Principal component analysis and agglomerative clustering of morphological variations in two different species of oil palm (Elaeis oleifera and Elaeis guineensis)

Introduction

Oil palm (Elaeis sp.) is a perennial diploid (2n=32) monocotyledon belonging to the family Arecaceae and sub-family Cocosoideae (Cronquist, 1981) [11] . Oil Palm has three accepted species E. guineensis (The African Oil palm), E. oleifera (The American Oil palm) and E. odora. Among these only E. guineensis and E. oleifera are commercially cultivated. At present the palm exists in a wild, semi-wild and cultivated state in three main areas of the equatorial tropics: Africa, South-east Asia and South & Central America, with Africa or South-American being probable center of origin of the palm (Corley and Tinker, 2016) [10] . As a crop, oil palm has gained wider acceptance than other oleaginous crops of the tropical belt, being by far the highest oil yielder per unit area (Sundram et al., 2003; Corley & Tinker, 2016) [28, 10] . Unlike traditional oilseeds, the fruit mesocarp of oil palm contains a large reserve of storage lipids yielding palm oil, while the nut of the fruit provides palm kernel oil (Yusoff, 2024; Basiron, 2007) [31,6] . The lipid composition, fatty acid profile, and vitamin content (A and E) of oil palm confer on the crop a distinct nutritional and commercial value (Sundram et al., 2003; Ng et al., 2012) [28,20] . Increasing attention is also being given to the use of palm oil as biodiesel, owing to its superior yield compared to other crops (Basiron, 2007; Ong et al., 2011) [6, 21] . Consequently, oil palm has been recently introduced in India as an irrigated crop, and systematic breeding programs have been initiated for its genetic improvement (Corley & Tinker, 2016) [10] . But, due to its allogamous and highly heterozygous nature, each oil palm is considered a distinct genotype, resulting in a relatively narrow genetic base (Rival & Parveez, 2010; Hayati et al., 2004) [25,13] . For a successful breeding programme, the assessment of genetic diversity is essential (Bakoumé et al., 2015; Cochard et al., 2009) [3,9] . Therefore, an attempt has been made to study morphological variations in 40 oil palms [23 [27,30] .

Materials and Methods

The experimental materials comprised 40 oil palms drawn from three different accessions: 240D × 281D and 80D × 281D (representing 17 E. guineensis palms) and oleifera-I (representing E. oleifera palms). These palms were selected from the oil palm gardens at the Indian Institute of Oil Palm Research (IIOPR), Pedavegi, Andhra Pradesh, and Palode, Kerala. All accessions were indigenously developed and maintained at Palode, Kerala. Each palm was evaluated for 14 morphological parameters: 1. palm height (cm), 2. annual plant height increment (cm/ year), 3. girth at 25 cm above ground level (AGL) (cm), 4. girth at 50 cm AGL (cm), 5. sex ratio (number of female inflorescences: total inflorescences), 6. number of developing fruit bunches (DFBs), 7. number of leaves, 8. number of spindles, 9. petiole depth (cm), 10. petiole width (cm), 11. number of leaflets, 12. rachis length (cm), 13. leaflet length (cm), and 14. leaflet width (cm). Observations were recorded following the standard descriptors for oil palm as outlined by the International Plant Genetic Resources Institute (IPGRI, 2000) [14] . The collected data were subjected to statistical analysis to assess the extent of morphological diversity among palms. Principal Component Analysis (PCA) was employed to identify the most informative traits contributing to total variation, while Agglomerative Hierarchical Clustering (AHC) was applied to group palms into distinct clusters based on morphological similarity. Both analyses were performed using XLSTAT (Addinsoft, 2022) [2] , an advanced statistical software integrated with Microsoft Excel. PCA was conducted following the methodology described by Jolliffe (2002) [15] , where eigenvalues and eigenvectors of the correlation matrix are used to extract the main components. Cluster analysis was performed using Euclidean distance and Ward's minimum variance method, following standard protocols in multivariate analysis (Kaufman & Rousseeuw, 2005) [17] . PCA and clustering are wellestablished tools for diversity studies in crops (Jolliffe & Cadima, 2016) [16] and have been effectively applied to oil palm germplasm (Sujadi et al., 2019; Yaakub et al., 2023) [27,30] .

Results and Discussion

Morphological Variability

The evaluated oil palm genotypes showed considerable variation in growth, reproductive, and leaf traits as is presented in Table 1. Palm height ranged from 134 to 471 cm (mean = 293.3 cm), with annual increments of 16.75-37.38 cm/year (mean = 24.29 cm/year), indicating differences in growth vigour. Stem girth averaged 304.97 cm (25 cm AGL) and 296.93 cm (50 cm AGL), reflecting robust vegetative development. Similar variation in height and girth has been reported in oil palm germplasm studies, highlighting their importance as indicators of growth and yield potential (Martin et al., 2022) [19] Reproductive traits were also highly variable: the sex ratio spanned 0-1 (mean = 0.21), and developing fruit bunches ranged 0-12 (mean = 3.38). Such variation is critical, as lower sex ratios generally favour higher proportions of female inflorescences, which directly enhances bunch production (Corley & Tinker, 2016) [10] . Leaf architecture also differed considerably, with palms averaging 34 leaves, 305.6 leaflets, and a mean rachis length of 527.9 cm. Leaflet size averaged 105.1 × 6.3 cm, while petiole depth and width averaged 5.2 cm and 8.6 cm, respectively. Variability in leaf number, rachis length, and petiole size has been linked to differences in canopy efficiency and light interception, which are key determinants of yield (Samsudin et al., 2025) [26] . [12,10,23] . The two girth measures were themselves highly correlated (r = 0.949), confirming measurement reliability. Both were also strongly linked to rachis length and leaflet width, underlining their importance as integrative indicators of vegetative robustness. Petiole width and depth were similarly associated with rachis and leaflet traits, supporting previous findings that petiole architecture underpins canopy expansion (Balakrishna et al., 2017; Barcelos et al., 2015) [4,5] . Leaflet traits were highly correlated with rachis length (r = 0.600-0.886), confirming their role as drivers of canopy photosynthetic capacity. Studies of germplasm collections have likewise emphasized leaflet number and size as central to canopy development and yield efficiency (Camacho-Villalobos et al., 2021) [8] .

Reproductive traits showed contrasting associations. Sex ratio was negatively correlated with palm height, girth, rachis length, and leaflet traits, while number of developing fruit bunches (DFBs) correlated negatively with height (r = -0.577), rachis length (r = -0.533), and leaflet size (r = -0.611 to -0.515). These patterns suggest a resource allocation trade-off, whereby greater vegetative vigour may occur at the expense of reproductive efficiency. Such trade-offs have been observed in perennial crops including oil palm (Breure & Corley, 1983; Weiner, 2004; Osorio-Guarín et al., 2019) [7,29,22] . Leaf number correlated positively with spindle number (r = 0.536) and DFBs (r = 0.432) but negatively with rachis length (r = -0.373), suggesting independence between leaf quantity and size. This distinction has been highlighted in crop physiology, where canopy productivity is shaped by both production rate and individual leaf dimensions (Legros et al., 2009) [18] . Plant height increment per year correlated modestly with sex ratio (r = 0.346), suggesting that faster-growing palms may invest proportionally more in reproductive effort, echoing developmental studies of palm flowering (Adam et al., 2011) [1] .

Principal Component Analysis of Trait Variability

The PCA was statistically supported (Bartlett's χ² = 516.85, df = 91, p < 0.0001; KMO = 0.774), confirming suitability of the dataset for multivariate analysis (Camacho-Villalobos et al., 2021) [8] . Four components with eigenvalues > 1 were retained, explaining 81.36% of the variance. Table-3 enumerates eigenvalues, variability% and cumulative variability explained by all 14 factors computed. Figure-1 represents the same graphically in the form of scree plot. PC1, accounting for 50.64% of the variance, was dominated by structural traits: palm height, girth, rachis length, petiole width, and leaflet size. These traits clustered as a "structural vigour" dimension, reflecting canopy and trunk robustness. Similar patterns were reported in oil palm and coconut, where leaf and stem architecture consistently dominate the first PCA axis (Perera et al., 2010; Popet et al., 2022) [, 23] . PC2 explained 13.53% of the variance and was defined by number of leaves and spindles. This "leaf production" axis separated palms with high canopy turnover from those with fewer leaves, reinforcing the importance of leaf production rate as a predictor of yield potential (Corley & Tinker, 2016; Legros et al., 2009) [10,18] . PC3 (9.74%) captured a growth-reproduction trade-off, with positive loadings for plant height increment and sex ratio. This axis reflects alternative allocation strategies among genotypes: those investing in rapid vegetative elongation exhibited distinct reproductive patterns, consistent with allocation theory in perennial crops (Weiner, 2004; Osorio-Guarín et al., 2019) [29,22] . PC4 (7.45%) highlighted variation in leaflet number and DFBs, suggesting that high leaflet proliferation may be linked with reduced reproductive efficiency. Comparable findings were observed in hybrid characterization studies, where foliage traits contributed to variation in bunch production (Research Square, 2025) [24] . [5,8] .

Overall, the integration of correlation and PCA analyses demonstrates that structural vigour traits dominate morphological variability, while reproductive parameters define important trade-offs. These findings align with multivariate studies in oil palm (Balakrishna et al., 2017; Popet et al., 2022) [4, 23] and support the use of PCA to identify elite genotypes combining canopy development with reproductive balance. By streamlining complex datasets into biologically meaningful dimensions, PCA reinforces its utility as a breeding tool, enabling the selection of palms with desirable trait combinations for yield improvement (Corley & Tinker, 2016; Research Square, 2025) [10,24] .

References

1. Adam H, Jouannic S, Morcillo F, Richaud F, Duval Y, Tregear JW. Determination of flowering in the oil palm: analysis of developmental processes and molecular regulation. Tree Physiol. 2011;31(4):335-350.
2. Addinsoft. XLSTAT statistical and data analysis solution. New York: Addinsoft; 2022.
3. Bakoumé C, Wickneswari R, Siju S, Rajanaidu N, Kushairi A, Billotte N. Genetic diversity of the world's largest oil palm (Elaeis guineensis Jacq.) field genebank accessions using microsatellite markers. Genet Resour Crop Evol. 2015;62(3):349-360.
4. Balakrishna P, Ravi M, Kumaran PM, Nandakumar PBA. Genetic diversity in oil palm genotypes assessed by multivariate analysis in India. Int J Curr Microbiol Appl Sci. 2017;6(8):2102-2112.
5. Barcelos E, Rios SA, Cunha RNV, Lopes R, Motoike SY, Babiychuk E, et al. Oil palm natural diversity and the potential for yield improvement. Front Plant Sci. 2015;6:190.
6. Basiron Y. Palm oil production through sustainable plantations. Eur J Lipid Sci Technol. 2007;109(4):289-295.
7. Breure CJ, Corley RHV. Fertilizer requirements of oil palm in Malaysia. Planter. 1983;59:439-459.
8. Camacho-Villalobos E, Espinoza-Mendoza E, VásquezVásquez C, Del Castillo C. Morphological and molecular characterization of an Elaeis oleifera germplasm collection in Ucayali, Peru. PLoS One. 2021;16(4):e0250445.
9. Cochard B, Amblard P, Durand-Gasselin T. Oil palm genetic improvement and sustainable development. OCL. 2009;16(1):49-54.
10. Corley RHV, Tinker PB. The Oil Palm. 5th ed. Oxford: Wiley-Blackwell; 2016.
11. Cronquist A. An integrated system of classification of flowering plants. New York: Columbia University Press; 1981. p. 1033-1035.
12. Hardon JJ, Turner PD, Tan GY. Breeding and selection of the oil palm. Biol J Linn Soc. 1985;23(1):11-19.
13. Hayati A, Wickneswari R, Maizura I, Rajanaidu N. Genetic diversity of oil palm (Elaeis guineensis Jacq.) germplasm collections from Africa: implications for improvement and conservation of genetic resources. Theor Appl Genet. 2004;108(7):1274-1284.
14. International Plant Genetic Resources Institute (IPGRI). Descriptors for oil palm (Elaeis guineensis). Rome: IPGRI; 2000.
15. Jolliffe IT. Principal Component Analysis. 2nd ed. New York: Springer; 2002.
16. Jolliffe IT, Cadima J. Principal component analysis: a review and recent developments. Philos Trans R Soc A Math Phys Eng Sci. 2016;374(2065):20150202.
17. Kaufman L, Rousseeuw PJ. Finding groups in data: an introduction to cluster analysis. New York: Wiley; 2005.
18. Legros S, Mialet-Serra I, Caliman JP, Siregar FA, ClémentVidal A, Fabre D, et al. Phenology and growth adjustments of oil palm (Elaeis guineensis) to photoperiod and climate variability. Ann Bot. 2009;104(6):1171-1182.
19. Martin G, et al. Morphological characterization for genetic diversity assessment in oil palm. Front Plant Sci. 2022;13:918304.
20. Ng TL, Choo YM, Ma AN. The prospects of increasing the value of palm oil for food. Lipids. 2012;47(4):349-355.
21. Ong HC, Mahlia TMI, Masjuki HH, Norhasyima RS. Comparison of palm oil, Jatropha curcas and Calophyllum inophyllum for biodiesel: a review. Renew Sustain Energy Rev. 2011;15(8):3501-3515.
22. Osorio-Guarín JA, Hormaza JI, Mejía JA, Alvarado G, de la Hoz JF, Torres GA. Genome-wide association study (GWAS) for morphological and yield-related traits in an oil palm OxG hybrid population. BMC Plant Biol. 2019;19:533.
23. Popet P, Amnouychokanant T, Chatchawankanphanich O, Maneechansook W. Principal component analysis of agronomic and yield traits of six tenera oil palm progenies in Southern Thailand. KMITL Sci Technol J. 2022;22(2):110-120.
24. Research Square. Morphological characterization and genetic divergence of 10 oil palm hybrids. Res Sq. 2025. https://doi.org/10.21203/rs.3.rs-6704891/v1
25. Rival A, Parveez GKA. Oil palm genetics and breeding. In: Corley RHV, Tinker PB, editors. The Oil Palm. Oxford: Wiley-Blackwell; 2010. p. 85-120.
26. Samsudin S, et al. Genome-wide association study identifies SNP markers linked to compactness traits in oil palm. Euphytica. 2025. https://link.springer.com/article/10.1007/s10681-02503502-x
27. Sujadi S, Wandita TS, Supena N, Yenni Y. Genetic distance of 47 accessions of oil palm (Elaeis guineensis Jacq.) germplasm from Cameroon based on morphological characteristics. J Penelit Kelapa Sawit. 2019;27(1):25-40.
28. Sundram K, Sambanthamurthi R, Tan Y-A. Palm fruit chemistry and nutrition. Asia Pac J Clin Nutr. 2003;12(3):355-362.
29. Weiner J. Allocation, plasticity and allometry in plants. Perspect Plant Ecol Evol Syst. 2004;6(4):207-215.
30. Yaakub Z, Marjuni M, Mustaffa S, Tun Mohd Salim WNS, Fatin Mohd Nasir A, Abdullah N, et al. Phenotypic variability of Elaeis oleifera germplasm revealed by principal component and cluster analyses. J Oil Palm Res. 2023;35(4):582-595.
31. Yusoff MM. Palm oil and palm kernel oil: chemistry, nutritional properties, and uses. Front Nutr. 2024;11:128145.