Depicting plant behaviour through human imprints

Depicting plant behaviour through human imprints: wild cereals‟ cultivation in the Holocene Sahara Anna Maria Mercuri1,*, Rita Fornaciari1, Marina Gallinaro2, Stefano Vanin3, Savino di Lernia2,4* 1 Laboratorio di Palinologia e Paleobotanica, Dipartimento di Scienze della Vita, Università degli Studi di Modena e Reggio Emilia, Viale Caduti in Guerra 127, 41121 Modena, Italy. 2 Dipartimento di Scienze dell'Antichità, Sapienza, Università di Roma, Via dei Volsci, 122 – 00185 Roma, Italy. 3 Department of Biological Sciences, School of Applied Sciences, University of Huddersfield, Queensgate, Huddersfield HD1 3DH, UK. 4 School of Geography, Archaeology & Environmental Sciences, University of the Witwatersrand, Johannesburg, Private Bag 3, Wits 2050, South Africa. * Corresponding authors: [email protected]; [email protected] Abstract The human selection of food plants cannot always have been aimed exclusively at isolating the traits typical of domesticated species today. Each phase of global change must have obliged plants and humans to cope with and develop innovative adaptive strategies. Hundreds of thousands of wild cereal seeds from the Holocene “green Sahara” tell a story of cultural trajectories and environmental instability revealing that a complex suite of weediness traits were preferred by both hunter-gatherers and pastoralists. The archaeobotanical record of Takarkori rockshelter in SW Libya covering four millennia of human occupation in the central Sahara gives us a unique insight into long-term plant manipulation and cultivation without domestication. The success of a number of millets was rooted in their invasive1 opportunistic behaviour, rewarded during their coexistence with people in Africa. These wild plants were selected for features that were precious in the past but pernicious for agriculture today. Reconnecting past practices with modern farming strategies can help us to seek out the best resources for the future. Understanding the long-lasting relationship between plants and humans is of crucial importance to sustainable development, and food is a key determinant of this multifaceted „alliance‟. Food plants are a central focus for various disciplines which often postulate a marked dichotomy between wild and domesticated species as playing different roles in cultural evolution. From this perspective, plant diversity decreases along an ideal, unilinear, human-controlled axis from resource exploitation to agricultural management1 2. The culmination of this process is the farming of the small number of domesticated species on which people largely rely for food today3. Such crops developed through human selection of desirable traits; by contrast, most wild plants are thought to have grown independently of human aims. Nevertheless, food plants include some minor wild cereals that, interestingly, present what is known as high „weediness‟: the ability to thrive in human environments due to their preference for disturbed soils4. Weedy plants are often prolific and their high rates of phenotypic plasticity and adaptation to agricultural systems, alongside the capacity to act as invaders in natural habitats, often lead to these plants being relegated to the undesirable „weed‟ category (but for the definition in this paper see below, the Methods)5 6. Here, based on combined archaeological and archaeobotanical data from Saharan Africa, we present the first evidence that non-agricultural societies selected plants for traits that overlap a considerable degree with traits that are characteristic of plants now considered as weeds. Examination of historical agricultural tradition suggests that several wild plants, defined as „weeds‟, possess key adaptive characteristics that were successful during past environmental changes. The sharp distinction between „unusefulness and usefulness‟, typically made on modern-day „crop and non-crop‟ organisms by agri-food actors, has no analogy in long-term archaeological/environmental records. We suggest that these modern concepts are no longer able to explain complex or intermediate conditions of cultural history such as the middle ground 7 between foraging and farming. Yet understanding plant management in foraging societies, prior to the development of agriculture in their regions, and when links between climate, environment and resources were deeply 2 embedded in socio-cultural systems, is crucial to our knowledge of what happened when cultivation did not lead to domestication. Here, we present archaeobotanical data which provides direct evidence of long-lasting plant management, with wild cereals persistently adapting to both human desires and environmental changes. The setting is the Holocene “green” Sahara and the period the four millennia from approximately 7500 to 3500 calBC. Numerous plant species have been gathered and manipulated but this has not always resulted in domestication8 9, a fact often explained as a consequence of the environment being sufficiently starch-protein productive10 or of plant genetics impeding the isolation of „domestic‟ traits3. Many partial or abandoned trajectories of trait selections have shown that domestication did not have a single cause or pattern but was a process with different pathways and species11. We can now add that, even when involuntary, selection was not „only and always‟ directed at isolating the features now considered typical of domesticated species. In the case of cereals, two main phenotypic changes occurring with domestication are usually modifications in seed morphology: the large seed size and the non-shattering rachis12 13 14. The harvesting for these characters is known to have had effects on loss of seed dormancy, reduced vegetative dispersal by rhizomes and changes in reproductive shoot architecture15. However, these morpho-physiological traits are not detectable in the archaeological record. We present, therefore, the first diachronic archaeobotanical evidence showing that, before clear changes in plants, the primary quality that attracted hunter-gatherers in Africa seeking the benefits of cultivation – implying a shift of labour investment to just a few food plant species16 – must have been the invasive and opportunistic behaviour of some wild grasses. The plant remains discovered at various archaeological sites worldwide testify to millennia of systematic gathering, whereas previously the early non-domestication cultivation of wild cereals for food and fodder had only been suggested in the Near East17 18. In this paper, we provide evidence for the systematic gathering, processing and cultivation of wild cereals at a time when the Sahara hosted extensive wet habitats and was covered by a savanna-like vegetation. Highly sophisticated relationships with the environment are apparent in many sites of the mountainous central Saharan regions19, and especially in the Takarkori rockshelter. This site lies on the first terrace of the Wadi 3 Takarkori in the southern Tadrart Acacus (Libya), about 100m above the wadi bed (Figure 1a). Hence, wild cereal stands were not available in the immediate vicinity of the site but were found closeby in the river plain, along with other plant foods, wild game and fish. The first occupation of Takarkori by Late Acacus hunter-gatherers started at c. 8200 calBC lasting until 6400 calBC. Later, Saharan Pastoral Neolithic groups, exploiting domesticate animals, used the site until c. 3000 calBC20 21. The archaeological deposit comprises an impressive quantity of mainly dried plant remains with excellent preservation, indicative of limited microbial degradation activity. This exceptional feature also characterizes other sites in the area22 and in the Sahara more generally. Human groups at Takarkori had sophisticated subsistence strategies, corralling Barbary sheep23, exploiting dairy products and cooking vegetables and the site also provides astounding evidence of the human ability to make art and weave fibres24 25 26 27. We now report the unique evidence of plant accumulations from the site, where loose wild cereals were clearly visible on the archaeological surfaces. These accumulations consist of thousands of grass florets/spikelets concentrated on top of the organic sand layers, and distributed around a few different sectors of the site (Figure 1b,c,d). The accumulations are evidence of different episodes of plant processing and food plant storage during the period c. 7500-3500 calBC (Figure 2). They demonstrate the broad diversity of wild cereals available during the „green‟ Sahara, though most of these species are now locally extinct in the central Sahara28 29. Results Wild cereals accumulated in the site - We identified 30 accumulations of dried seeds clearly visible on the organic sand layers uncovered during fieldwork (Table 1). As this record consists largely of wild cereals, the Takarkori seed accumulations are consistent with direct evidence of food at the site, though their use for fodder, and possibly bedding, cannot be discounted. Based on their morphological features, these plant remains are attributed to Brachiaria (including B. cf. leersioides-Blue signal grass), Bromus cf., Cenchrus (including C.biflorus-Kram kram, and C.ciliaris-African foxtail grass), Dactyloctenium (D. aegyptium-Crowfoot grass), Digitaria, 4 Echinochloa, Panicum, Pennisetum (P. cf. purpureum-Elephant grass), Setaria, Sorghum, Tragus cf., Urochloa (U. cf. panicoides-Liverseed), and to undifferentiated Andropogoneae. Combined morphological analysis and ancient DNA extraction made it possible to identify Panicum laetum (Wild fonio, or Desert panic), Echinochloa colona (Awnless barnyard grass, or Jungle rice) and Sorghum bicolor subsp. verticilliflorum (Wild Sudan grass) (for identification see Methods, and Supplementary Information 1). The ‘spot’ vs ‘mix’ wild cereal accumulations - The ratio of whole to fragmented plant remains allowed us to statistically distinguish between „spot‟ (>5; mainly whole spikelets; Figure 1c) and „mix‟ (<2; mainly spikelet fragments; Figure 1d) accumulations (see Methods, and Table 1). In total, the „spots‟ contained ~150,000 seeds, almost all attributable to the MP1 pastoralists and LA3 hunter-gatherers (71% and 16% of the total seeds in the „spots‟, respectively), while the „mix‟ accumulations with ~80,000 seeds belonged to the MP2-MP2/LP1 pastoralists (85% of the total seeds in the „mix‟) and to the LA2 hunter-gatherers (5%) (Table 1). „Spot‟ accumulations prevalently consist of whole spikelets (of the millets listed above) without external glumes. This may result from threshing and winnowing, which removed a number of by-products after millet harvesting30. By-products such as unripe grains, de-grained ears, plant stalks and rachis, leaves, spike parts and isolated glumes are fairly insignificant in „spot‟ accumulations. Winnowing by shaking, possibly with the help of baskets25, is the most plausible method given the limited space of the rockshelter terrace. The spikelets are unusually uniform in size (Panicum and Echinochloa: ~2.1±0.1 mm length; Sorghum: ~6.4±0.4 mm; see plots in Supplementary Information 2), probably as an effect of sieving, a common procedure in the processing of millets31. Slight modifications in the seed morphology of millets were observed (see below, and Methods). The cleaned spikelets were stored for consumption. Overall, this evidence is insufficient to tell us whether these seeds were eaten directly in the form of porridge27, stored without further processing for later consumption or kept until the next season for seeding. „Spots‟ were found both in a Late Acacus stone hut dated to approximately 7000 calBC (ID 8 and 10, respectively Sorghum/Cyperaceae and Urochloa/Brachiaria) and in a cluster of the Middle 5 Pastoral period (ID 13-18 and 20-22). Spatial analysis (Figure 1b) and study of the radiocarbon dates of these specimens suggest that almost all the MP spots may belong to a single storage facility dated to c. 4500 calBC (4546 <95.4%> 4463 calBC; Figure 2). Furthermore, the absence among the spot materials of remains of ants, which are capable of moving seeds, allows us to rule out the idea that these insects were responsible for these seed accumulations. Considering all the data, this archaeobotanical evidence can be interpreted as a residual form of low-level storage. This term was used for purpose-built granaries in an early Holocene pre-domestication context of the Jordan Valley, which was interpreted as possible cultivation of wild cereals32. In our case, probably, storage was made possible by the fact that most of these wild cereals exhibit strong seed dormancy, and would thus not have germinated until the following season (late spring and summer in the case of Liverseed and Jungle rice). The „mix‟ accumulations consist mainly of plant fragments and glumes, and their function is more ambiguous: depending on the archaeology and chronology, they may be food-processing waste or fodder residues (as in the case of stable floors). Focusing specifically on the „mix‟ with food processing residues, as the hulled cereals need dehusking, we found numerous split spikelets with opened external and inner glumes, a number of glumes representing plant processing by-products 30 31 , and some charred seeds as refuse from burning activities at the site (Figure 3). In summary, „spot‟ accumulations suggest the excess from a prevalent gathering activity while the „mix‟ type seems to originate from the mixed refuse produced by plant processing and animal breeding. Late Acacus vs Pastoral wild cereal accumulations - In a diachronic perspective (Figure 4), the data indicates a major botanical change in the passage from the Late Acacus hunting-gathering phases (12 seed accumulations from c. 7300 to c. 6400 calBC) to the Middle and Late Pastoral phases (11 seed accumulations from c. 4850 to c. 3970 calBC; 7 seed accumulations from c. 3970 to c. 3500 calBC). No accumulation of seeds is recorded from the Early Pastoral phase, probably for taphonomic reasons since these layers suffered heavy erosion in the site19. 6 Within the Late Acacus, ~ 800-1600 cereal seeds, with some remains of woody fruit plants such as Ficus and Ziziphus cf., and one „spot‟ of ~ 17,000 Setaria seeds were found in the earliest five LA2 accumulations, dated c. 7500-7100 calBC. Concentrations generally increase (up to ~10,000 cereal seeds) in the seven LA3 accumulations, dated c. 7100-6400 calBC. At this point, the focus of hunter-gatherers seems to have shifted from the earlier group of cereals (Pennisetum, Sorghum and other Andropogoneae, and to a lesser extent Panicum and Cenchrus) towards the tufted annual Urochloa-Liverseed, perhaps for a combination of cultural and ecological reasons (see below). Within the Pastoral Neolithic, relatively higher seed concentrations (~1500-45,000 seeds) are evident in the „spots‟ of the earliest ten MP1 accumulations, dated c. 4800-4300 calBC. A new cereal, Echinochloa, appears and immediately becomes abundant and recurrent in the record. Meanwhile, Urochloa continues to be common and sometimes prevalent (65-85% in 4 spots; Table 1); Panicum and Setaria also seem to have been intensively exploited at this time. Although about two thousand years of cultural and environmental changes33 34 separate these MP1 samples from those of the LA3 phase, the long-term continuity of the botanical discoveries made by previous hunting-gathering inhabitants is apparent. This can only be explained by the synergy between the adaptive behaviour of some wild cereals and perhaps by cultural transmission from generation to generation. Jungle rice, which probably did not grow in the area before, is a new arrival among the previous cereals. The archaeobotanical diagram (Figure 4) suggests that pastoralists brought Jungle rice with them and managed this plant alongside other millets already present in the area. This plant began to dominate whereas the importance of Liverseed decreased and almost disappeared from the MP2 and MP2/LP1 records. Discussion Plant processing at the site The whole spikelets and variable amounts of by-products in „spot‟ and „mix‟ accumulations reflect the different plant management and processing activities taking place at the site. Consistency 7 between the botanical evidence and the archaeological contexts – mostly spatial and chronological data – supports the idea that this type of archaeobotanical record is a strong cultural marker for Saharan societies. According to ethnographic information18 31, dispersed fruits fallen from wild plants such as Pennisetum, Cenchrus and Echinochloa may have been collected from the ground. Open baskets might have been used to gather seeds, as suggested by the basketry fragments, including some that preserve trapped grains (Pennisetum, Brachiaria, Dactyloctenium), unearthed at Takarkori25 and other sites in the region35 36. These plant accumulations show that cereals were processed at the site, and that selected seeds were probably stored after threshing, winnowing and sieving. The preferential location of „spot‟ accumulations in stone huts as can be seen in the northern sector of the site (Figure 1b), indicates a long occupation by the last Late Acacus hunter-gatherers focusing on abundant and predictable - albeit seasonal - plant resources. Our data show that they particularly harvested and transported Andropogoneae to be processed at the site for about 500 years. A low-level storage phase is evidenced by the spots accumulated after c. 7200-6700 calBC (from ID 7, Panicoideae 57% with predominance of whole spikelets). This early storage seems to have taken place before the „change‟ from Andropogoneae (Pennisetum, Sorghum) to Paniceae (Urochloa), with a transition dating to c. 7000-6600 calBC (from ID 8 to 9). The Sorghum bicolor species are known to have among the longest history of plant management in Saharan Africa37 38. Based on our record, we argue that sorghum was present but not widespread in the area, or alternatively that it grew far from the site, probably in mixed plots (e.g. with fonios39), and that the species was vulnerable to environmental instability. The spot of sorghum (72%, ID 8) also contains a significant amount of small-seed Cyperaceae (11%), perhaps an unwanted weed collected with the food plant when the stem was cut. This special spot suggests that gatherers not only stored sorghum seeds for consumption but also for possible seeding in the subsequent season. If this was for cultivation it appears to have failed, because sorghum did not spread enough to be recorded in other subsequent spots (Figure 4). Today, S. bicolor (subsp. bicolor) is considered a competitive crop but the species is actually less invasive than weeds like Johnson grass (Sorghum halepense (L.) Pers.) and several millets. The presence of the prolific weed Urochloa panicoides, for example, was found to reduce sorghum yields by 25-40% in modern uncontrolled farms40. In the early Holocene Sahara, 8 although sorghum produced larger seeds than other wild cereals, the weedy Liverseed, with its high seed productivity, offered greater advantages to people. Humans therefore shifted from exploiting cereals with larger seeds (sorghum, Figure 3) to the more prolific species (Liverseed), which in turn benefitted from anthropochory, or human dissemination. The adaptive behaviour of Jungle rice Archaeobotany suggests that a new influx of useful plants arrived at the site with Neolithic Pastoralists at around 4800 calBC (Figure 4). Less drought-tolerant than the sorghum of the Late Acacus phases, Echinochloa expanded in the area, likely also favoured by a middle Holocene wet and warm phase. At this time, the Sahara saw the development of wadi vegetation and semi-arid seasonal savanna with fragmented habitats and high lake levels in this area20 29. After a Neolithic MP1 phase of coexistence with Urochloa (U.panicoides), Jungle rice prevailed over the resident Liverseed during the MP2 (c. 4200-3900 calBC) and MP2/LP1 transition (c. 3970-3760 calBC), a time of great environmental instability in the central Sahara41. Echinochloa colona adapts easily to changing habitats in terms of soil water status42 and competitiveness. Its monotypic stands cause a decrease of plant species diversity with the increase of river distance and grazing intensity43. The abundance of Dactyloctenium during the Pastoral phases may represent an early phase of overgrazing already suggested by pollen20. Specifically, the intensive removal of plant biomass and trampling by small livestock – mostly ovicaprines – led to the establishment of fast-growing opportunistic species that expanded rapidly into native plant communities and vulnerable wetlands44. In this sense, these wild plants reacted as the dump-heap weeds associated with human-disturbed habitats that are thought to have thrived close to preagricultural dwellings45 (but the weedy origin of the domesticated major crops in the Near East is debated46). In our record, where we have weeds without domesticate species, the behaviour of Jungle rice may be only partly explained by its opportunistic-invasive nature. Our data suggest that in the competition between two weedy and very similar plants (Echinochloa vs. Urochloa), Jungle rice was also favoured by human selection. Cultivation without domestication in the central Sahara 9 Despite differences in the subsistence basis of peoples, our archaeobotanical record is evidence that some wild cereals (Urochloa, but also Panicum and Cenchrus) were continuously managed for four millennia, with systematic gathering and processing, possibly shifting towards some forms of cultivation more than once. Cultivation before domestication has been inferred from the presence of species outside their current natural habitats, the distance of natural plots from sites, overexploitation, the import of seeds and climate changes (see 9, e.g., the Middle Euphrates). Many of these features correspond with the long-term management of wild cereals in the Sahara. However, the Takarkori record is more analogous with a cultivation without domestication with anthropic pressure forcing slight modifications in the seed morphology of Panicum, which became plumper, and Echinochloa, which became more elongate (Figure 5). In eastern China, c. 4000 calBC, plumper and longer grains were found which reflect significant morphological shifts likely due “towards harvesting of mature panicles as opposed to immature panicles”38. In Syria, minimal changes in seed size suggested the use of plump-grained wild cereals (e.g. wild einkorn), favoured by cultivation without domestication9. Such changes are thought to precede the increase in seed size during the domestication process47 48. However, the Saharan pathway seems to have been quite different from that taken by the ancestors of major crops in the Near East, and the shape changes in our record can be regarded as “a directional change under selection by early cultivation practices”15. In Africa, the domesticated species of pearl millet47, t‟ef48 and African rice49 have smaller grains than the wild progenitors, and one major trait isolated in the domestication of pearl millet and t‟ef seems to have been the selection for higher grain numbers (a feature common in wild species having more seeds in individual spikes). From the archaeobotanical diagram of Takarkori (Figure 4), we infer some selection of „special traits‟ in African plants. First, it may have been the weediness of the wild grasses growing in the area (the ID1 to 12 accumulated by the LA2 and LA3), and then the increasing attention towards the collection of mature seeds (Panicum laetum) and panicles richer of seeds (Echinochloa colona). In this area of central Sahara, the long process from gathering to cultivation resulted in a pre-domesticate state for the two species, but the route seems to have continued (or was paralleled) elsewhere only for E. colona. Despite four millennia of human management of Saharan wild 10 grasses, domestication did not occur. Indeed, the species we identified are still largely reported as non-crop and weedy plants today (see Supplementary Information 3 for a list of citations in the scientific literature). The dual nature of wild cereals With few exceptions, the grasses we found in the Takarkori record are often described as aggressive colonizers, invaders, segetal, ruderal, sometimes mimic, weeds, and generalists growing wild in a broad range of habitats from swamps to open grasslands to semi-desert habitats. Many are annual species producing more seeds than perennials, and summer crops. Since they are droughttolerant and withstand heavy grazing, they are currently widely cultivated as food or forage grasses in tropical areas (e.g. Brachiaria and Urochloa 50). Our plant accumulations demonstrate that some Panicoideae (first Pennisetum, Setaria and Sorghum, then Brachiaria/Urochloa with Panicum, and Echinochloa) were continuously managed, and even actively propagated by seeding, so many of them survived environmental changes and lived in the area until desertification and water loss definitively limited their growth starting from c. 3500 calBC 22. Interestingly, in Eastern China, wild Echinochloa was processed with grinding stones but this became an agricultural weed once cultivated Asian rice became dominant in cultivation 51. In the absence of any domesticated crop, our African records show that Echinochloa colona was cultivated in association with Panicum laetum among the others. The early cultivation practices may have comprised a mixture of small-grained millets. High-value Panicoideae played an important role in human dietary starch-protein intake, and seem to have spread easily by moving into newlycreated human environments 1 4. Abundance and predictability were important factors in their evolutionary success, as the weedy traits were selected and appreciated first by early Holocene foragers for their food security and culinary practices, and later by Neolithic herders 27. The selection of weediness The Late Acacus hunter-gatherers settled near large stands of wild cereals to harvest huge amounts of seeds; the pastoralists continued this tradition, harvesting larger quantities of mature grains, likely not simply selecting only for 'large and non-shattering' seeds. Humans selected invasive and opportunistic plant behaviour before morphological traits of some grasses. 11 The archaeobotanical record that shows, with overlapping, the trend from gathering under prevalently natural selection to the cultivation under both human and natural selections52 (from ID 112 to ID 13-30, respectively; Figure 4). People selected weediness because it was an easier way to grow plants. By this process, we argue that humans elicited biological responses by these weeds, and the selected plants were rewarded as they improved their own fitness. Considering these human actions like parts of a 'long process' 15, our data show that the selection of weediness has continued during the transition from gathering to farming (Figure 4). The use of resilient plants (mainly Urochloa in our record) seems to have been a more conscious early action than the subsequent involuntary selection for the richest panicles searching for abundant grains (that likely caused the elongation of our Echinochloa spikelets). Efficient dispersal mechanisms, seed dormancy, and phenotypically plastic genotypes are common traits in the weed syndrome rooted in their capacity for rapid adaptation to environmental changes53. Interestingly, some wild cereals such as fonios are considered semi-domesticated species that have lost their efficient natural seed dispersal, a step towards domestication in many Poaceae. These plants present advantageous adaptations for cultivation, as they are able to grow with other crops and also highly competitive with weeds39 in subsistence agro-ecosystems. Although each weed species may have a diverse and unique history, our data suggest the revision of some commonly held opinions on these invasive field weeds. Specifically, we argue that: a) the selection of large natural stands was not a form of low-skilled pre-cultivation but a genuine human strategy based on the presence of opportunistic-invasive behaviour of some grasses; b) these plant species were the direct object of long-term human manipulation and not ancillary plants that spread after crops; and c) weedy millets had sometimes behaviours that, according to our data, make them preadapted to spread with domestic crops (including the aptitude to mimicry). This may have been effect of long-time human pressure on wild species. From a co-evolutionary perspective1, people selected for weediness - including abundance and predictability - thus ensuring that both plants and humans adapted successfully to environmental crises. The cultivation system, often regarded as beneficial only to humans, also proved highly advantageous to some „special‟ wild plants. Reconnecting past and future Wild cereals present both wild (weed) and cultivated (crop) behaviours and characteristics. The genera of millets found at Takarkori rockshelter include species that have been domesticated in 12 parallel numerous times. We argue that their tendency to evolve as crops is probably rooted in the very long history of manipulation and management the record of Takarkori shows for the African route to domestication. We agree with Viguera et alii 53 that our weeds “compose a unique evolutionary state, neither wild nor domesticated, that has developed in parallel to crop domestication”. Signs of selection for lost-value traits are hard to be recognized today but it is probable that these were not „aborted trajectories‟ as millennia of sophisticated plant management should have affected plant traits as well as human culture. The opportunistic behaviour of these wild cereals favoured their development in unstable environments while the human pressure did not cause severe limits11 on the chances of their autonomous survival. Today, based on their ecology 54, we should change our view of such weeds. The same behaviour that allowed these plants to survive in a changing environment in a remote past makes them some of the most likely possible candidates as staple resources in a coming future of global warming. They continue to be successfully exploited and cultivated in Africa today and are attracting the interest of scientists searching for new food resources. Our research suggests that, like modern gatherers, we should pay renewed attention to these plants in the hope of finding innovative responses to tackle desertification and biodiversity loss. 13 Methods Definition of the term „weediness‟ and the main traits of our weeds The term “weed” has been variously defined, depending on the different perspectives of ecologists, agronomists, archaeobotanists and the public, and we do not know a definition that has met a universal agreement55 (see Supplementary Information 3). The „weediness‟ described in this research on the Saharan prehistoric, and pre-domestication, landscape requires a definition of the term in an ecological perspective. Weeds possess many adaptive strategies to survive such as the phenotypic plasticity (i.e., the ability of an organism to express different phenotypes depending on the biotic or abiotic environment56), abundant seed production, prolonged viability of propagules with dormancy (so as to survive under unfavourable conditions), ability to fast growth of seedlings, spread at a high rate, high tolerance to environmental heterogeneity and to changes in environmental conditions, strong interspecific competition and resilience. Weeds are part of a dynamic ecosystem57 and all these traits make these weeds ideal for easy proliferation. Our weeds include invasive and opportunistic species. The term „invasive‟ is often used to indicate these plants (problematic today) as “they can successfully establish and spread following introduction into novel, often non-managed areas”. The authors of the previous definition reported also that “invasives are considered weeds in the broadest sense of the term; some authors (not the case of this paper) prefer to use the designation „weed‟ only for problematic plants in agriculture, and reserve the term „invasive‟ to indicate problematic plants growing in non-managed or wild areas” 55. Moreover, the species we identified in our archaeobotanical record are characterised by „opportunistic behaviour‟. In ecology, an opportunistic plant can thrive in different conditions, and rapidly takes advantage of environmental changes because the species is behaviourally flexible. Excavation and sampling Takarkori was excavated entirely by hand, using small trowels. The plant accumulations were identified by sight, mapped using an Electronic Total Station (ETS) and removed. The samples were then gently drysieved with a 0.6 mm screen in the laboratory, and the residue was examined to check for smaller parts (seeds were not found). Direct radiocarbon datings of each plant accumulation placed them in a chronological trend from the hunter-gatherer (Late Acacus: LA2, LA3) to Pastoral Neolithic phases (Middle and Late Pastoral: MP1, MP2, MP2/LP1), known for the stratigraphy of the site and the Tadrart Acacus region19 58 (see Table 1, with archaeological phases and calibrated ages59; Figure 2). Stereomicroscope examination showed that the plant accumulations consisted largely of dried yellow to brown spikelets of Panicoideae with a few other 14 plant taxa dispersed in the residual sand. These accumulations are consistent with one or a few plants entering the archaeological deposit contemporaneously. Entomological analyses demonstrated that insects played no role in transporting plant parts into the site. Insect search was carried out on 13 samples: seven spots (ID 2, 9, 10, 12, 17, 20, 25), one mix (ID 30), and five samples of soils external to the plant accumulations but belonging to the same stratigraphic unit and used as controls. The search was performed using a stereomicroscope and every arthropod fragment was isolated using micro-paintbrushes and identify at the lowest taxonomic level. In all the samples fragments of spider beetles (Coleoptera, Ptinidae) and of migratory locusts (Orthoptera, Acrididae) were detected. Few fragments (mainly heads) of ants (Hymenoptera, Formicidae) were found only in the control samples, whereas among the plant accumulation material no ant fragments were identified. Taxa identification and morphometry The botanical finds were hand-sorted and morphologically identified by observation under an Olympus SZ60 stereomicroscope at 40x-80x magnification, with the help of the specialized literature60 61 62 63 64 65 66. Whole spikelets (grain enclosed by lemma and palea) of all the species mentioned above dominate in the archaeobotanical record; spikelets consisting of the persistent papery tissues of external glumes (upper and lower glumes) and the smooth to verrucate inner glumes (lemma and palea) are also found in some accumulations. Since Brachiaria and Urochloa inter-hybridize, the boundary between the two genera is sometimes unclear and a variable number of intermediate phenotypic stages (from quasi-awnless to mucronate tip spikelets) were observed, suggesting that large stands hosted mixed populations. In the accumulations studied here, Pennisetum mainly takes the form of involucres of soft and plumose sessile bristles, and Cenchrus of the spiny involucres fused at the base. Alongside the fertile spikelets we observed sterile or immature spikelets of Pennisetum and Sorghum, both presenting the typical wild species morphology. The sorghum has hulled spikelets with the shattering wild type of spikelet basis, and its immature florets/spikelets are flattened and sometimes with persistent rachillae, probably reflecting the ripening variability of the panicle (see SEM images in Supplementary Information 1). Variable amounts of non-grass fruits/seeds including Ficus, Ziziphus cf., Cyperaceae, Persicaria and various few others (Asteraceae, Amaranthaceae, Apiaceae, and Colocynthis) were also present. Given the nature of the archaeological deposit, comprising mostly organic sands 19 , the seed accumulations also include small 15 amounts of stems (grasses and sedges), leaves and charcoal fragments, a few undetermined flowers and charred plant remains. From each plant accumulation, whole spikelets of Panicum laetum Kunth, Echinochloa colona (L.) Link. and Sorghum bicolor (L.) Moench subsp. verticilliflorum (Steud.) de Wet ex Wiersema & J. Dahlb. - syn. S. arundinaceum (Desv.) Stapf - were extracted under the stereomicroscope. The main morphological and morphometrical characters, on fertile, mature and intact spikelets, are: P. laetum: elliptic shape with pointed ends; bright spikelets, brown to reddish, with nuances of orange; fine longitudinal linear striations on lemma and palea; lemma: concave, lateral edges folded on the palea, 5-7 longitudinal veins are clearer than others; palea: almost flat, with two longitudinal veins more clear. The species is native to tropical Africa, and it has considerable importance in areas of marginal subsistence where it is often consumed as a famine food. Measurements on 690 specimens: 2.01 ± 0.09 mm (length) x 1.23 ± 0.10 mm (width) x 0.90 ± 0.08 mm (thickness). E. colona: elliptic to oblong shape; notched apex (three points, the central one is thicker and longer than the others), curved downwards; awnless, shiny spikelets, yellow to light brown; fine longitudinal linear striations on lemma and palea; lemma: concave, lateral edges folded on the palea, five longitudinal veins with three veins more evident than the others, and the middle one slightly larger; palea: flat. The species is hexaploid, native to several regions including Africa67, and the wild progenitor of Echinochloa frumentacea Link (syn. E. colona var. frumentacea (Link.) Ridl; sawa millet or Indian barnyard millet)68 69. The close DNA affinity between the two species is supported by their cross-compatibility70. Although, phylogenetic relationships within Echinochloa species are not fully understood, the ploidy differences between wild and crop species suggest that gene flow from cultivars has not a role in the weed‟s evolution53. Measurements on 750 specimens: 2.11 ± 0.13 mm (length) x 1.32 ± 0.09 mm (width) x 0.88 ± 0.10 mm (thickness). S. bicolor subsp. verticilliflorum: lanceolate shape; pointed apex (often covered with fine hairs, especially on the rachilla attachment point); shiny spikelets, dark brown with reddish hues; lemma: widely convex, without streaks, lateral edges folded on the palea; palea: convex; without streaks. The species is diploid, or tetraploid, native to Africa, and is the wild progenitor of S. bicolor subsp. bicolor (domesticated grain sorghum)71. Measurements on 100 specimens: 6.41 ± 0.35 mm (length) x 2.33 ± 0.14 mm (width) x 2.35 ± 0.17 mm (thickness). 16 Photos of the spikelets in dorsal and lateral view were taken using the Leica MC170 HD camera connected to the stereomicroscope, maintaining constant parameters (default options for white background), and the magnification of the stereoscope (9.0x) (Figure 3). The length, width and thickness of each spikelet were measured using the software package Leica Application Suite - LAS EZ version 3.0.0 (Leica Microsystems). These identifications were supported by molecular analyses. Ancient DNA was successfully recovered from the spikelets of Panicum, Echinochloa and Sorghum, allowing for the Polymerase Chain Reaction (PCR) amplification and sequencing of the barcode regions used for the taxonomic determination. Chloroplast marker rbcL provided the identification of the three genera. The DNA barcode sequence obtained from Echinochloa matched with maximum identity value (99%) seven samples in GenBank: the five accessions identified as E. colona, one accession of E. frumentacea [(Roxb.) Link] and one accession of E. ugandensis (Snowden & Hubbard). Moreover, the multi-locus approach (rbcL, matK, trnH-psbA and trnL markers) allowed for the species identification of Panicum laetum 72 73. Spikelet number estimations All the specimens in the accumulations of ≤ 15 ml were counted (ID 5, 9, 11, 13, 14, 15, 16, 20, 23, 24, 25, 27, and 28). To estimate the total number of macroremains in the other twenty-one accumulations, three separate 2, 4, 10 ml sub-samples were taken from each accumulation. The number of specimens in each subsample was used to compute the average number providing estimates according to the total volume. To account for differences in the counts, the weighted average and the absolute error were calculated for the final estimate in the total volume as shown in Table 1. The number of seeds in the 30 plant accumulations was used to calculate the total number of seeds in each cultural phase and in the archaeobotanical record. The number of broken spikelets was recovered by a visual validation according to the known size and morphology of the whole spikelet 74. A spikelet was determined by a lemna-palea association, or by 3-8 fragments depending on their size. The amount of spikelets recovered by inspecting the broken parts of Brachiaria, Urochloa, Panicum, Echinochloa, Digitaria, Dactyloctenium and Setaria was used to calculate the percentage ratio of whole to fragmented spikelets in each plant accumulation. This ratio was helpful in 17 describing what was evident on visual examination, i.e. the different status of plant accumulations with prevalent whole spikelets („spots‟) and those with prevalent chaff and processing by-products („mix‟). The morphology and fragmentation of the soft parts, and the general archaeological context, provide a better understanding of the nature of „mix‟ accumulations. For example, the most recent samples are more abundant in isolated glumes (e.g., of Dactyloctenium), accompanied by other cereal chaff and straw, and may be remains of fodder. Data availability The data that support the findings of this study are available from the corresponding authors upon request. References 1. Zeder, M.A. Core questions in domestication research. P. Natl. Acad. Sci. USA 112, 3191–3198 (2015). 2. García-Granero, J.J., Urem-Kotsou, D., Bogaard, A. & Kotsos, S. Cooking plant foods in the northern Aegean: Microbotanical evidence from Neolithic Stavroupoli (Thessaloniki, Greece). Quatern. Int. doi.org/10.1016/j.quaint.2017.04.007 (2017). 3. Diamond, J. Evolution, consequences and future of plant and animal domestication. Nature 418, 700–707 (2002). 4. Cox, S. in Plant Breeding and Farmer Participation (eds Ceccarelli, E. P. G. S. & Weltzien, E.) 1–26 (Rome, Italy: FAO, 2009). 5. Baucom, R. S. & Holt, J. S. Weeds of agricultural importance: bridging the gap between evolutionary ecology and crop and weed science. New Phytol. 184, 741–743 (2009). 6. Kuester, A., Conner, J. K., Culley, T., & Baucom, R. S. How weeds emerge: A taxonomic and trait-based examination using United States data. New Phytol. 202, 1055–1068 (2014). 7. Smith, B. D. Low-level food production. J. Archaeol. Res. 9(1), 1–43 (2001). 8. Weiss, E., Kislev, M. & Hartmann, A. Autonomous cultivation before domestication. Science 312, 1608–1610 (2006). 18 9. Willcox, G., Fornite, S. & Herveux, L. Early Holocene cultivation before domestication in northern Syria. Veg. Hist. Archaeobot. 17(3), 313–325 (2008). 10. Murphy, D. J. People, Plants & Genes: The Story Of Crops And Humanity (Oxford Univ. Press, 2007). 11. Fuller, D. Q., et al. Convergent evolution and parallelism in plant domestication revealed by an expanding archaeological record. P. Natl.l Acad. Sci. USA. 111(17), 6147–6152 (2014). 12. Meyer, R. S. & Purugganan, M. D. Evolution of crop species: genetics of domestication and diversification. Nat. Genet. 14, 840–852 (2013). 13. White, C. E. & Makarewicz, C. A. Harvesting practices and early Neolithic barley cultivation at elHemmeh, Jordan. Veg. Hist. Archaeobot. 21(2), 85–94 (2012). 14. Fuller, D. Q. & Allaby, R. G. Seed dispersal and crop domestication: shattering, germination, and seasonality in evolution under cultivation. Annu. Plant Rev. 38, 238–295 (2009). 15. Fuller, D. Q., Allaby, R. G., & Stevens, C. Domestication as innovation: the entanglement of techniques, technology and chance in the domestication of cereal crops. World Archaeol. 42(1), 13–28 (2010). 16. Purugganan, M. D. & Fuller D. Q. The nature of selection during domestication. Nature 457, 843– 848 (2009). 17. Willcox, G., Nesbitt, M. & Bittmann, F. From collecting to cultivation: transitions to a production economy in the Near East. Veget. Hist. Archaeobot. 21, 81–83 (2012). 18. Kislev, M.E., Weiss, E. & Hartmann, A. Impetus for sowing and the beginning of agriculture: ground collecting of wild cereals. P. Natl. Acad. Sci USA 101(9), 2692–2695 (2004). 19. Biagetti, S. & di Lernia, S. Holocene deposits of Saharan rock shelters: the case of Takarkori and other sites from the Tadrart Acacus Mountains (Southwest Libya). Afr. Archaeol. Rev. 30(3), 305–338 (2013). 20. Cremaschi, M. et al. Takarkori rock shelter (SW Libya): an archive of Holocene climate and environmental changes in the central Sahara. Quaternary Sci. Rev. 101, 36–60 (2014). 21. di Lernia, S. & Tafuri, M. A. Persistent deathplaces and mobile landmarks: The Holocene mortuary and isotopic record from Wadi Takarkori (SW Libya). J. Anthropol. Archaeol. 32(1), 1–15 (2013). 22. Mercuri, A. M. Plant exploitation and ethnopalynological evidence from the Wadi Teshuinat area (Tadrart Acacus, Libyan Sahara). J. Archaeol. Sci. 35(6), 1619–1642 (2008). 23. di Lernia, S. Dismantling Dung: Delayed Use of Food Resources among Early Holocene Foragers of the Libyan Sahara. J. Anthropol. Archaeol. 20, 408–441 (2001). 24. di Lernia, S. et al. Colour in context. Pigments and other coloured residues from the Early-Middle Holocene site of Takarkori (SW Libya). Archaeol. Anthrop. Sci. 8(2), 381–402(2016). 19 25. di Lernia, S., N'siala, I. M. & Mercuri, A. M. Saharan prehistoric basketry. Archaeological and archaeobotanical analysis of the early-middle Holocene assemblage from Takarkori (Acacus Mts., SW Libya). J. Archaeol. Sci. 39(6), 1837–1853 (2012). 26. Dunne, J. et al. First dairying in green Saharan Africa in the fifth millennium BC. Nature 486(7403), 390–394 (2012). 27. Dunne, J., Mercuri, A. M., Evershed, R. P., Bruni, S. & di Lernia, S. Earliest direct evidence of plant processing in prehistoric Saharan pottery. Nat. Plants 3, 16194 (2016). 28. Ozenda, P. Flore Et Végétation Du Sahara (Centre National de la Recherche Scientifique, 2000). 29. Mercuri, A.M. Human influence, plant landscape evolution and climate inferences from the archaeobotanical records of the Wadi Teshuinat area (Libyan Sahara). J. Arid Environ. 72(10), 1950– 1967 (2008). 30. Song, J., Zhao, Z. & Fuller, D. Q. The archaeobotanical significance of immature millet grains: an experimental case study of Chinese millet crop processing. Veg. Hist. Archaeobot. 22(2), 141–152 (2013). 31. Moreno-Larrazabal, A., Teira-Brión, A., Sopelana-Salcedo, I., Arranz-Otaegui, A. & Zapata, L. Ethnobotany of millet cultivation in the north of the Iberian Peninsula. Veg. Hist. Archaeobot. 24(4), 541–554 (2015). 32. Kuijt, I. & Finlayson, B. Evidence for food storage and predomestication granaries 11,000 years ago in the Jordan Valley. P. Natl. Acad. Sci.USA 106(27), 10966–10970 (2009). 33. Cremaschi, M. in Droughts, Food and Culture 65–81 (Springer US, 2002). 34. Kuper, R. & Kröpelin, S. Climate-controlled Holocene occupation in the Sahara: motor of Africa's evolution. Science 313(5788), 803–807 (2006). 35. Castelletti, L. et al. in The Uan Afuda Cave Hunter-gatherer Societes of Central Sahara (ed di Lernia, S.) 131-148 (AZA Monographs 1, All’Insegna del Giglio, Firenze, 1999). 36. Cremaschi, M. & Zerboni, A. in Landscape and Societies, Selected Cases (eds Martini, I. P., Chesworth, W.) 67–89 (Springer Science, 2011). 37. Wasylikowa, K., & Dahlberg, J. in The Exploitation of Plant Resources in Ancient Africa 11–31 (Springer US, 1999). 38. Fuller, D.Q. Contrasting patterns in crop domestication and domestication rates: recent archaeobotanical insights from the Old World. Ann. Bot. 100(5), 903–924 (2007). 39. De Wet, J. M. J. in Advisory Group Meeting on the Possible Use of Mutation Breeding for Rapid Domestication of New Crop Plants (Vienna, Austria, 1986). 40. Walker, S., Wu, H. & Bell, K. Emergence and Seed Persistence of Echinochloa colona, Urochloa panicoides and Hibiscus trionum in the Sub-tropical Environment of North-Eastern Australia. Plant Prot. Q. 25(3), 127 (2010). 20 41. Clarke, J. et al. Climatic changes and social transformations in the Near East and North Africa during the ‘long’ 4th millennium BC: A comparative study of environmental and archaeological evidence. Quaternary Sci. Rev. 136, 96–121 (2016). 42. Khedr, A., Serag, M., Shaaban, H. & Abogadallah, G. Differential responses of aquatic and aerobic forms of Echinochloa crus-galli (L.) Beauv. and E. colona (L.) Link. by morpho-physiological and molecular analysis. Environ. Earth Ecol. 1(1), 81–93 (2017). 43. Andrew, S.M., Totland, Ø. & Moe, S.R. Invasion of the cosmopolitan species Echinochloa colona into herbaceous vegetation of a tropical wetland system. Ecol. Res. 29, 969 (2014). 44. Zedler, J. B. & Kercher, S. Causes and consequences of invasive plants in wetlands: opportunities, opportunists, and outcomes. Crit. Rev. Plant Sci. 23(5), 431–452 (2004). 45. Smith, B. D. General patterns of niche construction and the management of ‘wild’ plant and animal resources by small-scale pre-industrial societies. Philos. T. Roy. Soc. B. 366(1566), 836–848 (2011). 46. Abbo, S., Gopher, A., Rubin, B., & Lev-Yadun, S. On the origin of Near Eastern founder crops and the ‘dump-heap hypothesis’. Genet. Resour. Crop Ev. 52(5), 491–495 (2005). 47. Fuller, D. Q. & Hildebrand, E. Domesticating Plants in Africa (eds Mitchell, P. & Lane, P.) 507–525 (Oxford Univ. Press, 2013). 48. D'Andrea, A.C., Klee, M. & Casey, J. Archaeobotanical evidence for pearl millet (Pennisetum glaucum) in sub-Saharan West Africa. Antiquity 75(288), 341–348 (2001). 49. Wu, W. et al. A single-nucleotide polymorphism causes smaller grain size and loss of seed shattering during African rice domestication. Nat. Plants 3, 17064 (2017). 50. González, A.T. & Morton, C.M. Molecular and morphological phylogenetic analysis of Brachiaria and Urochloa (Poaceae). Mol. Phylogenet. Evol. 37(1), 36–44(2005). 51. Yang, X., et al. Barnyard grasses were processed with rice around 10000 years ago. Sci. Rep-UK. 5, 16251 (2015). 52. Milla, R., Osborne, C.P., Turcotte, M.M. & Violle, C. Plant domestication through an ecological lens. Trends Ecol. Evol. 30(8), 463–469 (2015). 53. Vigueira, C.C., Olsen, K.M., & Caicedo, A.L. The red queen in the corn: agricultural weeds as models of rapid adaptive evolution. Heredity 110(4), 303–311 (2013). 54. Gurevitch, J., Scheiner, S.M. & Fox, G.A. The Ecology of Plants (A.D. Sinauer Associates, Inc.) 258–259 (Sunderland, Massachusetts, USA, 2002). 55. Kohli, R.K., Batish, D.R., & Singh, H.P. in Handbook of sustainable weed management (eds Singh, H.P. et al) 1–19 (Haworth Press Inc, New York, 2006). 21 56. Agrawal, A.A. Phenotypic plasticity in the interactions and evolution of species. Science. 294(5541), 321–326 (2001). 57. Aldrich, R.J. Weed-crop ecology: principles in weed management (New Breton Publ., North Scituate, Mass, 1984). 58. Cremaschi, M., & di Lernia, S. Holocene climatic changes and cultural dynamics in the Libyan Sahara. Afr. Archaeol. Rev. 16(4), 211–238 (1999). 59. Bronk Ramsey, C. Bayesian analysis of radiocarbon dates. Radiocarbon 51(1), 337–360 (2009). 60. Wasylikowa, K. Holocene flora of the Tadrart Acacus area, SW Libya, based on plant macrofossils from Uan Muhuggiag and Ti-n-Torha/Two Caves archaeological sites. Origini 16, 125–159 (1992). 61. Wasylikowa, K. Exploitation of wild plants by prehistoric peoples in the Sahara. Würzburger Geogr. Arbeiten 84, 247–262 (1992). 62. Mercuri, A. M. in UanTabu in the Settlement History of the Libyan Sahara, Arid Zone Archaeology, Monographs 2 (ed Garcea, E. A. A.) 161–188 (All’Insegna del Giglio, Firenze, 2001). 63. Olmi, L. et al. Cereali selvatici nel Tadrart Acacus - Sahara Centrale, durante l'Olocene iniziale e l’Olocene medio. Atti della Società dei Naturalisti e Matematici di Modena 137, 411–430 (2007). 64. Clayton, W. D. & Renvoize, S. A. Flora of Tropical East Africa: Gramineae (part 3) 451–898 (Rotterdam, A. A. Balkema for the East African Governments, 1982). 65. Sheriff, A. S. & Siddiqi, M. A. in Flora of Libya (ed El-Gadi, A. A., Tripoli, 1988). 66. Boulos, L. Flora of Egypt. Volume four. Monocotyledons (Alismataceae – Orchidiaceae) (Al Hadara Publishing, Cairo, 2005). 67. Lansdown, R.V. Echinochloa colona. The IUCN Red List of Threatened Species 2013: e.T164380A1047208. http://dx.doi.org/10.2305/IUCN.UK.2013-1.RLTS.T164380A1047208.en. Downloaded on 03 July 2017. 68. Hilu, K.W. Evidence for RAPD markers in the evolution of Echinochloa millets (Poaceae). Plant Syst. Evol. 189, 247–257 (1994). 69. Sood, S., et al. Barnyard millet–a potential food and feed crop of future. Plant Breeding. 134(2), 135–147 (2015). 70. Yamaguchi, H., Utano, A. Y. A., Yasuda, K., Yano, A., & Soejima, A. A molecular phylogeny of wild and cultivated Echinochloa in East Asia inferred from non‐coding region sequences of trnT‐L‐F. Weed Biol. Manag. 5(4), 210–218 (2005). 71. Carman, J.G., Jamison, M., Elliott, E., Dwivedi, K.K., Naumova, T.N. Apospory appears to accelerate onset of meiosis and sexual embryo sac formation in sorghum ovules. BMC Plant Biol. 11(1), 9 (2011). 72. Olmi, L. et al. in Windows on the African Past: Contemporary Approaches to African Archaeobotany. 175–184 (Africa Magna Verlag, Frankfurt, 2012). 22 73. Fornaciari, R., Fornaciari, S., Francia, E., Mercuri, A .M. & Arru, L. Panicum spikelets from the Early Holocene Takarkori rockshelter (SW Libya): Archaeo-molecular and-botanical investigations. Plant Biosyst. doi.org/10.1080/11263504.2016.1244117 (2016). 74. Martinoli, D. & Nesbitt, M. Plant Stores at Pottery Neolithic Höyücek, Southwest Turkey. Anatol. Stud. 53, 17–32 (2003). Correspondence and requests for materials should be addressed to S.d.L. and A.M.M. Acknowledgements This research is part of the activity of The Archaeological Mission in the Sahara, Sapienza University of Rome. Funds have been granted by Sapienza University of Rome (Grandi Scavi di Ateneo) and by the Italian Minister of Foreign Affairs (DGSP) entrusted to S.d.L. Libyan colleagues of the Department of Archaeology in Tripoli and Ghat are thanked. Funds for morphometrical and genetic analyses were provided by the project “SELCE – SELvaticiCEreali: il futuro nella risposta delle piante ai cambiamenti climatici”, sect. Scientific and Technological Research (Sime n.2015.0033), funded by the FCRMO-Fondazione Cassa di Risparmio di Modena, directed by A.M.M. We thank Erika Milburn and Julie Dunne who helped to clarify some expressions in English. The manuscript has benefited of the constructive comments of four anonymous referees and Simcha Lev-Yadun. Author contributions S.d.L. and A.M.M. conceived and planned the project. A.M.M. and S.d.L. wrote the paper. A.M.M. studied the archaeobotanical record and S.d.L. the stratigraphic and archaeological context. R.F. 23 performed morphometry and data analysis. M.G. made GIS analysis. S.V. made the entomological study. S.d.L. designed and directed the excavations and field sampling. All authors read and approved the final manuscript. 24 Legends Table 1 – The 30 plant accumulations from the archaeobotanical record of Takarkori rockshelter: archaeological context and archaeobotanical results. Cultural phases (*) = LA: Late Acacus; MP: Middle Pastoral; LP: Late Pastoral. From the oldest, they are: LA2: 8500-7900 uncal bp; 7600-6650 calBC; LA3: 7900-7400 uncal bp; 7050-6100 calBC; MP1: 6100-5500 uncal bp; 5200-4250 calBC; MP2: 5500-5000 uncal bp; 4450-3700 calBC; LP1: 5000-4000 uncal bp; 39502350 calBC. Calibrated years BC (**) = date calibration was carried out using OxCal Online v. 4.3 59 Figure 1 – Context of the archaeobotanical record of Takarkori rockshelter. a) The Tadrart Acacus Mountains, SW Libya, and location of the site (red circle); b) GIS map of the excavation and position of plant accumulations by type (mix, spot) and chronology (Late Acacus: yellow; Middle Pastoral: blue; Middle/Late Pastoral: grey); c) „spot‟ accumulation of the Middle Pastoral = MP1 phase (ID 17 in the map „b‟); d) „mix‟ accumulation of the Late Acacus = LA3 phase (ID 6 in the map „b‟). Figure 2 – Multiplot of calibrated (calBC) radiocarbon dates of the 30 archaeobotanical samples (Late Acacus: yellow; Middle Pastoral: blue; Middle/Late Pastotal: grey). See Table 1 for details of dates (generated using OxCal Online ver. 4.3 59). Figure 3 – “Spot” and “mix” plant accumulations. „Spot‟: a) overview (ID 15); dorsal and ventral view of the spikelets of: b) Panicum laetum (ID 15); c) Echinochloa colona (ID 15); d) uncharred grain (ID 14); e) Sorghum bicolor subsp. verticilliflorum (ID 8). „Mix‟: f) overview (ID 24); g) split spikelets of E. colona (ID 24); h) charred grains in dorsal and ventral view (ID 26); i) outer glumes of E. colona (ID 23). Figure 4 – The archaeobotanical record of plant accumulations from Takarkori rockshelter is unique in showing so clearly and gradually the data on the long transitional process from gathering to cultivation. Histograms show the percentages of seeds/fruits identified and counted in the plant accumulations, and their 25 total concentrations (in 10 ml); samples are plotted in chronological order from the oldest (bottom) to the youngest (top) accumulation. Figure 5 - Morphometry of ellipticity. The graphs show changes in the spikelet morphology of Wild fonio and Jungle rice during the occupation of the site. The ellipticity (the degree of deviation from sphericity) of Wild fonio decreases, with the spikelets tending towards an increasing roundness, especially visible when comparing length and thickness. The ellipticity of Jungle rice increases because the spikelets tend to lengthen. This is especially visible when comparing length and width (n = number of measured spikelets). 26 27