Dispersal traits in the hyper-arid hot desert of the United Arab Emirates

1Sharjah Seed Bank and Herbarium, Sharjah Research Academy, P.O. Box 60999, Sharjah, UAE 2Departamento de Biología Vegetal, Universidad de Málaga, P.O. Box 59, 29080, Málaga, Spain 3Department of Applied Biology, Faculty of Science, University of Sharjah, P.O. Box 27272, Sharjah, UAE 4Permanent address: Department of Biology, Faculty of Science, Al-Arish University, Al-Arish, Egypt *Author for correspondence: hatem_shabana@yahoo.com


INTRODUCTION
Plant dispersal plays an essential role in the colonization of new habitats, population dynamics as well as in species interactions and community structure and floral diversity (van der Pijl 1969, Howe & Smallwood 1982, Willson et al. 1990, Hughes et al. 1994, Morales & Carlo 2006).Plant dispersal mechanisms or dispersal modes (the movement or transport of seeds away from the parent plant) could also be considered as a key factor in conservation biology and restoration management (Strykstra et al. 2002, Navarro et al. 2009a, Jara-Guerrero et al. 2011).Seed size is an im-portant ecological trait for seedling establishment, formation of a persistent seed bank, and dispersal.In addition, seed size variation between or within species has important ecological and evolutionary significance in plant dispersal (Westoby et al. 1992).Generally, large seeds increase the chance of seedling survival and establishment, while small seeds contribute more to forming seed bank (Thompson et al. 1993).Likewise, large seed tends to be desiccation-tolerant under climate constraints in drylands (Pritchard et al. 2004).Seed size is often associated with growth form (Mazer 1989, Jurado et al. 1991, Lord et al. 1997), and dispersal mode (Ellner & Shmida 1981, Jurado et al. 1991, Navarro Shabana, Navarro & El-Keblawy, Dispersal in hyper-arid hot desert et al. 2009a, 2009b).For example, large-seeded species are more likely to be animal dispersed, while small-seeded species are more likely to have unassisted seed dispersal (Leishman et al. 1995, Jurado et al. 2001, Navarro et al. 2009a).
The 'dispersal spectrum' is the frequency distribution of dispersal modes in a particular vegetation type (Howe & Smallwood 1982, Hughes et al. 1994).For several decades, researchers have been interested in studying of the dispersal spectra in different ecosystems (van der Pijl 1969, Frankie et al. 1974, Gentry 1982, Howe & Smallwood 1982, Navarro et al. 2006).In most studies, researchers are interested in assessing the relationship between the relative presence of dispersal modes in a particular site and the prevailing environmental conditions, such as temperature, precipitation, soil nutrient status or canopy structure (Willson et al. 1990, Bullock 1995).
Anthropogenic disturbance has negative effects on forests and the activity of animal dispersers (Daniels et al. 1995).The dominance of species with structures that facilitate spatial dispersal by abiotic vectors (e.g.wind) tends to increase with growing disturbances (Ganeshaiah et al. 1998, Nama & Choudhary 2013).In addition, it has been commonly reported that long-distance seed dispersal is less frequent in arid and semi-arid habitats than in mesic habitats and traits constraining seed dispersal are common in arid and semiarid floras (van der Pijl 1972, Ellner & Shmida 1981, van Rheede van Oudtshoorn & van Rooyen 1999, Navarro et al. 2009b).In this way, restrictions to seed dispersal have been widely used to explain the origin and maintenance of patches in arid vegetation (Kefi et al. 2008).
Seed size is correlated with flowering time (Mazer 1989, Navarro et al. 2009b) and consequently, with dispersal phenology (Navarro et al. 2009b).In hyper-arid hot deserts, phenological events are mainly related to the amount and time of annual rainfall and soil moisture condition (El-Keblawy et al. 2015a).The earlier flowering allows for a longer development time of seeds, which would promote larger seeds (Mazer 1990).Conversely, the later flowering gives a good chance for using more resources (moisture conditions) for seed development, but shorter time available for seed maturation and dispersal (Roux et al. 2006).
To our knowledge, no studies exist that assess dispersal traits and dispersal phenology in the Arabian desert.This kind of studies is particularly important for comparative ecological studies and for developing scientific conservation initiatives (Finckh 2006).We expect that dispersal spectrum could be similar to those recorded in other arid floras with the particularity of the presence of Afro-Arabian floristic elements.We also expect variation of the dispersal traits in the different studied habitats.The aims of this study are to examine the variation of diaspore size (seed and fruit length) among: (1) growth forms, (1) absence/presence structures facilitating long distance dispersal (spatial dispersal) (3) dispersal modes (4) APG IV phylogenetic taxonomic groups (5) phytogeography and (6) dispersal phenology in different habitat studied from the United Arab Emirates.

Study area
The United Arab Emirates (UAE) are located in the southeastern part of the Arabian Peninsula, between latitudes 22°40′ and 26°00′ North and longitudes 51°00′ and 56°00′ East, covering a total surface of 83 600 km 2 .The UAE has a hyper-arid hot climate.The annual average temperature is approximately 27°C and humidity 45%.There are two main seasons; a rainy season with mild temperatures from November to March, with a mean temperature around 20°C and a dry hot season from April to October, with a mean temperature range between 35 and 40°C that can reach 47°C and humidity levels reaching more than 90%.The average amount of annual rainfall is around 110 mm with extreme variability in space and time and most of the rainfall occurring from January to April when temperatures are lowest (Böer 1997, UAE Ministry of Energy 2006, 2012, Sherif et al. 2009, UAE Ministry of Environment and Water 2015).
Our study is focused in inland forest and shrubland vegetation.Five different habitats were selected for this study: salt flats, gravel plains, sand sheets, mountains and high mountains (Jongbloed 2003, El-Keblawy et al. 2005, Karim & Fawzi 2007).The sites were selected to cover the main geomorphologic conditions characterizing the UAE (Böer & Gliddon 1997, Jongbloed 2003).Salt flats ("sabkha") separate the inland sandy areas and low Arabian Gulf coast.The salt flats region is dominated by Zygophyllum qatarense and Salsola imbricata with some frequent species such as Heliotropium kotschyi, Fagonia ovalifolia, Arthrocnemum macrostachyum, Suaeda vermiculata, Halopeplis perfoliata and Anabasis setifera.The gravel plains habitats are characterized by coarse sand and small rocky soils generally are dominated by Acacia tortilis mixed sometimes with Proso pis cineraria, cultivated palms (Phoenix dactylifera) and Calotropis procera.Most of the land surface of the UAE desert is covered with sand, which consists of a mixture of low dunes, high dunes and intervening sand flats.Sand sheets are dominated by Cyperus conglomeratus, Tribulus omanense and Zygophyllum qatarense.In the central desert, which is dominated by low sand dunes, the water table is deep and abounds in Haloxylon salicornicum, graminoids and some seasonal herbaceous plants such as Stipagrostis plumosa, Cleome amblyocarpa and Eremobium aegyptiacum.East of the Dubai region, the sand sheets abound in Leptadenia pyro technica, Calligonum comosum and Calotropis procera.
The mountain sites correspond to the Hajar Mountains series that extend from the UAE to the neighbouring Oman.Their elevations can reach more than 1000 m a.s.l.The common shrubby species in the mountains is Euphorbia larica.In the running wadis, trees such as Prosopis cineraria and Ficus cordata are dominant.On rocky hillsides (mountain), Moringa peregrina grows in addition to other graminoids and annuals (e.g.Cymbopogon commutatus, Launaea spi nosa, Filago desertorum and Reichardia tingitana).
The high mountain habitats correspond to the North Mountains and are geologically distinct from the Hajar Mountains.They are sedimentary rocks, mostly carbonate with a peak at 2000 m a.s.l.Ficus johannis grow in these mountains, while Pl.Ecol. Evol. 151 (2), 2018 the high plateaux are covered with bushes (e.g.Seriphidium herbaalba).Some annual species, such as Campanula erinus and Pentanema divaricata grow in silt among the rocks.

Species selection
Field sampling and phenological observations were carried out monthly between May 2014 and October 2016.The selected species represent a cross-section of the flora of the arid/hyper-arid hot desert with xeric shrubland vegetation; importance was given to cover the most common and some rare species that characterize the vegetation (Jongbloed 2003, El-Keblawy et al. 2005, Karim & Fawzi 2007).
We selected 30 sampling points (quadrates) per studied habitat which represent a total of 150 quadrates (fig. 1, electronic appendix 1).The quadrate size was 200 m 2 , in each quadrate the species dominance had been previously assessed with the relative importance values in which the abundance of the species has been estimated according to the Braun-Blanquet scale (Braun-Blanquet 1928).We discarded all the species that were present in less than 10% of all the relevés available.The selected dominant species probably represent c. 80% of the total cover in the five studied sites.Voucher specimens of the most studied species were kept in the Sharjah Seed Bank and Herbarium (Sharjah Research Academy).Botanical nomenclature follows The International Plant Names Index (IPNI 2016).The family and class affiliation of each species used APG IV (Angiosperm Phylogeny Group 2016).
Among dispersal traits, the dispersal mode and diaspore size (diaspora length) constitute the main regenerative traits according to Cornelissen et al. (2003) and Pérez-Harguindeguy et al. ( 2013).Dispersal modes follow van der Pijl (1982).Shabana, Navarro & El-Keblawy, Dispersal in hyper-arid hot desert Whenever a diaspore lent itself to more than one type of dispersal mode, the predominant form observed in more than 80% of sampling quadrates was generally used.
To measure diaspore size (diaspore length), diaspores were collected when ripe but before they started to fall off the plant.For each individual species, 100 diaspores were air-dried and their mean size was measured as the longest dimension (Cornelissen et al. 2003, Pérez-Harguindeguy et al. 2013).The achenes of Asteraceae were measured with the pappus.Exceptionally, when they were not available in the necessary quantities, the data were completed based on Flora databases (e.g.Flora of Pakistan 2016, http://www.tropicos.org/Project/Pakistan) and references (Miller & Cope 1996, Jongbloed 2003, Boulos 2009).
The term diaspore was used to name the dispersal unit (Weiher et al. 1999).We scored whether species had diaspores equipped with structures that facilitate spatial dispersal, such as pappi, barbs or wings for dispersal by abiotic vectors (e.g.wind) or fleshy fruits to dispersal by biotic vectors (e.g.vertebrates) (Ellner & Shmida 1981, Venable & Levin 1985, Cain et al. 2000).Species were considered as having restricted dispersal if diaspores lacked such dispersalenhancing characters (Willson 1993).Anemo-meteochoric species have structures facilitating long-distance dispersal by abiotic vectors (e.g.wind); barochory species have fleshy fruits or pods with nutrients structures facilitating long-distance dispersal by biotic vectors (e.g.vertebrates); ombrohydrochory, semachory and ballistic species lack structures facilitating long-distance dispersal.Diaspore appendage type were determined following Thomson et al. (2018) and based on the field observations.The colour categories follow (Wheelwright & Janson 1985); black (including dark red), red (including pink), yellow, orange, brown, grey, green, white and blue (including purple).
To establish dispersal phenology (rainy or dry season), phenological surveys were performed for at least 2.5 years.
In each year, the units of vegetation from each site were observed 10-12 times at approximately 30-45-day intervals.The onset of dispersal (the beginning of dispersal time) was described (electronic appendix 2).
Species were classified on the basis of their biogeographical distribution in desert regions of the world (Good 1964, Evenary et al. 1985).The major phytogeographical regions were: the central Australian region, the western and central Asian region (Irano-Turanian region, Eig 1931, 1932, Zohary 1973) and the African-Indian region (Saharo-Sindian region, Eig 1931Eig , 1932)), where our study area is located.Following these authors nine phytogeographical regions were recognized in order to establish relationships between dispersal traits and species biogeography.Species, family, habitats, dispersal traits, phytogeography and dispersal phenology of the studied species are presented in the electronic appendix 2.

Statistical analyses
One-way analysis of variance (ANOVA), followed by the post hoc comparisons (Tukey HSD test), was used to test the significance of differences (p < 0.01) of log diaspore size among APG IV groups and dispersal traits.Categories with less than seven species were excluded from the analyses.The association between nominal traits was determined with the Pearson χ 2 test-statistic.All statistical analyses were performed with SPSS 14 and SYSTAT 13.

RESULTS
A total of 302 species, belonging to 51 families and 192 genera, were studied (electronic appendix 2).Fabaceae (12.6% of total species), Asteraceae (9.9% of total species), Poaceae (9.6% of total species) and Amaranthaceae (7.6% of total species) were the richest families.Fabaceae species were recorded in all habitats.Asteraceae were absent in salt flats and Poaceae and Amaranthaceae were absent from high mountains.
A percentage of 53.6% of the studied species were woody species, 34.8% herbaceous and 11.6% graminoids species.The dominant growth form was dwarf shrubs (41.1%) followed by semi-basal herbs (21.9%) and tussocks (11.6%).The dispersal phenology of studied species was in the dry season (from April to October, 63.9% of total species) followed by the rainy season (from November to March, 32.1% of total species), and throughout the year (4% of total species) (electronic appendix 2).
Semachory was predominant in all studied habitats followed by anemo-meteochory and semachory, respectively (fig.3).Ballistic species were absent in salt flats, ombrohydrochorics were absent in high mountains and zoochoric species were represented only in mountains, sand sheets and gravel plains.
A total of 60% of studied species lacked diaspore appendages.48% of anemo-meteochoric species had winged appendages and 46.6% of them pappus.The anemo-meteochoric species with wings were dominant in sand sheets (33.3%) while species with pappus were dominant in gravel plains, mountains and sand sheets (27.6% in each habitat).
Diaspore length differed significantly according to the presence/absence of developed structures for long-distance dispersal (F (2, 299) = 42.590;p ≤ 0.001).The post hoc comparisons (Tukey HSD test) shows that the mean of diaspore length of species dispersed by developed biotic vectors (2.52±0.71cm) was significantly longer from the species dispersed by developed abiotic vectors (0.68±0.07 cm) and species with restricted spatial dispersal (0.47±0.06 cm).Species dispersed by biotic vectors (e.g.Prosopis cineraria, 18.8 cm and Calligonum comosum, 1.2 cm) had the larger diaspore size than those dispersed by abiotic vectors and species who lack structures facilitating long-distance dispersal (e.g.Reseda aucheri, 0.08 cm, fig.5B, table 1).

Relationships between presence/absence of developed structures facilitating long-distance dispersal, APG IV phylogenetic groups, growth forms, phytogeography and dispersal phenology
There were significant relationships between APG IV phylogenetic taxonomic groups and the presence/absence of developed structures for long-distance dispersal (χ 2 = 119.99;df = 10; p < 0.001, fig.9A).Absence of structures for long-distance dispersal was the most common in all phylogenetic taxonomic groups except for Campanulids, in which developed structures for dispersal by abiotic vectors was over-represented (fig.9A).Developed dispersal by biotic vectors was present only in Lamids, Malvids and Fabids (fig.9A).
Significant relationships were detected between the presence/absence of developed structures for long-distance dispersal and growth forms (χ 2 = 92.054;df = 12; p ≤ 0.001, fig.9B).Absence of structures for long-distance dispersal was overrepresented in herbaceous species and dwarf shrubs (fig.9B).Presence of structures for long-distance dispersal by biotic vectors was mainly prevalent in trees (fig.9B), while presence of structures for long-distance dispersal by abiotic vectors was over represented only in shrubs (fig.9B).Significant relationships were detected between the presence/ absence of developed structures for long-distance dispersal and species phytogeographical distribution (χ 2 = 48.268;df = 16; p ≤ 0.001, fig.10).Developed structures for dispersal by biotic vectors were mainly associated with Sudano-Decanian species (fig.10), which are species forming part of African vegetation in the study area.Saharo-Arabian species were exclusively restricted dispersal and they were dominant species in all regions except in the Saharo-Arabian and Irano-Turanian region.
Significant relationships were detected between the presence/absence of developed structures for long-distance dispersal and dispersal phenology (χ 2 = 15.644;df = 4; p ≤ 0.004).A total of 63.6% of the species with developed structures for long-distance dispersal dispersed in the dry season from May to October (fig.11).

DISCUSSION
The range of seed size variation in the studied species in the hot UAE deserts coincides with that found in other arid and semi-arid ecosystems (Jurado et al. 1991(Jurado et al. , 2001, Leish-      et al. 2000et al. , Navarro et al. 2006et al. , 2009a)).However, in the UAE deserts, there were no very small (log, 10 -4 ) or very large diaspores (log, 10 2 ).Harel et al. (2011) reported that seed size significantly decreased with increasing aridity and rainfall variability in seven out of fifteen sites in the hot Negev Desert.Our results are in agreement with previous studies that indicate that the largest diaspores were represented in the Fabaceae family while the smallest diaspores were found in Poaceae (Mazer 1989, Liu et al. 2014).The advantage of small seeds is a lower risk of being eaten by predators (Hulme 1994, van Rheede van Oudtshoorn & van Rooyen 1999); their small size enables them to fall into soil cracks and consequently avoid detection by predators (Gutterman 2002).This indicates that small seeds could form a longlived seed bank in the desert, which ensures species survival under the heterogeneous and unpredictable desert conditions.However, the small reserves in the small seeds would not help them to emerge from deep soil unless some kind of disturbance happens (El-Keblawy et al. 2015b, El-Keblawy & Gairola 2017, El-Keblawy 2017).On the other hand, small seeds tend to be dessication-sensitive seeds that germinate rapidly because delayed germination can induce seed death (Pritchard et al. 2004), as is the case of the ombro-hydrochoric species.
The seed dispersal spectrum of the studied hyper-arid hot desert in the UAE was characterized by the dominance of semachory (43.7%), followed by anemo-meteochory (28.8%) and barochory (23.8%).Nevertheless, semachory/ barochory is the largest group (67.5%).Other species in the Amaranthaceae, Zygophyllaceae, Brassicaeae and Poaceae families are semachores, just like other species in the Fabaceae and Zygophyllacae families are barochores with larger seeds that can take full advantage of the favorable surrounding conditions and germinate fast and to high levels (Liu et al. 2014).About 68.5% of the semachoric species were herbaceous, which is consistent with other desert ecosystems (van Rheede van Oudtshoorn & van Rooyen 1999).Zoochory has been described as the most dominant mechanism (more than 80%) in humid and dry tropical forests (Fleming 1979, Gentry 1982), but its presence decreases in ecosystems with dryer climates (Gentry 1982).In temperate areas, barochoric species are significantly more frequent than anemometeochoric and zoochoric species (Leishman et al. 1995, Wang et al. 2009).However, semachoric species lacking long-dispersal structures, are more frequent in our study area helping the seeds to stay near the mother plants, which could be considered as "safe sites" for seedling establishment (van Rheede van Oudtshoorn & van Rooyen 1999, Rodriguez et al. 2017).Only three of the studied species were zoochoric (the epizoochoric Tragus racemosus and Medicago sp.).The low frequency of epizoochory could reflect the low presence of mammals in the harsh desert conditions.
In the desert region, restricted spatial dispersal is more dominant than developed spatial dispersal and was more represented in herbaceous and than woody species (Ellner & Shmida 1981).These species can survive in situ, with regeneration and plant recovery in extreme and hard climatic conditions (Gutterman & Shem-Tov 1997, van Rheede van Oudtshoorn & van Rooyen 1999).The graminoids living preferably in the gravel plains and mountains having appendages which enable them to be effectively dispersed by the wind, pass through the cracks of the biological soil crusts, and settle down, such as Stipa sp., Stipagrostis sp. and Aristida sp., which could take a special way named "active drill" into soil cracks using mucrons (Schöning et al. 2004, García-Fayos et al. 2013).Brassicaceae, Lamiaceae and Scrophulariaceae herbaceous species are mainly restricted spatial-dispersal species with relatively small seeds, as found in other arid regions (Navarro et al. 2009a(Navarro et al. , 2009b)).Efficient spatial dispersal is associated with the presence of morphological structures favors dispersal long-distance by the wind such as hairy capsules (e.g.Aerva javanica), winged calyx (e.g.Astragalus squarrosus), wings (e.g.Tribulus sp.) or pappus in the Asteraceae species.
Dispersal mechanisms can be associated with climatic factors (Gentry 1982, Wikander 1984, Griz & Machado 2001, Navarro et al. 2009a, Jara-Guerrero et al. 2011), and most probably with the temporal patterns of water availability.These factors have the largest impact on plant propagation in hyper-arid hot desert as in other hot dry regions (Griz & Machado 2001).In the savannas of West Africa, strong relationships were found between fruiting and soil moisture, which may reflect a selection of dispersal periods that maximizes seed dispersion and germination (Seghieri et al. 2009).In our study, 25.1% of species disperse in the rainy seasons and they are mainly Sudano-Decanian barochoric species (Afro-Arabian species) (electronic appendix 2), which form part of the savanna vegetation.
The results of our study showed that species dispersed by abiotic vectors (wind) have a dispersal peak in synchronization with the dry season, as found in other regions with climatic restrictions, such as Mediterranean (Navarro et al. 2009a, 2009b), desert (van Rooyen et al. 1990) and dry tropical regions (Machado et al. 1997, Griz & Machado 2001).
Continuous patterns of dispersal throughout the year were observed in some of the studied species.This indicates an important seed supply to occur in the ecosystems that are characterized by seasonal variation and where a rainfall could be expected at any time of the year (Machado et al. 1997).These species correspond to barochoric species with large fleshy fruits (e.g.Ziziphus sp., Solanum sp.) or pods with nutrient structures (e.g.Senna italica, Indigofera sp., Tephrasia sp.).Interestingly, species that can disperse throughout the year have lower dormancy and are able to germinate at a wider range of environmental conditions; consequently, they can contribute to seedling recruitment at any time of the year (El-Keblawy 2017).
To conclude, the observed patterns suggest that the seed dispersal spectrum of the hyper arid hot desert of UAE follows the general patterns of arid and desert ecosystems with

Figure 1 -
Figure 1 -Illustration map for the 150 sampled quadrates in the study area from the hyper-arid hot desert of United Arab Emirates.

Figure 2 -
Figure 2 -Frequency of seed length (Ln) of studied species in the hyper-arid hot desert of United Arab Emirates.

Figure 3 -
Figure 3 -Frequency distribution of dispersal modes among species in the five studied habitats from the hyper-arid hot desert of United Arab Emirates.

Figure 4 -
Figure 4 -Diaspore length variation among A, APG IV phylogenetic taxonomic groups and B, growth forms of studied species from the hyper-arid hot desert of United Arab Emirates.(Cornelissen et al. 2003 and Pérez-Harguindeguy et al. 2013).

Figure 5 -
Figure 5 -Diaspore length variation among dispersal traits in the studied species from the hyper-arid hot desert of United Arab Emirates; A, dispersal mode; B, absence/presence structures facilitating long distance dispersal; C, absence/presence of appendage type; D, diaspore colour.

Figure 7 -
Figure 7 -Frequency distribution of dispersal modes among APG IV phylogenetic groups and growth forms of studied species from the hyper-arid hot desert of United Arab Emirates.

Figure 8 -
Figure 8 -Frequency distribution of dispersal modes among dispersal phenology of studied species from the hyper-arid hot desert of United Arab Emirates.

Figure 9 -
Figure 9 -Frequency distribution of absence/presence structures facilitating long distance dispersal among APG IV phylogenetic groups and growth forms of studied species from the hyper-arid hot desert of United Arab Emirates.

Figure 11 -
Figure 11 -Frequency distribution of absence/presence structures facilitating long distance dispersal among dispersal phenology of studied species from the hyper-arid hot desert of United Arab Emirates.

Table 1 -Mean and standard error (SE) of nine traits studied in the 302 plant species from the hyper-arid hot desert of United Arab Emirates.
Nine traits are: APG IV, growth forms, absence/presence structures facilitating long distance dispersal, dispersal mode, absence/presence of appendage type, diaspore colour, phytogeographical regions, habitats, dispersal phenology.