Localized management of non-indigenous animal domesticates in Northwestern China during the Bronze Age

The movements of ancient crop and animal domesticates across prehistoric Eurasia are well-documented in the archaeological record. What is less well understood are the precise mechanisms that farmers and herders employed to incorporate newly introduced domesticates into their long-standing husbandry and culinary traditions. This paper presents stable isotope values (δ13C, δ15N) of humans, animals, and a small number of plants from the Hexi Corridor, a key region that facilitated the movement of ancient crops between Central and East Asia. The data show that the role of animal products in human diets was more significant than previously thought. In addition, the diets of domestic herbivores (sheep/goat, and cattle) suggest that these two groups of domesticates were managed in distinct ways in the two main ecozones of the Hexi Corridor: the drier Northwestern region and the wetter Southeastern region. Whereas sheep and goat diets are consistent with consumption of naturally available vegetation, cattle exhibit a higher input of C4 plants in places where these plants contributed little to the natural vegetation. This suggests that cattle consumed diets that were more influenced by human provisioning, and may therefore have been reared closer to the human settlements, than sheep and goats.

Humans. As published previously 19,21 , humans pre-dating the 1900 cal BCE transition (WUB and MOZ) exhibit δ 13 C values that are strongly influenced by C 4 plant inputs (Fig. 3). Humans post-dating this transition (HUO, GAN/SAN, ZHQ) exhibit primarily mixed C 3 /C 4 diets, with a small number of individuals in all groups (except for Huoshaogou) exhibiting C 3 -dominated diets (Fig. 4). Table 1 shows the summary statistics for each site and Supplementary Figs. S1 and S2 present the bivariate plots with all data.
At Huoshaogou, the human δ 13  Plants. Bulk crop grains were measured in small numbers from both Northwestern and Southeastern regions (Figs. 3 and 4). Broomcorn millet (n = 3, from HUO, GAN/SAN, and MOG-C) exhibits a narrow range of δ 13 C values (− 10.5‰ to − 8.8‰) and a wide range of δ 15 N values (+ 1.2‰ at HUO to + 9.1‰ at GAN/SAN; with a sample from MOG-C at + 3.5‰). Foxtail millet (n = 2, from MOC-C and HUO) exhibits δ 13 C values of − 10.7‰ and − 9.7‰ and δ 15 N values of + 4.0‰ and + 6.7‰, respectively. Barley (n = 2, from MOG-C and GAN/SAN) exhibits lower δ 13 C values than wheat (n = 2, from GAN/SAN and MOG-C) (− 24.3‰ and − 23.8‰ for barley and − 21.5‰ and − 0.4‰ for wheat). The wheat δ 15 N values bracket those of the barley (+ 3.2‰ and + 4.1‰ for barley and + 2.9 and + 7.7‰ for wheat). With converted Δ 13 C values (cf. 42 ) of + 17.6‰ and + 18.1‰ and situated above the 'poorly watered band' , the barley samples appear to have been grown in water conditions that were not limiting to growth. The wheat, on the other hand, appears to have been grown in suboptimal watering conditions, with Δ 13

Discussion
Role of animal foods in human diets in the Hexi Corridor. The difference in mean δ 15 N values between the humans and animals in this study lie within the generally accepted 3-5‰ interval for trophic enrichment 48 . This suggests that, as far as protein intake was concerned, animal products had more than a minimal role in human diets. This disagrees with the suggestion (from an earlier study based on limited sample sizes) that animal products played a minor role in human diets in the Northwestern part of the Hexi Corridor 34 .
The results presented here suggest that animal products were consumed in sufficient amounts to drive protein intake, but not so much as to dominate both protein and energy intake. This stands in contrast to human diets in Central China, where it has been argued that historically, human subsistence was primarily based on grain consumption 49,50 .
To assess whether the protein component of human diets was driven by a particular domestic animal species, individual offsets in δ 15 N values were calculated between humans and the major animal taxa (sheep/goat, cattle, deer, dog, and pig). The data suggest that the offsets between humans and the omnivores (dogs and pigs) are lower than the 3-5‰ trophic enrichment interval, except at Huoshaogou. This indicates that animal protein intake was not limited to omnivore meat, but must have included additional sources.
The offsets between humans and herbivores (sheep/goat, cattle, deer) lie closer to the trophic enrichment interval, ranging from 2.5-6‰. However, the lack of any systematic patterns between the sites suggests that meat and dairy intake consisted of varied combinations of species at the different locations. Animal products were obtained either primarily from animals whose offsets lie close to 4‰ (i.e., sheep, goat and cattle), with smaller inputs from the remaining species (deer, dog, and pig), or in equal amounts from animals that lie above and below the 4‰ offset (for example, deer and pig at GAN/SAN and MOG-C). No two sites exhibit the same combination of δ 15 N offsets, making it unlikely that the communities in different locations followed the same dietary patterns. Instead, diets across the Hexi Corridor were heterogeneous: some communities and individuals may have chosen to consume diets dominated by sheep, goat, and cattle, while others preferred more deer, pig, and dog.
New insight into Bronze Age animal husbandry in the Hexi Corridor. The feeding patterns of dogs, pigs, and domestic herbivores at the study sites, as inferred from stable isotope analyses, provide insight into how   Table 2 for a breakdown of sample numbers. www.nature.com/scientificreports/ prehistoric populations in the Hexi Corridor integrated animals that had been domesticated in the region for millennia with those that had been recently introduced into their agricultural and social spheres.
Prior to the introduction of domesticates from Southwestern Asia in the 2nd millennium BCE, communities in Northern China practiced a subsistence economy based on millet cultivation and pig husbandry. These two spheres of the Neolithic economy were tightly integrated, as evidenced by the pigs' consumption of millet products and byproducts dating back to c. 5700 BCE 14,51 . Liu and Jones 52 argue that pigs were kept in social enclosures, which restricted their movement and reoriented their dependence on natural vegetation towards agricultural fodder and the leftovers of human food. Accordingly, the similarity of pig and human diets is evident both in early Neolithic contexts at the sites of Dadiwan and Xinglonggou 14,51 , and at Middle and Late Neolithic sites across North China 16,[52][53][54] .
In this study, the pig δ 13 C values lie within ~2‰ of the δ 13 C values of the humans, except at Huoshaogou, where pig δ 13 C values indicate a notably lower consumption of C 4 plants. This suggests that even within pig rearing-the agricultural sphere that had a long-standing tradition in the region-people across the Hexi Corridor made choices that broke with tradition and adopted a management strategy that enabled them to conceive of human and pig foods as separate.
With the arrival of the Southwestern Asian domesticates, farmers in the Hexi Corridor had a choice to either integrate the animals into the existing stall-feeding/millet-foddering system used for pig raising at Ganguya, Sanbadongzhi, Mogou and Zhanqi, or employ a local pastoral strategy, such as the one used for managing pigs at Huoshaogou. As a result of this choice, localised distinctions in animal herding strategies arose in different parts of the region, reflecting varying degrees to which the non-locally domesticated animals were integrated into the long-standing pig-millet economy.
Sheep and goats in both the Northwestern and the Southeastern regions exhibit diets primarily composed of naturally available plants, suggesting that they were managed within the local grazing landscape. In the wetter Southeastern Hexi Corridor, sheep/goat exhibit pure C 3 diets, while in the drier NW region, their carbon isotope values indicate consumption of both C 3 and C 4 vegetation. This is consistent with the natural spread of vegetation in these regions, with higher amounts of C 4 plants occurring in the NW Hexi Corridor 55 , owing to C 4 plants' higher proclivity to aridity and solar radiation 56,57 . Therefore, in the NW, sheep/goat diets reflect    Table 2 for a breakdown of sample numbers.  42,58,59 . These values are consistent with previously published data for herbivores from hot/ arid environments, whose elevated δ 13 C values have been explained by the composition of their diet rather than by a physiological response to hotter environments 60 . The diets of cattle exhibit the opposite patterns to the sheep/goat, with individuals from the drier NW Hexi Corridor exhibiting pure C 3 diets and those raised in the wetter SE region exhibiting δ 13 C values up to − 14‰, indicative of C 4 plant consumption. At the Northwestern sites of Huoshaogou, Ganguya and Sanbadongzhi, cattle subsisted on more restricted diets compared to the sheep and goats, probably because cattle lacked access to the arid-adapted C 4 or water-stressed C 3 plants.
In the Southeastern region, cattle were only available from Mogou, which is situated at an altitude of 2200 masl (meters above sea level). Distribution of C 4 vegetation declines at high elevations 61 , with C 3 -dominated landscapes occurring above ~1500-3000 masl, depending on the continent. This suggests that any C 4 signatures in cattle raised at Mogou are unlikely to be the result of grazing on naturally occurring C 4 plants in the immediate surroundings of the settlement. Instead, they may have either been fed cultivated C 4 crop products/ byproducts or moved seasonally to lower altitudes with native C 4 grasses. Ongoing research of sequential tooth enamel carbonate isotope analysis will help clarify the seasonal dietary and mobility patterns of these animals.
A parallel can be drawn between the results of the current study and inferences of similar practices from the southern Levant. At the Chalcolithic site of Marj Rabba 62 , in Jordan, wider variation in cattle δ 13 C values compared to sheep/goat δ 13 C values is interpreted to indicate that two foddering/pasturing strategies were employed for cattle: one which involved local grazing and another which may have involved winter foddering on C 3 /C 4 fodder and/or mobility outside of the local region. Similar isotopic distinctions between sheep/goat and cattle can be seen in Northern China 53,63 . At Xinzhai, stable isotope analysis of tooth enamel carbonate shows that, in contrast to sheep/goat diets, cattle diets included high amounts of seasonal fodder (likely C 4 millet), which was interpreted to indicate that these animals were managed closer to the settlements than the sheep/goats 64 66 . These distinctions in herding practices are likely a result of either physiobehavioral differences between the two animals or varying assignment of culinary/cultural values. Sheep and goats tolerate rocky, frosty, and arid environments, whereas cattle require more water and are less resilient to extreme temperatures 66 . In Bronze Age China, cattle likely enjoyed high sacrificial value for ritual activities, as documented by early textual records 67 . The prestige and power associated with this status may have been a reason that cattle were kept closer to settlements rather than allowed to roam in more distant pastures with sheep and goats.
In summary, a contrast can be drawn between herding practices in the NW and SE regions of the Hexi Corridor. In the NW, sheep and goats may have been taken to graze on the fringes of the Gobi Desert and the foothills/hillsides of the Qilian Mountains, where they would have had access to C 4 plants and water-stressed C 3 plants. Cattle would have tolerated these landscapes less easily and would therefore have needed to graze closer to the oases or rivers where the settlements were located. This type of pastoral system is evidenced with the recent discovery of a large corral (over 200 m 2 ) at Xihetan 68 . In the SE, on the other hand, Mogou is situated in a spatially constrained valley between the highlands and multiple lower-elevation catchment zones along the Tao River. In a location with limited grazing lands, allowing cattle to roam on land otherwise suitable for farming activities would have created significant socio-political tensions. A strategy that relied on seasonal pasturing of sheep and goats and stall-feeding of cattle would have provided farmers with an optimal solution representing continuity with the long-lasting Neolithic tradition of pig rearing 52 .

Implications
The insight gained into animal management in Bronze Age Hexi Corridor has implications for: 1. the role of animal products in local human diets, 2. the heterogeneous nature of human diets, and 3. the relationship between newly introduced domestic animals and local rearing traditions.
The well-documented human dietary changes in the early 2nd millennium BCE in Northwestern China were partly driven by the consumption of newly introduced domestic herbivores. Previous discussions have primarily focused on the shifts in staple grain consumption from millet towards SW Asian cereal crops. This paper argues www.nature.com/scientificreports/ for increased role of animal products in human diets. While wheat and barley were gaining the status of staple crops, the consumption of sheep, goats, and cattle was also increasing.
In the broader regional context, the Hexi Corridor and the Loess Plateau present opposing dietary patterns in the 2nd millennium BCE. People in the Hexi Corridor adopted Western grains rapidly, while those in the Loess Plateau neglected them. Although the results of this study constitute only a local assessment, it appears that the wide range of subsistence activities employed in the Hexi Corridor contrasts with the unified milletbased agrarian practice widespread across the Loess Plateau. From this perspective, the regional differences can partially be explained by differences in subsistence economies: sedentary farming in the Loess Plateau versus multi-resource agro-pastoralism in the Hexi Corridor. This in turn challenges the traditional narrative of 'modes' of subsistence-hunting, foraging, pastoralism, and farming-progressing and developing in a linear evolutionary framework. The results from the Hexi Corridor show that people moved fairly readily between varying modes of subsistence and coexisted with neighboring populations that employed different modes. It has been demonstrated, ethnographically and archaeologically, that the same people may have practiced more than one subsistence mode in a single lifetime, reflecting the choices of people under certain social and environmental conditions rather than a proscribed stage of 'economic development' 21,69,70 .
The third inference concerns the relationship between non-locally domesticated plants/animals and indigenous culinary and rearing traditions. The social context in which agricultural and culinary innovations occurred across prehistoric Eurasia has been heavily debated 19,71 . Emphases have been placed on the reaction of an existing social and culinary system to novel crops or the role of technology as a key driver of their adoption and translocation. In the context of Southeastern dispersal of metallurgical traditions, for example, Rawson 31 suggested that when foreign innovations were adopted in ancient Central China, they were transformed within highly organized social and cultural systems, and this was particularly pronounced in the adoption of bronze casting technique within Eastern ritualistic contexts. In the case of eastward expansion of wheat, it has been demonstrated that this process likely exerted selection for phenotypic traits that were particularly suited to the eastern boiling and steaming tradition 23 . Both these 'transformations' initially occurred in an area that included the Southeastern Hexi Corridor. Our results suggest a similar process in the adoption of cattle: in locations with limited grazing lands suitable for pasturing of cattle, people adapted the local pig rearing economy towards cattle stall-feeding.

Materials and methods
Archaeological human (n = 194), animal (n = 366), and plant (n = 9) samples were obtained from nine sites in the Hexi Corridor of Gansu Province and measured for δ 13 C and δ 15 N values (Fig. 1, Table 2). Most of the human data (n = 189) has been published previously 19 . The summary statistics of a portion of the animal dataset (n = 167) and values of three plant samples (from Huoshaogou) were used in previous human diet modelling 23 . This is the first time that all the data is reported in full. Forty-eight human samples from Mogou Cemetery measured by Ma et al. 22 are included here for comparison, bringing the total of human samples to 242.
Six sites (Wuba, Mozuizi, Xihetan, Ganguya, Sanbadongzi, and Huoshaogou) are located in the drier Northwestern part of the region, while three (Mogou Settlement, Mogou Cemetery, and Zhanqi) are located in the wetter Southeastern region. Due to their geographic and temporal proximity, Ganguya and Sanbadongzi are grouped together and treated as one analytical unit. All sites have been assigned to time periods of either 'pre-1900 cal BCE' or 'post-1900 cal BCE' using established cultural chronologies and radiocarbon dates 19 . www.nature.com/scientificreports/ Animal samples were identified to the lowest taxonomic level possible based on qualitative morphological traits. The current study was aimed at identifying samples suitable for isotopic analysis and did not provide a full overview of the composition of the zooarchaeological assemblage; more comprehensive analysis is still ongoing and will provide complementary information. Domestic taxa include sheep (Ovis aries), domestic goat (Capra hircus), domestic cattle (Bos taurus), domestic dog (Canis familiaris), domestic pig (Sus scrofa), and equids (Equus sp., horse/donkey). Wild animals included deer (taxon unspecified), monkeys (macaques, Macaca sp.), cat (Felis sp.), hare (taxon unspecified), rodent (taxon unspecified), badger (Meles sp.) and/or beaver (Castor fiber), bear (Ursus sp.), and multiple birds (taxon unspecified).
Bone collagen was isolated using a modified Longin 72 protocol 73 . Only samples with acceptable collagen C/N ratios (2.9-3.6 74 ) are reported. Plant samples represent homogenous mixtures of 2-8 grains (for barley, Hordeum vulgare, and wheat, Triticum aestivum) and 15-22 grains (for broomcorn millet, Panicum miliaceum, and foxtail millet, Setaria italica) per archaeological context (see Supplementary Table S1). The samples represent the bestpreserved grains at the site. The plant isotopic values were corrected for a charring offset of + 0.1‰ in δ 13 C values and + 0.3‰ in δ 15 N values following Nitsch et al. 75 . The barley and wheat δ 13 C values were converted to Δ 13 C values to enable comparison of the carbon isotope discrimination of the samples to the 'watering thresholds' established by Wallace et al. 59 .

Data availability
All new data discussed in this paper are presented in the "Results" and Supplementary Materials. Data that have been published previously (Liu et al. 19,23 ) are acknowledged in the "Materials and methods" and in the Supplementary Materials.