Celtis sinensis Pers . ( Ulmaceae ) naturalised in northern South Africa and keys to distinguish between Celtis species commonly cultivated in urban environments

Invasive organisms are considered to be one of the most important threats to biodiversity worldwide (Vitousek et al. 1997) and costs South Africa millions of rands annually to control (Le Maitre et al. 2002). Southern Africa’s current concerns with invasive plants were already predicted in the 1960s and the 1970s (Stirton 1978; Wild 1961), fully recognised and documented in the 1980s (Henderson et al. 1987; Wells et al. 1986) and flagged, and their invasiveness categorised, since the 1990s (Henderson 2001; Henderson & Cilliers 2002).


Introduction
Invasive organisms are considered to be one of the most important threats to biodiversity worldwide (Vitousek et al. 1997) and costs South Africa millions of rands annually to control (Le Maitre et al. 2002). Southern Africa's current concerns with invasive plants were already predicted in the 1960s and the 1970s (Stirton 1978;Wild 1961), fully recognised and documented in the 1980s (Henderson et al. 1987;Wells et al. 1986) and flagged, and their invasiveness categorised, since the 1990s (Henderson 2001;Henderson & Cilliers 2002).
Urban areas are seed pools for potentially invasive alien plants, as these species are cultivated as garden subjects in mass and propagules escape from there to become naturalised in natural areas, but especially along disturbed water courses and roadsides (Horvitz et al. 2014;Kalwij et al. 2008). Naturalised species are capable of reproducing and spreading without direct assistance by people, and the most aggressive species become invaders capable of penetrating and replacing indigenous vegetation and threatening indigenous biodiversity (Ndhlovu, Milton & Esler 2016). Hence, the number of species becoming invasive increases annually (Seebens et al. 2017) and threatens agricultural productivity, water resources and natural ecosystems, with temperate agricultural or urban sites among the areas most invaded by alien plants (Lonsdale 1999).
Phytosociological studies of urban areas conducted in the 1990s hardly reported any individuals of C. sinensis (Cilliers & Bredenkamp 1999), and subsequently, it was highlighted as an emerging invader by Mgidi et al. (2007). In recent years, C. sinensis (which originates from China, Japan and Korea) has become very conspicuous in urban open spaces, roadsides and riparian areas on the Highveld. Celtis sinensis also poses a major threat to the indigenous C. africana Burm.f., as it and the closely related C. australis hybridise readily with the former in Gauteng (Glen & Van Wyk 2016). In a study by Whittemore and Townsend (2007) on the hybridisation potential of Celtis species, C. sinensis was one of the species that tested positive for this ability. Buck and Bidlack (1998) found that data sets of morphological and isozymic data demonstrate congruity among Celtis species, which is ascribable to hybridisation.
Three of the eight African Celtis species are indigenous to southern Africa, namely the widespread C. africana (white stinkwood), C. gomphophylla Baker (false white stinkwood) in the Eastern Cape and KwaZulu-Natal, and C. mildbraedii Engl. (Natal white stinkwood) in KwaZulu-Natal and Swaziland (Pooley 1997 Celtis sinensis is not a declared alien invasive tree in South Africa, and this article compares its population data with that of an indigenous (C. africana) and declared alien tree species (Melia azedarach L.) from similar urban and agricultural environments in the Tlokwe Municipal area. A key is provided to distinguish between the indigenous and cultivated Celtis species of northern South Africa, and a simple key for quick identification of juvenile Celtis individuals sold at nurseries. A taxonomic treatment of C. sinensis is provided for the species to be officially included in the southern African flora as naturalised.

Distribution and population data
Distribution records were obtained from the National Museum in Bloemfontein (NMB), National Herbarium in Pretoria (PRE), H.G.W.J. Schweickerdt Herbarium (PRU), A.P. Goossens Herbarium (PUC) and the South African Plant Invaders Atlas database (Henderson 1998).
Cultivation records of trees were extracted from a garden database comprising frequency data for 256 gardens surveyed in Ganyesa, Potchefstroom and Roodepoort (Davoren et al. 2016). Three land-cover types (riparian, rangelands and urban open space) were identified within a 60 km radius of Potchefstroom, and 16 belt transects per land-cover were sampled. Transects were 50 m × 5 m, and every woody species within each transect was identified up to species level, counted and the height measured with a clinometer. Land-cover type data were supplemented with frequency data from 16 belt transects from domestic gardens (Davoren et al. 2016) and included in the study for comparative purposes.
Height class distributions were compiled for each species. Sturge's rule was applied (Scott 2009) to determine the sizeclass distribution and frequency intervals for the population graphs. Stability between successive size classes was determined with quotient analysis (Botha, Witkowski & Shackleton 2004). The likelihood that any two trees within a population originate from the same size-class was assessed by Simpson's Index of Dominance (SDI) (Venter & Witkowski 2010). Permutation Index (PI) was calculated for each species to calculate the degree of deviation from a monotonic decline or from an inverse-J curve (Wiegand et al. 2000).

Juvenile morphology
Seeds were collected from correctly identified C. africana, C. australis, C. occidentalis and C. sinensis in the field and germinated at the North-West University Botanical Garden. Plants were maintained under identical nursery conditions for 3 years in terms of watering, fertiliser additions and percentage shade. The length and width of the petiole, leaf and fruit of six plants per species were measured (plants < 1.5 m in height).

Cultivation and naturalisation patterns
Celtis sinensis occurs as a garden subject throughout South Africa and is one of the ten most widely cultivated species in our sample of Highveld gardens (third overall, Table 1). It is also very popular as a street tree. Celtis sinensis is the only alien tree in the top five that is not a fruit tree, suggesting that its popularity lies elsewhere, probably its hardiness, fast growth, dense shade, resistance to pollution and being deciduous (letting the sun through in the winter) (Glen & Van Wyk 2016). These are also the characteristics that make Vachellia karroo a popular indigenous garden subject (Joffe 2001; Table 1). It is important to note that the indigenous C. africana is less popular as a garden subject compared to C. sinensis (Table 1).
Celtis sinensis is more popular as a garden subject than all but one declared alien invasive species, namely Morus alba (Table 1).
The survey of land-cover types (urban open space, riparian areas and rangelands) revealed that M. alba, despite its wide cultivation, was not the most common alien of non-cultivated areas, but instead C. sinensis, Ligustrum lucidum and Melia azedarach were recorded more frequently (Table 2). Celtis sinensis and M. azedarach were the only alien trees in both the top ten lists for most frequently cultivated (Table 1) and naturally occurring (Table 2) species. Melia azedarach is a declared invasive alien and is no longer available in the nursery trade because of its weedy nature and public awareness (Henderson 2007).
Celtis africana, one of the only three indigenous tree species that naturally occur more frequently than M. azedarach and one of the five tree species found more frequently than C. sinensis, is commonly found across all the sampled land-cover types (Table 2). Celtis africana is not common in gardens because of C. sinensis being most available from nurseries. Celtis africana is very common in natural areas, with > 60% of the surveyed population recorded from rangeland, hills and ridges ( Figure 1). Melia azedarach is also not commonly cultivated and is most commonly found in urban open space (> 40% of the population). More than 60% of the recorded C. sinensis individuals were from gardens, with the rest found in urban open space, riparian areas and rangelands in descending order ( Figure 1). Celtis sinensis is definitely spreading from urban areas outward, typically following the predicted pathway along watercourses and roads (Milton & Dean 1998). This adaptability can be ascribed to its ability to withstand heavy winter frosts, fleshy fruits that are prone to bird dispersal and lack of competition in a predominantly treeless Highveld landscape.

Population structure
Limited variation in the quotient (0.2-0.8) is indicative of a stable population for C. africana ( Figure 2a) (Botha et al. 2004;Shackleton 1993). The fluctuating quotient for C. sinensis (0.1-1.2) is interpreted as instability within the height classes of the specific population probably because of erratic colonisation and survival events ( Figure 2b). However, the comparatively little variation (0-0.8) detected for M. azedarach indicates that the population structure is stable because of the longer time it has had to establish since naturalisation ( Figure 2c).
None of the populations had SDI values below 0.1, revealing that size frequency is steeper than would be expected from an exponentially declining population (Botha et al. 2004). In this case, the two Celtis species have the same SDI (0.66) and lower than M. azedarach (0.86). In all three cases, this is an indication that individuals are not evenly distributed among the size classes, but this needs to be considered with PI values that are based on the assumption that an ideal undisturbed population should represent a monotonic decline (PI = 0). This is typical for C. africana, but not for C. sinensis (PI = 4) and M. azedarach (PI = 2), which have discontinuous populations, where larger individuals show a higher frequency than a previous size-class (Botha et al. 2004;Venter & Witkowski 2010).
Overall, the population structure of C. sinensis is most similar to that of the invader (M. azedarach) and not the related, indigenous species (C. africana). The C. sinensis population   has more individuals than C. africana in the taller height classes, which is indicative of successful establishment, and the high numbers of saplings show the ability to reproduce (Table 3). This is an indication that C. sinensis is successfully increasing by reproducing and establishing in the environment.

Land-cover
In the urban open spaces, one can argue that the seeds are originating from cultivated individuals or garden refuge; however, the fact that it also occurs in rangelands and riparian areas in agricultural areas indicates that it can reproduce independently and that it has acquired a successful dispersal vector.
Etymology: Celtis is a classical Latin name for the unrelated African species, Ziziphus lotus L., and sinensis indicates that the species originates from China (Clarke & Charters 2016;Glen 2004).
Common names used for the tree in South Africa include Chinese nettle tree, Japanese hackberry, Chinese netelboom (Afr), po shu (Chinese), modutu (Setswana), lesika, mohatlakgomo, mogatakgomo (Sesotho). The Tswana name, modutu, means boredom and loneliness, suggesting that the dense shade of this tree is sought by people feeling bored.
Distribution: Celtis sinensis occurs on the Highveld of central South Africa (Figure 3) as a garden subject and has become naturalised in urban open spaces and riparian areas. It is generally sparsely distributed, but can form dense infestations in riparian areas and on roadsides. It is gradually colonising natural areas such as rangelands.
Phenology: Small, greenish flowers are present for a brief period during early spring. Fruits are produced in summer, but dried fruit can be present on the trees till late autumn.
Dispersal: Celtis sinensis reproduces with seeds, but also produces suckers. Its fleshy seeds are usually dispersed by animals (birds and bats) and by water run-off, or are dumped in urban open space as garden waste.

Diagnostic characters
Most Celtis species exhibit juvenile leaf morphology (Whittemore & Townsend 2007) and long persistence thereof, which is different from the adult leaf morphology and makes it difficult to distinguish between young plants of C. africana, C. australis, C. occidentalis and C. sinensis. The shape and size of the leaves, number of teeth on the toothed margin, the extent to which the teeth extend along the margin to the leaf base, petiole length and the colour and size of the drupe show enough variability to distinguish between species in their juvenile state.
Celtis sinensis has dark green, ovate leaves (30 mm-100 mm × 35 mm-60 mm) that are glossy above and softly hairy underneath. The leaves of C. australis are ovate-elliptic (50 mm-120 mm × 15 mm-35 mm), dark green and rough on the upper surface, but the lower surface is a lighter green and minutely hairy. Celtis africana leaves are ovate (15 mm-100 mm × 10 mm-50 mm), very hairy on both surfaces and dark green when young, but adult leaves are dull green. Celtis occidentalis leaves are broadly ovate to cordate (60 mm-130 mm × 40-80 mm), darker green and glossy above and softly hairy underneath.
The most conspicuous character by which juvenile plants of C. sinensis can be distinguished from C. africana, C. australis and C. occidentalis is the extent to which leaf margin is serrated (Figure 4). This variability is also contained in the adult leaf morphology. Celtis sinensis leaves are only toothed in the upper third towards the apex and have 8-16 teeth per side on the leaf margin. The leaf margin of C. africana is toothed more than two-thirds from the apex to the leaf base and generally has 14-22 teeth per side. Celtis australis has more than 20 teeth per side, and the margin is nearly completely toothed from the apex to near the base. Celtis occidentalis is also completely toothed down to the leaf base and commonly has 30-40 teeth per side.
The petiole length ranges from 3 mm to 7 mm for C. africana, from 3 mm to 10 mm for C. occidentalis and from 5 mm to 15 mm for both C. sinensis and C. australis. 3: Number of individuals > 1.5 m and saplings (< 1.5 m) of C. sinensis compared to a co-occurring indigenous relative and an invasive alien across land-cover types.  Gardens  7  18  34  58  145  300  12  30  43   Open space  11  28  30  31  78  72  125  313  1146   Riparian  27  68  165  17  43  64  27  68  846   Rangeland  105  263  398  14  35  56  39  98  511   Total  150  -627  120  -492  Fruit morphology and colour are also variable ( Figure 5), but not a character of value to distinguish between saplings. Fruit of C. sinensis are glabrous and an ovoid (5 mm-8 mm) drupe, dark orange in colour and presented on short, stout stalks (4 mm-6 mm). Celtis australis fruit (10 mm-12 mm) are also ovoid and hairless, yellow-white turning purple-black when ripe and presented on long stalks (12 mm-14 mm). Celtis africana fruit (4 mm-6 mm) are ovoid and glabrous, yellow turning brown and presented on long thin stalks (12 mm-14 mm). Celtis occidentalis has obovate or subglobose and dark orange fruit (8 mm-11 mm) turning purple-black when ripe and presented on long stalks.

Land-cover types
Celtis sinensis is easily distinguishable from the indigenous C. gomphophylla and C. mildbreadii in that the leaf apex of C. gomphophylla is elongated into a distinct drip tip, which is almost a third of the leaf length, and from C. mildbraedii, which has a symmetric leaf base (apex of C. sinensis is elongated, but not into a distinctive drip tip, and the leaf base is asymmetrical).

Key to adult specimens of Celtis in South Africa
1A. Leaf apex elongate into a distinct drip tip, up to onethird of the leaf length … 2 1B. Leaf apex elongate pointed, but not into a distinct drip tip … 3 2A. Margin of leaf entire, base slightly asymmetric, leaf covered in minute and stiff hairs giving it a roughish feel; drupe conical-ovoid, 5 mm in diameter, mature fruit dark yellowish; distributed in KwaZulu-Natal and Eastern Cape in lowland and upland forests … C. gomphophylla 2B. Margin of leaf sharply serrated from base to apex or along the upper two-thirds of the leaf, base strongly asymmetric, leaf covered in spreading hairs giving it a velvety feel; drupe subglobose to pyriform, 10 mm-12 mm in diameter, mature fruit dark brown almost black; distributed mainly in urban areas throughout South Africa … C. australis 3A. Leaf base cuneate and symmetric; fruit ovoid-ellipsoid, mature fruit red; distributed in Swaziland and KwaZulu-Natal in isolated patches of lowland forests… C. mildbreadii 3B. Leaf base rounded to cuneate or cordate and asymmetric; fruit globose or subglobose, mature fruit yellow to orange, brown or purplish … 4 4A. Margin of leaf subentire to crenate along upper half of leaf, glossy dark green, ovate to ovate-elliptic, adaxial surface glabrous, but with hairs abaxially scattered along major veins and tufted in vein axils, leaf has a leathery, smooth surface; fruit 5 mm-8 mm in diameter; distributed mainly in urban areas throughout South Africa … C. sinensis 4B. Margin of leaf serrated along upper two-thirds of the leaf; dark green on both surfaces or paler beneath, narrowly to broadly ovate; velvety or roughish surface... 5 5A. Leaves broadly to narrowly ovate to ovate-elliptic, dark green in colour, sparingly to densely covered with hairs, leaf has a velvety surface; fruit 4 mm-6 mm in diameter; distributed throughout South Africa in all forests types, mesic savanna and wooded grassland … C. africana 5B. Leaves ovate lanceolate to broadly ovate, dark green above, paler beneath, rough on the upper surface, smooth below; fruit 8 mm-11 mm in diameter; distributed mainly in urban areas throughout South Africa … C. occidentalis

Conclusion
Celtis sinensis is commonly cultivated on the Highveld of central South Africa as a garden subject or street tree and is naturalised in urban open spaces and riparian areas. It is colonising rangelands of agricultural areas and should therefore be considered as a potential invasive species. It shows a similar population structure and dispersal corridors as the declared invasive, M. azedarach. Vegetative characters of the leaves and fruits distinguish accurately between indigenous and alien Celtis species, both for juvenile and adult trees. The findings of this study will sensitise the horticultural industry to the potential invasive threat of C. sinensis and will inform the public on how to ascertain that white stinkwood purchased from nurseries are indeed C. africana.