Past rainfall patterns in Southeast Asia revealed by microanalysis of δ 18 O values in human teeth

Variations in human subsistence and settlement patterns have been documented at Holocene archaeological

Bioarchaeological research on the Holocene of Southeast Asia has focused primarily on mobility and settlement patterns of Neolithic coastal societies, particularly around and after the 4.2 ka event that saw the aridification of a number of regions around the world (Park et al., 2019; The Routledge Handbook of Bioarchaeology, 2016).Scholars have also directed considerable effort to studying the spread of farming, using combined linguistic, genetic, and archaeological evidence to infer long-distance mobility of island dwelling communities and movement of their material goods (Reepmeyer et al., 2011;Huffer et al., 2016;Lloyd-Smith et al., 2016;Vallée et al., 2016;Higham, 2021;Pawlik, 2021).An influential model by Bellwood, 2005Bellwood, , 2011 posited that migrating farmers moved from modern-day Taiwan into the northern Philippines by 2000 BCE and subsequently westwards into Indonesia and southeastwards into the Bismark archipelago by 1350 BCE (Oxenham et al., 2016).Research that followed focused on tracing archaeological evidence to support or refute this model, but direct inferences of climatic patterns and changes that would have hindered or facilitated such movements are still lacking.

Rainfall regimes
Rainfall is a key climatic factor that shapes the structure of tropical ecosystems (Brockman and van Schaik, 2005;Schwartz et al., 2020).Understanding changing rainfall patterns is not only useful for tracking the impact of past and present climate variation, but also for understanding human adaptations to environments and cultural experiences.A common approach to describe rainfall regimes, Mean Annual Precipitation (MAP), provides only partial insight into the degree to which rainfall patterns differ across geographical areas.Rainfall seasonality indices (Walsh and Lawler, 1981;Feng et al., 2013) have been developed to assess relative seasonality (defined as a dimension of seasonality encompassing contrasts between seasons) across geographical regions, with higher indices characterizing areas where rainfall is more concentrated in a given month/season and lower indices describing regions with a more uniform distribution of rainfall.These indices have been found to be better correlated with functional traits and life histories of tropical organisms than MAP or the length of the dry season (Schwartz et al., 2020).
Similarly, stable oxygen isotope compositions of precipitation (δ 18 O precip values)-which are controlled by air temperature, rainfall amount, latitude, altitude, and movements of water masses-show different degrees of seasonal variability in different regions (Epstein and Sharp, 1959;Dansgaard, 1964;Sharp, 2017).In regions with pronounced seasonal δ 18 O precip variability, large contrasts exist between summer and winter δ 18 O precip values.In regions with 'dampened' seasonal δ 18 O precip variability, these differences are reduced.For example, in a continent-wide study of rainfall across Australia, seasonal δ 18 O precip variability as high as ~12 ‰ was identified at Mount Isa in the central inland region, while dampened seasonal δ 18 O precip variability of ~2 ‰ was identified at Perth on the west coast (Hollins et al., 2018).In the tropics, rainfall amount is the primary determinant of rainfall stable isotope composition (Dansgaard, 1964;Gat, 1996;Bowen, 2010).

Assessment of ancient climate and seasonality
To understand the impact of shifting climate on the sub-annual scale, a range of methodologies has been used to extract seasonality patterns from trees, corals, ocean sediments, and speleothems.Poussart et al. (2004) measured the δ 18 O values of cellulose from seasonal tree-rings and extracted a record of distinct seasonal variability in δ 18 O precip values across Southeast Asia (18 ‰ in Thailand and 4 ‰ in Indonesia).Yang et al. (2016) assessed the factors that drive inter-annual variability in δ 18 O values of speleothems at four cave sites affected by the Asian monsoon (Oman, India, Laos, China).These authors argued that while precipitation amounts might correlate with annual seasonal patterns in rainfall, they cannot be used to document multi-annual phenomena, such as the El Niño-Southern Oscillation (ENSO).
Stable oxygen isotope measurements of tooth enamel (δ 18 O enamel ), which reflect the composition of water, food, and air consumed by the individual during tooth formation (Bryant and Froelich, 1995;Kohn, 1996), provide a means to infer aspects of past climatic conditions.Despite the contribution of food and air to metabolically produced water and hard tissue δ 18 O values, foundational studies in oxygen isotope systematics have shown a strong relationship between bioapatite (the mineral component of tooth enamel) and local meteoric water oxygen isotope compositions (Kohn, 1996;Longinelli, 1984;Green et al., 2022;Longinelli and Peretti Padalino, 1983;d'Angela and Longinelli, 1990).For example, (Green et al., 2022 established an association between local monthly rainfall amounts in Ethiopia, δ 18 O measurements from an upstream water source, and δ 18 O enamel values from modern baboon (P.hamadryas) tooth enamel formed concurrently.Moreover, their comparisons of δ 18 O enamel values between sympatric Ugandan primate species with different diets revealed highly overlapping ranges with relatively minor average differences (~1 ‰).
The majority of previous studies in paleoecology have applied either a 'bulk' or a macroscopic 'sequential' sampling method for extracting δ 18 O enamel values (Pederzani and Britton, 2019;Vaiglova et al., 2022).Bulk sampling involves the extraction of one homogenized sample per individual, often drilled along the entire length of the preserved tooth.For example, King et al. (2013) analyzed bulk enamel samples from 118 human burials at Ban Non Wat (Thailand) spanning the Neolithic (1750 BCE) through to the Iron Age (400 AD), and reported evidence for a shift from lower to higher aridity that may have negatively impacted rice cultivation.Macroscopic sequential sampling uses enamel samples drilled along the axis of tooth development to assess isotopic changes in diet and water intake during the period of tooth formation.Kubat et al. (2023) reported three δ 18 O enamel values obtained from an Indonesian Homo erectus premolar, suggesting that it obtained water from a largely invariant riverine water source.However, because such macroscopic samples cut across incrementally forming enamel, their δ 18 O enamel values provide an averaged, and dampened, record of the isotopic composition of the individual's body water and associated food and drinking water.Thus, these samples need to be considered as arbitrary units homogenizing a loosely known amount of isotopic input signal (Balasse, 2003).Nevertheless, sequential isotopic data extracted from teeth have the potential to inform us about broad-scale climate patterns in the past in order to understand people's experiences of, and adaptation to, their local environments.
Secondary ion mass spectrometry (SIMS) enables measurement of the innermost enamel in situ at a scale of a few microns.Blumenthal et al. (2014) first demonstrated the use of this technique through measurement of modern tooth enamel of captive woodrats.Most recently, a Sensitive High-Resolution Ion Microprobe (SHRIMP) was used to analyze histological sections of hominin and primate teeth, enabling successive measurements of δ 18 O enamel at a spatial resolution equivalent to approximately one week of enamel extension (Green et al., 2022;Smith et al., 2018).For example, Smith et al. (2018) used this technique to compare climate seasonality recorded in the teeth of two Neanderthals (~250 ka) and one Holocene individual (5 ka) from Payre, France.Here we apply the SHRIMP (SIMS) method to extract sequential δ 18 O enamel values from 15 ancient human teeth recovered at three archaeological sites in Southeast Asia: Con Co Ngua (CCN), Vietnam, ~6800-6200 cal BP; Pain Haka (PH), Indonesia, ~3000-2100 cal BP; Napa (NP), Philippines, ~2000-1500 cal BP (Fig. 1).Bulk strontium isotope ratios ( 87 Sr/ 86 Sr) (one per archaeological tooth), commonly used to identify non-local individuals by comparison to local bedrock 87 Sr/ 86 Sr ratios (Bentley, 2006), are used to assess the likelihood that the studied individuals were born away from the location where they were buried and thus record non-local rainfall.Data obtained from the Global Network of Isotopes in Precipitation (GNIP) database are used to assess the seasonal variability of δ 18 O precip values measured near the study locations in the 20th-21st century.The recorded seasonality in the archaeological teeth is compared with the expected patterns of seasonal variability of modern δ 18 O precip at the study locations.In addition, two modern human teeth from Australasia (Southeastern Queensland, Australia; Dunedin, New Zealand: Fig. 1) are used to investigate the suitability of postindustrial human teeth for inferring seasonal rainfall patterns.
Fig. 1.Map of Southeast Asia showing locations of archaeological sites and GNIP stations.Created using ArcGIS Pro v 2.9.5 and basemap, 1:10 m Natural Earth I with Shaded Relief and Water, provided permission free by Natural Earth.

Archaeological and modern samples
The archaeological sites in this study represent coastal settlements and burial grounds that were in use across Southeast Asia over a period of ~5000 years.Con Co Ngua, Vietnam, is a large open-air huntergatherer site dating to the Da But cultural complex, which stretched across Northern Vietnam and Southern China between ~6800 and 6200 cal BCE (Oxenham et al., 2018).The inhabitants continued to maintain a foraging subsistence strategy during a time when farmers further to the north (in Southern China) were becoming more and more reliant on agriculture; thus representing an alternative adaptation to the Holocene Climate Optimum (Oxenham et al., 2018).The five individuals included here (all first molars, M1s) represent individuals recovered from burials from Layer 2 (older) and Layer 1 (younger) (Huffer et al., 2016).
Pain Haka, Indonesia, is located on the island of Flores and represents a coastal cemetery, stretching over 700 m along the shoreline, that seems to have been most heavily used around 2500 BP 59,60 .No occupation layers have been uncovered at the site, raising speculation that the cemetery may have been used to inter people from across wider reaches of the island region.All five individuals included in this study were represented by M1s.Four individuals (PH10, PH22, PH45, PH47B) were directly dated in this study using AMS radiocarbon dates to 3004-2363 cal BP (see Table S2).
Napa (Location 1), originally referred to as Catanauan, Philippines, represents a collection of two coastal cemeteries situated on a beach within a sheltered cove on the western part of the Bondoc peninsula, Philippines.The main cemetery dates to the first half of the first millennium AD (Metal Period) and was probably used for a period of 100-200 years, a period characterized by distinct pottery and burial grave goods including metal objects (short swords/daggers, iron blades/ knives), beads (made of shell, glass, carnelian), and other shell objects (both worked and unworked).A large number of human remains has been uncovered from varying depositional contexts, including jar burials, extended burials, and pit burials (Paz et al., 2018).Four individuals from Napa, Philippines (1 M1, 1 M2, 2 M3s) date to the Metal Period (~2000-1500 cal BP) (Vlok et al., 2017) and one individual (C507A; M2), recovered ~100 m from the site, represents a more recent (although not modern) individual.
Two living human M1s were also examined: one from the east coast of Australia (Southeastern Queensland, T1) born in 1990 and one from the South Island of New Zealand (MS5) born in 2011.See Table S2 for contextual information.

Modern rainfall data
Modern (20th-21st century) rainfall data were obtained from the Global Network of Isotopes in Precipitation (GNIP, International Atomic Energy Agency (Global, 2022)).In Vietnam, the closest stations to Con Co Ngua (20  ; 1962-1997).Belgamann et al. (Belgaman et al., 2017) obtained rainfall δ 18 O values from locations closer to Pain Haka that cover a period of 3 years (2010-2012), however, because these locations were identified to fall into the same cluster as Jakarta (according to the variability in seasonal δ 18 O precip values) and because the GNIP database provides a longer temporal coverage, the GNIP data are used in this study as a comparative dataset.In the Philippines, the closest stations to Napa (13  (Frew et al., 2011).

High-resolution sequential tooth δ 18 O analyses
The tooth enamel δ 18 O values were measured by ion microprobe (SHRIMP-SI) at the Australian National University, as detailed in Smith et al. (2018).In brief, tooth enamel in thin sections was first imaged with transmitted light microscopy, accentuated growth lines were mapped, daily secretion rates were measured, and enamel extension rates were calculated to guide spot placement at approximately weekly intervals of growth from the dentine horn tip to the enamel cervix.Following the removal of cover slips with xylene, the sections were cleaned with petroleum spirit, warm RBS35 detergent solution and millipore water, dried for at least 24 h in a 60 Distances of SHRIMP δ 18 O measurements along the innermost enamel from the cusp to cervix of each tooth were converted to time in days using incremental features of growth observed in the teeth through transmitted light microscopy.This was accomplished through two steps.First, the timing of enamel extension at the innermost enamel was determined at each of a series of accentuated lines, such that the distance of any given accentuated line at its intersection with the EDJ was associated with a day of extension relative to the onset of enamel formation in that cusp.Second, a polynomial regression relating distances to days was created using the distance-day measurements from accentuated lines at the enamel-dentine junction, and this regression was applied to calculate the timing of secretory deposition at every SHRIMP spot location.
To estimate the season of birth, we hypothesized seasonal climatic variation that results in annual oscillations in oxygen isotope values within teeth.We further predicted that these oscillations have a sinusoidal form, that their amplitudes of variation are equal to the range of variation observed in each M1, but that the offset of the sine wave (that is, the day the tooth begins forming in the seasonal cycle) is unknown.We defined the middle of the monsoon and the middle of the dry season isotopically, as the lowest and highest points in the sinusoidal curve, following known patterns of isotopic and climatic variability in the region.Then, we selected a seasonal offset for our hypothetical sine curve P. Vaiglova et al. that best matched observations using the curve fit function in Python's SciPy package, using a least-squares metric of fit.This procedure describes when a tooth most likely began forming over multiple seasons, which was then related to the neonatal (birth line) and added to prenatal formation time (i.e., relating the predicted seasonal patterns to the day of birth.)This approach was applied to the four Vietnamese individuals that showed the strongest annual isotopic patterns over multiple years.
2.4. 87Sr/ 86 Sr ratio analyses 87 Sr/ 86 Sr ratios of the ancient individuals were used to determine whether any of them were non-local to the site where they were buried (and thus recording non-local seasonality signals).Samples from Napa and Con Co Ngua were analyzed at the Vrije Universiteit Brussel (VUB), Belgium.Tooth enamel was drilled from the cusp of the tooth avoiding regions identified as containing diagenetic overprint.This resulted in sampling of a latitudinal transect of surface enamel extending from the occlusal surface down to up to three-quarters distance to the cervix.The tooth enamel powder was pre-treated with 0.1 M acetic acid (CH 3 COOH) solution in excess for 30 min followed by three milliQ TM water rinses and dried overnight at 50 • C (Pellegrini and Snoeck, 2016).10 mg of sample was then acid digested in 1 mL 14 M nitric acid overnight and evaporated to dryness, before strontium was extracted through column chemistry using ion exchange resin (Eichrom Sr Spec).In short, the columns and resin were rinsed 2 × with 2 M HNO 3 then 2 × 7 M HNO 3 , samples loaded in 7 M HNO 3 , charged 4 × 1 mL 7 M HNO 3 , and columns were eluted and Sr collected with 6 × 1 mL 0.05 M HNO 3 .Samples were measured on a Nu Plasma 3 MC-ICP Mass Spectrometer (PD017 from Nu Instruments, Wrexham, UK) at the Vrije Universiteit Brussel.During the course of this study, repeated measurements of the NBS987 standard yielded 87 Sr/ 86 Sr = 0.710240 ± 24 (2σ for 60 analyses), which is, for our purposes, sufficiently consistent with the mean value of 0.710252 ± 13 (2σ for 88 analyses) obtained by TIMS (Thermal Ionization Mass Spectrometry) instrumentation (Weis et al., 2006).Data were corrected for mass fractionation by internal normalization to 86 Sr/ 88 Sr = 0.1194.All the sample measurements were normalized using a standard bracketing method with the recommended value of 87 Sr/ 86 Sr = 0.710248 (Weis et al., 2006).Procedural blanks were considered negligible (total Sr (V) of max 0.02 versus 9-10 V for analyses; i.e. ≈ 0.2%).For each sample the 87 Sr/ 86 Sr ratio is reported with a 2σ error.
Strontium concentrations were measured on the five tooth enamel samples from Napa.Samples were pre-treated as above and digested in Teflon beakers (Savillex) using subboiled 14 M HNO 3 at 120 • C for 24 h, evaporated to near-dryness and subsequently digested with a drop of concentrated HNO 3 .Following dilution with 2% HNO 3 , Sr and Ca concentrations in the sample digests were determined using a ATTOM ES HR-ICP mass spectrometer from Nu Instruments at the Vrije Universiteit Brussel (VUB), Belgium, in low ( 88 Sr) and medium ( 44 Ca) resolution using Indium (In) as an internal standard and external calibration versus various reference materials (SRM1400, SRM1486, SRM1515).The actual strontium concentrations were then calculated by normalizing the calcium data to 40% wt.Accuracy was evaluated by the analysis of an internal bioapatite standard (CBA).Based on repeated digestion and measurement of these reference materials, the analytical precision of the procedure outlined above is estimated to be better than 5% (1σ, n = 10 for CBA).
Samples from Pain Haka were analyzed at the Max Planck Institute for Evolutionary Biology (MPI) as part the analysis of the entire Pain Haka skeletal assemblage.To remove possible surface contaminants, the tooth surfaces were abraded using a sonicated Dremel® rotary tool with a diamond tipped burr.Then, using a Dremel® diamond cutting wheel, a 10-20 mg sample of enamel was removed from the tooth and adhering particles, such as organic matter or dentine, were also removed.The enamel samples were weighed, then 2 mL of 65% HNO 3 was pipetted into a Savillex beaker.The beakers were capped and the samples were digested for 2 h at a temperature of 120 • C on a hot plate.Once fully digested, lids were removed from the beakers and the samples were dried for 8 h at 120 • C.After evaporation, samples were dissolved in 1 mL of 3 M HNO 3 .Each batch contained a blank and a standard (NIST SRM 1486, bone meal) to act as controls during the sample preparation.Strontium was purified using established processes of column chemistry (Deniel and Pin, 2001) fully described in Kramer et al. (2021).After the purification was complete, samples were dissolved in 3% HNO 3 and analyzed using a Thermo Fisher Neptune MC-ICP-MS instrument. 87Sr/ 86 Sr ratios were normalized using repeated measurements of the NIST-SRM 987 standard.The five bone meal standards (SRM 1486) gave consistent results with the expected value: average 87 Sr/ 86 Sr = 0.70929 ± 0.00002 for five runs total (long term value measured at this lab: 0.709300 ± 0.000026 1σ, n = 88).

Radiocarbon dating
Four individuals from Pain Haka included in this study (PH10, PH22, PH45, PH47B) were radiocarbon dated using Accelerator Mass Spectrometry (AMS) at the Radiocarbon Dating Laboratory, University of Waikato.Collagen was extracted from bone samples using a modified Longin (1971) protocol.Samples were demineralized in 1 M HCl for up to 4 days, rinsed in MilliQ™ water, and gelatinized by heating in weakly acidic water (pH3 at 90 • C for 4 h).They were subsequently ultrafiltered (using pre-cleaned Centriprep®, Ultracel YM-30 filters) and freeze-dried (using a Labconco FreeZone Triad freeze-dryer backed by an Edwards nXDS10i series oil-free pump) for a minimum of 48 h before being turned into graphite for AMS dating.
To assess annual seasonal variability, the sequences were broken down into year 1 (day 1-365), year 2 (day 366-731) and year 3 (day 732-1096) of tooth formation, and the remaining measurements discarded.The annual variability in δ 18 O enamel values for each year represented in the dataset is expressed as Δ 18 O year (Fig. 2).A full 3rd year was recorded in 11/15 archaeological samples and in the modern samples from Australia and New Zealand.In the modern sample from New Zealand, the year 1 portion of enamel was affected by a carious lesion, so only measurements for years 2 and 3 are considered.The Δ 18 O year values range between 3.3-6.0‰ (Vietnam), 2.1-4.4 ‰ (Philippines), 2.0-4.1 ‰ (Indonesia), 1.9-3.7 ‰ (Australia), and 1.5-2.2‰ (New Zealand) (Table 1).
Monthly δ 18 O values of modern precipitation (δ 18 O precip ) measured at the station closest to the study site in Vietnam (Hanoi, 2004(Hanoi, -2018;;GNIP database) show pronounced differences between maximum (April) and minimum (August) δ 18 O values.This highly seasonal trend correlates most closely with the trend in amounts of monthly rainfall, exhibiting an inverse correlation between the amount of rainfall and δ 18 O precip values (Dansgaard, 1964) (Fig. 3a and b).At the closest stations in the Philippines (Manila andDiliman Quezon, 1961-2016), the seasonal variability in δ 18 O precip was less pronounced than in Hanoi, but maximum values were detected in January-April and minimum values were detected in July-September.This trend also correlates most closely P. Vaiglova et al. (and inversely) with monthly amounts of precipitation (Fig. 3c and d).In Indonesia (Jakarta, 1962(Jakarta, -1997) ) seasonal variability was even less pronounced, nevertheless showing slightly higher δ 18 O precip values between July-September (also see Belgaman et al., 2017).This lack of pronounced seasonality is most consistent with the rather flat trends in air temperature (Fig. 3f).
To better characterize annual seasonality in modern precipitation, the annual variability in δ 18 O precip values was calculated for each year represented in the modern dataset (Table S1).Only years where at least 9 monthly measurements were made were included.The rainfall in Vietnam showed the highest amplitude of seasonal variation (Δ 18 O precip = 10.7 ± 2.0 ‰, 1σ; over 7 years), followed by the Philippines (Δ 18 O precip = 8.2 ± 1.2 ‰, 1σ; over 21 years) and Indonesia (Δ 18 O precip = 5.9 ± 1.5 ‰, 1σ; over 15 years).
Considering the modern individuals, the closest GNIP station to where T1 spent their childhood is in Brisbane (Australia, 1970(Australia, -2014)).Although the decadal trends there do not show strong seasonality in the 1970s, 1980s, 2000s, and 2010s (Fig. S1), in the first two years of T1's life (1990)(1991), the maximum and minimum δ 18 O precip values differed by 5.7 ‰ and 9.2 ‰, respectively (Table S1).Contrary to the pattern in the δ 18 O precip values, however, the amplitude of seasonal variation recorded in T1's first molar was smaller in year 2 (1991, 1.9 ‰) compared to year 1 (1990, 3.7 ‰; Table 1), likely a result of drinking attenuated tap water drawn from municipal reservoirs.In Dunedin, where MS5 spent their childhood and also drank tap water from rain tanks and municipal sources, δ 18 O precip data were available only between 2007 and 2009, prior to when MS5 was born (2011).The available data do not show strong seasonality in δ 18 O precip values (Fig. S1), explained by the predominantly marine climate (Yang et al., 2020).River δ 18 O values have also been found to exhibit dampened seasonality on the South Island (where Dunedin is located) due to the blurring of composition as a result of the large contribution of snowmelt to rivers (Yang et al., 2020).Correspondingly, the annual seasonality recorded in the second and third year of MS5's life was equally low (2012: 2.2 ‰; 2013: 1.5 ‰; Table 1).

Tooth enamel 87 Sr/ 86 Sr ratios
Strontium isotope ratios were measured for all archaeological samples (one sample per individual) and show the greatest variation in the Philippines (0.7052-0.7086), and lowest variation in Vietnam (0.7091-0.7096) (Fig. 4).One sample from Indonesia (PH47B) displays  a 87 Sr/ 86 Sr ratio 0.0012 lower compared to the average of the remaining four samples (averaging 0.7076 ± 0.0001).One sample from Vietnam (M80) has a 87 Sr/ 86 Sr ratio 0.0004 lower compared to the average of the remaining four samples (averaging 0.7095 ± 0.0001).The samples from Indonesia (0.7073 ± 0.0006, 1σ) fit within the larger range of 87 Sr/ 86 Sr ratios of the samples from the Philippines (0.7071 ± 0.0013; 1σ), with both groups displaying values well below the 87 Sr/ 86 Sr ratio of modern seawater (0.7092, McArthur et al., 2001).The samples from Vietnam fall into a higher range of ratios (0.7094 ± 0.0002, 1σ), which includes modern seawater.

Tooth enamel, precipitation, and seasonality
Comparison of the archaeological individuals and the modern rainfall datasets shows that the seasonal variabilities in archaeological δ 18 O enamel resemble the seasonal variabilities in modern δ 18 O precip across the distinct latitudes (highest in Vietnam, lowest in Indonesia, intermediate in the Philippines).It is possible to estimate the season of birth from first molars of individuals from highly seasonal environments, as was done for a Neanderthal individual from Payre, France from its annually variable SHRIMP measurements (Smith et al., 2018).Four of five Vietnamese individuals were born six months prior to the peak of the monsoon (range ~155-225 days), consistent with reports of human birth seasonality in tropical subsistence societies where food availability correlates with rainfall (Bronson, 1995).Rainfall patterns in central Vietnam are driven by the shift in moisture source from the Indian Ocean (summer) to the South China Sea (winter), while most rainfall precipitates during the autumn season (Wolf et al., 2020).Modern rainfall patterns in Indonesia are controlled by the Asian monsoon (in the wet season) and the Australian monsoon (in the dry season) (Belgaman et al., 2017).A large proportion of rainfall in the Philippines is induced by tropical cyclones, with the highest contribution in the northern part of the country, attributed to the enhancement of the Asian monsoon (Bagtasa, 2017).
In the modern human teeth from Southeastern Australia and the South Island of New Zealand, the recorded patterns are largely flattened, exhibiting smaller annual amplitudes in δ 18 O enamel values (1.9-3.7 ‰, and 1.5-2.2‰, respectively; Table 1).This is consistent with the fact that the overall water intake of these individuals came from a range of sources not restricted to local rainfall, including city dams/reservoirs, rivers, large water tanks, and processed, boiled, or bottled water.While  P. Vaiglova et al. showing that these data are sensitive enough to pick out these modern behaviors, the findings highlight that inferring the causal mechanisms of body/ingested water δ 18 O values of contemporary individuals is highly complex (Ueda and Bell, 2021), and furthermore that we should not expect modern and ancient samples to reflect local δ 18 O precip values to the same degree.
Although previous studies have suggested that breastfeeding or physiological changes may lead to isotopic enrichment of bone mineral δ 18 O values (reviewed in Pederzani and Britton, 2019), neither the M1s in the current study (Fig. 2), nor a similar number of wild primate M1s also measured with the SHRIMP-SI (Green et al., 2022) show evidence of isotopic enrichment during the first year, when milk would constitute the great majority of ingested liquids.Moreover, a recent study of serially-forming wild orangutan molars did not find noteworthy differences between M1s and M2s/M3s in the same four individuals (Smith et al., 2023).It appears that M1 enamel returns values that are as informative of seasonal δ 18 O amplitudes as later forming teeth.

Isotopic inference of childhood residence
87 Sr/ 86 Sr ratios of the archaeological teeth were used to determine whether any individuals may have been non-local to the site where they were buried.The 87 Sr/ 86 Sr ratios of archaeological humans from Con Co Ngua (Vietnam) measured here lie in the same range as the ratios of archaeological humans (n = 40) previously measured from the same site (0.7095 ± 0.0002) (Huffer et al., 2022).The narrow range of values reflects the well mixed composition of the surrounding alluvial landscape composed of sediment originating from substrates with lower ratios (limestones and Permian to Early Triassic marine sediment; ~0.7070-0.7085,Huffer et al., 2022) and higher ratios (igneous outcrops and granites, dacites and rhyolites, ~0.7080-0.720).Thus, the results suggest that the diets of the humans were composed of food items sourced from the surrounding alluvial catchment as well as from marine resources.
The site of Napa (Philippines) sits on Pliocene-Pleistocene Limestones, and the Bondoc Peninsula consists almost entirely of sedimentary rocks (Pratt, 1914;Aurelio et al., 2013).Due to its coastal location and underlying geology, the immediate baseline is expected to be ~0.7092,but the proximity of volcanic activity may lower the regional 87 Sr/ 86 Sr baseline. 87Sr/ 86 Sr ratios of rocks, lava and volcanic dust measured from locations close to Napa, Philippines record ratios between 0.7037 and 0.7048 (Mukasa et al., 1994;Castillo and Newhall, 2004;Maussen et al., 2018); all lower compared to the 87 Sr/ 86 Sr ratios of the archaeological humans measured in this study.This suggests that the human diets were composed of a mixture of food sources originating from coastal locations influenced by strontium sea-spray as well as food sources growing in closer proximity to volcanic geology.
On the island of Flores, Indonesia, the geology is composed of Neogene and intermediate basic volcanic rocks (O'Sullivan et al., 2001).Thus, the less radiogenic values of the archaeological humans measured from Pain Haka are consistent with consumption of food items from the immediate landscape.However, we also cannot exclude the possibility that they originated from distant locations that have the same underlying geology and 87 Sr/ 86 Sr ratio composition.
There is a general lack of evidence to suggest that any of the individuals in this study record non-local climate and rainfall patterns.Most samples at each site form a tight cluster, likely reflecting diets obtained from the local landscape.Those that fall outside the sitespecific clusters (M80 in CCN and PH47B at PH) may possibly be nonlocal.However, if they are non-local, they do not have disparate δ 18 O sequences compared to the remaining individuals at each site, which may be the result of originating from a similar latitude or place with similar regional weather patterns.

Implications for future isotopic studies
It is important to note that seasonal variability in δ 18 O precip values does not always correlate to indices of relative seasonality of rainfall (Schwartz et al., 2020;Walsh and Lawler, 1981;Feng et al., 2013).Oxygen isotope fractionation in rainfall is affected by factors including air temperature and source of airmasses bringing in rainfall with varying stable oxygen isotope compositions (Dansgaard, 1964) that do not necessarily impact on the timing, duration, and intensity of rainfall events.Using the R script provided in Schwartz et al. (2020) and the GNIP data used in this study, the seasonality indices for the study locations are as follows: 1) Feng et al. ( 2013) index: 0.133 (Indonesia), 0.135 (Vietnam), 0.240 (Philippines) 2) Walsh and Lawler (1981) index: 0.591 (Indonesia), 0.770 (Vietnam), 0.814 (Philippines) These indices suggest that the differences between the summer amounts of rainfall and the winter amounts of rainfall are highest in the Philippines and lowest in Indonesia, consistent with the pattern shown in Fig. 3.However, Fig. 3 also shows that the seasonal differences in rainfall amounts do not map directly onto the amplitudes of seasonal δ 18 O precip variability.This underscores the need to acknowledge that while the seasonal differences in δ 18 O precip provide useful information about the factors impacting the δ 18 O composition of rainfall (such as temperature and air mass composition), they cannot be used to provide a full picture of the relative seasonality of rainfall in a given location.As a result, the variability in δ 18 O enamel values is best employed in paleoclimate research to assess possible shifts in rainfall patterns in one location (e.g., through an assessment of changing amplitudes of seasonal δ 18 O precip variation through time) rather than to compare the degrees of seasonality between different regions.
Furthermore, the measured δ 18 O enamel values can only be used to infer aspects of past rainfall seasonality, and not as a technique for determining with certainty where any given individuals lived during tooth formation (also see Pederzani and Britton, 2019;Ueda and Bell, 2021;Pryor et al., 2014).A mean intra-tooth value of 16 ‰ recorded from a human tooth would be consistent with residence at any of the 3 studied locations; it is only when full annual seasonality amplitudes are considered for larger numbers of samples that a non-local origin of an individual might be more reliably inferred.The SHRIMP method provides more accurate estimates of seasonal variability in δ 18 O of water intake than lower-resolution methods, which-by sampling across larger areas of sequentially forming enamel-produce more averaged δ 18 O values of water intake (Balasse, 2003).
In conclusion, fine-scale δ 18 O enamel sequences measured on archaeological teeth can provide useful information for identifying climate variation experienced by past individuals.We caution that these results cannot be used to determine exact geographic origins of the studied individuals (see also Pryor et al. (2014).Future studies may focus on assessing chronological changes in the seasonal variability of δ 18 O enamel from individuals recovered at the same site, while tempering conclusions about the degrees of relative rainfall seasonality across different environments.In regions like Southeast Asia, where δ 18 O values of precipitation are predominantly driven by amount of precipitation and shifting water sources (Wolf et al., 2020;Araguás-Araguás et al., 1998;Modon Valappil et al., 2022), changes to seasonal variation in δ 18 O values of past precipitation can be used to assess shifts in the behaviors of tropical monsoons.This can provide a more nuanced climatic background for investigating aspects of human adaptation to these landscapes.

Ethics declaration
Modern teeth were not extracted for the purposes of this study, but P. Vaiglova et al. shared with the researchers after their clinical removal.Consent was obtained from the individuals and/or their legal guardians.

P
. Vaiglova et al.They argued that the studied Neanderthals experienced a cooler and more seasonal climate compared to the Holocene individual from the same location; consistent with climate reconstructions obtained from speleothem records in southern France, northern Italy, and southern Ireland (McDermott et al., 1999).Similarly, Smith et al. (2023) measured δ 18 O enamel values in orangutans from the islands of Borneo and Sumatra, finding annual and bimodal trends in recent Sumatra individuals that are consistent with modern rainfall data.Comparisons of recent Bornean orangutans with two late Pleistocene/early Holocene individuals from Niah Cave suggest drier and more open environments with reduced monsoon intensity during this earlier period, consistent with other Niah Caves studies and long-term speleothem δ 18 O records in the broader region.

Fig. 2 .
Fig. 2. Annual variability in δ 18 O enamel values for individual years represented in the archaeological dataset.The individual sequences were truncated into year 1 (day 1-365), year 2 (day 366-731) and year 3 (day 732-1096) of tooth formation and the measured values are summarized as maximum, minimum and mean δ 18 O enamel values for each year.On the bottom, the amplitudes of intra-tooth variation (δ 18 O maxδ 18 O min ) are shown individually, and their range per geographical location is indicated with dashed lines.

Fig. 3 .
Fig.3.Climate data obtained from the GNIP database summarizing rainfall collected at stations located closest to the archaeological sites (Vietnam: Hanoi; Philippines: Manila; Indonesia: Jakarta).The top panels show the annual variability in δ 18 O precip values and the bottom panels show trends in precipitation amounts and air temperature.

Table 1
Summary statistics of δ 18 O enamel values and 87 Sr/ 86 Sr ratios of archaeological individuals from Con Co Ngua (Vietnam), Pain Haka (Indonesia), Philippines (Napa), and modern individuals from Australia (Southeastern Queensland) and New Zealand (Dunedin) measured in this study.VUB = Vrije Universiteit Brussel, Belgium; MPI = Max Planck Institute, Leipzig, Germany.