Neolithic cultivation of water chestnuts (Trapa L.) at Tianluoshan (7000-6300 cal BP), Zhejiang Province, China

Water chestnuts (Trapa) are frequently recovered at Neolithic sites along the Lower Yangtze River Valley and have been important components of the diets of prehistoric people. However, little systematic research has been conducted to determine their cultural and dietary importance. Excavations at the Tianluoshan site produced large quantities of well-preserved specimens, which provide an excellent collection for studying morphological changes with time. Using modern wild and domesticated water chestnuts (n = 447) as a reference, we find Neolithic samples (n = 481) at Tianluoshan are similar in shape but smaller in size compared to the domesticated species Trapa bispinosa. In particular, the Tianluoshan water chestnuts have bigger seeds than the wild species Trapa incisa. Further, water chestnuts diachronically increased in size at the Tianluoshan site with significant differences (one-way, ANOVA) observed for length (p = 7.85E-08), height (p = 3.19E-06), thickness (p = 1.2E-13), top diameter (p = 5.04E-08) and bottom diameter (p = 1.75E-05) between layers 7 (6700-6500 cal BP) and 6 (6500–6300 cal BP). These results suggest that water chestnuts were actively selected based on size (big), shape (full fruit, two round horns, wide base, etc.) and were an important non-cereal crop to the agricultural practices at the Tianluoshan site.

where plant, animal and human remains were recovered. Since 2004, eight cultural levels were excavated, and in the lower part of the site below layer 6, the artifacts are sealed in a waterlogged and anaerobic environment that results in exceptional preservation of organic material 18 . Abundant plant remains have been recovered such as: rice (Oryza rufipogon/sativa), water chestnut (Trapa bispinosa), acorn (Quercus spp. (sensu lato)), foxnut (Euryale ferox), peach (Prunus persica), bottle gourd (Lagenaria siceraria), etc 19 . Many of the water chestnuts found during the excavation are fragments that represent refuse from human consumption with a comparatively small number of intact specimens. However, in pits where the water chestnuts were stored, they were found completely intact. This large number of waterlogged and well-preserved water chestnuts recovered from the different cultural layers at Tianluoshan (Fig. 1e) permits the investigation of how these plants may have been selected for advantageous traits for human consumption through time.
While past archaeobotanical studies have examined the use of wild and domesticated plant resources in many regions such as the Near East [20][21][22] and Central Europe [23][24][25] (where the preservation of specimens can also be waterlogged), there has been less research attention focused on non-grain crop domestication processes in China. In order to address this topic in more detail, we measure the morphological characteristics of water chestnuts at the Tianluoshan site and compare these results with modern wild and domestic varieties from eastern China. The goal of this research is to investigate possible diachronic changes in the size, shape and type (wild or domestic) of the water chestnuts. By comparing the Neolithic and modern water chestnuts, we can examine the process of water chestnut cultivation at Tianluoshan, which will provide new information about how non-cereal crops were influenced by humans in China.

Results
All sample information and measurements are listed and summarized in the Supplementary Tables 1 and 2.
Comparison of the waterlogged and preserved samples (see Methods section below) from pit H69 (layer 6) found no statistical differences for the five measured parameters except for the top diameter (p = 0.017, one-way ANOVA), which was larger in the preserved specimens (Table 2). This suggests that caution should be applied in the interpretation of this parameter. However, the effects of preservation were not significantly different for the other measurements and the two groups were treated equally. The water chestnut measurements are grouped according to archaeological layer in Table 3, and are found to increase in size from layers 8 to 6. No statistical differences were found between the measurements of layers 8 and 7. However, significant differences (one-way, ANOVA) were observed for length (p = 7.85E-08), height (p = 3.19E-06), thickness (p = 1.2E-13), top diameter (p = 5.04E-08) and bottom diameter (p = 1.75E-05) between layers 7 and 6. In terms of shape characteristics (Table 4) (classification methods can be found in the Method section), the majority of the Tianluoshan water chestnuts (~77%) belong to Type III. They have a relatively large size in comparison to the other types, and have the shape of an inverted triangle (Fig. 2a). The fruits are plump and the shoulder horns are short, round and not spinous. The beak is not distinct and the base connected to the stem is wide. The Type II specimens (~20%) have the shape of a diamond and possess a big crown and neck. In addition, few Type II specimens possess mastoids on their abdomen. The two shoulder horns point horizontally or at a slight angle upwards and have short, sharp and hard tips (usually together with big horns). The Type I water chestnuts or the wild variety are found in the smallest numbers (~4%) and are generally smaller in size with a narrow-inverted-triangle shape and a crown on the top. The shoulder horns slant upward and have hard tips while the base is generally shorter than the Type II samples. Most of the Type I specimens have mastoids on their abdomen, and bases that are similar to those of Type II.

Discussion
Water chestnuts have a long and stable presence in the archaeobotanical assemblage at Tianluoshan 13,26 . The large quantities of intact water chestnuts from storage pits, as well as the broken pieces from human consumption, reflect the important role that this plant played in the human diet.
Diachronic increases in the size of the Tianluoshan water chestnuts from layers 8 to 6 are shown in Fig. 3. These archaeological specimens are smaller than the modern domestic types but remarkably bigger than the modern wild types. In terms of shape, Type III (full fruit, two round horns, wide base, etc.) occupied the dominant position at Tianluoshan for at least 700 years. Comparing the Neolithic water chestnuts with the modern Chinese varieties, specimens at Tianluoshan are relatively similar in shape with the domesticated species Trapa bispinosa Roxb., but distinctly different from the wild species, Trapa incisa Sieb. and Zucc. (Fig. 4).
According to modern botanical research genetic polymorphisms are found in water chestnuts, and they are capable of generating new variants in the wild 27,28 . In the absence of human intervention, they will not consistently produce the same size and shape of fruit if subjected to variable and/or unfavorable growing conditions [29][30][31] . For example, in order to maintain the desirable traits of the Nanhu water chestnuts (a highly domesticated species without sharp horns), modern farmers need to accomplish a series of activities at specific growing times such as: picking the water chestnuts, selecting the good ones for conservation and sowing, maintaining and cleaning the water ponds of invasive plants and manuring 32 . If these tasks are not properly completed, cross-breeding with wild species or self-pollination can occur as water chestnut seeds can lie dormant for years on the bottom of a body of water such as a lake or pond 30 . Thus, the finding that the majority of the Neolithic water chestnuts are attributed to Type III, and increase in size from layer 8 to 6, is strong evidence that they were already in the process of domestication and being actively selected and cultivated for desirable traits (size and shape of the fruit).
The Hemudu Culture, to which the Tianluoshan site is attributed, represents the transitional period from hunting-gathering to agriculture 10,33 . Like water chestnuts, vast quantities of rice were found at Tianluoshan and previous research determined that it was also under the process of domestication [34][35][36] . In particular, the discovery of rice fields around the dwellings indicates that the people had knowledge about the practices of rice agriculture such as: burning straw, weeding and watering the land 35 . The development and advances of agricultural practices during this period likely reflects an understanding of the conditions and activities that were beneficial to the cultivation of a wide variety of plants. Besides water chestnuts and rice, excavations at the Middle and Late Neolithic sites on the Lower Yangtze River Valley found the remains of many other edible plants (eg. peach, tea (Camellia sinensis L.) and melon (Cucumis melo L.)), and these are also believed to have been cultivated and possibly in the process of domestication 19,26,[36][37][38][39] . Thus, being one of the essential food resources of the Hemudu Culture, it is not difficult to understand that water chestnuts were actively selected and cultivated by the Tianluoshan inhabitants, and constitute an important component to the early agricultural practices of the site.
Since the majority of the Tianluoshan water chestnuts unearthed from layer 8 already show a similar type with Trapa bispinosa, while only a minority are approximately close to Trapa incisa, this suggests that the domestication of water chestnuts in this region had commenced sometime earlier than approximately 7000 years ago. Thus, the Tianluoshan water chestnuts were already in the process of domestication but had not yet reached the modern version in terms of size and shape of the domesticated species. So far among the sites of the Shangshan Culture (11000-8600 cal BP), which is one of the oldest Neolithic cultures discovered in the Lower Yangtze River Valley, only starch grains of water chestnuts have been found 40 . However, the macro-remains of water chestnuts have been recovered at the Kuahuqiao site (8000-7000 cal BP) 41 (Fig. 1b,c). Future work is planned to investigate the size and shape of the water chestnuts from Kuahuqiao, and these results will be compared to those of Tianluoshan to better understand the timing and the process of water chestnut cultivation and management in the Lower Yangtze River Valleys of China.  Table 3. Results of one-way ANOVA on Neolithic water chestnuts summed by layers (unit: mm). Estimated ages of the layers are based on radiocarbon dates from Wu et al. 68 and Jin et al. 52,69 . Note: a,b,c are used to show the outcome of the one-way ANOVA, the same letter under the same column means no significant difference.

Conclusions
In spite of being previously considered as a wild dietary resource that was gathered during the Neolithic 15,16 , the water chestnuts at Tianluoshan demonstrate that they were already under intensive domestication and cultivation by approximately 7000 BP. This finding supports the view that Neolithic humans had a good command and understanding of natural resources such as the growth cycles and requirements of various plant species. These findings at Tianluoshan support past archaeobotanical research [20][21][22]43,44 from other parts of the world that agricultural intervention was applied to a series of plant resources, rather than to cereals only. Thus, humans had well-adapted and developed subsistence strategies based on local environmental and climate conditions. This viewpoint is now being recognized by more scholars in China 38,[45][46][47][48] , and contributes to a deeper understanding of how non-cereal plants were influenced by human activity and selection in the Lower Yangtze River Valleys of China.

Methods
Two types of methods were used to collect the water chestnuts at Tianluoshan. 1) In storage pits, the majority of water chestnuts were preserved intact and hand shovels were used to excavate and collect the remains. 2) In the cultural layers, the water chestnut remains were generally husk fragments. However, after manually washing the soil, some complete water chestnuts were selected from the other botanical remains with the use of a mesh screen (0.34 mm).  Complete or nearly complete water chestnuts (n = 481) from five different excavated grids (layers 6 to 8) of the Tianluoshan site were collected and measured with an electronic digital caliper (brand: Guanglu, range: 0-200 mm) to two significant figures (Fig. 2a,b). Since most of the Neolithic water chestnuts do not have obvious necks and humps, which are diagnostic characteristics of modern species, we decided to take measurements of length, height, thickness, top diameter and bottom diameter [49][50][51] to quantify the shapes and sizes of the specimens (Fig. 2c).
The majority of the Neolithic samples (n = 384) were preserved in their original waterlogged condition. However, some of the water chestnuts (n = 97) were previously conserved in a solution of trehalose and water according to the protocol of Nagahama et al. 52 . In order to investigate possible differences related to these preservation conditions between waterlogged (n = 261) and trehalose preserved water chestnuts (n = 52), measurements from both types of samples from pit H69 (layer 6) are directly compared. In addition, eight different modern water chestnuts (n = 447) were also measured including three wild varieties (n = 268) and five domesticated types (n = 179) (Fig. 3). One of these wild populations grows freely inside of the Hemudu site park (~7 km from Tianluoshan), and the other two are from Hubei Province and Shandong Province. The domesticated species were purchased from markets in Yuyao, Jiaxing and Hangzhou, modern cities located in eastern and northern Zhejiang Province.
Water chestnuts were also sorted into three different shape classifications. These were characterized as: Type I (wild), Type II (intermediate) and Type III (domesticated) (Fig. 2a) 1,49-51 . The classification system developed here was based on past research which found that the size of the fruit is an extremely important criteria when classifying water chestnuts 49,50 . In addition, the shape and number of horns 2,49,50 , the existence and height of the mastoid 50 and the shape of fruit 27,31 can all function as features to separate water chestnuts into different groups. First, the water chestnuts at Tianluoshan were placed into two groups according to the size of the fruit. The first group (containing Type I and II specimens) is smaller in size, with an average length of less than 30 mm. In addition, the Type I and II water chestnuts display certain wild type characteristics, such as comparatively sharp horns (defense against predators and to help the water chestnuts insert into the bottom of lakes for reproduction), a short bottom diameter (allowing it to break from the stem easier), and a bigger crown (Fig. 2a-c). In contrast, the second group is composed of the Type III specimens, and these are bigger in size with the majority of samples longer than 30 mm. Further, the Type III water chestnuts display certain domesticated features such as relatively shorter, wider and rounded horns, a wide bottom diameter and smaller or not obvious crowns (Fig. 2a-c).
After this initial classification, some additional differences in the first group suggest that they could be further divided into two additional types: Type I and Type II specimens. This was based on the work of Kadono 27 who pointed out that the shape of the fruit is a critical feature when classifying Trapa and Wan 51 who argued that the mastoid plays an important role in the classification of water chestnuts. Visual observation of the water chestnuts in the first group found that some of the samples had horns that slant upward, making the whole water chestnut a narrow inverted triangle shape (Fig. 2a (wild)). In addition, these specimens have mastoids on their abdomen. As these characteristics are quite similar to the modern wild water chestnuts collected in this paper, these specimens were classified as Type I (wild). The rest of the samples were observed to have horizontal horns, a diamond shape and few mastoids on their abdomen. These specimens display differences compared to the modern wild varieties, but have not reached the shape and size of the domesticated samples. Therefore, these specimens were classified as Type II or and intermediate stage of development.
Data availability. All data generated or analysed during this study are included in this published article (and its Supplementary Information files).