Ten couples of parental and regenerated Tall coconut accessions planted at 143 trees per hectare have been used (Table 1). Adult palms aged between 9 and 23 years from International Coconut Genebank for Africa and Indian Ocean (ICG-AIO) located in Marc Delorme research station were assessed. Genebank is located between 5˚14' and 5˚15' of North latitude and between 3˚54' and 3˚55' of West longitude.

The parental plant material (G0), initially introduced in ICG-AIO from 1968 to 1975 consisted of 10 tall coconut accessions planted in six collection plots (N˚081 planted in 1968, plot N˚091 planted in 1969, plot N˚101 planted in 1970, plot N˚102 planted in 1975, plot N˚111 planted in 1970 and plot N˚112 planted in 1972). The parental accessions constitute the reference population in the study and the ancestors of the regenerates. The accessions were introduced into Côte d’Ivoire as seed from five regions: Africa, South Pacific and Far East.

The regenerated plant material (G1) consisted of 10 regenerated accessions of Grand coconut. It was obtained by the controlled pollination method from parental accessions already present in Côte d’Ivoire and planted on 2 plots (plot N˚081 planted in 2002 and plot N˚091 planted in 1988) [11]. The crosses carried out within each of the parental accessions to obtain the regenerated ones are of type G0 × G0. The WAT accession was used as control in the trials at each parental (G0) and regenerated (G1) generation.

The climate is equatorial type with two humid and dry seasons (Figure 1). Monthly mean of rainfall distribution during decades that preceded field genebank regeneration and from 1988 to 2010 remains similar (Figure 1).

For each character, 30 or 300 coconut palms were observed respectively per accession and generation. Data about parental accessions were provided from Coconut Data Management (CDM) software of the Marc Delorme station. Regenerated coconut palms were observed from July 2009 to July 2011. Agromorphological data was recorded according to methods described [12] [13]. Twenty-six quantitative characters appearing in the coconut’s descriptor and selected in long list of coconut descriptor study in detail by Yao were assessed on the stem (3), leaf (6), inflorescence (5) and the fruit (12) [14] [15].

On the stem, the stem girth at 20 cm above soil level (C20), the stem girth at 150 cm from the ground (C150) and the number of leaf-scars between 1 and 2 m from the ground (NLS) were measured.

Leaf measures concerned the leaves of rank 24. So the rachis length (RL), petiole length (PL), number of leaflets down one side of the leaf (NL), leaflet length (LL), leaflet width (LW) were measured on the leaf. The theoretical leaf area was calculated according to the mathematical formula (LA = 2 × NL × LL × LW) stated by [16].

Floral measures were carried out in inflorescences raised by rank 24 leaves. The length of inflorescence axis (ILA), peduncle length (ILP), spikelet length (ILS), distance between point of spikelet insertion and insertion of the first fruit (ISFFD) and the number of spikelets (INS) were assessed.

Concerning fruits traits, Fruit Components Analysis (FCA) and yield characters were assessed. On the Four mature nuts were collected per bunch from one palm and stored for 15 days to achieve maturation stages before analysis of their components at laboratory. The weights of the whole fruit (FW), husk (HW), husked nut (HNW), shell (SW), endosperm (EW), water (WW) and the ratio of copra per nut (Q) were analyzed. Yield characters were also determined including the copra weight per nut (CNW = dry endosperm weight × 100/94). Fresh meat oil content (FMO) and dry meat oil content (DMO) were assessed based on weight of total oil extracted using Soxhlet method. While, the number of bunches per palm per year (NBPY) and the number of fruits harvested per palm per year (NFPY) were determined at the field.

Weighted mean data of all characters collected from the Tall coconut accessions were used in all analyses. Means of raw data of each character were transformed from geometric weighting method [17] [18]. This transformation based on the performance of the experimental control West African Tall (WAT) was performed per generation. It permitted to minimize morphological gaps due to environment effects appeared between accessions planted on distinct plots during one generation. Indeed, WAT is considered as control in all experimental plots of Tall coconut palms at the Marc Delorme station. For each character, weighted means were done following relation [18]:

X ¯ i , k = ( X ¯ C , k × X ¯ i j , k ) / X ¯ C i j , k (1)

where X ¯ i , k is weighted mean of the character i in generation k for one accession, X ¯ C , k mean of the character i of the WAT control of all plots in generation k, X ¯ i j , k raw value of the character i in plot j and generation k for one accession and X ¯ C i j , k mean of the WAT control for the character i in plot j and generation k.

Principal Component Analysis (PCA) and Multiple Analysis of Variance (MANOVA) were performed using Statistica software v.7.1 [19]. PCA was conducted to identify divergence factors between G0 and G1 accessions generation. Likewise, MANOVA was achieved to reveal characters that discriminate generations G0 and G1 at 5% likelihood. Prior, the equality of variances of studied characters in accessions per generation was verified.

Principal Component Analysis (PCA) achieved from the 26 agromorphological traits identified 3 main factors of divergence between parental and regenerated accessions (Table 2). The first three factors of PCA expressed together 69.95% of total variation.

Factor 1 explained 35.22% of the total variability (Table 2). Considering correlation values between axes and traits exceeding 0.7, Factor 1 was negatively

Variables mostly correlated to the components are shown in bold.

correlated with petiole length (PL; r = −0.82) and weights of whole fruit (FW; r = −0.93), husk (HW; r = −0.76), husked nut (HNW; r = −0.82), shell (SW; r = −0.93), water (WW; r = −0.74), endosperm (EW; r = −0.89) and coprah per nut (CNW; r = −0.8). Oil content in fresh meat (FMO) was positively correlated (r = 0.72) with Factor 1. Axe 1 was interpreted as divergence factor following components of the nut.

Factor 2 described 20.17% of the variance (Table 2). It was positively correlated with stem girth at 20 cm above soil level (C20; r = 0.73) and stem girth at 150 cm height (C150; r = 0.8) and negatively with number of leaf-scars (NLS; r = −0.81). It was defined as the divergence factor describing height growth and stem vigour.

Factor 3 explained 14.56% of the total variance (Table 2). It was positively correlated with number of bunches per palm per year (NBPY; r = 0.73) and number of fruits harvested per palm per year (NFPY; r = 0.79). Axe 3 was considered as divergence factor according to bunch and fruit yields.

The analysis of the parental and regenerated accession couples distribution has been achieved on the two plans formed by the Factors 1 - 2 and 1 - 3. The divergence Factors 1 and 2 explained together maximum of variability of 55.39% (Figure 2). They permitted to distinguish the two accessions generations. In the plan 1-2, majority of the parental accessions G0 were located in the negative part of the axis 1 and positive part of the axis 2.

Regenerated accessions G1 were located in the positive part of the axis 1 and negative part of the axis 2. So, the regenerated accessions diverge to the parental ones according to a low weight of nut components. Likewise, the regenerated accessions G1 are characterized by low stem vigour and less quickly growth.

The divergence Factors 1 and 3 explained together 49.78% of the total variability (Figure 3). Differentiation between the two Tall coconut generations G0 and G1 was less clear in plan 1-3.

Multiple analysis of variance (MANOVA) showed that only 8 (31%) of the 26 characters assessed, discriminate (F = 254.3; p = 0.049) G0 and G1 generations. Thus, characters that allowed a complete distinction of the two generations of Tall coconut palms are those describing bulb and stem girths, number of leaf-scars, petiole length, leaflet width, theoretical leaf area, the annual yield of bunches and nuts per palm (Table 3). That represents 100%, 50% and 17% of characters measured respectively on the stem, leaf and nut. In the majority, characters estimated essentially on the inflorescence and the fruit components didn’t vary significantly after one regeneration cycle.

N: number of accession per generation; p: probability value of t-test; t: t-test statistical value; G0: parental population of coconut palm; G1: first cycle of regenerated population of the coconut palm.