aplin 2014; Table 1). Matja Kuru 2 also produced a complex bone artefact from the Late Pleistocene units, marine shell beads, and a dog burial dated to the Holocene (O’Connor 2006; González et al. 2013; O’Connor, Robertson, and Aplin 2014). The faunal assemblages from these inland sites contrast with those recovered in the Pleistocene coastal sites, which show a focus on aquatic resources.

2.1.3. Tron Bon Lei (Alor, Indonesia)

Tron Bon Lei are two adjacent rock shelters located ca. 33 m asl and 160 m inland in a volcanic ridge on the southwest coast of Alor Island, near Lerabain village. In 2014, three test pits were excavated in Tron Bon Lei, squares A and C in the west-facing shelter and square B in the southerly facing shelter, with two more squares (D and E) opened adjacent to square B in 2018 (Samper Carro et al. 2016, 2017, 2019, 2022). The squares in the southern shelter (B, D and E) were excavated down to bedrock at a depth of 3.2 metres, yielding 24 radiocarbon dates from charcoal and marine shell. Modelled dates suggest four occupational phases: a Late Pleistocene phase dated to 21–18.9 ka cal BP, a terminal Pleistocene-early Holocene phase (12.5–9.5 ka cal BP), a middle Holocene phase (8.3–7.2 ka cal BP) and a late Holocene phase (5.4–3 ka cal BP) (Samper Carro et al. 2016, 2019).

The faunal assemblage from square B in Tron Bon Lei produced a total of 49,984 remains (NR), largely dominated by marine fish, and distributed unevenly throughout the sequence (Samper Carro, Louys, and O’Connor 2017; Table 1; S1). Phase I (Late Holocene) yielded few remains (NR = 1,321) compared to Phase II (NR = 19,843), date to the middle Holocene. Phase III, dated to the terminal Pleistocene-early Holocene, produced the largest number of remains overall (NR = 21,994). Phase IV (Late Pleistocene) yielded a total of 6,740 remains.

2.1.4. Here Sorot Entapa (Kisar, Indonesia)

The island of Kisar, 25 km northeast of Timor-Leste, is the smallest island of Wallacea known to have been occupied in the Pleistocene (O’Connor et al. 2018). There, Here Sorot Entapa shelter, located about 24 m asl and approximately 80 m from the coast, has evidence of human occupation in two chronological phases: Phase 1, ranging from the terminal Pleistocene to the early Holocene (15.5–9.5 ka cal BP); and Phase 2 dated to the Middle-Late Holocene (4.9–1.8 ka cal BP). The archaeological deposits contained stone artefacts, animal bone and a large quantity of molluscs, urchin and crab, as well as, fishhooks and beads (O’Connor et al. 2018).

Fish bones dominate the faunal assemblage, with 26 families identified (Table 1; S1). Phase 1 contains a larger number of remains, as well as a higher taxonomic diversity than Phase 2 (O’Connor et al. 2018). The rest of the faunal assemblage includes small numbers of reptiles, large and small murid rodents, birds, fruit bats and some marine turtles, with the introduction of Pacific rat, black rat and domestic dogs during the Late Holocene.

2.2. Methods

2.2.1. Analysis of vertebrate remains

The complete archaeological assemblages from one of the excavated 1x1m squares from Laili, Tron Bon Lei and Here Sorot Entapa have been published elsewhere (Hawkins et al. 2017; Samper Carro, Louys, and O’Connor 2017; O’Connor et al. 2018). For the purpose of this study, only basic quantitative units for vertebrate remains (i.e. Number of identifiable specimens; NISP) were considered.

For Matja Kuru 2, the results from the initial zooarchaeological analysis of four of the six 1 × 1 squares excavated are presented here for the first time. Bones were sorted into major categories (Class) and identified to the lowest taxonomic level by reference to the ANU Archaeology and Natural History osteological collection. Vertebrate remains were quantified by recording the total number of remains, including identifiable and unidentifiable identifiable specimens (NR). For Muridae, specimens were classified into three size categories: giant, large and small murid.

To reassure that differences in the relative abundance of taxa among sites are not related to inter-observer errors, the categories and counts published for the assemblages from Laili and Here Sorot Entapa (originally not sorted and classified by the same researcher) were revisited by the author.

2.2.2. Density index calculation

In order to be able to compare the data obtained from the four sites object of study, a standardisation procedure was applied to control whether peaks on faunal remains were due to differences in the excavated volume per spit or otherwise, were the effect of significant variations in the density of material documented. Following methods previously published (e.g. Geiling et al. 2018), a density index (DI) was calculated by dividing the number of artefacts (NISP or NR) in each spit by the spit depth in cm (Zspit), which was calculated as the difference between the top (Ztop) and bottom (Zbottom) depth measured in cm, following the formula:

$DI=\frac{Nspit}{Zspit}$

Values for the indexes were later correlated with the chronostratigraphic sequence of the site, in order to provide an overview of occupation dynamics in each of the sites during the different occupational phases. Finally, the indexes were compared between sites to provide a general view of occupation dynamics in the Lesser Sunda Islands during the Pleistocene/early Holocene.

2.2.3. Statistical testing

Statistical testing was conducted to examine (a) samples biodiversity in each site across different chronological periods, (b) distribution of taxa between sites and chronological periods, and (c) effect of climatic events and inter-site variation on the calculated density indexes for fauna. These tests were run separately for tetrapods and ichthys (i.e. cartilaginous and bony fishes), unless otherwise indicated.

Samples biodiversity (including tetrapods and ichthys) was calculated through a dominance index (Simpson index, D) and an information statistic index (Shannon index, H). Unbiased options were selected for both indexes as they are less sensitive to sample size (Hurlbert 1971; Chao and Shen 2003), and computed as:

$D_u=\frac{\sum _i{n}_i\left(n_i-1\right)}{n\left(n-1\right)}$

$H_u=H-\left(S-1\right)/\left(2n\right)$

Where n is the total number of individuals and S the number of taxa. For each index, 95% centred confidence intervals were computed with a bootstrap procedure. In this procedure, a given number of random samples (9,999) are produced, each with the same total number of individuals as in the original sample. For each individual in the random sample, the taxon is chosen with probabilities proportional to the original abundances (Hammer, Harper, and Ryan 2001).

Two nonparametric Kruskal-Wallis H test (i.e. one-way ANOVA on ranks) were conducted to test whether there was a similar distribution of taxa between (a) sites and (b) chronological periods. Correspondence analysis (CA) was chosen as an appropriate algorithm to compare associations between NISP and difference occupational phases (Legendre and Legendre 1998). The CA finds the eigenvalues and eigenvectors of a matrix containing the Chi-squared distances between samples. CA also gives the percentages of similarity accounted for by these components (Hammer, Harper, and Ryan 2001).

A two–way analysis of variance (ANOVA) was performed to examine the effect of climatic events in different chronological periods (i.e pre-LGM, LGM, and post-LGM) and inter-site variations on density indexes of faunal remains (including tetrapods and ichthys) as an effect of occupation densities. Residual analysis was performed to test for two-way ANOVA’s assumptions. Outliers were assessed by inspection of a boxplot, normality was assessed using Shapiro-Wilk’s normality test and homogeneity of variances was assessed by Levene’s test. Statistical tests were performed on PAST v.4.09 (Hammer, Harper, and Ryan 2001) and IBM SPSS Statistics v.28 (IBM SPSS Statistics for Windows 2021).

3. Results

3.1. Taxonomic composition

3.1.1. Pre-LGM (45,000-27,000 cal BP)

Occupations dating prior to the peak of the Last Glacial Maximum are documented in phase III of Laili (27,000–44,600 cal BP) and phase III of Matja Kuru 2 (31,000–42,000 cal BP). The tetrapod composition in both sites is dominated by, at least, two genera of small Muridae (i.e. Komodomys sp., Melomys sp.) with different size individuals for both genus (Hawkins et al. 2017). Giant Muridae, consisting of at least two different genera, were identified in both sites, while the presence of large Muridae in Laili has not been confirmed, although this could be an effect of analysis bias (i.e. skeletal elements misclassified as small or giant murids). The terrestrial or arboreal habitat of the giant murids could not be defined.

The rest of tetrapods identified consist of Chiroptera, including mega- and microbats, Lacertilia (likely skink or gecko species), Serpentes, Anura, and Aves (Table 1). The avifauna in Laili included two species of quail, characteristic of open grasslands or wetland, as well as three species of songbirds (Passeriformes), which are likely to be forest or woodland species (Hawkins et al. 2017). Specimens from the order Testudines were identified in both sites, although there are a notably larger number of remains in the Matja Kuru 2 (Figure 2). Some of the skeletal elements identified in Matja Kuru 2 might be classified as Chelodina mccordi timorensis (Kuchling et al. 2007; Eisemberg et al. 2016), an endemic freshwater species, while no evidence of marine turtles was recorded in Laili and Matja Kuru 2. The ichthyofauna identified in both sites consist of a small number of mostly unidentified bony fish remains (vertebrae, spines, ribs and cranial fragments). Identified taxa include freshwater eel (Anguillidae), sea breams (Sparidae) and wrasses (Labridae).

Figure 2. Distribution of number of tetrapod remains per site in occupational phases dated to before the Last Glacial Maximum. Abbreviations: MK2 = Matja Kuru 2, followed by square (B, C, D or DD).

Diversity indices indicate that the taxonomic composition in these occupations is relatively uneven. Laili yielded a dominance score 0.68 (95% CI, 0.65–0.70), with similar scores for squares B and C in Matja Kuru 2, while square DD in this site scored 0.52 (95% CI, 0.49–0.56). Shannon indices range from 0.74 in Laili (95% CI, 0.69–0.78) to 1.10 (95% CI, 1.02–1.17) in square DD from Matja Kuru 2 (Table 2; Figure 3).

Table 2. Diversity indices calculated for the sites analysed. 95% Confidence interval in brackets.

	Laili			Tron Bon Lei		Here Sorot Entapa
	pre-LGM	LGM	post-LGM	LGM	post-LGM	post-LGM
Taxa_S	9	8	8	19	22	35
Individuals	2788	1965	1166	495	1133	6425
Dominance_D	0.68 (0.65–0.70)	0.57 (0.55–0.60)	0.49 (0.46–0.53)	0.22 (0.20–0.25)	0.18 (0.16–0.19)	0.17 (0.16–0.18)
Shannon_H	0.74 (0.69–0.78)	0.91 (0.86–0.97)	1.11 (1.04–1.17)	1.96 (1.86–2.07)	2.19 (2.12–2.25)	2.33 (2.30–2.37)
	MK2_B		MK2_C		MK2_DD		MK2_D
	pre-LGM	post-LGM	pre-LGM	post-LGM	pre-LGM	post-LGM	pre-LGM	post-LGM
Taxa_S	11	11	10	9	10	8	9	8
Individuals	2104	1865	1288	4776	1180	2706	3951	5234
Dominance_D	0.68 (0.66–0.71)	0.29 (0.27–0.31)	0.67 (0.63–0.70)	0.43 (0.41–0.44)	0.52 (0.49–0.56)	0.57 (0.55–0.59)	0.61 (0.59–0.63)	0.51 (0.49–0.53)
Shannon_H	0.81 (0.75–0.86)	1.66 (1.61–1.71)	0.84 (0.77–0.91)	1.29 (1.26–1.32)	1.1 (1.02–1.17)	1.02 (0.97–1.07)	0.92 (0.88–0.96)	1.15 (1.11–1.18)

Figure 3. Diversity Indices calculated per site in occupations dated to before (pre-) and after (post-) the Last Glacial Maximum. Abbreviations: MK2 = Matja Kuru 2, followed by square (B, C, D, DD); TBL = Tron Bon Lei; HSE = Here Sorot Entapa.

3.1.2. LGM (22,000-18,500 cal BP)

Occupations coinciding with the Last Glacial Maximum are documented in phase II from Laili (18,500–22,000 cal BP) and phase IV from Tron Bon Lei (18,900–21,000 cal BP). In Laili, taxonomic composition is dominated by small murids (genus Komodomys and Melomys), followed by Aves (songbirds and quails), and two genera of giant murids. In Tron Bon Lei, there is a large amount of snake remains, followed by small murids and small numbers of other taxa (Figure 4). However, tetrapod fauna represents 2% of the total NR, with abundant remains of cartilaginous and bony fish specimens. Identified fish families include groupers (Serranidae), triggerfish (Balistidae), wrasses (Labridae) and snappers (Lutjanidae) (Table 1; S1).

Figure 4. Distribution of number of tetrapod remains per site in occupational phases dated to the Last Glacial Maximum. Abbreviations: TBL = Tron Bon Lei.

Simpson index produced a score 0.57 (95%CI, 0.55–0.60) for Laili, while the dominance score in Tron Bon Lei is 0.22 (95%CI, 0.20–0.25), indicating a more diverse taxonomic composition (Table 2). This result is supported by lower values of the Shannon index for Laili (H = 0.91 95%CI, 0.86–0.97) compared to Tron Bon Lei (H = 1.96 95%CI, 1.86–2.07).

3.1.3. Post-LGM (17,500–9,500 cal BP)

Every site included in this analysis preserved occupational phases dated to after the Last Glacial Maximum, with chronologically overlapping occupations. The tetrapod fauna is dominated by small murids – two genera identified in Laili (Hawkins et al. 2017) – with a peak on snake remains in Here Sorot Entapa (Figure 5). The number of Chiropteras remains in every site increase from previous occupational phases, including the presence of large fruit bats (Pteropodidae) and several genera of microbats (Yangochiroptera). Regarding Testudines, remains of marine turtles (Chelonioidea) are documented in Tron Bon Lei and Here Sorot Entapa, with the presence of freshwater turtles (Chelodina) only in Matja Kuru 2.

Figure 5. Distribution of number of tetrapod remains per site in occupational phases dated to before the Last Glacial Maximum. Abbreviations: MK2 = Matja Kuru 2, followed by the square label (B, C, D or DD); TBL = Tron Bon Lei; HSE = Here Sorot Entapa.

There is a large amount of ichthyofauna documented in Tron Bon Lei and Here Sorot Entapa, representing a total of 26 fish families (Table 1; S1). Identified families include a majority of species from reef/near shore habitats (neritic zone), although some families identified in both assemblages, such as Carangidae (jacks and trevallies) and Scombridae (tuna and mackerels), contain open ocean/pelagic species.

Diversity indices show differences between the sites analysed, with inland sites scoring larger D values (0.29 in square B to 0.57 in square DD of Matja Kuru 2) than Tron Bon Lei (D = 0.18, 95%CI, 0.16–0.19) and Here Sorot Entapa (D = 0.17, 95% CI, 0.16–0.18) (Fig. X). Shannon indices indicate diverse taxonomic composition in every site, with Here Sorot Entapa (H = 2.33) and Tron Bon Lei (H = 2.19) yielding the higher scores (Table 2; Figure 3).

3.2. Statistical testing taxonomic distribution

A Kruskal-Wallis H test was run to determine if there were differences in the taxonomic distribution (i.e. species representation) between the different sites analysed. Distribution of species representation scores for tetrapods was similar for all sites, as assessed by visual inspection of the boxplots generated (Figure 6). Medians and distribution of tetrapods representation scores were not statistically significantly different between sites (Table 2). Conversely, medians and distribution of ichthys representation scores were statistically significantly different for several fish families, which representation is restricted to the coastal sites of Tron Bon Lei and Here Sorot Entapa (Table 3). Distribution was also dissimilar in these sites.

Table 3. Results of the Kruskal-Wallis test (X2) for the median values per species across the sites analysed.

	N	X2	d.f.	Asymptotic Sig. (2-sided test)
Small murids	14	9.85	6	0.13
Large murids	11	6.41	5	0.27
Giant murids	13	8.73	6	0.19
Chiroptera	14	6.52	6	0.37
Lacertilia	14	10.5	6	0.11
Testudines	14	10.45	6	0.11
Serpentes	11	8.35	6	0.21
Anura	14	7.75	6	0.26
Aves	13	10.6	6	0.1
Carcharhinidae	14	12.95	6	0.04
Anguillidae	14	4.15	6	0.66
Muraenidae	14	13	6	0.04
Belonidae	14	9.75	6	0.14
Holocentridae	14	12.95	6	0.04
Acanthuridae	14	12.78	6	0.05
Carangidae	14	12.95	6	0.04
Cirrhitidae	14	13	6	0.04
Gerreidae	14	13	6	0.04
Haemulidae	14	13	6	0.04
Kyphosidae	14	13	6	0.04
Labridae	14	12.01	6	0.06
Lethrinidae	14	12.95	6	0.04
Lutjanidae	14	12.78	6	0.05
Mullidae	14	13	6	0.04
Nemipteridae	14	13	6	0.04
Pempheridae	14	13	6	0.04
Priacanthidae	14	13	6	0.04
Scaridae	14	10.21	6	0.12
Sciaenidae	14	13	6	0.04
Scombridae	14	12.78	6	0.05
Serranidae	14	12.95	6	0.04
Sparidae	14	9.75	6	0.14
Balistidae	14	12.95	6	0.04
Balistoidei	14	13	6	0.04
Diodontidae	14	12.88	6	0.05
Ostraciidae	14	9.75	6	0.14
Tetraodontidae	14	13	6	0.04

Figure 6. Boxplots showing the distribution representation scores for tetrapods across sites. For site.

The second Kruskal-Wallis H test, aimed at investigating differences in species representation between the different chronological events analysed, yielded similar results. Taxonomic distribution of scores in tetrapods and ichthys was similar when comparing before, during and after Last Glacial Maximum occupations (Figure 7). Medians and mean ranks of species distribution scores were not statistically significantly different between chronological periods (Table 4).

Table 4. Results of the Kruskal-Wallis test (X2) for the median values per species across the chronological periods analysed.

	N	X2	d.f.	Asymptotic Sig. (2-sided test)
Small murids	14	1.41	2	0.5
Large murids	11	2.1	2	0.35
Giant murids	13	0.04	2	0.98
Chiroptera	14	4.91	2	0.09
Lacertilia	14	0.75	2	0.69
Testudines	14	2.82	2	0.24
Serpentes	14	1.19	2	0.55
Anura	14	1.3	2	0.52
Aves	13	0.78	2	0.68
Carcharhinidae	14	2.1	2	0.36
Anguillidae	14	3.9	2	0.14
Muraenidae	14	1	2	0.61
Belonidae	14	2.15	2	0.34
Holocentridae	14	2.06	2	0.36
Acanthuridae	14	2.06	2	0.36
Carangidae	14	2.06	2	0.36
Cirrhitidae	14	1	2	0.61
Gerreidae	14	1	2	0.61
Haemulidae	14	1	2	0.61
Kyphosidae	14	1	2	0.61
Labridae	14	0.55	2	0.76
Lethrinidae	14	2.06	2	0.36
Lutjanidae	14	2.06	2	0.36
Mullidae	14	2.06	2	0.36
Nemipteridae	14	2.06	2	0.36
Pempheridae	14	2.06	2	0.36
Priacanthidae	14	2.06	2	0.36
Scaridae	14	3.81	2	0.15
Sciaenidae	14	1	2	0.61
Scombridae	14	2.06	2	0.36
Serranidae	14	2.06	2	0.36
Sparidae	14	0.38	2	0.83
Balistidae	14	2.06	2	0.36
Balistoidei	14	1	2	0.61
Diodontidae	14	2.19	2	0.34
Ostraciidae	14	2.15	2	0.34
Tetraodontidae	14	1	2	0.61

Figure 7. Boxplot showing the distribution representation scores for tetrapods across chronological periods, indicating the sites that produced extreme outliers (stars) and regular outliers (dots). For site abbreviations, see previous figures.

Axis 1 and 2 of the correspondence analysis (CA) explain 90.55% and 7.43% of the variance in the abundances of taxa, respectively (Figure 8). The different occupational phases documented in each site group together, indicating inter-site instead of intra-site changes in taxonomic compositions. There is a clear division in axis 1 between inland and coastal sites, with negative values for Tron Bon Lei and Here Sorot Entapa due to the dominance of fish bones (98% of NR in Tron Bon Lei, 96% in Here Sorot Entapa) versus tetrapods in these sites. Laili and Matja Kuru 2 appear clustered together, with similar positive values in axis 1 due to the large number of murids in these sites. Along axis 2, Laili show more positive scores than Matja Kuru 2. Differences in the values between these two sites resides in the differences in numbers between bird remains (6%-10% of total NR in Laili) and turtle remains (2%-3% of total NR in Matja Kuru 2). Overall, the CA demonstrates a clear division between inland and coastal sites based on taxonomic composition, with site-specific variations related to frequency of species but regardless of timing of the occupational phases.

Figure 8. Plot of the correspondence analysis comparing abundance of taxa per site across chronological periods. Yellow dots = pre-LGM; Black dots = LGM; Purple dots = post-LGM. For site abbreviations, see previous figures.

3.3. Statistical testing of density indexes

Outliers were identified in most of the sites for the different chronological periods of occupation (Figure 9). These outliers could be an effect of differences in occupation densities (i.e. larger indexes for more intense occupations and vice versa). Nevertheless, to evaluate whether these outliers had an appreciable effect in the analysis, the same tests were conducted for datasets including and excluding the outliers. The results for the two-way ANOVA conducted excluding outliers did not yield statistically significant results. As such, the analysis continued including outliers. The density indexes for most of the sites considered violated the assumption of normality, as assessed by Shapiro-Wilk’s test (p < .05) (Table 5).

Table 5. Results of the Shapiro-Wilk’s normality test (W) for the fauna density indices.

		W	df	Sig.
pre-LGM	Laili	0.795	14	0.00
	Matja Kuru 2_B	0.539	38	0.00
	Matja Kuru 2_C	0.85	35	0.00
	Matja Kuru 2_D	0.741	15	0.00
	Matja Kuru 2_DD	0.831	38	0.00
LGM	Tron Bon Lei	0.709	17	0.00
	Laili	0.922	11	0.34
post-LGM	Tron Bon Lei	0.927	17	0.20
	Laili	0.836	15	0.01
	Matja Kuru 2_B	0.795	31	0.00
	Matja Kuru 2_C	0.938	31	0.08
	Matja Kuru 2_D	0.837	31	0.00
	Matja Kuru 2_DD	0.91	31	0.01
	Here Sorot Entapa	0.899	19	0.05

Figure 9. Boxplot showing outliers in the distribution of density indices for each site (numbers correspond to spit numbers). For site abbreviations, see previous figures.

As ANOVAs are considered fairly ‘robust’ to deviations from normality (Maxwell and Delaney 2004), the analysis proceeded to evaluating the homogeneity of variances. The assumption of homogeneity of variances was violated, as assessed by the Levene’s test for equality of variances (p < .001). Nevertheless, as data transformation through variance stabilizing transformations (ln, log10 and square root) did not sorted the heterogeneity of variances (scatter plots residuals vs predicted, studentized vs predicted), the analysis continued by the determining the interaction effect between the estimated marginal means and the focal (chronological period) and moderator (site) variables with profile plots (Figure 10) and the test of between-subjects effects.

Figure 10. Profile plots showing the estimated marginal means of the density indices calculated for each site per chronological period.

Results indicate that there is no statistically significant interaction effect between the site and chronological period for the density indexes score (F (5,329) = 1.46, p = .204, partial η² = .022). Therefore, an analysis of the main effect for chronological period and site was performed. There is a statistically significant different in mean density indexes score (i.e. main effect) for chronological period (Fv(2,329) = 7.90, p < .001, partial η² = .046) and site (F (6,329) = 56.79, p < .001, partial η² = .509). All pairwise comparisons were run where reported 95% confidence intervals and p-values are Bonferroni adjusted. The unweighted marginal means for pre-LGM, LGM and post-LGM periods were 42.31 (SE = 22.49), 38.96 (SE = 46.45), and 131.46 (SE = 19.01), respectively. Regarding site unweighted marginal means, scores were: Tron Bon Lei 148.9 (SE = 41.18); Laili 50.57 (SE = 38.3); Matja Kuru 2, square B 146.58 (SE = 29.06); Matja Kuru 2, square C 62.2 (SE = 29.61); Matja Kuru 2, square D 118.01 (SE = 37.76); Matja Kuru 2, square DD 76.6 (SE = 29.05); Here Sorot Entapa 1227.4 (SE = 55.08).

No statistically significant differences were observed between pre-LGM occupations, associated with a mean density index score 3.35 (95% CI, −120.84, 127.54) points higher than during the LGM (p > .05). LGM occupation yielded a mean density score 274.94 (95% CI, 395.27, 153.72) points lower than post-LGM occupations, p < .001. Post-LGM occupations are associated with a score 271.15 (95%CI, 200.29, 342) points higher than pre-LGM occupations, statistically significantly different p < .001. Geographically, Here Sorot Entapa is the only site that scores statistically significant differences (p < .001) with every other site tested.

4. Discussion

The analysis of the vertebrate assemblages documented in Laili, Matja Kuru 2, Tron Bon Lei and Here Sorot Entapa provides valuable data to investigate environmental and biodiversity changes related to the Last Glacial Maximum. Additionally, it provides insights into human occupation intensity during the Late Pleistocene-early Holocene in the Lesser Sunda Islands.

Taxonomic composition of tetrapods in the four sites analysed, during every chronological period, follows a similar pattern without statistically significant differences: large number of small murids and varied representation of a diversity of other taxa, including large and giant murids, bats, reptiles, turtles and birds. Laili and Matja Kuru 2 sites produce remains of giant and large murids, with abundant bird remains in Laili and turtle specimens in Matja Kuru 2. As shown by correspondence analyses (Figure 8), there are no clear differences regarding changes in foraging strategies between diachronic occupational phases. As such, site-specific ‘specialisation’ on distinct taxa could instead be the result of ease of access to specific niches, as lake Ira Lalaro would have been accessible for turtle hunting, and it is likely that arboreal habitats were accessible from Laili. Interestingly, no evidence of LGM occupations has been recorded in MK2; therein lies the question whether this occupational hiatus could be related to changes in water levels in the lake and consequently loss of this resource.

Previous taphonomical analysis on natural (i.e. owl roost) and archaeological assemblages from Tron Bon Lei and Laili documented direct evidence of owl predation in small murids and small birds (Hawkins et al. 2018; Hawkins, O’Connor, and Louys 2019). However, in these sites, larger animals such as giant murids, freshwater and marine turtles, large fruit bats and medium-large birds are likely to comprise the result of human foraging activities (Hawkins et al. 2018; Hawkins, O’Connor, and Louys 2019). In Matja Kuru 2, a taphonomic analysis of larger murids, freshwater turtle, birds and fish is ongoing, although preliminary observations suggest an anthropogenic accumulation, with sporadic contributions from avian predators (Samper Carro et al. 2022).

A remarkable difference in terms of taxonomic composition is observed between inland and coastal sites. The abundance of fish remains in Tron Bon Lei and Here Sorot Entapa far exceeds the number of any other taxa identified, including small murids. Predictably, a similar pattern is observed when comparing coastal with inland sites within Timor, with fish remains in Lene Hara and Asitau Kuru (formerly Jerimalai) outnumbering terrestrial vertebrates (O’Connor and Aplin 2007; O’Connor, Ono, and Clarkson 2011).

Similar percentage of NISP values are observed for the different fish families in Here Sorot Entapa and Tron Bon Lei. On the contrary, the percentage of offshore fish, especially Scombridae (tuna and mackerel) in Asitau Kuru, is remarkably higher that in any of the other two sites (Figure 11). These statistically significant differences (p > 0.05) could be explained in terms of coastal topography, sea levels and structure of fish communities in marine environments. Although the coastal profile in the three islands shows a similar profile, with a nearshore abrupt drop, the distance between the neritic and pelagic zones in each of the coasts varies. The east coast of Timor, in the vicinity of Asitau Kuru, has greater potential for offshore fishing, with a deep channel immediately offshore and opening to the Timor Sea. Hence, preferences in fishing strategies related to coastal and marine current dynamics might have explained the differences observed between these assemblages.

Figure 11. Percentage of number of specimens (NISP) per fish family in Tron Bon Lei, Here Sorot Entapa and Asitau Kuru (former Jerimalai).

In terms of palaeoecological reconstructions, the results from the earliest occupations documented in Timor Leste (Laili and Matja Kuru 2) confirmed previous observations (Hawkins et al. 2017). Simpson and Shannon indices indicate a diverse fauna, without single taxa dominating any of the assemblages. There are no significant differences between the different sites and occupational phases, although, as expected, Here Sorot Entapa and Tron Bon Lei yielded higher diversity scores, as a result of the large number of different fish families identified in these assemblages.

Paleoecological inferences from the fauna recovered in Laili support the hypothesis that a mosaic of heterogeneous habitats was present in Timor prior and after the LGM event (Hawkins et al. 2017). In relation to the ‘savanna corridor’ hypothesis and the potential absence of forest in the Sunda Islands (Heaney 1991; Bird, Taylor, and Hunt 2005), the results from Laili and Matja Kuru 2 suggest the presence of forested habitats. Future analysis of the possible habitat of the different genera of giant murids identified in Matja Kuru 2 will increase the available data to assess local forest cover before and after the LGM.

The statistical testing of standardised fauna values based on excavated volumes yielded useful data to assess occupation dynamics. The presence of outliers indicates accumulation peaks at different depth of the deposit. Whether these peaks correspond to the presence of more intense human occupations or otherwise, periods of cave use by avian predators need to be tested in the future. As such, comparing density indexes from vertebrate remains with those obtained from invertebrate remains (i.e. shellfish), stone artefacts and other human by-products (i.e. ochre, shell beads) will provide data to evaluate the degree of anthropogenic contribution to the accumulation of the vertebrate assemblages.

Although no statistically significant differences were observed between before and during LGM occupational phases, differences in the mean density score indicate a larger number of remains in the pre-LGM occupations. Moreover, post-LGM occupations yielded the highest mean density scores among occupational phases. Consequently, occupation dynamics in the Lesser Sunda Islands during these periods show a decrease on intensity during the LGM, bouncing back after this event, as sea levels rapidly rise. Geographically, these changes in density of vertebrate remains accumulation are observed in Here Sorot Entapa, which assemblage yielded an enormous quantity of vertebrate remains compared to the other sites, supporting the hypothesis that small islands in Wallacea were suitable for human inhabitation, at least, from the LGM onwards (O’Connor et al. 2018).

The results presented in this paper, even though significant, entail a substantial caveat: the fauna analysis does not include invertebrate remains. Shellfish and other marine invertebrates were a major resource in Wallacean coastal sites during the Pleistocene (e.g. Ono, Soegondho, and Yoneda 2009; Ono et al. 2020; Szabo and Amesbury 2011). In the Lesser Sunda Islands, molluscs and other invertebrates (i.e. barnacles, crabs and sea urchin) usually comprise the larger number of remains documented in coastal sites, such as Here Sorot Entapa and Tron Bon Lei. However, none of these sites provides quantitative data comparable to that obtained from the vertebrate remains, as NISP has not been reported for any of these sites. The invertebrate assemblage from Tron Bon Lei is unpublished, while only a total weight of the invertebrate remains have been reported in Here Sorot Entapa (O’Connor et al. 2018). For the Timor sites analysed in this paper, the invertebrate assemblage from Matja Kuru 2 has not been published, with available quantitative data (i.e.NISP) from Laili (Hawkins et al. 2017).

Molluscs data from Laili indicate the exploitation of marine, estuarine and freshwater zones during the Pleistocene, with an intensification on harvesting of mangrove molluscs post-LGM (Hawkins et al. 2017). During the peak of the LGM, there is a decline in freshwater species, likely reflecting a drop in precipitation during this event, consequently affecting the water flow in the Laleia River and loss or diminishment of freshwater habitats (Hawkins et al. 2017). The intensification on the exploitation of marine/brackish environments after the LGM observed in Laili might also be observed in Here Sorot Entapa, where a peak on the presence of marine molluscs is documented in units dated post-LGM (O’Connor et al. 2018). Nevertheless, the lack of more specific invertebrate data for these sites hampers correlating these results with those obtained from the vertebrate assemblages in the four sites of study.

5. Conclusions

The analysis of vertebrate remains from sites in Timor, Alor and Kisar provides remarkable data to investigate human occupation dynamics and living conditions in these insular environments before, during and after the Last Glacial Maximum.

Results from the comparison of taxonomic composition in the four sites analysed suggest that this climatic event did not dramatically affect vertebrate diversity, which shows the presence of persistent heterogeneous habitats throughout the Late Pleistocene-early Holocene period. Additionally, fauna diversity suggest the adaptability of human populations in these environments, implying their ability to hunt a myriad of vertebrates, from terrestrial, arboreal and avian species to freshwater and marine taxa. Site-specific changes in species representation can be explained as a result of resource availability and accessibility, instead of defined by climatic constraints.

Although occupation intensity based on fauna density indices shows slight differences among chronological periods and sites, these variations do not seem to indicate periods of extreme resource depletion. Consequently, fauna density indices suggest that the Lesser Sunda Islands were not dramatically affected by the environmental shifts brought by the Last Glacial Maximum and the subsequent rapid sea level rise. Future analyses, aimed at comparing the data obtained from the vertebrate assemblage with that of standardised values from stone artefacts, shellfish and other material culture (i.e. ochre, beads), as well as vertebrate taphonomy, will allow us to investigate the correlation between faunal assemblages and other material providing direct evidence of human involvement in the formation of these archaeological assemblages.

Acknowledgments

Fieldwork and analysis in Laili, Matja Kuru 2, Tron Bon Lei and Here Sorot Entapa was funded by an ARC Discovery Grant (DP0878543), an ARC Laureate Fellowship awarded to Prof. Sue O’Connor (FL120100156), and the ARC Centre of Excellence for Australian Biodiversity and Heritage (CE170100015). Fieldwork in Timor-Leste was undertaken with the assistance of Cecilia Assis and the Ministério do Turismo, Artes e Cultura from Timor-Leste. We thank the representative of the Department of Culture (Mr. Gil), the Department of Forestry (Mr. Marvaos), the Chef de Suco of Mehara (Mr. Antonio) and the Ratu Heads from Poros and Mkury villages and the local villagers for their assistance during fieldwork in Matja Kuru 2. Fieldwork in Indonesia was conducted with the assistance of Dr. Mahirta and staff and students from Universitas Gadja Mada and the Balai Arkeologi Organisation. Fieldwork in Tron Bon Lei was carried out on a FRP visa granted by RISTEK (O’Connor 1304/FRP/SM/V/2014). Fieldwork in Here Sorot Entapa was carried out on a FRP visa granted by RISTEK (O’Connor 1456/FRP/SM/VII/2015).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Australian Research Council [CE170100015, DP0878543, FL120100156].

Notes on contributors

Sofia C. Samper Carro

Sofia C. Samper Carro is currently a Research Fellow at the School of Culture, History and Language (CHL) at the Australian National University and a Legacy project manager at the ARC Centre of Excellence for Australian Biodiversity and Heritage (CABAH). Dr. Samper Carro a is zooarchaeologist and taphonomist with a background on Pleistocene and Holocene faunal assemblages. She has a wide experience researching Neanderthal and AMH subsistence strategies, completing taphonomical analysis of human and non-human accumulations, studying human occupation dynamics through density analysis, geometric morphometrics and photogrammetry and 3D modelling. She has also experience on bioanthropological and archaeothanatological analyses of human remains from Island Southeast Asia. Her current projects aim at investigating the possibilities of palaeoproteomic studies to improve our knowledge about Neanderthal and AMH diet and behaviour.