Cannabis in Asia: its center of origin and early cultivation, based on a synthesis of subfossil pollen and archaeobotanical studies

Abstract

Biogeographers assign the Cannabis centre of origin to “Central Asia”, mostly based on wild-type plant distribution data. We sought greater precision by adding new data: 155 fossil pollen studies (FPSs) in Asia. Many FPSs assign pollen of either Cannabis or Humulus (C–H) to collective names (e.g. Cannabis/Humulus or Cannabaceae). To dissect these aggregate data, we used ecological proxies. C–H pollen in a steppe assemblage (with Poaceae, Artemisia, Chenopodiaceae) was identified as wild-type Cannabis. C–H pollen in a forest assemblage (Alnus, Salix, Quercus, Robinia, Juglans) was identified as Humulus. C–H pollen curves that upsurged alongside crop pollen were identified as cultivated hemp. Subfossil seeds (fruits) at archaeological sites also served as evidence of cultivation. All sites were mapped using geographic information system software. The oldest C–H pollen consistent with Cannabis dated to 19.6 ago (Ma), in northwestern China. However, Cannabis and Humulus diverged 27.8 Ma, estimated by a molecular clock analysis. We bridged the temporal gap between the divergence date and the oldest pollen by mapping the earliest appearance of Artemisia. These data converge on the northeastern Tibetan Plateau, which we deduce as the Cannabis centre of origin, in the general vicinity of Qinghai Lake. This co-localizes with the first steppe community that evolved in Asia. From there, Cannabis first dispersed west (Europe by 6 Ma) then east (eastern China by 1.2 Ma). Cannabis pollen in India appeared by 32.6 thousand years (ka) ago. The earliest archaeological evidence was found in Japan, 10,000 bce, followed by China.

Full text

Vol.:(0123456789)
1 3
Vegetation History and Archaeobotany
https://doi.org/10.1007/s00334-019-00731-8
REVIEW
Cannabis inAsia: its center oforigin andearly cultivation, based
onasynthesis ofsubfossil pollen andarchaeobotanical studies
JohnM.McPartland1,2· WilliamHegman3· TengwenLong4
Received: 20 December 2018 / Accepted: 6 May 2019
© Springer-Verlag GmbH Germany, part of Springer Nature 2019
Abstract
Biogeographers assign the Cannabis centre of origin to “Central Asia”, mostly based on wild-type plant distribution data.
We sought greater precision by adding new data: 155 fossil pollen studies (FPSs) in Asia. Many FPSs assign pollen of either
Cannabis or Humulus (C–H) to collective names (e.g. Cannabis/Humulus or Cannabaceae). To dissect these aggregate data,
we used ecological proxies. C–H pollen in a steppe assemblage (with Poaceae, Artemisia, Chenopodiaceae) was identified as
wild-type Cannabis. C–H pollen in a forest assemblage (Alnus, Salix, Quercus, Robinia, Juglans) was identified as Humulus.
C–H pollen curves that upsurged alongside crop pollen were identified as cultivated hemp. Subfossil seeds (fruits) at archaeo-
logical sites also served as evidence of cultivation. All sites were mapped using geographic information system software.
The oldest C–H pollen consistent with Cannabis dated to 19.6 ago (Ma), in northwestern China. However, Cannabis and
Humulus diverged 27.8Ma, estimated by a molecular clock analysis. We bridged the temporal gap between the divergence
date and the oldest pollen by mapping the earliest appearance of Artemisia. These data converge on the northeastern Tibetan
Plateau, which we deduce as the Cannabis centre of origin, in the general vicinity of Qinghai Lake. This co-localizes with the
first steppe community that evolved in Asia. From there, Cannabis first dispersed west (Europe by 6Ma) then east (eastern
China by 1.2Ma). Cannabis pollen in India appeared by 32.6 thousand years (ka) ago. The earliest archaeological evidence
was found in Japan, 10,000 bce, followed by China.
Keywords Cannabis sativa· Humulus lupulus· Cannabaceae· Biogeography· Centre of origin· GIS
Introduction
Cannabis holds significance in human history and life today
as a triple-use crop. First, its fruits (seeds) provide valuable
protein and essential fatty acids. Archaeological evidence in
a food context dates back to 10,000 bp, in Japan (Kobayashi
etal. 2008). Its bast cells supply fibres, for cordage and tex-
tiles. Carbonized hemp fibres, found with silk and spinning
wheels, date to 5,600 bp, in Henan Province, China (Zhang
and Gao 1999). Its flowering tops produce cannabinoids,
which have been used for medicinal, shamanic, and recrea-
tional purposes. Archaeological evidence of drug use dates
to 2,700 bp, in Xinjiang region (Russo etal. 2008; Jiang
etal. 2016).
Despite a voluminous literature emerging in the last three
decades, the classification of Cannabis and its centre of ori-
gin remains under debate. A single species concept (Small
and Cronquist 1976) has support from measures of popula-
tion differentiation, such as FST (Sawler etal. 2015; Lynch
etal. 2016) and barcode gaps (McPartland 2018). Other bot-
anists recognize several Cannabis species (Hillig and Mahl-
berg 2004; Clarke and Merlin 2013). Hypotheses regarding
the Cannabis centre of origin began with Ibn Wahshīyah in
930 ce. He proposed that šāhdānaj was brought to Babylon
from India or perhaps China (Hämeen-Anttila 2006). De
Communicated by F. Bittmann.
Electronic supplementary material The online version of this
article (https ://doi.org/10.1007/s0033 4-019-00731 -8) contains
supplementary material, which is available to authorized users.
* John M. McPartland
mcpruitt@myfairpoint.net
1 University ofVermont, Burlington, VT05405, USA
2 GW Pharmaceuticals, Sovereign House, Histon,
CambridgeCB249BZ, UK
3 Department ofGeography, Middlebury College, Middlebury,
VT05753, USA
4 School ofGeographical Sciences, University ofNottingham
Ningbo China, Ningbo315100, China

Vegetation History and Archaeobotany
1 3
Candolle (1883) offered Central Asia as the centre of origin
of C. sativa. His biogeographical theories were based upon
the distribution of wild-type plants, as well as linguistic,
historical, archaeological, and fossil data.
Print fossils (i.e. impressions of leaves or fruits in rocks)
of Cannabis are limited to only two collections: Friedrich
(1883a, b) found fossil leaves in Germany that he named
Cannabis oligocaenica. His species epithet indicates the
Oligocene epoch, 33.9–23.03 million years (Ma) ago.
Palamarev (1982) identified a fossil seed (fruit) as “Can-
nabis sp.” in Bulgaria. He dated the find to the late Miocene
(“Pontian age”, 7.3–5.3Ma). Dorofeev (1969) reported a
fossil seed, “Cannabis sp.”, from the Miocene in Siberia.
He subsequently re-identified the fossil as an extinct species,
Humulus irtyshensis (Dorofeev 1982). Humulus and Can-
nabis are sister genera, forming a phylogenetic clade within
the family Cannabaceae.
Contrary to the paucity of print fossils, hundreds of fossil
pollen studies (FPSs) have identified subfossil Cannabis pol-
len. Fröman (1939) first used pollen analysis to reconstruct
the history of Cannabis. Following analytical refinements
by Fries (1958), dozens of papers have been published.
This culminated with the elaborate meta-analysis by Dör-
fler (1990), followed by our own meta-analyses (Long etal.
2017; McPartland etal. 2018).
Cannabis pollen is often straightforward to recognize
(Mercuri etal. 2002), however, morphological similari-
ties between Cannabis and Humulus pollen grains have
prompted palynologists to use collective labels, e.g. Canna-
bis/Humulus or Cannabaceae. Significant labelling bias also
arises: Chinese FPS palynologists assign Cannabis pollen to
“Humulus” (Li 1974), and northern European palynologists
often assign Humulus pollen to “Cannabis” (Wilson 1975).
Parsing Cannabis from Humulus may be more difficult
in Asia than in Europe. Five studies of Asian Cannabis (C.
sativa ssp. indica) indicate that pollen grain diameter is
smaller than that of European hemp (C. sativa ssp. sativa),
so the size of Asian Cannabis pollen falls within the range
of Humulus (literature in McPartland etal. 2018).
FPSs resorting to collective labels (hereafter abbreviated
C–H pollen) have thwarted the study of Cannabis. Clarke
and Merlin (2013) reviewed dozens of Asian FPSs, and they
were flummoxed by FPSs that lumped data as C–H pollen.
Long etal. (2017) synthesized 46 FPSs in their pan-Eurasian
study. They resolved the C–H dilemma by limiting FPSs
to studies that explicitly identified pollen as Cannabis—a
strategy that excluded a lot of C–H data.
McPartland et al. (2018) used ecological proxies,
instead of grain morphology, to differentiate C–H pollen
as either Cannabis or Humulus pollen. Cannabis flourishes
in steppe—an open, treeless habitat. European phytosoci-
ologists and other field botanists report wild-type C. sativa
cohabitating with Poaceae, Artemisia, and Chenopodiaceae
(hereafter abbreviated PAC). Thus C–H pollen in a steppe
assemblage (with PAC pollen) was identified as wild-type
Cannabis. Conversely, Humulus lupulus, a perennial herba-
ceous scandent, requires trees to climb. European phytosoci-
ologists and other field botanists report H. lupulus associat-
ing with Alnus, Populus, and Salix spp. (abbreviated ASP).
Thus C–H pollen in a mesophytic forest assemblage (with
ASP pollen) was identified as Humulus.
Palynologists have noted correlations between Cannna-
bis and PAC, and between Humulus and ASP, and extended
these associations into the past. For supporting palynologi-
cal and phytosociological literature, see McPartland etal.
(2018). In this study, we aim to further develop this method,
focusing in particular on the Asia context. The ecological
proxies method will be applied to a multilingual collection
of Asian FPSs, to reconstruct the evolutionary and human-
related history of Cannabis in Asia.
Methods
FPS search strategy anddata analysis
We collated an FPS database using several internet search
engines (European Pollen Database, Web of Science, Google
Scholar) using keywords and Boolean operators: Asia AND
(palynology OR pollen) AND (Cannabis OR Humulus OR
Cannabaceae). This was repeated with Chinese characters
when searching Chinese-based references. Additional FPSs
were obtained through citation tracking—references in
retrieved publications were searched for antecedent sources,
and these were retrieved. To map pollen in space and time,
retrieved publications had to meet three inclusion criteria:
(1) precise geographical coordinates, (2) accurate chronol-
ogy, (3) a minimal threshold amount of pollen grains.
1. Precise geographical coordinates were localized to
within a hundredth degree of latitude and longitude.
Some studies did not provide geographical coordinates.
We obtained coordinates of those sites via Google Earth,
which uses World Geodetic System of 1984 (WGS84)
datum. Several FPSs were conducted at deep-water sites,
which explains sites located in seas and oceans. Geogra-
phers have debated the border between eastern Europe
and western Asia. We included three studies on the edge:
a deep-water site in the Black Sea off the coast of Tur-
key, a site in Georgia, and a cis-Ural site in Russia.
2. Accurate chronology necessitated restricting data to
FPSs with absolute dating, such as radiocarbon (14C),
optically stimulated luminescence (OSL), or magne-
tostratigraphical methods. We excluded studies that
assigned “relative dates” within sediments cores (i.e.

Vegetation History and Archaeobotany
1 3
dates inferred by changes in vegetation, such as the start
of the Holocene).
3. Palynologists debate the minimal amount of pollen
required to determine the local presence of a plant spe-
cies at a study site (versus pollen at a study site that
arrived via long distance transport). Bottema et al.
(2003) specifically mentioned C–H pollen in the “prob-
lem of long-distance transport”. Several FPSs located
in south-western Asia had three or fewer C–H pollen
grains. These rare grains likely blew in from Europe.
The Etesian winds blow from the Balkans (harbouring
endemic Cannabis) into Israel, Palestine, Lebanon, and
Egypt (Zecchetto and De Blasio 2007). For a retrieved
publication to be included in our study, C–H pollen had
to appear in a minimum of five separate strata within
a stratigraphic core (for details see McPartland etal.
2018).
Ecological proxies intheAsia context
In Central and East Asia, Cannabis seems to have a stronger
alliance with Artemisia, less so with Poaceae and Chenopo-
diaceae. We took a closer look at Artemisia, to better pin-
point the Cannabis centre of origin. Artemisia and Cannabis
share parallel evolutionary patterns: In phylogenetic stud-
ies, Artemisia nests within the Antemideae subfamily (Zhao
etal. 2010), and Cannabis nests within the Cannabaceae
family (Yang etal. 2013). Artemisia evolved in Central/East
Asia during the late Eocene, ca. 36–33.9Ma, out of the
worldwide Antemideae (Miao etal. 2011). Cannabis also
evolved in Central/East Asia, ca. 27.8Ma, out of the world-
wide Cannabaceae (McPartland 2018).
Artemisia and Cannabis are wind-pollinated and dioe-
cious. Both genera exhibit phenotypic plasticity, with adap-
tive phenotypes that respond to environmental changes,
enabling them to colonize new geographic locations. A
meta-analysis of FPSs in Europe showed that Cannabis and
Artemisia were fellow travellers; their geographic ranges
expanded and contracted in unison during, respectively,
glacial periods and warmer periods (McPartland etal. 2018).
Humulus pollen, in Central and East Asia, correlates with
ASP pollen, as well as Quercus, Betula, Juglans, Camellia,
and Robinia (abbreviated ASP +). Analysing Humulus pol-
len in East Asia is complicated by two additional species, H.
yunnanensis and H. japonicus (= H. scandans). H. yunnan-
ensis is a tree-climbing species limited to Yunnan province
(Zhou and Bartholomew 2003). It is a rare and endangered
plant (Hu and Wu 1992). No phytosociological studies have
been published that include H. yunnanensis.
H. japonicus is native to Japan, Korea, and eastern China
(Zhou and Bartholomew 2003), the Russian Far East (Maxi-
movich 1859), Vietnam and Laos (Pételot 1954). Phytosoci-
ological and field studies report H. japonicus in communities
dominated by trees—specifically Alnus (Lee etal. 1976; Kim
etal. 2010; Jeong etal. 2012; Lee etal. 2013), Salix (Kolbek
and Karolímek 2008; Kim etal. 2010; Oh etal. 2010; Jeong
etal. 2012; Lee etal. 2013), Quercus (Lee etal. 1976; Kol-
bek and Karolímek 2008; Kim etal. 2010; Oh etal. 2010),
Robinia (Lee etal. 1976; Kolbek and Karolímek 2008; Kim
etal. 2010; Jeong etal. 2012; Lee etal. 2013; Lee and Ahn
2014), and Camellia (Lee etal. 1976; Kim etal. 2010; Lee
and Ahn 2014; Eom and Kim 2017).
However, H. japonicus sometimes colonizes habitats
that overlap with those of Cannabis. These include ruderal
(Kolbek and Sádlo 1996; Oh etal. 2008) and riverside com-
munities (Jung and Kim 1998). Some riverside communi-
ties include Poaceae, Artemisia, or Chenopodiaceae in the
ground layer (Song and Song 1996; Jarolímek and Kolbek
2006; Balogh and Dancza 2008; Andrek etal. 2010). One
phytosociological study reports H. japonicus co-localizing
with C. sativa—a ruderal community in Korea (Kolbek
and Sádlo 1996). Maximovich (1859) reported H. japoni-
cus growing with C. sativa in the Amur region of the Rus-
sian Far East, and Clarke and Merlin (2013) photographed
feral hemp and H. japonicus growing together in Shandong
Province.
Despite these intermittent associations with herbaceous
plants, Asian palaeobotanists characterize Humulus as a
drought-intolerant climber of trees (Ni etal. 2010). They
treat Humulus pollen as a botanical marker of deciduous
broadleaved forests (Zhou etal. 2007), or tropical evergreen
forests (Lee and Liew 2010).
The ratio of non-arboreal pollen (NAP, pollen from
grasses, forbs, and sedges) and arboreal pollen (AP, tree pol-
len) serves as an indicator of landscape openness. NAP and
AP percentages are oppositional—when one goes down, the
other goes up. Similarly, palynologist have shown that Alnus
and PAC demonstrate oppositional characters (literature in
McPartland etal. 2018).
C–H pollen can be identified as that of cultivated hemp
when its pollen count surges or becomes a continuous curve
in synchrony with pollen from other crop plants. This met-
ric was first adopted by Fries (1958) and used in ten other
palynological studies. Other crop plants include Avena
(oats), Hordeum (barley), Secale (rye), Triticum (wheat),
and Cerealia-type (undifferentiated cereal pollen).
The presence of Cannabis pollen in very high percentages
indicates a former hemp-retting site. Retting is a technical
term for rotting, a process that separates fibres from the rest
of the stalk. Soaking hemp stalks in water encourages bac-
terial growth and retting. When flowering male plants are
soaked in a retting-pond, large quantities of pollen settle
into pond sediments. Cannabis pollen ≥ 15% of TLP (total
land pollen) is usually considered evidence of hemp retting,
and percentages up to 97% have been reported (literature in
McPartland etal. 2018).

Vegetation History and Archaeobotany
1 3
Pollen algorithm
To differentiate C–H pollen, the algorithm by McPartland
etal. (2018) was adjusted to account for Asian conditions.
We identified C–H pollen as that of cultivated Canna-
bis when it appeared de novo along with crop pollen, or
increased at least twofold over earlier pre-Neolithic counts.
Several FPSs in South Asia report Cerealia pollen, attrib-
uted to agriculture, in ancient strata that clearly predate
archaeological evidence of grain cultivation (e.g. 12,000
bp, Quamar and Bera 2017). In these FPSs we looked for
twofold increases and continuous curves of C–H pollen in
the presence of Cerealia pollen.
To differentiate C–H pollen in pre-agricultural strata,
we used ecological proxies. When C–H occurred in
a pollen assemblage where the NAP-to-AP ratio ≥ 2
(i.e. ≥ 66/33%), dominated by steppe vegetation (PAC), we
identified it as wild-type Cannabis. When C–H occurred
in a pollen assemblage where the NAP/AP ratio ≤ 0.5
(i.e. ≤ 33/66%), in the presence of ASP+, we identified it
as Humulus. In some ambiguous FPSs, pollen counts of
PAC and ASP + rise and fall in near-synchrony, and the
NAP/AP ratio approaches 1:1 (i.e. 50/50%). At these sites,
we classified C–H pollen as unresolved C/H.
Archaeobotanical evidence
We also included botanical evidence from archaeological
sites. Previous studies have collated archaeological reports
of hemp seeds, phytoliths, stalk fragments, fibre, cord-
age, or textiles, and pottery impressions of those materi-
als (Clarke and Merlin 2013; Long etal. 2017). McPart-
land and Hegman (2018) stratified the relative robustness
or validity of these materials. Microscopically-analysed
seeds, phytoliths, and stem fragments were considered the
most robust evidence. Fibre, cordage, and textiles were
problematic. For example, Song etal. (2017) unearthed
a few plant fibres identified as hemp, C. sativa, at a site
occupied by Palaeolithic hunter-gatherers that dates to
28.5ka. This could be the oldest hemp fibre ever found,
but their photomicrograph of a “hemp” fibre is by no
means convincing. Song etal. (2017) report plentiful flax
fibres at their site. McPartland and Hegman (2018) detail
the difficulties in differentiating hemp from flax fibre.
Regarding pottery impressions, identifying the plant spe-
cies that made the cord impression is even more difficult.
At least 14 plant species were utilized for pottery cord
impressions in Neolithic China (Kuhn 1988). In this cur-
rent study, we limited archaeological findings to micro-
scopic analyses of seeds, phytoliths, and stem fragments.
GIS mapping andbinning strategy
Latitude and longitude (referencing WGS84 datum) of each
FPS was plotted, using geographic information system (GIS)
software, ArcGISPro 2.2. The FPS sites were plotted on
three maps, corresponding to three binned time slices. Each
FPS site was notated with a symbol indicating pollen inter-
pretation—either wild-type Cannabis, Humulus, cultivated
Cannabis, or unresolved C–H pollen. Archaeological sites
with hemp seeds or phytoliths were notated with another
symbol.
Stratigraphical data were binned intothree time
slices
Bin 1. This period includes the Oligocene (33.9–23.03Ma),
Miocene (23.03–5.3Ma), Pliocene (5.3–2.58Ma), and
Pleistocene (2.58Ma–11.6 ka) epochs. The symbols for
these sites are sized according to the age of the pollen. A
weighted centroid for Cannabis pollen data was also calcu-
lated (weighted by geographical location and age).
Bin 2. Early- to Mid-Holocene, 11.5–5.0ka, a period of
improved climate, re-emerging forests, and the Mid-Holo-
cene Climatic Optimum. This period includes the onset of
agriculture in Asia and the earliest archaeobotanical evi-
dence of Cannabis usage.
Bin 3. Late-Holocene, 4.5–0ka, a period of profound
anthropogenic impact on landscapes, and the earliest
recorded Asian history.
Extrapolating intotheOligocene
McPartland (2018) constructed a maximum likelihood phy-
logenetic tree (PAUP* version 4.0b10) of 11 Rosales gen-
era, using rbcL + trnL-trnF sequences, and a nonparametric
variable rate-smoothing algorithm (r8s version 1.70), cali-
brated with four fossil date intervals (Boehmeria 60–34Ma,
Morus 56–34Ma, Celtis 65–56Ma, Humulus 28–16Ma).
The molecular clock estimated that Cannabis and Humu-
lus evolved (diverged) 27.8Ma. This date is bracketed by
other estimates: 87.3Ma (Boutain 2014), 24.8Ma (Wu etal.
2018), and 18.2Ma (Zhang etal. 2018b). Most of these
estimates predate the oldest Cannabis fossil pollen found in
this study (19.6Ma). Molecular-based divergence dates are
expected to predate fossil evidence, because fossil records
are fragmentary and incomplete (Parham etal. 2012). To
deduce the location of Cannabis during the temporal “miss-
ing link” between the molecular divergence date and the
oldest fossil pollen, we used two sets of indirect data.
Bosboom etal. (2011) mapped an “aridification zone”
arising in Central Asia at the Eocene–Oligocene bound-
ary, 34Ma. Orogenic changes at that time—the rise of the
Tibetan Plateau and the retreat of the Tarim Sea—forced

Vegetation History and Archaeobotany
1 3
concomitant changes in climate, resulting in the evolution
of steppe vegetation. China’s first steppe communities origi-
nated at the Eocene–Oligocene boundary (Sun etal. 2014),
and continued to develop through the Oligocene and Mio-
cene (Wang 1996), spanning the temporal missing link. We
transferred a best approximation of the palaeogeographic
map by Bosboom and colleagues onto a modern map of
Asia.
Secondly, we applied a previously reconstructed history
of Artemisia in Asia. Miao etal. (2011) tracked the spa-
tiotemporal appearance of Artemisia pollen, beginning at
the Eocene–Oligocene boundary, in a meta-analysis of 122
FPSs. As a proxy for the distribution of Cannabis prior to
our oldest pollen, we mapped their distribution of Artemi-
sia pollen, in six binned time slices used by Miao and col-
leagues: late Eocene (38–34Ma), latest Eocene (34Ma),
early Oligocene (34–28Ma), Oligocene (34–23Ma), early
Miocene (23–20Ma), and mid-Miocene (14Ma).
Results
The search strategy identified 173 FPSs that included C–H
pollen or archaeological studies with seeds, phytoliths, or
stem fragments. Seven studies did not meet our inclusion
criteria, and another 11 studies reported duplicate data. The
remaining 155 studies were tabulated, each with a citation
number, study location, details regarding application of the
algorithm, and duplicate reports (ESM Table1). Excluded
studies were also tabulated, with exclusion criteria (ESM
Table2).
Bin 1 (19.6Ma–11.6ka, Fig.1). The oldest C–H pollen
consistent with Cannabis dated to 19.6Ma (early Miocene).
The site is located in Ningxia, China, on the border between
the Tibetan Plateau and the Loess Plateau. For scholars
interested in the original publication, see ESM Table1,
citation #1. During the Pliocene, Cannabis pollen occurs
in Northwest China (Ningxia, 2.6Ma, ESM Table1, #2).
During the first half of the Pleistocene (i.e. the Gelasian
and Calabrian ages, 2.58Ma-781ka), Cannabis pollen is
located in Bashkorostan, Russia (1.5Ma #156), Hebei Prov-
ince (1.2Ma #3), and the Russian Altai (787ka, #4).
During the latter half of the Pleistocene, Cannabis pollen
appears in four of six Chinese regions: Northwest China
(Shaanxi, 342ka, #5; Gansu, 145ka, #6; Shaanxi, 50ka, #8;
Shaanxi, 25ka, #12; Tibet, 20ka, #13; Xinjiang, 14.5ka,
#20; Gansu, 12ka, #28). North China (Inner Mongolia,
35ka, #9). Northeast China (Liaoning, 16Ma, #16; Jilin,
13ka, #24). East China (Shanxi, 15ka, #19). FPSs in two
remaining provinces, South Central and Southwest China,
have Humulus pollen but no Cannabis pollen.
FPSs from latter half of the Pleistocene show sites with
Cannabis pollen elsewhere in East Asia (Korea, 32ka, ESM
Table1 #10); as well as Kazakhstan (130ka, #152), South-
west Asia (Turkey, 111ka, #7; Black Sea near Turkey, 17ka,
#152; Caspian Sea near Iran, 14ka, #23; Georgia, 13ka,
#154; Syria, 11.9ka, #30); South Asia (India, 32.6ka, #11;
Sri Lanka, 18ka, #14; India, 12.8–11.9ka, #25, #26, #27,
#29); and Central Siberia (Altai, 16–15ka, #15, #17, #18).
One FPS in Bin 1 (ESM Table1, #9) conflicted with our
algorithm’s designation of a retting-site (Cannabis pol-
len ≥ 15% of TLP). That study reported C–H pollen reaching
Fig. 1 Bin 1 (19.6Ma–11.6ka).
Age-weighted geographical
centroid for Cannabis data is
marked by a star. Background
base map by Natural Earth, free
open-source map data, https ://
www.natur alear thdat a.com

Vegetation History and Archaeobotany
1 3
61%, but it was clearly not a retting-site, because it occurred
34ka, long before agriculture began. The short-lived C–H
pollen spike was superseded by Artemisia reaching 80% of
TLP, another unusual finding.
Bin 2 (11.5–5ka, Fig.2). During the early Holocene,
C–H pollen consistent with wild-type Cannabis occurs
across Asia, from the Syrian and Anatolian steppes in the
west to the Liaoning plains in the east, and from the Altai
steppe in the north to the Central Highlands of India in the
south, even the Horton Plains of Sri Lanka. No C–H pollen
consistent with cultivated Cannabis is detected in Bin 2.
However, several archaeological sites with Cannabis arti-
facts fall into Bin 2. The oldest sites are in Japan (Chiba,
10,000 b p, ESM Table1 #142) and China (Henan, 7,850
bp, #132). Somewhat younger artifacts are found in Japan
(Fukui, 7,200 bp, #144; Aomori, 5,900–4,300 bp, #146), and
in China (Hunan, 6,400–5,300 bp, #104; Gansu, 5,000 bp,
#109; and Inner Mongolia 5,000 bp, #134).
Bin 3 (4.5–0ka, Fig.3) During the late Holocene, C–H
pollen consistent with wild-type Cannabis occurs across the
same range as Bin 2, excepting the loss of the Sri Lankan
site. Pollen consistent with cultivated Cannabis appears at 35
sites. At 11 of those sites, Humulus in a forested environment
gave way to land clearance and Cannabis cultivation, result-
ing in two symbols at the same sitewithin this time slice.
The oldest sites with pollen suggestingcultivated Canna-
bisare located in the lower Yangtze River basin, dating to
5,330 bp (#125) and 5,000 bp (a retting-site, with 15% TLP,
#130). In Northwest China, cultivated Cannabis is located in
Xinjiang (3,720 bp, #20; 2,600 bp, #129) and Qinghai (3,000
bp, #97). In East China, it appears in Shanxi (2,800 bp, #19).
Outside of China, pollen consistent with cultivated Cannabis
appears in Russia (3,700 bp, #155), Turkey (3,200 bp, #45;
2,300 bp, #153), Korea (3,150 bp, a retting-site, with 25% TLP,
#139), and India (2,500 bp, a retting-site, with 28% TLP, #71).
The oldest archaeological sites within this time slice are
located in Gansu (5,000–4,700 bp, #108, #110), Qinghai
(4,200–3,500 bp, #115, #116), Inner Mongolia (3,900–3,400
bp, #133), Shandong (3,600–3,000 bp, #124), and Xinjiang
(2,800–2,500 bp, #135, #136). Outside of China, relatively
old sites are found in central Korea (4,590–4,240 bp, #138),
the Ganges River basin in India (4,600–3,200 bp, #62–65),
Japan (3,500–3,000 bp, #147, #148), Nepal (2,400 bp, #76),
and the Russian Far East (2,500 bp, #79).
Extrapolating intotheOligocene
Next we deduced the location of Cannabis during the tem-
poral “missing link” between its divergence date (27.8Ma)
and oldest pollen (19.6Ma), using two sets of indirect data.
Bosboom etal. (2011) mapped an aridification zone aris-
ing at the Eocene–Oligocene boundary. The periphery of
their zone is demarcated by the central Tarim Basin, south-
ern Mongolia, and southeast of the Xining Basin. A best
approximation of this palaeogeographic zone was transferred
to a modern map in Fig.4. The oldest C–H pollen consist-
ent with Cannabis (site #1) is located at the south-eastern
perimeter of the aridification zone (Fig.4).
Artemisia pollen during the late Eocene, Oligocene,
and early/mid Miocene, based on data from Miao etal.
(2011) is also mapped in Fig.4. The age-weighted centroid
for these data is located within Bosboom’s aridification
Fig. 2 Bin 2 (11.5–5ka). Back-
ground base map by Natural
Earth, free open-source map
data, https ://www.natur alear
thdat a.com/

Vegetation History and Archaeobotany
1 3
zone. The Artemisia centroid, at 38.621°N, 102.205°E, is
located 60km north of the Cannabis centroid shown in
Fig.1 (38.186°N, 101.910°E). This amazing proximity
is somewhat spurious, because the data set of Artemisia
and Cannabis pollen somewhat reflects the distribution
of palynologists, and accessibility to related strata, rather
than a theoretically complete distribution of pollen.
Discussion
The use of ecological proxies is an inferential method of
differentiating Cannabis and Humulus pollen (McPartland
etal. 2018). It offers a way to dissect Cannabis/Humulus
and other collective names assigned by palynologists, due
Fig. 3 Bin 3 (4.5–0ka). Back-
ground base map by Natural
Earth, free open-source map
data, https ://www.natur alear
thdat a.com/
Fig. 4 Black ellipse: aridifica-
tion zone arising at the Eocene–
Oligocene boundary (Bosboom
etal. 2011) transferred to a
modern map. Age-weighted
geographical centroid for
Artemisia data is marked by a
star. Background base map by
Natural Earth, free open-source
map data, https ://www.natur
alear thdat a.com/

Vegetation History and Archaeobotany
1 3
to the morphological similarities between Cannabis and
Humulus pollen. However, the method is inherently proba-
bilistic. The method also assumes that climatic require-
ments of modern Cannabis, Humulus, PAC, and ASP + can
be extrapolated to past populations—the nearest living
relative method (Mosbrugger and Utescher 1997).
Cannabis pollen has an age-weighted geographical cen-
troid located in the north-eastern Tibetan Plateau (Fig.1).
However, the evolution of Cannabis predates these data. We
applied Miao’s Artemisia data as a proxy for the location
of Cannabis during this temporal “missing link”. The age-
weighted geographical centroid for Artemisia pollen is also
located in the north-eastern Tibetan Plateau (Fig.4). Both
centroids fall within Bosboom’s aridification zone. The old-
est Cannabis pollen lies at the zone’s periphery. Thus, we
deduce the centre of origin of Cannabis, along with that
of Artemisia, as the north-eastern Tibetan Plateau, in the
general vicinity of Qinghai Lake.
Cannabis expanded westward from its centre of origin.
C–H pollen consistent with Cannabis appeared in central
Russia (the cis-Ural region; ESM Table1, #156) by 1.5Ma.
Further west, a fossil seed (fruit) assigned to Cannabis in
Bulgaria dated to 7.3–5.3Ma (Palamarev 1982). Cannabis
expanded eastward from its centre of origin. By the end of
the Pleistocene, all regions of China except South Central
and Southwest China showed evidence of Cannabis.
FPSs in South Asia merit special attention: Humulus is
not native to India; H. lupulus was introduced as a culti-
gen by British colonists in the 1840s (Hooker 1890; Bakshi
and Atal 1985; Khuroo etal. 2007). The introduction of
non-native Humulus was also reported in Pakistan (Steward
1971), Nepal (Sood and Thakur 2015), and Burma (Kress
etal. 2003). Floras of Bangladesh and Thailand omit Humu-
lus (Khan and Halim 1990; Santisuk and Balslev 2015).
The absence of Humulus in South India provided a “beta
test” of our algorithm—all C–H pollen in South Asia should
appear in pollen assemblages with a NAP/AP ratio ≥ 2, con-
sistent with Cannabis. In fact, several sites showed NAP/AP
ratios approaching 1:1 (ESM Table1, #42, #57, #59, #66,
#67). The C–H pollen at these sites may be due to long-
distance transport, or represent small steppe communities
surrounded by forests. One South Asian study found C–H
pollen in an assemblage with a NAP/AP ratio ≤ 0.5 (#71).
The C–H percentage surged to 28%, in the presence of crop
pollen (Cerealia, Fagopyrum), which the algorithm identi-
fied as a hemp retting-site, as did the original authors.
South Asian studies that recorded “Cannabis” in pol-
len assemblages with high AP values may have misiden-
tified Celtis tree pollen. Notably, few South Asian FPSs
included Celtis (family Cannabaceae) in their pollen
diagrams. Three Celtis species are distributed in South
Asia, C. australis, C. tetrandra, and C. wightii (Watt
1889; Hooker 1890). Celtis pollen grains resemble those
of Cannabis—they are circular to elliptic, triporate with
circular pores surrounded by an annulus, a thin exine, and
smooth to verrucate surfaces. They have grain diameters
whose sizes fall into the range of Cannabis and Humulus
(see notes in ESM Table1, #60).
Whether Cannabis is native to South Asia, versus an
introduced species, is a long-standing debate (Watt 1889;
Hooker 1890). Experts still argue whether Himalayan Can-
nabis is indigenous (Zhou and Bartholomew 2003) or a natu-
ralized alien (Khuroo etal. 2007). We found pollen con-
sistent with Cannabis appearing in India by 32.6ka, which
suggests an indigenous species.
Cannabis in South Asia by 32.6ka begs the question of
when it actually arrived. Early floristic exchanges between
India and Asia were shaped by plate tectonics. As the Indian
plate migrated towards the Asian plate, it made a “glanc-
ing contact” with Sumatra 57Ma, followed by Burma, and
then a “hard collision” with Tibet 35Ma (Ali and Aitchison
2008). The glancing contact between continents resulted in
floristic exchanges during the Eocene (Bande 1992; Morley
and Dick 2003). The extant flora of India is often termed
Indo-Malayan. Cannabis arriving via Southeast Asia, during
the Eocene, seems unlikely. Southeast Asian FPSs are bereft
of C–H pollen, and the Indo-Malayan exchange occurred
before Cannabis evolved.
The migrating Indian plate initiated uplift of the Tibet
plateau ca. 40–35Ma, which was the primary cause of Bos-
boom’s aridification zone. The Tibet uplift, followed by the
Pamir uplift (35Ma) created a dispersal barrier between
Central Asia and India. Perhaps the biogeographical dis-
persal patterns of related plants, which we discuss below,
might reveal how Cannabis hurdled the dispersal barrier
into South Asia.
In contrast to the north-eastern Tibetan Plateau as the
centre of origin of Cannabis, Zhang etal. (2018b) offered
“low latitude” China as the centre of origin. Their estimate
was based on genetic variation in cpDNA sequences of 52
extant Cannabis accessions. Spatial analysis of molecular
variance (SAMOVA) determined the optimal number of
haplogroups to divide the accessions; the data best fit K = 3.
Phylogenetic tree topology placed the “L” (low latitude) hap-
logroup basal to the other haplogroups, whose accessions
came from Tibet (n = 7) and Yunnan (6), as well as Inner
Mongolia (3), Gansu (1), Guangxi (1), and Shandong (1).
Tibet and Yunnan embrace the south-eastern fringe of the
Tibetan Plateau, so our estimation of the Cannabis centre of
origin differs little from that of Zhang etal. (2018b).
The south-eastern fringe of the Tibetan Plateau, how-
ever, was the last sector uplifted, ca. 8Ma (Xing and Ree
2017). Prior to that, low altitude China was warmer and
wetter than today, and covered by subtropical broad-leaved
forests (Sun etal. 2011; Huang etal. 2016). We suggest
this ecosystem would not have driven the evolution of a

Vegetation History and Archaeobotany
1 3
steppe plant like Cannabis, according to the nearest living
relative method (Mosbrugger and Utescher 1997).
The south-eastern fringe of the Tibetan Plateau is
called the Henduan Mountains. Clarke and Merlin (2013)
proposed a Pleistocene glacial refugium in the Henduan
Mountains. Congruent with this, Zhang and colleagues
estimated a Cannabis crown date (divergence of the three
haplogroups) of 2.24Ma—the early Pleistocene. Given
this scenario, the “L” haplogroup should have expressed
the greatest π (nucleotide diversity). Instead, greatest π
was expressed by the “M” haplogroup, whose accessions
primarily came from Xinjiang (n = 5) and Tibet/Qing-
hai (n = 4). Disparity between stem age and crown age is
exampled by Ginkgo biloba, with uncertain range dynam-
ics arising from cpDNA sequences of extant relictual
populations (Hohmann etal. 2018).
Turning to cultivated Cannabis, the oldest pollen signal
identified by our methods dated to 5,330 bp (ESM Table1,
#125). At some sites dominated by AP pollen, we may
have misidentified pollen signals of Cannabis cultivation
as Humulus pollen. Pollen-based detection of early agri-
culture in densely forested sites often shares this conun-
drum (e.g. Tarasov etal. 2018).
Archaeological findings (seeds) predate 5,330 bp. Per-
haps the seeds were collected from wild-type Cannabis. A
Jōmon Culture site yielded the oldest seeds (ESM Table1
#142). The seeds were found with other edible nuts and
fruits, indicating food use. The Jōmon people made pot-
tery but did not farm—their economy was based on wild
resources and nut harvesting (Bleed and Matsui 2010).
Contrary to this wild-type hypothesis, however, photomi-
crographs of the Jōmon seeds do not show wild-type traits,
suggesting they were not only cultivated, but domesticated.
Serviceable fibre is not easily extracted from wild-
type growth, which branches excessively. Densely sown
crops, with minimal branching, yield the best fibre. The
clearest palynological signal of fibre use comes from
FPSs with Cannabis pollen ≥ 15% of TLP, indicative of a
retting-site. The oldest retting-site, dating to 5,000 bp, is
located in the Yangtze River delta (ESM Table1, #130).
The pollen identification is problematic, yet the authors
discuss hemp-retting as the source of their pollen surge,
as do other authors (TableS1, #71). Older non-palyno-
logical evidence—carbonized hemp fibres (excluded in
this study)—dates to 5,600 bp, in Henan (Zhang and Gao
1999). Intact rope and cloth dates to 5,000 bp, in the Yang-
tze River delta (Zhou 1980).
Archaeological evidence of ceremonial or drug use
dates to 2,700 bp, in Xinjiang (ESM Table1, #135, #136).
This interpretation is secured by the presence of processed
leaves and female flowering tops, stored in a leather bas-
ket, wooden bowl, or earthenware pot.
Cannabaceae biogeography
The geographic ranges of related species may help locate an
organism’s centre of origin (Crisci etal. 2003). The Humu-
lus centre of diversity lies in southwest China: H. lupulus
occurs in Sichuan; H. japonicus occurs in Sichuan, Yunnan,
Chongqing, Guangxi, and Guizhou; H. yunnanensis is lim-
ited to Yunnan (Zhou and Bartholomew 2003). Small (1978)
divided H. lupulus into five varieties, which collectively cir-
cumnutate the Northern Hemisphere at temperate latitudes.
Gray (1859) noted the Humulus connection between East
Asia and North America, foretelling the Bering land bridge
theory.
Pteroceltis tatarinowii, the only extant species of that
genus, is a tall canopy tree limited to China and Mongolia.
A phylogeographical reconstruction of 28 extant populations
identified southern China as the species’s centre of origin (Li
etal. 2012). However, fossils indicate that Pteroceltis is a
relictual lineage that once had a wider geographic range, and
similar to that of H. lupulus. Fossils of Pteroceltis tertiaria
were found in Germany, and fossils of Pteroceltis knowltonii
were found in the USA (Manchester etal. 2009).
Celtis is the largest genus in the Cannabaceae, with about
70 species, and 13 grow in China. Celtis also has the larg-
est native range in the Cannabaceae, spanning temperate
as well as tropical latitudes. In the Northern Hemisphere,
Gray (1859) included Celtis in his proto-Bering land bridge
theory. In the Southern Hemisphere, Celtis occurs in South
America, Africa, and Australia (Stevens 2008). The oldest
Celtis fossils (of C. aspera, 64–56Ma) have been found in
the Russian Far East and the western USA (Manchester etal.
2002). This hints at an East Asian origin for the genus.
Trema and Parasponia are sometimes synonymized; a
molecular study found both Trema and Parasponia para-
phyletic, with no clear basal lineages (Yesson etal. 2004).
Twelve Trema species grow in subtropical and tropical
regions, in southern China, India, Southeast Asia, Africa,
and Central America. China is the Trema centre of diversity,
with eight species growing there. Parasponia species do not
occupy continental Asia; they are found in Indonesia, Papua
New Guinea, Philippines, northern Australia, and some Mel-
anesian and Polynesian islands.
Aphananthe species occur in China, Japan, Korea, India,
Sri Lanka, Southeast Asia (including Philippines and Papua
New Guinea), Mexico, Madagascar, and Australia. Fossils
indicate a wider distribution, with finds in western Siberia,
Germany, and the USA (Yang etal. 2017). Phylogenetic
studies place Aphananthe basal to the rest of the Canna-
baceae, hence it is the oldest genus (Yang etal. 2013; Zhang
etal. 2018a). A fossil from Germany, Aphananthe cretacea,
dates to 66–72.1Ma (Knobloch and Mai 1986).
Yang etal. (2017) reconstructed the biogeography of
Aphananthe using DNA sequences and other molecular

Vegetation History and Archaeobotany
1 3
methods. They identified East Asia as the ancestral area
of extant Aphananthe species. Dispersal took Aphananthe
across the Bering land bridge into North America around
19.1Ma, and into South Asia by 18.1Ma. A route from East
Asia to South Asia was not delineated. A. cuspidata occupies
contiguous areas spanning East Asia and South Asia—Yun-
nan, Burma, and India—so that is the likely route.
In summary, most Cannabaceae genera have origins in
East Asia. The ancestral distribution of Cannabis resembles
that of H. lupulus, except Cannabis dispersed to South Asia,
and H. lupulus crossed the Bering land bridge. We propose
that the progenitor of Cannabis dispersed from East Asia to
the north-eastern Tibetan Plateau, where it underwent para-
patric speciation in Bosboom’s aridification zone, amidst
Asia’s first steppe community.
Cannabis dispersed from the Tibetan Plateau, first to
the west (Russia and Europe) and then to the east (China).
By the end of the Pleistocene, Cannabis spread through-
out Asia, except for Southeast Asia. Thus, wild-type Can-
nabis was available for people across Eurasia to bring into
cultivation and domesticate. Several sites in south-eastern
Europe, mostly associated with the Yamnaya Culture, sug-
gest autochthonous domestication in Europe (Clarke and
Merlin 2013; Long etal. 2017; McPartland and Hegman
2018). Vavilov (1926) would have agreed, “it is probable
that the cultivation of hemp arose simultaneously and inde-
pendently in several places”.
References
Ali JR, Aitchison JC (2008) Gondwana to Asia: plate tectonics, paleo-
geography and the biological connectivity of the Indian sub-conti-
nent from the Middle Jurassic through latest Eocene (166-35Ma).
Earth-Sci Rev 88:145–166
Andrek HY, Balog L, Sheer MV (2010) Humulus japonicus Siebold et
Zucc. (Cannabaceae)—нoвий aдвeнтивний вид флopи Укpaїни.
Укp бoтaн жypн 67:438–445
Bakshi SK, Atal CK (1985) Hops in India. Council of Scientific and
Industrial Research, Jammu-Tawi
Balogh L, Dancza I (2008) Humulus japonicus, an emerging invader
in Hungary. In: Tokarska-Guzik B etal (eds) Plant invasions:
human perception, ecological impacts and management. Backhuys
Publishers, Leiden, pp 73–79
Bande MB (1992) The Palaeogene vegetation of peninsular India
(megafossil evidence). Palaeobotanist 40:275–284
Bleed P, Matsui A (2010) Why didn’t agriculture develop in Japan?
A consideration of Jomon ecological style, niche construction,
and the origins of domestication. J Archaeol Method Theory
17:356–370
Bosboom RE, Dupont-Nivet G, Houben AJP etal (2011) Late Eocene
sea retreat from the Tarim Basin and concomitant Asian paleoen-
vironmental change. Palaeogeogr Palaeoclimatol Palaeoecol
299:385–398
Bottema S, Kopaka K, Alexopoulos A (2003) The Late-Holocene vege-
tation history of Gavdos (Crete) in relation to long distance pollen
dispersal: the Trypiti pollen diagram. In: Tonkov S (ed) Aspects
of palynology and palaeoecology. Pensoft, Moscow, pp 199–212
Boutain JR (2014) On the origin of hops: genetic variability, phy-
logenetic relationships, and ecological plasticity of Humulus
(Cannabaceae). Doctoral dissertation, University of Hawai’i,
Honolulu, HI
Clarke RC, Merlin MD (2013) Cannabis: evolution and ethnobotany.
University of California Press, Berkeley
Crisci JV, Katinas L, Posadas P (2003) Historical biogeography. Har-
vard University Press, Cambridge
De Candolle AP (1883) Origine des Plantes Cultivées. Baillière,
Paris
Dörfler W (1990) Die Geschichte des Hanfanbaus in Mitteleuropa auf-
grund palynologischer Untersuchungen und von Großrestnach-
weisen. Prähist Z 65:218–244
Dorofeev PI (1969) Mиoцeнoвaя флopa Maмoнтoвoй гopы нa Aлдaнe
(Miocene Flora of the Mammoth Mountain on the Aldan). Izd-vo
Akademia nauk SSSR, Leningrad
Dorofeev PI (1982) Cannabaceae. In: Takhtajan AL (ed) Иcкoпaeмыe
цвeткoвыe pacтeния Poccии и coпpeдeльныx гocyдapcтв, T. 2
(Fossil Flowering Plants of Russia and Neighboring States, Vol
2). Izd-vo Nauka, Leningrad, pp 43–48
Eom BC, Kim JW (2017) Phytocoenosen and distribution of a wild
tea (Camellia sinensis (L.) Kuntze) population in South Korea.
Korean J Plant Res 30:176–190
Friedrich PA (1883a) “Cannabis oligocaenica nov. spec.”, Beiträge zur
Kenntnis der Tertiärflora der Provinz Sachsen. Schropp, Berlin,
pp 165–166
Friedrich PA (1883b) Atlas zu den Abhandlungen zur geologischen
Specialkarte von Preussen den Thüringischen Staaten, Band IV,
Heft 3. Schropp, Berlin
Fries M (1958) Vegetationsutveckling och odlingshistoria i Varnhem-
strakten: en pollenanalytisk undersökning i Västergötland. Acta
Phytogeogr Suec 39:1–63
Fröman I (1939) Die Hölzer des Rades und der Hopfenfund. In: von
Post L, Oldeberg A, Fröman I (eds) Ein eisenzeitliches Rad aus
dem Filaren-See in Södermanland, Schweden. Wahlström & Wid-
strand, Stockholm, pp 89–98
Gray A (1859) Diagnostic characters of new species of phanerogamous
plants collected in Japan by Charles Wright, Botanist of the U.S.
North Pacific Exploring Expedition. With observation upon the
relations of the Japanese flora of that of North America. Mem Am
Acad Arts Sci 6:377–453
Hämeen-Anttila J (2006) The last pagans of Iraq: Ibn Wahshīyah and
his Nabatean Agriculture. Brill, Leiden
Hillig KW, Mahlberg PG (2004) A chemotaxonomic analysis of
cannabinoid variation in Cannabis (Cannabaceae). Am J Bot
91:966–975
Hohmann N, Wolf EM, Rigault P etal (2018) Ginkgo biloba’s footprint
of dynamic Pleistocene history dates back only 390,000 years.
BMC Genomics 19:2999
Hooker JD (1890) The Flora of British India, Vol 5: Chenopodiaceae
to Orchideae. L. Reeve & Co., London
Hu Z, Wu QA (1992) Studies of the rare and endangered plant species
in the Yunnan region of China. In: Adams RP, Adams JE (eds)
Conservation of plant genes: DNA banking and invitro biotech-
nology. Academic Press, New York, pp 267–272
Huang YJ, Jia LB, Wang Q, Mosbrugger V etal (2016) Cenozoic plant
diversity of Yunnan: a review. Plant Divers 38:271–282
Jarolímek I, Kolbek J (2006) Plant communities dominated by Salix
gracilis in Korean peninsula and Japan. Biol Bratisl 61:63–70
Jeong HR, Kim HJ, Choi K etal (2012) Vegetation structure and distri-
bution of forested wetland at public and private forests in Daegu
City. J Agric Life Sci 46:69–84
Jiang HE, Wang L, Merlin MD etal (2016) Ancient Cannabis burial
shroud in a Central Eurasian cemetery. Econ Bot 70:213–221
Jung YK, Kim JW (1998) Syntaxonomy of mantle communities in
South Korea. Korean J Ecol 21:739–750

Vegetation History and Archaeobotany
1 3
Khan MS, Halim M (1990) Flora of Bangladesh, No. 14: Cannabi-
daceae. Bangladesh Agricultural Research Council, Dacca
Khuroo AA, Rashid I, Reshi Z etal (2007) The alien flora of Kashmir
Himalaya. Biol Invas 9:269–292
Kim SS, Kim YS, Ha SG, Shin HT (2010) Dispersion of vascular plant
in Daepyeong swamp, Korea. J Korean Nat 3:187–198
Knobloch AH, Mai DH (1986) Monographie der Früchte und Samen in
der Kreide von Mitteleuropa. Rozpr ustred ustavu Geol 47:1–219
Kobayashi M, Momohara A, Okitsu S etal (2008) Fossil hemp fruits
in the earliest Jomon period from the Okinoshima site, Chiba Pre-
fecture. Shokuseishi kenkyū 16:11–18
Kolbek J, Karolímek I (2008) Man-influenced vegetation of North
Korea. Linzer Biol Beitr 40:381–404
Kolbek J, Sádlo J (1996) Some short-lived ruderal plant communi-
ties of non-trampled habitats in North Korea. Folia Geobot
31:207–217
Kress WJ, DeFilipps RA, Farr E etal (2003) A checklist of the trees,
shrubs, herbs, and climbers of Myanmar. Smithsonian Institution,
Washington, DC
Kuhn D (1988) Textile Technology: Spinning and reeling. In: Needham
J, Wang L (eds) Science and civilisation in China, vol 5. Part 9.
Cambridge University Press, Cambridge, pp 1–520
Lee SJ, Ahn YH (2014) Study of vegetation structure about shrine
forest in Jirisan National Park with regard to global warming. J
Environ Sci Int 23:1,863–1,879
Lee CY, Liew PM (2010) Late quaternary vegetation and climate
changes inferred from a pollen record of Dongyuan Lake in south-
ern Taiwan. Palaeogeogr Palaeoclimatol Palaeoecol 287:58–66
Lee HJ, Kim JH, Chun YM, Choung HL (1976) Synecology of the for-
est vegetation of Yeongjongo. Korean J Ecol 26:223–236
Lee KS, Cho MG, Moon HS, Jeon KS (2013) The list of vascular
plants at Junam wetland in Changwon City. Korean J Agric For
Meteorol 15:67–75
Li HL (1974) An archaeological and historical account of cannabis in
China. Econ Bot 28:437–448
Li XH, Shao JW, Lu C etal (2012) Chloroplast phylogeography of a
temperate tree Pteroceltis tatarinowii (Ulmaceae) in China. J Syst
Evol 50:325–333
Long T, Wagner M, Demske D, Leipe C, Tarasov PE (2017) Cannabis
in Eurasia: origin of human use and Bronze Age trans-continental
connections. Veget Hist Archaeobot 26:245–258
Lynch RC, Vergara D, Tittes S etal (2016) Genomic and chemical
diversity in Cannabis. Crit Rev Plant Sci 35:349–363
Manchester SR, Akhmetiev MA, Kodrul TM (2002) Leaves and fruits
of Celtis aspera (Newberry) comb. nov. (Celtidaceae) from the
Paleocene of North America and Eastern Asia. Int J Plant Sci
163:725–736
Manchester SR, Chen ZD, Lu AM etal (2009) Eastern Asian endemic
seed plant genera and their paleogeographic history throughout
the northern hemisphere. J Syst Evol 47:1–42
Maximovich CJ (1859) Primitiae florae Amurensis. Versuch einer Flora
des Amur-Landes, Kaiserliche Akademie der Wissenschaften
McPartland JM (2018) Cannabis systematics at the levels of family,
genus, and species. Cannabis Cannabinoid Res 3:203–212
McPartland JM, Hegman W (2018) Cannabis utilization and diffusion
patterns in prehistoric Europe: a critical analysis of archaeological
evidence. Veget Hist Archaeobot 27:627–634
McPartland JM, Guy GW, Hegman W (2018) Cannabis is indigenous
to Europe and cultivation began during the Copper or Bronze
age: a probabilistic synthesis of fossil pollen studies. Veget Hist
Archaeobot 27:635–648
Mercuri AM, Accorsi CA, Mazzanti MB (2002) The long history
of Cannabis and its cultivation by the Romans in central Italy,
shown by pollen records from Lago Albano and Lago di Nemi.
Veget Hist Archaeobot 11:263–276
Miao YF, Meng QQ, Fang XM etal (2011) Origin and develop-
ment of Artemisia (Asteraceae) in Asia and its implications for
the uplift history of the Tibetan Plateau: a review. Quat Int
236:3–12
Morley RJ, Dick CW (2003) Missing fossils, molecular clocks, and
the origin of the Melastomataceae. Am J Bot 90:1,638–1,644
Mosbrugger V, Utescher T (1997) The co-existence approach—a
method for quantitative reconstructions of Tertiary terrestrial
palaeoclimate data using plant fossils. Palaeogeogr Palaeocli-
matol Palaeoecol 134:61–86
Ni J, Yu G, Harrison SP, Prentice IC (2010) Palaeovegetation in
China during the late quaternary: Biome reconstructions based
on a global scheme of plant functional types. Palaeogeogr Pal-
aeoclimatol Palaeoecol 289:44–61
Oh YJ, Yoo JH, Moon BC etal (2008) Habitat characteristic and
community structures of Humulus japonicus in Korea’s middle
region. Korean J Environ Agric 27:72–79
Oh HK, Beon MS, Kim YH (2010) Classification by plants commu-
nities of the Wi-do (Island), Buan—focused on Jilli evergeen
forest, Chido wetland, and Seokgeum. J Korean Nat 3:159–169
Palamarev E (1982) Heoгeнcкaтa кapпoфлopa нa Meлнишкия
бaceйн. Paleontol Stratigr Lithol 16:3–43
Parham JF, Donoghue PCJ, Bell CJ etal (2012) Best practices for
justifying fossil calibrations. Syst Biol 61:346–359
Pételot PA (1954) Les plantes médicinales du Cambodge, du Laos
et du Viêtnam. Centre de recherches scientifiques et techniques,
Saigon
Quamar MF, Bera SK (2017) Pollen records related to vegetation and
climate change from northern Chhattisgarh, central India during
the late Quaternary. Palynology 41:17–30
Russo EB, Jiang HE, Li X etal (2008) Phytochemical and genetic
analyses of ancient cannabis from Central Asia. J Exper Bot
59:4,171–4,182
Santisuk T, Balslev H (2015) Flora of Thailand, Vol 13, Part 1: Acha-
riaceae, Adoxaceae, Cannabaceae, Caprifoliaceae, Ericaceae,
Salicaceae & Ulmaceae. Forest Herbarium, Depart of National
Parks, Wildlife and Plant Conservation, Bangkok
Sawler J, Stout JM, Gardner KM etal (2015) The genetic structure of
marijuana and hemp. PLoS ONE 10:e0133292
Small E (1978) A numerical and nomenclatural analysis of morpho-
geographic taxa of Humulus. Syst Bot 3:37–76
Small E, Cronquist A (1976) A practical and natural taxonomy for
Cannabis. Taxon 25:405–435
Song JS, Song SD (1996) A phytosociological study on the riverside
vegetation around Hanchon an upper stream of Nak-tong River.
Korean J Ecol 19:431–451
Song YH, Cohen DJ, Shi JM etal (2017) Environmental reconstruction
and dating of Shizitan 29, Shanxi Province: an early microblade
site in north China. J Archaeol Sci 79:19–35
Sood SK, Thakur R (2015) Herbal resources of India and Nepal. Sci-
entific Publishers, Jadhpur
Stevens PF (2008) Angiosperm phylogeny website, Version 9.
Accesssed at http://www.mobot .org/MOBOT /resea rch/APweb /
Steward RR (1971) Flora of West Pakistan. Fakhri, Karachi
Sun BN, Wu JY, Liu YS etal (2011) Reconstructing Neogene vegeta-
tion and climates to infer tectonic uplift in western Yunnan, China.
Palaeogeogr Palaeoclimatol Palaeoecol 304:328–336
Sun JM, Ni XJ, Bi SD etal (2014) Synchronous turnover of flora,
fauna, and climate at the Eocene-Oligocene boundary in Asia.
Sci Rep 4:7,463
Tarasov PE, Savelieva LA, Long T etal (2018) Postglacial vegetation
and climate history and traces of early human impact and agricul-
ture in the present-day cool mixed forest zone of European Russia.
Quat Int. https ://doi.org/10.1016/j.quain t.2018.02.029

Vegetation History and Archaeobotany
1 3
Vavilov NI (1926) The origin of the cultivation of “primary” crops,
in particular cultivated hemp. Tpyды пo пpиклaднoй бoтaникe,
гeнeтикe и ceлeкции 16:221–233
Wang WM (1996) On the origin and development of steppe vegetation
in China. Palaeobotanist 45:447–456
Watt G (1889) A dictionary of the economic products of India, vol 2.
Calcutta Office of the Superintendent of Government Printing,
Allen
Wilson DG (1975) Plant remains from the Graveney boat and the early
history of Humulus lupulus L. in W, Europe. N Phytol 75:627–648
Wu JY, Liu J, Provan J etal (2018) Testing Darwin’s transoceanic dis-
persal hypothesis for the inland nettle family (Urticaceae). Ecol
Lett 21(1515):1529
Xing YW, Ree RH (2017) Uplift-driven diversification in the Heng-
duan Mountains, a temperate biodiversity hotspot. PNAS
114:E3,444–E3,451
Yang MQ, van Velzen R, Bakker FT etal (2013) Molecular phylogenet-
ics and character evolution of Cannabaceae. Taxon 62:458–472
Yang MQ, Li DZ, Wen J etal (2017) Phylogeny and biogeography of
the amphi-Pacific genus Aphananthe. PLoS ONE 12:e0171405
Yesson C, Russell SJ, Parrish T etal (2004) Phylogenetic framework
for Trema (Celtidaceae). Plant Syst Evol 248:85–109
Zecchetto S, De Blasio F (2007) Sea surface winds over the Mediter-
ranean Basin of satellite data (2000-04): Meso- and Local-scale
features on annual and seasonal time scales. J Appl Meteorol Cli-
matol 46:814–827
Zhang SL, Gao HY (1999) 荥阳青台遗址出土的丝麻品观察与研究
(Observation and study of silk and hemp recovered from Qingtai
archaeological site, Xingyang). Zhōngyuán Wénwù 3:10–16
Zhang HL, Jin JJ, Moore MJ etal (2018a) Plastome characteristics of
Cannabaceae. Plant Divers 40:127–137
Zhang QY, Chen X, Gou HY etal (2018b) Latitudinal adaptation and
genetic insights into the origins of Cannabis sativa L. Front Plant
Sci 9:1876
Zhao HB, Chen FD, Chen SM etal (2010) Molecular phylogeny of
Chrysanthemum, Ajania and its allies (Anthemideae, Asteraceae)
as inferred from nuclear ribosomal ITS and chloroplast trnL-F
IGS sequences. Plant Syst Evol 284:153–169
Zhou YM (1980) 钱山漾残绢片出土的启示 (Revelations of the exca-
vation of the silk tabby remnant from Qianshanyang). Wénwù
1980(1):74–77
Zhou Z, Bartholomew B (2003) Cannabaceae. In: Wu ZY, Raven PH,
Hong DY (eds) Flora of China, vol 5. Science Press, Beijing, pp
74–75
Zhou B, Shen CD, Sun WD etal (2007) Elemental carbon record of
paleofire history on the Chinese Loess Plateau during the last
420ka and its response to environmental and climate changes.
Palaeogeogr Palaeoclimatol Palaeoecol 252:617–625
Publisher’s Note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.