P. erinaceus is a leafy arborescent species about 20 m high in a straight shaft. Compound leaves, alternate and imparipinnate consist of 7 to 15 leaflets from oval to elliptical shapes [12] . The inflorescence is a panicle made up of golden yellow fragrant hermaphroditic flowers. The fruit is an indehiscent pods garlic at maturity, rounded and flat in shape. Straw-coloured fruits contain one or two seeds. The seeds are kidney-shaped and dark brown in colour [12] .

The study was carried out in the savannahs of Côte d’Ivoire (northern part of the country), the natural distribution area of P. erinaceus. According to variation in climatic factors and vegetations, Sudanian savannah, sub-Sudanian savannah and Guinean savannah can be distinguished [13] . In each of these savannahs, two sites were selected to characterize the morphological variability of natural populations of P. erinaceus (Figure 1).

In Sudanian savannah, the sampling sites are the Ouarigué protected area (58,000 ha) located between 9˚61' and 9˚65' North latitude and between 4˚89' and 4˚96' West longitude and the Téguirdouo site (village to 5 kms from Bouna) located between 9˚25' and 9˚28' North latitude and between 2˚92' and 2˚95' West longitude. The Téguirdouo site is subjected to many pressures including illegal grazing and bushfires. The Ouarigué protected area is under pressure from human activities such as uncontrolled land clearing and overgrazing. The climate in the area is Sudanese and typified by a long dry season (October to June) and a short rainy season (July to September). The annual rainfall varies from 700 to 1200 mm [14] with the intermittent presence of harmattan (fresh and dry wind) from December to February. Temperatures vary from 28˚C to 32˚C [13] . The vegetation consists of shrubby and grassy savannah with forest galleries along the watercourses [13] . The soils are of tropical ferralitic and ferruginous types [15] .

In sub-Sudanian savannah, the protected areas of Kahanso (7400 ha) located at latitude 9˚17' North and longitude 7˚37' West, and Kouassi-Ndawa located at latitude 8˚13' North and longitude 2˚90' West (300 ha), constitute the sampling sites. The Kahanso site, which is poorly preserved, is typified by the presence of grazing. On the other hand, Kouassi-Ndawa site deprived of all human activities is a well preserved site. These sites are characterized by open forests and wooded savannahs that stretch over ferralitic soils [15] . The climate is Sudano-Guinean, under the influence of the harmattan, with two seasons, a rainy (4 to 5 months) and a dry (7 to 8 months). Annual precipitation ranges from 1200 to 1600 mm [14] and occurs mostly during July to September. The temperature varies from 16˚C to 36˚C with an average of 27˚C [16] .

In Guinean savannah, the Yalo protected area (7˚80' - 8˚10' North latitude and 7˚70' - 8˚12' West longitude) with an area of 26,800 ha and the Moyenne Marahoué protected area in the department of Zuenoula (7˚52' - 7˚80' North latitude and 6˚12' - 6˚30' West longitude) with an area of 38,800 ha were investigated to this study. The protected area of Yalo is an open anthropized forest (Teak reforestation (Tectona grandis) and grazing). The Moyenne Marahoué site is also subject to pressures such as farming, illegal tree logging and carbonization wich is the important activity practiced by the population in this area. The Guinean savannah is the area of the pre-forest sector made of mesophilic (or semi-deciduous) forests and savannahs. The climate is equatorial with a bimodal regime; it consists of two rainy seasons spaced by two dry seasons [13] . The average annual rainfall is 1200 mm with an average annual temperature of 27˚C [17] . The soils are ferralitic type [18] .

Individuals of P. erinaceus from the same site were considered as one population, provenance or origin. In each population, trees separate at least 100 m from each other, with a diameter of 1.30 m from the ground (dbh) greater than 30 cm [1] [19] were selected for organs collection. Each individual is numbered with paint to prevent it from being inventoried twice, then georeferenced using a GPS system (Geographical Positioning System). A total of 144 fruit-trees of P. erinaceus from the six sites following the climatic gradient were sampled in this study (Table 1).

Fourteen quantitative traits were measured on trees, leaves, fruits and seeds of P. erinaceus [1] [19] [20] . The parameters measured on the productive trees are the diameter at breast height (dbh), the height of the bole and the total height. The heights (total and bole) were measured using a graduated pole. The circumference of the trees was measured with a tape measure. The mensurations on leaves concerned ten leaves per tree. The variables evaluated on the leaves are the length and width of the leaf and the number of leaflets per leaf. On each leaf, three leaflets (basal, medial and apical) were collected for measuring the length and width of the leaflet. The mensurations on leaves and leaflets were made with a graduated ruler and an electronic caliper with an accuracy of 0.01 cm. Fruit characterization was performed on 150 fruits per tree. At the fruit level, height, width and weight were measured. The number of seeds per fruit was counted. For each seed, length, thickness and weight were measured. The length and width of the fruit and seed were measured using an electronic caliper with an accuracy of 0.01 cm. Fruits and seeds were weighed with a 0.1 mg electronic precision scale. The total traits which is using in this study are summarized in Table 2.

Several statistical methods were used to evaluate and analyse the structure and distribution of phenotypic variability of P. erinaceus from the collected data. The modalities of the different parameters have been codified for the statistical analyses. The coded data were analyzed using XLSTAT Ecology software version 2018.5, according to the six sites of the three types of savannah. A descriptive statistic was performed to analyze all the variables of the six sites in order to highlight the mean, standard deviation, coefficient of variation (CV), maximum and minimum. The relationships between the quantitative traits were analyzed by submitting a data at a matrix correlation in order to determine the coefficients of correlation. A transformation of the reduced centred variables was performed for multifactor analysis followed by hierarchical ascendant classification (HAC). The discriminant function analysis (DFA) was carried out to structure the P. erinaceus of the trees according to morphological traits. This analysis also revealed the most discriminating traits. A dendrogram from the HAC was constructed using the Unweighted Pair-Group Method with arithmetic Average (UPGMA). A multiple analysis of variance was performed to compare all discriminating morphological traits together then one way analysis of variance (ANOVA 1) were carried out using the means of the discriminating variables to compare the different groups constituted. Significant ANOVA tests were followed by the multiple pair comparison (LSD) Fisher test at the 5% significance threshold.

The results of the descriptive statistics carried out with all the morphological traits are in Table 3. Of the fourteen morphological traits, seven (diameter at breast height, bole height, leaf length and width, leaflet length and seed length) showed a fairly high variability with CV varying from 15% to 30%. The fruit weight is the most variable trait with CV upper than 45% whereas the fruit height and width are the less variable with CV lower than 10%. The number of leaflets per leaf and seed thickness showed moderate variability (10 < CV < 15).

The multivariate analysis of variance (MANOVA) for the parameters of the identified groups showed very highly significant difference (F = 145.2; p = 0.038). The results of the analysis of variance (ANOVA) for each of the four discriminating traits by factorial weight showed a very highly significant difference (Table 6). Of the four distinctive traits, two (leaf width and the number of leaflet per leaf) revealed a complete distinction among the three groups. Regarding leaf width, group 1 has wide leaves (19.61 ± 2.14 cm) than those in group 3 (16.44 ± 2.23 cm) which are wide than those in group 1 (14.87 ± 1.32 cm). In terms of the number of leaflets per leaf, trees in group 2 have a high number of leaflets (10.48 ± 0.90) than those in group 3 (9.64 ± 1.25) which in turn are smaller than those in group 2 (8.90 ± 0.94). Two characteristics (fruit weight and height) revealed a partial distinction among the three groups. Thus, trees in group 1 and 2 are larger (0.65 ± 0.04 g and 0.65 ± 0.07 g respectively) than those in group 3 (0.27 ± 0.04 g). For fruit height, Group 1 and 2 trees have big fruits (69.48 ± 2.82 mm; 68.44 ± 2.68 mm respectively) than those in group 3 (64.64 ± 4.29 mm).

Descriptive statistics made from the fourteen morphological traits of P. erinaceus from savannahs of Côte d’Ivoire revealed significant variability for some of the traits studied such as height of the drum, diameter at breast height, weight of the fruit and weight of the seed. This variability could be related mainly to genotypic factors and to a lesser extent to anthropogenic and environmental factors. There is a partial correlation between morphological variables and genetic data in trees [19] [20] [21] . Indeed, morphological characters are markers that reveal important variations between individuals of the same trees population. Thus, they enable to highlight genetic variation related to geographical origin [22] . The variation in morphological traits should be taken into account in the implementation of conservation measures for trees species. Phenotypic variability is essential for the selection of attractive genotypes for domestication purposes [23] . Intra-specific genetic variation is needed for “plus-trees” based on a large gene pool, assurance of species’ ability to adapt to changes in the environment, to avoid the loss of interesting phenotypes for both local populations’ uses and for wood machining. No scientific large-scale reforestation studies with P. erinaceus have yet been undertaken in Côte d’Ivoire. Through present study, some P. erinaceus trees of attractive and contrasting phenotypes should be used as “plus-trees” for the production of seedlings for the implementation of the species in agroforestry systems for its conservation.

Discrimant factorial analysis and Ward’s ascending hierarchical classification grouped the P. erinaceus individuals sampled in the six studied sites into three morphological groups. Of the fourteen traits used for the distinction among the 144 samples, four were found to be discriminatory. Of the four, two (fruit weight and height) enabled partial distinction among the three groups and the other two (leaf width and number of the leaflet) enabled complete distinction among the three groups. These four morphological descriptors could be used as markers to identify phenotypic groups among individuals of P. erinaceus. This differentiation of P. erinaceus trees could be due to hereditary (genetic) and environmental (origin) factors. These results are consistent with those obtained by [24] on P. santalinus and [25] on Adansonia digitata. Apart from the probable effect of ecological factors on the differential vegetative development of the analyzed population of P. erinaceus, the spatial structuring of morphological diversity could be accentuated by the impact of human pressures (farms and pasture) [20] . In addition, leaf width and number of the leaflet traits can be used as a selection index for an improvement program to increase leaf size for forage production. Based on these results, it is preferable to make a selection from group 1 of P. erinaceus consisting of Kouassi-Ndawa trees for a valuation of the species. Also, individuals from the three contrasting phenotypic groups could be chosen as seed producers to produce plants of different origins and phenotypes for provenance tests in the three types of savannah for varietal selection adapted to each type of savannah for different uses for local populations and industrial exploitation.

The discriminant factorial analysis identified the most discriminating traits for the morphological characterization of the P. erinaceus populations studied. The first component is defined by the fruit weight and height. Leaf width and number of leaflet contributed to the formation of the second component. The analysis of variance of morphological characteristics indicated a differentiation of the populations studied. It results from these comparative analyses that group 1 trees (Kouassi-Ndawa) in the sub-Sudanian savannah have the phenotypic characteristics of high modalities. However, group 2 trees (Moyenne Marahoué) have small leaves with a reduced number of leaflets and large fruits. Group 3 trees (Yalo, Kahanso, Ouarigué and Téguirdouo) have the smallest morphological character values. These results could be explained in part by the difference in anthropogenic action on rural populations and to a lesser extent by climatic conditions [25] [26] . It should be noted that the most sought-after aspect of P. erinaceus in Côte d’Ivoire is its production of wood (construction and export) and leaves (livestock feed). The discriminating traits identified constitute selection markers for routine identification of phenotypes adapted to different savannah ecosystems in Côte d’Ivoire.