Validating earliest rice farming in the Indonesian Archipelago

Preserved ancient botanical evidence in the form of rice phytoliths has confirmed that people farmed domesticated rice (Oryza sativa) in the interior of Sulawesi Island, Indonesia, by at least 3,500 years ago. This discovery helps to resolve a mystery about one of the region’s major events in natural and cultural history, by documenting when rice farming spread into Indonesia, ultimately from a source in mainland China. At the Minanga Sipakko site in Sulawesi, preserved leaf and husk phytoliths of rice show the diagnostic morphology of domesticated varieties, and the discarded husks indicate on-site processing of the crops. The phytoliths were contained within an undisturbed, subsurface archaeological layer of red-slipped pottery, a marker for an evidently sudden cultural change in the region that multiple radiocarbon results extend back to 3,500 years ago. The results from Minanga Sipakko allow factual evaluation of previously untested hypotheses about the timing, geographic pattern, and cultural context of the spread of rice farming into Indonesia, as well as the contribution of external immigrants in this process.

www.nature.com/scientificreports/ horizontal unit within the alluvial terrace, situated between older and younger layers of the natural alluvium of the Karama River. In June 2017, the non-backfilled excavation trench profiles from 2004 to 2005 were cleaned by our team, allowing the collection of 22 point-specific phytolith samples from securely defined positions within the stratigraphic layers. From those samples, microscopic examination enabled the identification of preserved phytoliths of rice and other botanical taxa, as well as charcoal particles that could be submitted for direct radiocarbon dating.

Results
Refining the site chronology and cultural sequence. Previous excavations dated Minanga Sipakko rather broadly, possibly as old as 3,800 ybp or as late as 2,500 ybp 22 . In the lowest and oldest portion of the archaeological layer, the pottery fragments came mostly from red-slipped thin-walled earthenware vessels, resembling the oldest pottery horizon seen in the Philippines and connected with origins in Taiwan 17,23-25 . These findings suggested that Minanga Sipakko contained one of the earliest pottery assemblages in the Indonesian Archipelago, and therefore the site was considered ideal for this current research.
In order to refine the dating of the early pottery horizon at Minanga Sipakko, five new radiocarbon samples from our study confirm the first occupation as early as 3,500 ybp (Figs. 2 and 3; Supplementary Table S1 online). At 240 cm depth, the base of the pottery horizon overlaid a natural riverbank surface, and a sample at 235-240 cm provided a radiocarbon date of 3,562-3,400 ybp. From this base of the pottery horizon, the red-slipped tradition continued upwards through about 190 cm depth of the deposit that encompassed three remarkably similar radiocarbon dates of 3,321-3,061ybp, 3,230-3,007ybp, and 3,326-3,075 ybp. Above 190 cm, small amounts of plain pottery were noticed within the primarily red-slipped assemblage, and then the next sedimentary unit above 185 cm contained rapidly increasing amounts of plain pottery, post-dating 3,057-2,866ybp. In total, the sampled profile provided a sequential record of human activity from at least 3,500 ybp and continuing through 2,800 ybp.  Among the rice phytoliths, the leaf bulliform (flabellate) types were scrutinized for the numbers of scales around their convex ends, utilizing the observation that the proportion of bulliform phytoliths with nine or more such scales can identify domesticated (versus wild) varieties [26][27][28] . In the early occupation layers at Minanga Sipakko, those pre-dating 3,057-2,866 ybp, 50-55% of the leaf bulliform phytoliths consistently displayed nine or more convex-edge scales (Figs. 5 and 6). These consistent proportions are above the 33.33% maximum for convex-edge scales that correlates with wild varieties, and within the expected range for domesticated varieties ( Fig. 6; Supplementary Table S3 online). The consistent proportions of these convex-edge scales in all of the early layer samples indicate that domesticated rice was grown under conditions that were consistent throughout the time range of the site.

Identification of rice remains at
In addition to the leaf bulliform (flabellate) phytolith morphologies, the rice evidence at Minanga Sipakko suggests the deliberate processing of rice harvests at the site, as reflected in the relative proportions of leaf/stem (bulliform and parallel bilobate) and husk (double-peaked) morphologies (see Fig. 3). According to the cropprocessing model based on phytoliths, the early stages of rice processing, such as threshing and winnowing, will produce a much higher proportion of bulliform and bilobate phytoliths than double-peaked phytoliths. At the terminal stage, the phytolith assemblages will be dominated by the double-peaked type, as the end byproduct is mainly rice husk 29 . In most samples from lower layers, the double-peaked type was dominant, especially at depths of 230-235 cm and 200-205 cm, where they accounted respectively for 11.38% and 38.63% of all phytoliths. This suggests that rice processing through to the final stage of husk discard was undertaken in the site during the time of the initial red-slipped pottery horizon.

Discussion
Prior to this study in Sulawesi, only limited numbers of ancient sites in ISEA had yielded datable but controversial rice-related remains prior to 3,000 ybp, although abundant evidence of rice domestication has been reported as early as 4,800-4,200 ybp from Taiwan 9,24,30 . The absence of compelling evidence in ISEA, until now, has allowed numerous plausible yet ultimately untested notions about the origins of rice farming in the region, as well as about possible links with larger cross-regional population movements.
The new findings from Minanga Sipakko constitute the oldest and the most securely dated evidence of rice farming in Indonesia, in this case co-occurring with the first pottery-bearing horizon and dated as early as 3,562-3,400 ybp. The abruptness of the red-slipped pottery-bearing horizon in the region suggests a sudden appearance of a new cultural tradition.  Table S1 for details and counts) (Created by Z. Deng www.nature.com/scientificreports/   www.nature.com/scientificreports/ At a larger scale, the spread of this pottery-bearing horizon largely matches the modern geographic distribution of Austronesian-speaking populations in Taiwan, the Philippines, eastern Indonesia, and the western fringe of Remote Oceania. Austronesian was the most widespread language family in the world prior to the colonial era, representing one of the most impressive records of population dispersal in the history of humankind 38,39 . Questions have been debated for nearly a century about the location of the Austronesian homeland, the motives of the ancient migrants, and the nature of the dispersal process 40-46 . The results presented here provide general support for the Farming-language Dispersal Hypothesis and the closely related Out of Taiwan Hypothesis as applied to ISEA 8,9,38,47,48 . These hypotheses refer to Taiwan as the source of Austronesian-speaking rice farmers prior to 4,000 ybp, and then some of their descendants migrated by sea into ISEA in search of new lands. Eventually, these migrants expanded into new territories across more than one third of the globe's surface, occupying places from Madagascar to Easter Island.
Our conclusions are consistent with recent studies of cranial morphology 49 and ancient DNA lineages 50 , both suggesting that external biological populations came from East Asia, spread into ISEA after 4,500-4,000 ybp, and became the majority resident groups of ISEA by 3,000 ybp. The linguistic and biological findings appear compatible with the archaeological manifestation of an early pottery-bearing horizon that now can be confirmed as including domesticated rice, at least as far south as Sulawesi.

Materials and methods
Sample collection. In 2017, the authors revisited Minanga Sipakko. After cleaning the profile of the east wall of square M2, sediment samples were retrieved from the cultural layer at five-centimeter intervals, starting at the base and continuing upward. In total, 22 samples were collected, starting with the deepest and oldest at a depth of 235-240 cm below modern surface then continuing through the uppermost and latest sample at a depth of 130-135 cm. For each sample, roughly 150 g bulk sediment sample was taken, and around 20 g from each sample then was sub-packaged for phytolith analysis. The remaining samples were sorted under a binocular stereomicroscope at × 15 magnification, and charcoal fragments larger than 0.9 mm were selected for radiocarbon dating. A research permit for the Minanga Sipakko site was granted to T. Simanjuntak (Chief Investigator), A.A. Oktaviana, and B. Hakim to undertake the field research. H.-C. Hung, Z. Deng, and M.T. Carson were invited for laboratory phytolith study, cross-regional comparison, and manuscript preparation.
Radiocarbon dating. In total, 11 charcoal samples were sent to Beta Analytic Testing Laboratory for accelerator mass spectrometry (AMS) radiocarbon dating. Four samples were too small to yield radiocarbon dates (Beta-508804, 523845, 523846 and 528110), and two were apparently intrusive samples as their radiocarbon dates were older than 43,500 years BP (Beta-508805 and 523847). Therefore, in total five samples yielded relevant radiocarbon dates. These results were calibrated by OxCal 4.3.2 51 , using the IntCal13 atmospheric curve 52 , and they are presented in Fig. 2 and Supplementary Table S1, together with the previously published dates.
Phytolith extraction and identification. For each sample, 2 g of sediment was processed to extract phytoliths according to established procedures [53][54][55] . The weighed samples first were placed inside 50 ml centrifuge tubes and treated with 30% H 2 O 2 . Next, they were left for 12 h for full reaction and to ensure that all organic matter was removed. Three distilled water rinses were performed to remove the residual reagent. Then, carbonate aggregates and certain oxides were removed by adding 15%HCl. After becoming fully reacted, another three distilled water rinses were performed. Phytoliths were floated from the sediments in heavy liquid (ZnBr 2 , density 2.35 g/cm 3 ). The suspension with extracted phytoliths was moved into 10 ml centrifuge tubes. Distilled water was added to separate heavy liquid by centrifugation. The concentrated phytoliths then were washed again with 30% ethyl alcohol. In the last step, phytoliths were removed by pipette and mounted on a slide with Canada Balsam.
After air drying, the slides were observed under an optical microscope at 400 × magnification to identify and count all phytolith morphotypes. For each sample, at least 500 phytoliths were identified and recorded according to modern references and published criteria [56][57][58] . All slides have been browsed thoroughly to search for all well-preserved rice bulliform phytoliths, even after more than 500 phytoliths were recorded. The numbers of scales around the convex end of each rice bulliform phytolith were counted under an optical microscope at 630 × magnification, referring to the published criteria 26,55 .

Data availability
All raw data of this study are available in the Supplementary Files (Tables S1-S3), available online with this publication.