Assessment of plant species diversity associated with the carob tree (Ceratonia siliqua, Fabaceae) at the Mediterranean scale

1Institut Méditerranéen de Biodiversité et d’Ecologie marine et continentale (IMBE) [IMBE is sponsored by Aix Marseille University, Avignon University, CNRS and IRD], Station marine d’Endoume, Chemin de la Batterie des Lions, FR-13007 Marseille, France 2Royal Botanic Gardens, Kew, Richmond, Surrey, TW9 3DS, United Kingdom 3Université Saint-Joseph, Faculté des sciences, Laboratoire Caractérisation Génomique des Plantes, B.P. 11-514 Riad El Solh, Beyrouth 1107 2050, Lebanon 4Università degli Studi di Catania, Dipartimento di Agricoltura, Alimentazione e Ambiente (Di3A) Via Valdisavoia 5 – IT-95123 Catania, Italy 5Université Cadi Ayyad Marrakech, Faculté des Sciences Semlalia, Laboratoire d’Ecologie et Environnement, Morocco 6Conservatoire Botanique National Méditerranéen de Porquerolles (CBNMed), 34 avenue Gambetta, FR-83400 Hyères, France 7Laboratoire des Symbioses Tropicales et Méditerranéennes (LSTM) [LSTM is sponsored by Univ Montpellier, CIRAD, IRD, INRA, Montpellier SupAgro], TA A-82/J Campus International de Baillarguet FR-34398 Montpellier cedex 5, France *Author for correspondence: alex.baumel@imbe.fr


INTRODUCTION
The Mediterranean thermophilous woodlands shelter several relictual taxa that were more widespread during the Miocene warmer climate (de Saporta 1888, Quézel & Médail 2003, Thompson 2005). Due to their higher incidence along the coasts and within the lowlands, thermophilous forests were very early impacted by human activities, mainly for forage and fruits harvest (Grove & Rackham 2001, Blondel 2006. The recurring exchanges between natural ecosystems and cultivated areas constituted a pivotal aspect in the long-term process of Mediterranean fruit tree domestication (Miller & Gross 2011, Weiss 2015, Aumeeruddy-Thomas et al. 2017 and they still play an important role for genetic resources in sight of global change (Besnard 2016). Natural and semi-natural forests, as well as traditional Mediterranean agroecosystems, harbor fruit tree populations throughout environmental gradients, and thereby they constitute reservoirs of evolutionary and genetic resources of crop wild relatives.
The carob tree (Ceratonia siliqua L.), is a Mediterranean thermophilous fruit tree widely exploited for food and forage since antiquity and currently for industrial, agricultural and soil restoration purposes (Battlle & Tous 1997). Despite its contribution to Mediterranean agroecosystems, the native status of the carob tree is still source of debate. On the basis of archaeological and etymological evidences several authors consider that after its domestication in the Middle-East around 6000-4000 BC the carob tree was disseminated by humans to the West (de Candolle 1883, Battlle & Tous 1997, Ramón-Laca & Mabberley 2004. As a consequence wild carob trees are a "semi-natural" or "semi-domestic" component, beside other fruit tree species, such as oaks, pistachios and almonds that were also extensively exploited since millennia (Blondel 2006). On the other hand, phytosociological studies usually included the carob in spontaneous vegetation as a native (i.e. as opposed to feral), considering it as a bioindicator of Mediterranean thermophilous maquis and forest communities (Zohary & Orshan 1959, de Bolòs 1970, Quézel & Médail 2003. However, a comprehensive survey of plant diversity associated to carob tree is still lacking despite the constantly emphasized importance of the tree in the Mediterranean human culture and the long-standing interest of botanists (e.g. de Candolle 1883, Hillcoat et al. 1980, Batlle & Tous 1997, Ramón-Laca & Mabberley 2004.
A comprehensive conservation of carob tree habitats requires to define the main patterns of plant species diversity through its range. After decades of investigation the phytosociological literature potentially contains this knowledge. In this study we use available phytosociological data to examine geographical patterns of plant species diversity associated with carob tree all over the Mediterranean region.

Collection of floristic relevés
A set of 1683 phytosociological relevés performed according to the method of Braun-Blanquet (1964) and including C. siliqua were gathered from databases (56.5%) and from a large literature survey (43.5%). The sources of this information are detailed in electronic appendix 1B. The nomenclatural homogenization was performed according to the following steps: first the homogenization was done progressively, country by country, and according to the Euro+Med PlantBase (2015) and the African Plant Database version 3.4.0 (2015). For a final correction of synonyms and plant name authorships the files were merged and submitted to The Plant list (2013) using the function TPL() of the Taxonstand R package (Cayuela et al. 2012). The list of all names of taxa with their authorships is available in electronic appendix 3. After removing those relevés with a lack of precision to assign geographical coordinates, the final data set included 1542 relevés and 1823 taxa (species and subspecies). Information about the plot size of each relevé was available for 961 of them, and the 75% ranged from 100 to 400 m 2 whereas the 25% remaining were distributed from 4 to 90 m 2 with a dominance of relevés of 50 m 2 (11%). The correlation between plot size and species richness was low and nonsignificant (Kendall τ = -0.04, p = 0.09) and therefore was not considered as a bias in the analyses. Moreover relevés done on plot size below 100 m 2 were not concentrated in one region and where found in areas where a lot of relevés were collected. The table of species occurrences with their abundance and geographical coordinates of relevés is available in electronic appendix 3.

Multivariate analyses of floristic structure
Multivariate analyses of floristic structure were carried out with a reduced data set of 1542 relevés for 1141 species, in which singletons (i.e. species observed in only one relevé) were removed. To investigate the geographical patterns of plant diversity, the relevés were gathered into areas containing at least 25 relevés; 18 areas were defined, 11 in the Western and 7 in the Eastern Mediterranean. These areas were selected to best reflect the Mediterranean geography (table 1, see map of the 18 areas in electronic appendix 1A). The geographical split between Western and Eastern Mediterranean was based on biogeographical literature (Finnie et al. 2007). A principal coordinate analysis (PCoA; dudi.pco() function, ade4 R package, Dray & Dufour 2007) was performed based on the Bray-Curtis dissimilarity index (vegdist() function, vegan R package, Oksanen et al. 2016). Species abundance was obtained from Braun-Blanquet's abundance-dominance codes using the scores of 0.5, 1, 2, 3, 4 and 5. The first 100 axes of the PCoA were kept, as they included 75% of total variance, and were submitted to a between-class analysis (BCA, bca() function, ade4 R package) using geographical area as the explanatory variable to test for geographical trends (randomization test, 999 iterations).

Plant species composition and diversity associated to the carob tree
Analyses of plant species diversity were done on the data set with all species, i.e. including singletons (1542 relevés for 1823 taxa). A constancy table reporting the frequency of species in each of the 18 areas was built to summarize the composition of main floristic groups at the scale of the study and only the species present in at least 25% of the relevés of one area were kept. To summarize the pattern of species composition at a coarser level, the 18 areas were gathered in five groups according to their floristic similarity within the constancy table, by a hierarchical cluster analysis (HCA; Bray Curtis distance and Ward algorithm; functions vegdist(), vegan R package, and hclust()). The dendrogram was truncated to retain five main clusters or floristic groups, which we consider a good compromise to describe the main geographical pattern at the Mediterranean level. To characterize the composition of these floristic groups the functions indval() and const() (labdsv R package, Roberts 2016) were used to list the indicator and most frequent plant taxa. These lists were used to assign a phytosociological alliance to each floristic group (species indicator values available in electronic appendix 4). The floristic group were assigned to phytosociologi-cal alliances following Quézel & Médail (2003) and Tsiourlis et al. (2007).
Patterns of plant species diversity were summarized for each area and compared according to the five main floristic group obtained by the HCA and finally according to Western and Eastern Mediterranean. At the level of the 18 areas, Pearson correlation tests revealed a strong and significant bias caused by the number of relevés in each area for gamma species richness (r = 0.93, p < 0.001), but not for alpha species richness (r = 0.35, p = 0.15). To take into consideration the bias due to different sampling efforts, the comparison of species diversity among the five floristic groups or between the Western and Eastern Mediterranean was done with the sample-size based interpolation and extrapolation (R/E) of Hill numbers developed by Chao & Jost (2012). The rarefaction and prediction analysis of Hill numbers was applied to the three orders of taxonomic diversity (species richness, Shannon diversity and Simpson diversity) with the iNEXT() function (iNEXT R package, Hsieh et al. 2016); confidence intervals were based on 999 iterations and R/E curves were done with ggiNEXT() function. Rarefaction (interpolation) and extrapolation (prediction) of Hill numbers are a unified method to account for differences in sample size when comparing species diversity across several assemblages.

Geographical pattern of plant species diversity associated with carob tree
The plant diversity analyses revealed common association between carob and plant species of Mediterranean forest environments such as Pistacia lentiscus, Olea europaea, Smilax aspera, Quercus coccifera, Rhamnus spp. and Phillyrea spp. (table 2, electronic appendix 4). The BCA analysis indicated that a large part of the floristic variance was due to the within-area level: the between-area inertia percentage was 19% (p = 0.001, 999 permutations). However, the BCA ordination revealed a strong floristic differentiation between the Western and the Eastern basins, followed by the opposition between Southern Morocco and the French Riviera ( fig. 1A  & B). By contrast plant species diversity of the areas of East- Diversity estimates produced by rarefaction analyses (iNEXT R package) to compare plant species diversity associated to carob tree among five geographical areas. LCL= lower confidence limit, UCL= upper confidence limit. ern is weakly structured; only the area z17 gathering relevés from Lebanon and Syria appeared to be different and only from the fourth axis of the BCA (not shown).

Main plant species assemblages associated to carob tree
Floristic composition among the 18 areas was partitioned by a HCA in five floristic groups (     Table 3 -Summary of the indicator plant taxa associated with carob tree according to five floristic groups. Plant taxa were listed in decreasing order of their indicator value in each floristic group and the first twenty are shown here. The whole list is available in electronic appendix 4. The last row indicates the phytosociological alliance of the floristic group (see the text for the details).
spectively (table 1). Although Shannon diversity index was similar across basins, Simpson diversity index was slightly but significantly higher in the Eastern than in the Western Mediterranean (table 1), indicating higher cover-abundance scores of species in the relevés of the Eastern basin. Besides, this result also indicates that the higher species richness in the Western basin is related to the recording of more numerous rare species (table 1).

DISCUSSION
The thermophilous woodlands of the Mediterranean region constitute reservoirs of genetic resources for several fruit trees. To our knowledge the ecological range of wild relatives of domesticated fruit trees has been overlooked and very rarely examined at the scale of the Mediterranean region. This study illustrates the potential of phytosociological literature to characterize the plant species biodiversity in a broad geographic range and confirms that phytosociological literature shelters invaluable resources for ecological and environmental surveys. An initiative of a systematic collection of phytosociological data as done in Mediterranean databases SIVIM (Font et al. 2012), SILENE (2015 and VEGHEL-LAS (Dimopoulos et al. 2012) is indispensable for the ecological scientific community. By means of data from these three databases (SIVIM, SILENE, VEGHELLAS) and of an extensive literature investigation the plant species diversity associated with carob tree was described throughout its current distribution in the Mediterranean based on 1542 floristic relevés. All the plant assemblages associated to carob tree described here have in common the overall high frequency of Olea europea and Pistacia lentiscus and the dominance of trees, shrubs or lianas such as Smilax aspera. Alongside these forest and maquis elements our analyses revealed that the vegetation associated with carob trees is diversified at the scale of the Mediterranean; the Oleo sylvestris-Ceratonion siliquae phytosociological alliance being only one of the five alliances described in this study (table 3 and see Results). The range of floristic differentiation is higher in the Western Mediterranean where three poles of diversity are shown in the BCA plot ( fig. 1): Southern Morocco (an endemic phytosociological alliance), France and Southern Spain and Portugal. By contrast the areas of the Eastern basin are less differentiated and more overlapping ( fig. 1) indicating a more homogeneous floristic diversity. Rarefaction/extrapolation sampling analyses of Hill numbers (table 1, fig. 2) converged to show that the plant assemblages associated with the carob tree contained higher species-richness in the Western than in the Eastern basin with more infrequent species in the Western (see differences in Simpson diversity, table 1).
In summary, carob tree habitats are characterized in the Western basin by a higher species richness due to higher number of rare species and a stronger floristic differentiation due to the strong latitudinal gradient of the Western range. In the Eastern basin, the floristic differentiation is less pronounced but Simpson diversity is higher indicating a greater diversity of abundant species. However, when geographical patterns were summarized according to five floristic groups a more nuanced pattern was revealed with a level of species richness in Greece similar or higher to those observed in the West and highest Simpson diversity in the East (table 1). The lowest estimates of plant species richness (extrapolation curve fig.  2A) are found for the southern and eastern limits of carob tree range. This pattern might be regarded as the consequence of either biogeography per se or different anthropogenic activities (Quézel & Médail 2003). For example in Greece, the carob tree is extensively exploited in multi-use agrosystems  (Blondel 2006) and it is also very abundant in rocky outcrops, cliffs and canyons where it is spread by the profusion of goats in landscapes. Thus, in Greece carob trees spread in an open vegetation, the Sarcopoterion spinosi alliance, associated to a large diversity of plant species (table 1). By contrast, toward the Eastern range limit, such as in Lebanon, the vegetation with carob tree that was investigated in the literature correspond in general to abandoned fields and orchards that are colonized by thermophilous shrubs (Talhouk et al. 2005).

Conclusions
The debate about the status of the carob tree -native, or introduced for cultivation -in the Western Mediterranean (Ramón-Laca & Mabberley 2004) will remain open until adequate investigations such as phylogeographical and paleoecological studies are conducted. In the context of forthcoming research on the history of the carob tree in the Mediterranean, the extensive data survey reported here bore evidence of the presence of carob tree along a large ecological gradient in the Western Mediterranean, where it is very well integrated to spontaneous vegetation within hotspots of plant biodiversity well known to be also biogeographical refugia (Benabid & Cuzin 1997, Médail & Quézel 1999, Molina-Venegas et al. 2017. Within the context of the aggravation of global changes, the ecological resilience capabilities have become a major issue in Mediterranean basin (see Thiébault & Moatti 2016). This issue is crucial, notably for Mediterranean orchards based on fruit trees, for which high productivity practices lead to a drastic decrease and homogenization of millenary biodiversity. The large ecological range of the Mediterranean carob trees is potentially an important evolutionary legacy for the conservation of genetic resources and seed sourcing for new uses such as restoration ecology (e.g. Domínguez et al. 2010), and also to develop a more water-efficient fruit farming in the drier areas of the Mediterranean basin.

ACKNOWLEDGMENTS
We thank Xavier Font and Panayotis Dimopoulos for their help and the data they provided from SIVIM and VEGHEL-LAS databases respectively. This study is part of an international project on carob conservation, biogeography and ecology (https://dynamic.cirad.fr) supported by French national agency of research (ANR, contract n°14-CE02-0016). Marwa Moakhar was supported by ANR and Mednet (Mediterranean Network of environmental training sites supporting OSU Institut Pythéas master's degrees and international academic partnerships).