The southern Central Asian mountains as an ancient agricultural mixing zone: new archaeobotanical data from Barikot in the Swat valley of Pakistan

The mountain foothills of inner Asia have served as a corridor of communication and exchange for at least five millennia, using historically documented trade routes such as the Silk Road and the Tea-Horse Road. Recent research has illustrated the important role that this mountain corridor played in the dispersal of crops and farming technology between northeast and southwest Asia 5,000 to 1,000 years ago. However, the role of the mountain valleys along the southern rim of the Pamirs and Himalaya in facilitating crop dispersals has not yet been fully explored. Notably, ongoing debates over secondary dispersals of Hordeum (barley) and Triticum (wheat) into China and the routes of dispersal for the East Asian crops Oryza sativa (rice), Prunus persica (peach) and P. armeniaca (apricot) into northern India are continuing topics of inquiry. In this article, we add to these discussions by focusing on archaeobotanical remains from the Barikot site (ca. 1200 bce–50 ce) in the Swat valley of northern Pakistan. The Swat valley is an ancient settlement zone in the Hindu Kush-Karakoram foothills, whose cultural features have always had a strong link with inner Asia. The archaeobotanical assemblage illustrates that a diverse array of crops, with origins across Asia, were cultivated around the same settlement. Additionally, these farmers likely implemented seasonal cropping cycles and irrigation that required various labour inputs and water management regimes.


Introduction
Human populations around the world congregate in biologically rich ecotonal zones; as such, mountain foothills often represent cultural hotspots (Turner et al. 2011).
Archaeological research across inner Asia has consistently shown high human population density throughout these ecotones between hills and lowlands, especially in river valleys or alluvial floodplains Frachetti et al. 2018). These river valleys have served as agricultural experimentation zones for millennia . Farmers grew both southwest and northeast Asian crops in these foothills for the first time in the late 3rd millennium bce, representing some of the earliest agriculture in Central Asia (Spengler et al. 2014;Spate et al. 2017;Zhou et al. 2020).

3
The even earlier first agriculture in South Asia has been identified in Balochistan, southwest Pakistan, at the sites of Mehrgarh, Sherikhan Teraki, Sarai Khola and Kili Gui Muhammad, all on alluvial fans (Petrie and Thomas 2012). The ease of irrigation along the alluvial slopes facilitated experimentation with farming practices, such as the development of crop rotation cycles (Miller et al. 2016;. Over the past few years, archaeobotanical investigations along the inner Asian mountain corridor from the Pamirs north to the Altai (Kuzmina 2008;Frachetti 2012), has greatly expanded our understanding of crop diversification (Liu et al. 2016;Motuzaite Matuzeviciute et al. 2018;Spengler et al. 2018) and technological advances in cultivation (Miller et al. 2016). Yet, further south along the southern rim of the Pamirs, Hindu Kush, Kunlun and Himalaya, there are still questions regarding when and from where certain crops arrived. Historically, farmers there have grown a mixture of crops of East, South and southwest Asian origins together with North African crops, creating a unique economy across northern India and Pakistan. These crops all have differing cultivation requirements, including water needs, growing day lengths and minimal season durations. In this paper, we contribute to the understanding of cultivation systems in the river valleys and alluvial plains of the southern Central Asian mountains, by adding archaeobotanical data from the site of Barikot (ca. 1200 bce to 50 ce) in the Swat valley of northern Pakistan (Fig. 1). We further explore three questions: 1, what water-demanding crops were cultivated in this semi-arid region; 2, were crop rotation cycles used; and 3, what role did arboriculture play in the economy.
Sharp transitions between altitudinal ecoclines (ecological gradients determined by the elevation level on a mountain) cause greater biodiversity than the flat steppe or desert regions. The ability to access seasonally available resources at different points on the ecoclines or to shift the elevation of grazing pastures or cultivated fields allows people to adapt better to ecological changes and to use complex foraging or farming strategies. Ethnobotanical studies of wild plant foraging in the Swat valley demonstrate that people use seasonally specific plants from different ecotonal zones and are Fig. 1 A map of southern Central Eurasia, with selected agricultural villages or urban sites from the 2nd and 1st millennia bce with archaeobotanical results. 1, Murghab sites (ca. 3000-1000 bce); 2, Sarazm (ca. 3500-2000 bce); 3, Djarkutan (ca. 2500-1500 bce); 4, Shortughai (ca. 2500-1500 bce); 5, Mundigak (ca. 4500-2200 bce); 6, Barikot (ca. 1200 bce-50 ce); 7, Kanispur (ca. 2700Kanispur (ca. -2000 and Qasim Bagh (ca. 1800-1400 bce); 8, Burzahom (ca. 2400-600 bce); 9, Gufkral (ca. 2400-1600 bce) and Semthan (ca. 1500-1 bce); 10, Kunai (ca. 3000-2000 bce); 11, Bhirrana (ca. 3300-2800 bce); 12, Siswal (ca. 3800-1500 bce); 13, Hastinapur (ca. 900-500 bce); 14, Atranjikhera (ca. 2000-200  aware of resource distribution in time and space (Ahmad et al. 2011;Zeb et al. 2013). Many Central Asian mountain ranges have high glacial peaks, which can represent challenges to cross them, especially during the colder season. As a result, mountain passes have historically been areas of great attention, not only providing routes of passage, but also water and forage for grazing livestock, and they also provide useful archaeological deposits (Frachetti et al. 2018;Taylor et al. 2019). The High Terek pass historically connected Xinjiang to Central Asia through the Alay mountains of Kyrgyzstan, as the Jiptik, Torugart, and Irkeshtam passes did through the Pamir mountains. Further south, the valleys of Kashmir, Khyber and Swat in the western Himalayan Hindu Kush region of Pakistan possess dense archaeological deposits, reflecting the funnelling of human populations through them from the 3rd millennium bce until today (Bandey 2009;Han 2012;Coningham and Young 2015;Vidale et al. 2015;Olivieri 2017;Olivieri et al. 2019). These valleys have a long history of sedentary agropastoral occupation, relying on a variety of different crops and large deep conical storage pits (Costantini 1987;Lone et al. 1993;Spate et al. 2017). They also witnessed a tumultuous succession of changing empires between ca 500 bce and 500 ce (Olivieri et al. 2019).
Interestingly, despite strong cultural dispersal through these mountain valleys, there is limited evidence for human gene flow (Jeong et al. 2016). The Swat valley in particular shows an impressive genetic continuity over three millennia from 1200 bce to 1200 ce (Narasimhan et al. 2019), which can tentatively be read as the consequence of the exceptional stability of farming in Swat. Most of the spread of human genetic material onto the plateau came from the north, where ecoclinal gradients are not as pronounced. In this sense, the mountains of Hindu Kush and southern Himalaya may have served as a genetic barrier to populations at low elevations, who were not physiologically adapted to the thin air and other ecological constraints of the higher land (Jeong et al. 2016). Archaeological research in the Kali Gandahki valley and other neighbouring regions in the southern Himalaya illustrate a system of exchange that connected people in the high mountains to populations lower down, in the upper Indus and Ganges (Ganga) drainage basins (Aldenderfer and Eng 2016;Jeong et al. 2016). Evidence for systematic exchange comes from finds of carnelian and glass beads, serving vessels, cups, bamboo mats and cotton cloth in 1st millennium bce burials (Eng and Aldenderfer 2017), textiles of cotton, linen and wool, produced in tabbies and twills (Gleba et al. 2016), metal artefacts from the mid 1st millennium ce (Massa et al. 2019) and salt (Dhungel 2002). Additionally, the well-documented Tea-Horse Road ran from the southern China lowlands, in what is now Yunnan, to the high plateau during the late 1st millennium ce (Mair and Hoh 2009). These patterns of exchange between high and low regions seem to exist for the river valleys spanning the entire periphery of the Central Asian montane zone, and this study places the data from Barikot in this cultural context. Therefore, the Barikot site provides an ideal case study for better understanding the movement of material culture, including crops, between South, Central and East Asia.

The Barikot site
Barikot is located on a silty terrace below a rocky outcrop (34° 40′ N, 72° 12′ E; 799 m a.s.l.) in the middle of the Swat valley, Khyber Pakhtunkhwa Province, Pakistan, at the confluence of two tributary rivers, the Kandag and Karakar (Vidale and Olivieri 2002). The first evidence of human occupation at Barikot goes back to ca. 1700-1400 bce (Swat Period IV; Stacul 1987), although occupation in the area dates back to the early 3rd millennium bce, as evidenced at the nearby cave of Galeghai (Stacul 1987). From 1200 to 800 bce, archaeological data show the presence of a large settlement (roughly 15 ha), including a hilltop acropolis (Callieri et al. 2000), an inner stronghold (Olivieri et al. 2019) and an extensive outer graveyard (Narasimhan et al. 2019). After a phase of abandonment with scarce archaeological data, the construction of an urban centre began at ca. 500 bce (Olivieri and Iori 2020;Olivieri et al. 2019). The site was already fortified, before the construction of a long urban wall with square towers every 28 m which was built after 150 bce at the time of the successors of the Indo-Greek ruler, Menander. At that time the city included an acropolis on the rocky outcrop and covered ca. 12 ha. The city grew in the following centuries, reaching its maximum area at the time of the Kushans around the second century ce. After a series of earthquakes combined with the collapse of the Kushan Empire, the lower city was abandoned towards the end of the third century or slightly later (Olivieri 2011(Olivieri , 2012 and reconstructed along the flanks of the acropolis (Olivieri 2003). The site has been excavated by the Italian archaeological mission in Pakistan, Istituto Italiano per il Medio ed Estremo Oriente (IsMEO), Italian Institute for the Middle and Far East (now ISMEO, International Association of Mediterranean and Oriental Studies), since the 1970s in sondages by G. Stacul (with important contributions by L. Costantini and other specialists) and it has been consistently excavated since 1984, under the directorship of P. Callieri and L.M. Olivieri.
Barikot is located on the highest and most strategic point in the valley. It is characterized by a rocky hill overlooking the Swat river to the north, and a vast terrace sloping down to the south. The area is marked by two tributary valleys, of the rivers Karakar and Kandak, which flow through about 100 km 2 of fertile silty plains surrounded by mountains. This area represents the ager (catchment area including agricultural fields) of Barikot (Fig. 2). The city is known in Classical sources as Bazira or Beira (Beira = Pakistani and Indo-Iranian va(y)ira and Bazira = Sanskrit vajra, Baums 2019) and in medieval sources as Vajirasthana (von Hinüber 2020). It was an Assakenian stronghold that was besieged and conquered by Alexander the Great in 327 bce. The area was renowned in antiquity for its fertility; in historical sources the abundance of crops and the size of herds in this area were praised by Arrian of Nicomedia in the early second century ce (Brunt and Robson 1976-83, IV 25, 4). Curtius Rufus) defined it as an urbs opulenta (rich city) in the first century ce (Atkinson and Yardley 2016, VIII 10, 22), the latter being a term clearly indicating agricultural wealth, which is otherwise very rarely used by the Roman historian and only applied to Tarsos, Babylonia, Persepolis and to Bactria (L. Prandi, personal communication). Previous archaeobotanical research in the Swat valley by Costantini (1987) included a section on earlier occupation layers at Barikot (ca. 1700-1400 bce). The present work on Barikot represents the first large scale and systematic archaeobotanical study in this region; the data presented here reveal the agricultural system used across the mountain valleys and alluvial fans in the past; they also helps link ongoing discussions about South and Central Asia. By working on this site we can begin to fill in the gaps in our understanding of the spread of crops across the ancient world.

Methods
Sediment samples were collected from the site during excavations in 2016 and 2017, from all distinct features and occupation phases in two excavation trenches, 12W and 11K. The first trench was opened outside the Indo-Greek urban wall, the second roughly 100 m inside the area marked by that wall. The quantity of soil varied between samples, mostly about 7 L which was the size of the sediment bag used. Since the flotation of sediments was done during the excavation, we decided to keep the time gap between excavation and floating very short by not storing sediment samples, so to minimize the risk of contamination or mixing of samples. The samples were floated in the field using a water overflow machine, as described by Fritz (2005), Pearsall (2000) and Watson (1976). Flotation was organized by L. M. Olivieri and performed by M. Vidale, E. Iori, M. Fantoni, with Mr. Abid, ISMEO field officer at Barikot. The minimum sieve size used for the light fraction was ca. 1.4 mm, and heavy fraction material was not collected. The floated samples were then dried, packaged, and, once licensed by the archaeological authorities in Pakistan, were exported to the archaeobotanical laboratory in Jena, Germany. The export license was granted to L.M. Olivieri as director of the ISMEO mission Fig. 2 The site of Barikot, its ager (surrounding land), and the middle Swat valley seen from the southwest (photo by Carla Biagioli, courtesy ACT, Italian Archaeological Mission in Pakistan) by the Department of Archaeology and Museums, Government of Pakistan. Similarly, the non-objection certificate was issued by the Directorate of Archaeology and Museums, Government of Khyber Pakhtunkhwa.
In the laboratory, the samples were systematically sorted and all botanical remains were identified and quantified (ESM 1). All of the identified botanical remains were carbonized except for a few mineralized Celtis fruitstones, which are prone to mineralization. We identified the botanical remains using a personal comparative collection of plant remains and a number of identification keys, including Cappers and Bekker (2012) and Neef et al. (2012). All domesticated crops are individually reported, with density, ubiquity and abundance counts presented in Table 1. Seeds of wild plants appeared at extremely low densities and all possible weed seeds or dung burning remains are mentioned below. Mean number of individuals (MNI) estimates are not used, but we liberally applied the category of Cerealia to include all domesticated grain fragments that were too broken or damaged to differentiate between wheat and barley. In most cases, if less than 50% of the complete grain was present, we included it in Cerealia in an attempt not to artificially inflate the abundance counts. The category of Legume was used to describe fragments of domesticated legumes that were too damaged or fragmented to assign to a type. Due to the coarse sieve mesh used, millet grains and most seeds of wild plants were lost during flotation, if they were present in the first place. The loss of any millet grains has significant implications for interpreting seasonal crop cycles and crop diversification. The limitations of these data for interpreting crop complexes and cultivation practices are discussed below.

Results
The archaeobotanical assemblage from Barikot consists of 159 sediment samples representing 140 individual layers or contexts, which were collected over two separate field seasons. These samples represent a comprehensive blanket sampling of the excavated units, and they should evenly reflect the food plants present over the duration of occupation. In total, the assemblage consists of 5,450 morphologically identified seeds; 934 of those come from the 2016 field season excavations in Trench 12W, 2,642 of them came from the 2016 excavations in Trench 11K, and 1,874 of them came from the 2017 excavations in Trench 11K (ESM 1). In addition to the total sum of identified seeds, there were 596 fragments of Cerealia, and 226 Legume fragments. We do not include Cerealia or Legumes in the total counts, so as not to artificially inflate the crop counts. Surprisingly, there were only 510 fragments of wood larger than 2.0 mm, but these were not identified to species, as there was not a suitable comparative collection of the trees in this region. Of the total 5,450 seeds, 5,301 of those were domesticated grains or legumes. There were also 49 Gossypium arboreum/herbaceum (cotton) seeds. Additionally, there were several fruits and nuts, including 25 Vitis vinifera (grape) pips, 14 Celtis caucasica (hackberry) pips (mostly mineralized), seven Ziziphus nummularia fruitstones, six possible Prunus fruitstone fragments, two each of Juglans regia (walnut) nutshell fragments, Prunus cf. armeniaca (possible apricot) and cf. Daphne, one Rosa sp., and a possible Cucurbita sp. (Fig. 3). The remaining seeds of wild plants include nine Cannabis seeds (a wild morphotype), nine Pooid grass seeds, two each of Cyperaceae, Poaceae, Fabaceae, and one each of Galium sp. and Euphorbia. Of the 5,350 grains and legumes, the most dominant crop in the assemblage was Hordeum (barley) (n = 2,235). It appeared to be primarily H. vulgare var. vulgare (a hulled six-rowed form), which is typically found in Central and South Asia for this time period (Spengler 2015). Some naked barley morphotypes may have been present, but we did not systematically quantify them. There were also 410 Triticum (wheat) grains, which all appeared to be a freethreshing form-presumably T. aestivum ssp. aestivum (bread wheat). Some of these were highly compact, similar to modern sphaerococcoid types, but, as with most wheat landraces, a wide range of variation was present. The wheats mostly appeared to be hexaploid, but there were a few that might be tetraploid durum wheats. As crops spread north from India and northeast from the Iranian plateau into Central Asia, the hulled wheats were left behind, although, T. turgidum ssp. durum (free-threshing tetraploid wheat) from the 3rd millennium bce has been reported from Kashmir (Stevens et al. 2016). Therefore, the possibility should be considered that the few morphologically distinct grains from Barikot could be from glume wheats.
The lack of millets in the assemblage is most likely a result of the coarse sieve meshes used during recovery in the field, although future research will possibly modify this assumption. There were also 1,406 Oryza sativa (rice) grains, which varied immensely in morphology and range from elongated to compact, but mostly overlapped with measurements for O. sativa ssp. japonica (Chen et al. 2020). The most prominent legume is Lens culinaris (lentil). There were 678 lentils or lentil fragments that were intact enough to clearly differentiate them from other legume crops. There were also 287 Pisum sativum (pea) or pea fragments and 60 cf. Vicia sativa (vetch/pea). The possible V. sativa seeds can either represent minor secondary crops or weeds in cultivated fields; it is also possible that they are simply immature or underdeveloped peas. Three possible Vicia ervilia (bitter vetch) might also be immature peas. There were 150 seeds of Vigna mungo/radiata (mung beans or black gram); in this study, we did not attempt to differentiate between different species of cultivated Vigna. There were also 64 Macrotyloma uniflorum (horse gram) and two Lathyrus sativus (grass pea). No legume fragments resembled Cicer arietinum (chickpea). A single cf. Linum sp. (possible flax) seed was recovered from Sample 2016.121; an unidentified large legume was recovered from 2016.118 and a few cf. Avena sp. (possible oat) were recovered, but they may have been field weeds.
Collectively, there were nine clearly domesticated field crops in the assemblage and an additional four possible minor crops or wild relatives that were not morphologically distinct enough for reliable identification. In addition, Vitis vinifera was cultivated. The Ziziphus nummularia fruitstones, Juglans regia and Prunus cf. armeniaca nutshell fragments may represent cultivation of local trees, or Fig. 3 A few key non-grain crops from the Barikot assemblage. a three views of a Pisum sativum; b mineralized Celtis caucasica fruitstone; c Gossypium sp. seed; d the single possible cucurbit seed from the assemblage; e Ziziphus nummularia fruitstone they may represent the collection of wild resources. This is a diversity of crops and intensity of agricultural systems unlike anything that has been identified at any archaeological site further north in Central Asia at this time. Many of these crops, notably the warm-weather legumes, are absent from archaeobotanical results from Central Asia, but they are present further south in the Indus valley. The lack of many of these crops further north is likely connected to ecological limitations; legumes, such as Pisum sativum and Lens culinaris, which are more cold tolerant, spread into Central Asia earlier than other legumes. Likewise, many of these crops, notably Oryza sativa and Gossypium sp., are extremely water demanding and would have required irrigation, as they are traditionally grown in wet fields in this region.

Agricultural diversification
The earliest cultivation system to spread into these valleys and across northern India and Pakistan relied on cereals, notably Triticum aestivum and both naked and hulled forms of Hordeum vulgare, possibly also emmer and free-threshing tetraploid wheat (Fuller 2002;Salunkhe et al. 2012;Stevens et al. 2016). In addition to the cereals, other aspects of the southwest Asian repertoire arrived in northern India early, including Lens culinaris, Lathyrus sativa, Pisum sativum and Cicer arietinum, as well as sheep, goat and zebu cattle (Stevens et al. 2016;Petrie and Bates 2017;Bates 2020). Cereal crops arrived in the Bolan pass in Balochistan at the sites of Mehrgarh by 5000 bce (Costantini 2008), Mundigak in Afghanistan around the same time, Sheri Khan Tarakai by 4000 bce (Thomas and Cartwright 2010), and on the upper Punjab plain by 2500 bce (Vishnu-Mittre 1972;Saraswat 1986;Saraswat and Pokharia 2003;Fuller 2006;Stevens et al. 2016). Liu et al. (2016) recently published direct dates from Hordeum vulgare grains at Damdama (2500-2400 bce) and Lahuradewa (2300-2000 bce), both on the Ganges plain (Saraswat 2005) and from sites along the Indus dating to 2850-2450 bce. Both Triticum aestivum and T. turgidum ssp. dicoccum have been reported from sites in northern India and Kashmir (Lone et al. 1993;Fuller 2002;Salunkhe et al. 2012). As these cereals spread further north and east, the glume wheats were grown less and there is currently no evidence for T. turgidum in Central Asia until the Islamic period (see discussion in Spengler et al. 2018). Farmers integrated other southwest Asian crops into the cultivation system in Kashmir, including Lens culinaris and Pisum sativum (Saraswat and Pokharia 2003).
The Swat and Kashmir valleys are corridors of exchange and communication through the Hindu Kush mountains and they have dense archaeological remains, including farming village sites that date back to roughly 2500 cal bce (Vidale and Olivieri 2002;Bandey 2009). Historically, the Jhelum river valley in Baramulla District served as a key corridor connecting the Kashmir valley to Punjab. Interestingly, this region has many rivers and can easily provide irrigation, but it sits in a rain shadow outside the effects of the Indian summer monsoons. At this time in Kashmir, the southwest Asian crops were grown together with crops that were domesticated in India, such as Vigna mungo and V. radiata (Saraswat 1986;Saraswat and Pokharia 2003;Fuller 2006;Stevens et al. 2016). Archaeologists have argued that occupation and farming intensified in the Swat valley in the 2nd millennium bce, and geometric rock art depictions have been interpreted as field cultivation scenes (Vidale and Olivieri 2002;Vidale et al. 2011).
Archaeologists have noted that early sedentary occupation in northern India and Pakistan is culturally distinct from that of the Indus valley and further south (Mani 2008). Much of the current interpretation of what this cultural system looked like comes from the sites of Burzahom and Gufkral in Kashmir (ca. 3000-1500 bce), where cereal grains and sheep and goat bones were recovered (Sharma 2000(Sharma , 2013Bandey 2009). Oryza sativa and Prunus fruitstones, likely almonds, have been reported in the archaeobotanical assemblages from the later phases at these sites dating between 1800-1000 bce (Kajale 1982;Lone et al. 1993). Although Pokharia et al. (2018) do note that there is some question about the exact stratigraphic context of the Oryza sativa grains, they do mention directly dated T. aestivum (2290-2051 cal bce) and H. vulgare (2467-2236 cal bce) grains from Kanishkapura (Kanispur). Other southwest Asian crops appear to have spread into Kashmir at this time as well, attested by the finds of T. turgidum ssp. dicoccum, Lens culinaris, Lathyrus sativa, Pisum sativum and Vicia sativa from Kanispur. A later occupation layer at Kanispur, between 100 and 300 ce, showed that East Asian millets, Panicum miliaceum and Setaria italica (and possibly Echinochloa sp.), as well as Oryza sativa, had spread into the region by this period (Mani 2008). The site of Qasim Bagh, in the same area, provided direct dates on wheat and millet grains spanning 1800 to 1400 bce (Spate et al. 2017). The site is located at about 1,500 m, tucked into the Kashmir valley with the Himalaya to the east and the Pir Panjal mountains to the west. Also, single specimens each of Vitis vinifera and Sesamum indicum, and two specimens each of Linum usitatissimum and Macrotyloma uniflorum collectively indicate the increasing diversity in cultivated crops through time (Pokharia et al. 2018).
While this paper is primarily focused on the ancient connected cultural realm of the Swat and Kashmir valleys, it is important to acknowledge the interconnected nature of the broader region. Archaeobotanical research on the northern Ganges plain, from cultural contexts spanning 2500-200 bce, has led to the identification of many of the same domesticated crops as we see in the Barikot assemblage. Ancient farming systems along the Indus utilized an astonishing 1 3 array of crops, arguably a greater diversity than anywhere else in the ancient world (Weber 1991(Weber , 1999Fuller and Madella 2002). Pokharia et al. (2017) summarized archaeobotanical remains from the sites of Hetapatti (ca. 2500-1500 bce), Raja-Nal-ka-Tila (ca. 1300-700 bce), and Rajdhani (ca. 800-200 bce). The abundant archaeobotanical data from the Indus region was recently synthesized by Bates (2019), although many other syntheses have been presented Table 2 Summary of crop introductions and geographical areas of origin, covering southern Central and South Asia from the 3rd millennium bce to the 1st millennium ce (Table 2; for example, Misra and Kajale 2003;Weber and Fuller 2008;Pokharia et al. 2014). Several studies have also illustrated shifting preference for millets over time, from primarily Eleusine sp. in the Mature Harappan to a mix of Eleusine sp., Setaria sp. and Panicum sp. in the Later Harappan (Reddy 2003). Table 2 illustrates a gradual process of increasing crop diversification through time, as crops from different regions spread into northern South Asia and cultivation systems expanded to incorporate them.
There have been a handful of archaeobotanical studies of the Kushan period, ca. in the first 300 years ce, after organized trade routes and reliable markets were fully established across inner Asia (for instance, Pokharia and Saraswat 1999). As an example, biological remains from a trench in the northeastern part of the Balaji temple at Naimisharanya, in Sitapur District of Uttar Pradesh, illustrate that there was a complex cultivation system. As in other areas of Central Asia at this time, barley continued to dominate the cultivation system (Srivastava and Tewari 2011) and both hulled and naked forms were present -likely all six-rowed (Saraswat 1980(Saraswat , 1992. Four types of pulses were also reported from Naimisharanya (Srivastava and Tewari 2011). While Echinochloa crus-galli is a common agricultural weed, it was also grown as a crop in parts of Central Asia, likewise, seeds of Vicia sativa at the site could either represent a minor legume crop or weeds in winter crop fields. Interestingly, ancient remains of Azadirachta indica (neem) fruits were also recovered from Naimisharanya (Srivastava and Tewari 2011).

Water-demanding crops
Much of northern Pakistan lies in the rain shadow of the Pamirs and receives limited precipitation, with the majority coming during the winter months. The cultivation of waterintensive crops would have required ways of managing water resources. In exploring the prominence of crops such as Gossypium sp. and Oryza sativa, we can understand better the labour investment needed in farming. Additionally, T. aestivum and most legumes can be water-demanding if they are grown in open fields rather than small gardens.
Traditionally, rice cultivation across West Asia relied on wet paddy fields; but in some cases, this highly waterdemanding crop is/was grown in very arid regions. By the late medieval period, it appears to have been present across West Asia Chen et al. 2020). Understanding the role of rice in the cultivation system of the Swat valley helps illustrate how the crop rose to significance across areas further north and west, despite the evident ecological constraints. The origins and timing of its cultivation in South Asia and the process of hybridization between East and South Asian rice varieties have been topics of debate for roughly a decade (such as Fuller and Madella 2002;Fuller 2005Fuller , 2006Fuller , 2011Saraswat 2005;Tewari et al. 2008;Fuller et al. 2010;Petrie et al. 2016). There is evidence for people collecting wild O. sativa across northern India for several millennia before the introduction of East Asian domesticated O. sativa ssp. japonica (Saraswat 2004(Saraswat , 2005Pokharia et al. 2017). While there is a considerable range of morphological variability among O. sativa grains from Barikot, and there is a significant range of overlap between O. sativa ssp. indica and ssp. japonica, the Barikot grains are generally more compact, resembling more closely ssp. japonica. Following this rule of thumb, the O. sativa at Barikot, like early examples of rice in the southern Himalaya and Central Asia, may have had East and not South Asian origins (Chen et al. 2020). Vidale et al. (2011) suggest that East Asian rice originally dispersed into northwest India through the mountain passes, referencing early O. sativa in pottery impressions reported by Costantini (1987). The prominence of rice in the Indus valley seems to have largely arisen after the introduction of East Asian varieties, which hybridized with local cultivated O. sativa (Stevens et al. 2016). For example, at the site of Hetapatti, on the Ganges plain, in the 2nd millennium bce, Hordeum vulgare is the dominant crop, but by the 1st millennium bce at the site of Rajdhani, rice grains are 250 times more abundant than barley (Pokharia et al. 2017).
Gossypium arboreum/herbaceum is a second waterdemanding crop recovered from Barikot, with a total of 49 seeds identified. While Gossypium production in most of Central Asia and India is for commercial purposes today and therefore extremely water demanding, cotton growing was originally for small-scale consumption. Hirth (2009) notes that small-scale Gossypium cultivation for domestic craft production does not require major water or labour inputs. Most reports of early Gossypium remains from archaeological sites prior to large-scale irrigation and organized cultivation strategies in the medieval period, are restricted to the moist tropical or semitropical zones of South Asia (Brubaker et al. 1999). The dispersal of Gossypium into more northern regions is often linked to the Islamic Green Revolution (Watson 1983), and there may have been photoperiod sensitivity constraints to its northern dispersal (Brite and Marston 2013). The Barikot Gossypium sp. illustrates that cotton at least started dispersing north well before the Islamic period. Likewise, Brite and Marston (2013) report Gossypium sp. seeds from Kara-tepe in Uzbekistan dating to 300-500 ce. Interestingly, traces of Gossypium sp. fabric and threads were identified from inside two reliquaries from the Buddhist sanctuary of Butkara I in Swat, roughly 20 km north of Barikot, belonging to a first and second century ce archaeological context (Costantini et al. 2002). Gossypium sp. was clearly a prominent crop across the Indus region from the mid 3rd millennium bce onwards (Fuller 2008;Bouchaud et al. 2011).

Crop rotation cycles
South Asia, broadly speaking, provides an exceptional region for studying human adaptation to variable and dynamic environments. Within the area of northern India and Pakistan there are widely differing rainfall patterns, with some areas receiving summer monsoonal rains and others better suited for sowing in autumn (Petrie et al. 2016(Petrie et al. , 2017Petrie and Bates 2017). A myriad of possible ancient cultivation strategies have been proposed to accompany these differing ecological constraints (Fuller 2011;Kingwell-Banham et al. 2015). Botanical investigations at the site of Rojdi, led by Weber (1991Weber ( , 1999 suggest that many of the summer crops were millets, Eleusine sp., Panicum sumatrense, Setaria pumila and S. italica. Similarly, work at Harappa has led to discussions of the nature of crop rotation cycles in the Indus valley (Weber 2003). Recently, Petrie and Bates (2017) synthesized this extensive body of literature, further illustrating how complex northern Indian cultivation systems were. The site of Barikot is located in a winter rainfall zone in the rain shadow of the Pamirs, leading to unique cultivation conditions and the ability to grow warm weather crops further north than usual.
The intensification of agricultural strategies appears to be directly connected with trans-Eurasian crop exchange (Fuller and Boivin 2009;. For example, Pokharia et al. (2018) and Lone et al. (1993) suggested that the presence of East Asian millets and rice in Kashmir may represent cereal farming in two seasons, with millet sown in spring and harvested in the dry season. Miller et al. (2016) suggest that the adoption of East Asian millets in urban contexts across Central and southwest Asia may represent the introduction of summer crops into a previously winter crop system. This argument is supported by the almost complete lack of millets in big urban centres until the early 1st millennium bce, whereas they were prominent in the foothills of Central Asia for millennia before then, likely linked to low investment rain-fed farming. Secondary crops often play key roles in crop rotation systems, notably as summer vegetables, legumes, or oil and fibre crops (Sherratt 1983). Oryza sativa was undoubtedly an important summer crop at Barikot; additionally, a few of the South Asian legumes could have been summer crops, including Vigna mungo/ radiata and Macrotyloma uniflorum. Pisum sativum and Lens culinaris are grown most extensively in cool climates, either at high altitudes in South Asia or as winter crops in some hotter regions. The prominent summer crops across the more arid regions of southwest and Central Asia include the East Asian millets, which were not identified at Barikot, either illustrating a largely winter crop-based cultivation system or a loss of data due to the coarse sieve. The unique geography of the Swat valley makes it particularly suitable for a double cropping system with wet and dry season crops sown on the same plot of land in spring and autumn, that was developed at much lower elevations and latitudes. Notably there is high thermal insulation in the valley, because of its unusual East-West orientation, seen in a few other valleys in this region, including Baghlan and Kapisa on the two sides of the Hindu Kush and the Nangarhar, Manshera, Kashmir, and possibly the valley of Almora (Uttarakhand) to the south of the Karakoram Himalayas. These are all examples of what has been defined by one of the authors as "double-crop economic pocket zones" as river valleys particularly good for farming, where mild conditions allow South Asian crops to be grown further north than usual and for two crops per year on a plot of land (Olivieri 2021). In this regard, drought tolerant crops, such as the millets, were less important, and rice played a more prominent role at Barikot.
According to Lone et al. (1993), the East Asian millets arrived late to the southern Himalayan rim and northern India, Panicum miliaceum at ca 100 bce and Setaria italica at 250 ce, although Spate et al. (2017) recently pushed those dates back to the early 2nd millennium bce with the discovery of millet at Qasim Bagh. Setaria italica is well attested in the Indus valley by the late Harappan period (ca. 2000 bce), at Surkotada, Ojiyana, andHulas (cf. Fuller 2003, 2006;Pokharia 2007), and it increases in prominence at later sites (Petrie and Bates 2017;Bates 2019). The dispersal of East Asian millets through the inner Asian mountain corridor is well known . However, a secondary spread of East Asian cultural traits into northern India and Pakistan during the early 2nd millennium bce is also recognized. Fuller and Boivin (2009) note the appearance of Cannabis sativa, East Asian Oryza sativa, Prunus persica (peach), and P. armeniaca (apricot) in northern India as signs of a "Chinese horizon", signs of influence from the northeast. This is present in the valleys of Kashmir and Swat by 1800 bce (Lone et al. 1993;Stevens et al. 2016). Finds of harvesting knives, jade and other stone objects, and tripod ceramic vessels are all signs of cultural dispersal from the northeast (Fairservis 1975;Stacul 1976Stacul , 1993Sharif and Thapar 1992;Han 2012;Coningham and Young 2015).

Arboriculture
Understanding the process of domestication of fruit and nut trees, long-generation perennials, can also provide much information on concepts of land tenure and investment in future generations. It can also inform on scheduling in the economy, resource protection practices, mobility and exchange. It is clear that people have been collecting fruits and nuts from the once-wooded foothills of Central Eurasia for millennia (Spengler and Willcox 2013); however, we know little about when people started actively maintaining these trees and breeding them. In the northern Indus valley during the 2nd millennium bce, the cultivation of long-lived perennials is indicated by archaeobotanical remains of Vitis vinifera, Gossypium arboreum (tree cotton), Phoenix dactylifera (date palm), and Ziziphus spp. (Fuller and Madella 2002;Bates 2020). Slightly later, Prunus persica and P. armeniaca were introduced to the Indus valley; although neither grows particularly well in such hot moist regions. Further south in India, a much wider variety of fruit trees has been archaeobotanically identified by the mid 2nd millennium bce; however, most of these would not have been productive further north, including Artocarpus heterophyllus (jackfruit), Mangifera indica (mango), and Citrus spp. (oranges, lemons etc.) (Fuller 2009;Stevens et al. 2016).
Most of the fruit and nut remains recovered in the Barikot study represent plants that would have grown wild in the foothills, making it impossible to determine to what degree people were cultivating, maintaining, or manipulating them. Only 25 Vitis vinifera pips were recovered from Barikot, but grapes are far less likely to be carbonized than parched cereal grains, especially if the seeds were removed during wine making. Grapes are the best represented fruit of ancient Central Asia, and they played a prominent role in the economy of people in the foothills and desert oases for at least three millennia . Miller (2008) discusses the early appearance of viticulture in the Kopet Dag mountains of Turkmenistan, roughly four millennia ago, and Spengler et al. (2017) discuss the spread of Vitis into the northern foothills by two and a half millennia ago. A single grape pip from Qasim Bagh (Spate et al. 2017) and two from Loebanr III (ca. 1700-1400 bce) were recovered (Costantini 1987). Vitis pips are also found from sites across the Indus valley dating back to the early Harappan period (Bates 2019). Wine making in Swat is archaeologically confirmed by at least as early as the early 1st millennium ce, through the discovery of several wine presses, particularly the area of Barikot (Olivieri et al. 2006), as well as several pottery devices for distillation at Barikot (Allchin 1979) and by textual evidence (Falk 2009).
Ziziphus nummularia and Z. mauritiana are both found at sites across northern India and Pakistan, especially in more arid regions (Bates 2019;Pokharia et al. 2017). A specific small-fruited morphotype of Z. nummularia grows wild throughout the hills of the Swat valley (Ahmad et al. 2011). Therefore, the small Ziziphus fruitstones that we recovered at Barikot may have been collected from the wild. Alternatively, Ziziphus trees are often planted at field edges and only lightly maintained; they often mark boundaries or protect fields, while also providing fruit. It is likely that low investment maintenance of this long-generation perennial was practised by the occupants of Barikot. Similarly, Juglans regia (walnut) woods once covered much of the southern Central Asian foothills; however, these trees were eventually cultivated and the wild stands were lost. Pleistocene deposits at Karewa in Kashmir illustrate how long the walnuts have been present in the mountain foothills and that people have been collecting the nuts for tens of thousands of years (Vishnu-Mittre 1965, 1984. Juglans regia and Prunus dulcis (almond) nutshells were also recovered from Kanispur (3000-2000 bce), illustrating a long history of human use in the region (Pokharia et al. 2018). It is likely that these trees were planted and maintained near the settlement in the Swat valley. It is also possible that the 14 Celtis caucasica seeds and small fragments of nutshell all represent either foraging in the wild or the maintenance of such useful trees near the settlement. Ultimately, none of the arboreal seed/ fruit remains recovered at Barikot definitively indicate the cultivation of trees. However, the possibility that people were loosely maintaining and manipulating the populations of Ziziphus and Juglans regia trees in spaces between fields is highly likely. This likelihood is further supported by the evidence of viticulture, showing that the population was already familiar with perennial crops.

Conclusions
The Barikot settlement site is located in the Swat valley of northern Pakistan and it was occupied during the 1st millennium bce. Historically the valley represented a key pass through the mountains, and these biologically diverse passes where arable farming was possible were areas of human settlement for millennia. The archaeobotanical evidence that we present here shows the complex agricultural practices that were underway in the high mountain pass in a semi-arid ecological zone three millennia ago. The farmers at the site undoubtedly used some form of irrigation or water channelling in order to maintain water-demanding crops in the rain shadow of the Pamirs. Crops, such as Oryza sativa and Gossypium sp., are particularly water-demanding and would have required great labour and time investments. However Gossypium sp. can be grown on a smaller scale in restricted field plots and therefore requires less water input than modern industrial cotton production. Additionally, simple irrigation is particularly easy in mountain valleys with steep slopes and steady glacial melt streams in the summer which can be diverted for this purpose.
The Barikot data help link the archaeobotany of South and Central Asia, allowing further understanding of the processes of crop exchange across the ancient world. The farmers at Barikot grew crops that originally came from central East Asia (O. sativa ssp. japonica), southwest Asia (cereals and legumes), and central South Asia (legumes). These farmers brought together crops that had evolved in different ecological settings and were adapted to different ecological conditions. Due to the particular evolutionary courses of the progenitors for these crops and their different growing habits, all of these crops required specific growing practices. This allowed for the formation of complex seasonal cycles of agriculture, and several summer legumes and rice appear to have complemented winter wheat and barley. Furthermore, there is reason to believe that some limited maintenance of trees for fruits and nuts was practised, although all of the identified ones represent local taxa. The Barikot archaeobotanical assemblage illustrates how complex and regionally specific cultivation practices were across South Asia. This study also adds to our understanding of the prominent role of the river valleys, and "double-crop pocket zones", across the Pamirs, Hindu Kush and southern Himalaya in the spread of crops and cultivation technology.