CHAPTER III 
MATERIALS AND METHODS 
3.1 Breeding Materials and Site Location 
The germplasm used in this study originated from Senegal and Gambia. A random 
sample of five families from 13 sites originating from Senegal and also Gambia 
(population; designated as SEN 01, SEN 02, SEN 03, SEN 04, SEN 05, SEN 05, SEN 
06, SEN 07, SEN 08, SEN 09, SEN 10, SEN 11, SEN 12, SEN 13 GAM05.02 and 
GAM05.08) were collected in July-August 1993 by researchers from Malaysian Palm 
Oil Board (MPOB) from the two countries. The germplasm materials were planted at 
the MPOB Research Station Kluang, Johor in 1996. The palms were derived from the 
Independent Completely Randomized Design, where a total number of 415 open- 
pollinated palms were planted in Trial 0.352 (Senegal materials) and Trial 0.357 
(Gambia materials) with two replicates each (41 progenies in replication I and 15 
progenies in replication 2). For the purpose of this study, available quantitative data 
on the Senegal and Gambia germplasm were collected at the MPOB headquarter. 
3.2 Data Collection 
The pertönnance of progenies for yield and oil yield components was assessed based 
on data on fresh fruit hunch (FFB) yield records, bunch number (BNO) and average 
hunch weight (ABW), bunch components, vegetative and physiological characters and 
fatty acid characters. 
28 
3.2.1 Yield and Yield Components 
Harvesting of oil palm usually begins at 36 months after field planting with 
subsequent operations carried out at regular intervals of seven to ten days, i. e. three 
rounds in a month. Data on yield and yield components were evaluated from year 
2000 - 2007. 
Procedure for Yield Recording 
Recordings of bunch weight (BWT) and bunch number (BNO) on individual are 
carried out during the harvesting rounds in oil palm breeding. Fresh bunch weight 
(FFB) is the sum of the bunch weight (BWT) while BNO is the total of all the bunch 
counts and average hunch weight (ABWT) is the quotient between FFB and BNO. 
The yield components were derived as follows: 
FFB (kg/p/yr) = 
r; '=1 BjVT, 
BNO (bunches/p/yr) = 
r; '=1 BNO; 
ABWT (kg/p/yr) = FFB/BNO 
Where n is the number of harvesting rounds. 
3.2.2 Bunch Analysis 
Quantitative data on the bunch components were evaluated in the 2001 until 2006. 
Bunch and fruit components were determined using the bunch analysis technique 
developed by Blaak et al. (1963). 
29 
Procedure. for Bunch Analysis 
Samples of three to five bunches from each palm were analysed. Each bunch was 
weighed and fruit hearing spikelets chopped off the stalk. The spikelets were 
randomly segmented for analysis of the fruit to bunch (F/B) and fruit compositions 
(FC). To ease picking, the F/B portions were kept for two days and the weight of 
empty spikelets. fertile and parthenocarpic fruits were recorded. Analysis of the fruit 
compositions was continued following spikelet sampling on the same day. Fruits were 
detached and randomly sampled, counted, weighed and scraped. The nuts were crack 
opened and kernel weighed. A 5g sample of minced mesocarp was oil extracted in 
hexane for 18-24 hours using soxhlet apparatus. The amount of oil in the sample was 
calculated on weight difference before and after extraction. The components of the 
hunch were calculated using the below formulae: 
FB = Fruit to Bunch (%) = [(FFWT + PFWT) / SWT] x[ (BWT -STKWT) / BWT] x 100 
P/F = Parthenocarpic to Fruit (%) = PFWT / (FFWT + PFWT) x 100 
M/F = Mesocarp to Fruit (%) _ [(FSWT - FNWT) / FSWT] x 100 
MC = Moisture Content of mesocarp(%) =100 -[{(TDMWT-TWT)/(FSWT-FNWT); x 100] 
O/DM = Oil to Dry Mesocarp (%) 
O/WM = Oil to Wet Mesocarp (%) 
O/B = Oil to Bunch (%) 
O'Fi = Oil to Fibre (%) 
K/F = Kernel to Fruit (%) 
MNW = Mean Nut Weight (g) 
MFW =Mean Fruit Weight (g) 
P/B = Parthenocarpic to Bunch (%) 
OPY = Oil per Palm per Year (kg) 
_ [(ETMWT - ETFWT) / (ETMWT - ETWT)] x 100 
=[(100-MC)xO/DM]/100 
_ (F/B x M/F x O/WM) / 10,000 
_ [(ETMWT -ETFWT) / (ETFWT -ETWT)] x 100 
= (KWT / FSWT) x 100 
= FNWT /NOFNUT 
= FSWT / NOFNUT 
= (P/F x F/B) / 100 
= (O/B x FFB) / 100 
30 
KPY = Kernel per Palm per Year (kg) 
Where; 
BWT = Bunch weight 
SWT = Spikelet weight 
PFWT = Parthenocarpic fruit weight 
FSWT = Fruit sub sample weight 
NOFNUT = Number of fresh nut 
TDMWT = Tin + dry mesocarp weight 
ETWT = Extraction thimble weight 
= (K/B x FFB) / 100 
STKWT = Stalk weight 
FFWT = Fertile fruit weight 
ESPKWT = Empty spikelet weight 
FNWT = Fresh nut weight 
KWT = Kernel weight 
TWT = Tin weight 
ETFWT = Extraction thimble + fibre weight 
ETMWT = Extraction thimble + mesocarp weight FFB = Fresh fruit bunch 
3.2.3 Vegetative Measurements and Physiological Characters 
Vegetative characters of the oil palm germplasm were pooled eight years after field 
planting. Frond production was first calculated after the 7th year before other 
parameters were taken a year after, i. e. year 2004. The physiological characters on the 
other hand were assessed based on measurements on collective data of bunch yields 
and hunch quality components, following the methods developed by Squire (1986). 
Data on the physiological characters were assessed in the year 2007. The various 
components of the vegetative and physiological parameters were calculated using the 
below formulae: 
FP = Frond production (no/p/yr) 
PCS = Petiole Cross Sectional Area (em 
RI. = Rachis length (m) 
LL = Leaflet Length 
LW = Lcatlet Width 
LN = Leaflet number (no/p/yr) 
D= Trunk diameter (cm) 
11T = Trunk height (m) 
IITI = Ilcight increment 
LA Leaf Area (m ) 
= number of fronds produced in one year 
= petiole depth x petiole width 
= length from tip of rachis to the first ligule 
=(LLI +LL2+...... LL6)/6 
=(LWI +LW2+.... LW6)/6 
= (number of fronds on one side of rachis)x2 
= diameter of trunk at one meter fromground 
= height of trunk of ground to base of frond 41 
_ (HT at year t) / (age at year t- 2) 
_ [(2; (LL; x LW; ) x LN x 0.57) / 6] / 10,000 
31 
LAR = LeafArea Ratio 
LAI = Leaf Index Ratio 
LDW = Leaf Dry Weight (kg) 
TDW = Trunk Dry weight (kg) 
FDW = Frond Dry weight (kg) 
FI = Frond Index 
I = Fractional interception 
C= Conversion Efficiency (g/MJ) 
VDM = Vegetative Dry Matter (kg/p/yr) 
BDM = Bunch Dry Matter (kg/p/yr) 
TDM = Total Dry Matter 
BI = Busich Index 
NAR = Net Assimilation Rate 
TOIL = Total Oil (k(, /p/yr) 
TEP = Total Economic Product (kg/p/yr) 
Where: 
_ (FP x LA) / VDM 
= (40 fronds x LA x PD )/ 10,000 
= (0.1023 x PCS) + 0.2062 
= 3.142 x (D/2)2 x (I-1T / Age in year) x 1,000 x 0.17 
= FP x LDW 
= LA/ LDW 
=1- exp (_0.47) - (LM -0 
3) 
=TDM/(31 x f) 
_ (FDW + TDW) 
= 0.53 x FFB 
_ [(VDM + BDM) x PD'] / 1,000 
= BDM / (BDM + VDM) 
= TDM / (0.52 x LAI) 
= OPY + (KPY x 0.5*) 
= OPY + (KPY x 0.6") 
PD = Planting Density 
*0.5 = Kernel Oil Extraction Rate 
"0.6 = Relative Price of Palm Kernel Oil to Price of Palm Oil 
FFB = Fresh Fruit Bunch (kg/p/yr) 
OPY = Oil per Palm Year (kg) 
KPY = Kernel per Palm per Year (kg) 
3.2.4 Fatty Acid Traits 
Fatty acid traits were pooled between years 2000 - 2007. The fatty acid composition 
was evaluated using the method proposed by Timms (1978) for routine analysis using 
gas chromatography. 
Procedure. for Fatty Acid Analysis 
The tatty acid analysis was carried out using MPOB method Timms (1978). Fresh 
bunches were collected from the palm, weighed and chopped. Spikelets samples were 
selected one each from apical, middle and basal portion of the bunch. The samples 
32 
were sterilized and fi-uits from the sterilized spikelets separated. The mesocarps were 
separated and oven dried at 105 °C for a minimum of 3 hours. The dried mesocarp 
were minced and dried. After mincing, the dried mesocarp was diluted with 300m1 
hexane and filtered with a spoonful of sodium sulphate. The filtrate was sealed and 
kept in dry place. Oil and hexane mixture was later separated into oil and hexane, 
using rotary evaporator by distilling the hexane vapors under vacuum. The extracted 
oil is then poured in vials, labeled and stored for FAC and carotene analysis. 
For the fatty acid methyl determination, 0.05g of the extracted oil was measured and 
1.9m1 of solvent hexane added and homogenized. Sodium methoxide of 0.11111 was 
also added until a cloudy solution is formed. After 10 minutes, the clear portion of the 
methyl ester was pipetted into gas chromatograph vials and covered with Teflon vial 
caps for analysis. The separation of the FAME by the GC machine was under 
programmed conditions. The fatty acids components were represented by the peaks on 
the chromatograpll. 
3.3 Statistical Analyses 
3.3.1 Variability Profile 
The quantitative morphological data collected was arranged in Excel Microsoft word. 
The oil palm accessions were organized according to their family codes. Simple 
descriptive statistics such as mean, standard deviation, standard error, minimum, 
maximum and variance for each collected traits were calculated using SPSS statistical 
tool. This was done in order to know the extent of variation in the germplasm 
accessions. The average of the quantitative data was also standardized to give equal 
33 
weight to all measurements using the following formula (Microsoft Office Excel, 
2007): 
Z=X-, u/Q (1) 
Where, X= Value to standardize 
p= Arithmetic nlcan 
6= standard deviation of the distribution 
3.3.2 Principal Components Analysis (PCA) 
PCA simplifies the complex data by transforming number of correlated variables into 
a smaller number of variables called principal components. The first principal 
component accounts for maximum variability in the data as compared to each 
succeeding component. PCA was analyzed using "THE UNSCRAMBLERRX" 
software (CAMO software version 10.1). Mathematically, PCA involved in the 
decomposition of original data matrix, X, into a structure part and noise part. In matrix 
representation, the model with a given number of components as follows: 
X= 7P1 +E (2) 
where T is the scores matrix, P the loadings matrix (transposed) and E the error matrix. 
The structured part of data is the combination of scores and loadings in which focused 
by user in interpretation of PCA results while the remaining part is called error or 
residual matrix. The uth column of Tand ath row of P is represented by vectors of 1 
and p,, respectively and both are the vector representations of the uth PC. The number 
of PCs is denoted by A. while a is the number of PC such as 1,2,3 up to A. The 
maximum number of PCs (: 1) determined is either 1- I (number of objects - 1) or J 
34 
(number of variables) depending on which give smaller value. Thus, the first scores 
vector and the first loadings vector are called eigenvectors of the first principal 
component. Therefore, each successive component is characterized by a pair of 
eigenvectors for both the scores and loadings (CAMO, 2014). 
In contrast, the residual matrix, E, represent the fraction of variation that cannot be 
modeled well. It other words, it cannot be explained by available PCs, yet, useful in 
the lack-of-fit measurement of model to the original data where small value of E 
indicated the good model and vice-versa (CAMO, 2014). Usually after PCA, the size 
of each component can be measured and represented by eigenvalue. In PCA, the more 
significant the components indicated the larger of their size, thus have larger 
eigenvalue. Therefore, the eigenvalue of a PC is the sum of squares of the scores and 
represented as follows: 
g'I tij2 (3) 
where g<< is the ath eigenvalue. The sum of all nonzero eigenvalues for a data matrix 
equals the sum of squares of the entire data-matrix, so that 
za=1 9" zi=i zj=1 xij, 
where K is the smaller of 1 or J (Brereton, 2003). 
3.3.3 Cluster Analysis 
(4) 
Cluster analysis identities variable which were further clustered into main groups and 
subgroups using Ward's method through the "THE UNSCRAMBLERK'X" software 
(CAMO software version 10.1). The general purpose of cluster analysis is to group 
similar objects or samples into respective classes based on their specified 
;ý 
characteristics or variables and it includes different type of algorithms such as joining 
(tree clustering), two-way joining (block clustering) and k-means clustering. The aim 
of the tree clustering algorithm was to join the objects or samples in each class of 
themselves together using specified distance measures and linkage rules, thus, lornung 
larger cluster by connecting all the objects at the last step known as hierarchical tree. 
The Ward's method (Ward, 1963) used in this study optimizes an objective function; 
that is, it minimizes the sum of squares within groups and maximizes the sum 01' 
squares between groups. Ward's method is similar to the linkage methods in that it 
begins with N clusters, each containing one object, it differs in that it does not use 
cluster distances to group objects. Instead, the total within-cluster suns of' squares 
(SSE) is computed to determine the next two groups merged at each step of' the 
algorithm. The error sum of squares (SSE) is defined (for multivariate data) as: 
I1 
SSE 
ý=i ; -i 
Výý - V)I (5) 
where vii is the jh' object in the i°i cluster and #i is the number of objects in the i'l' 
cluster. 

CHEMOMETRIC ANALYSIS OF OIL PALM (Elaeisguineensis 
Jacq. ) GERMPLASM FROM SENEGAL AND GAMBIA 
Hana Saleh Alhadi Abdarhman 
(Matric No. 3130043) 
Dissertation submitted in partial fulfillment for the degree of 
MASTER OF SCIENCE 
(FOOD BIOTECHNOLOGY) 
Faculty of Science and Technology 
UNIV'ERSITI SAINS ISLAM MALAYSIA 
Nilai 
May 2015 
i 
1p If N, ýý 
DECLARATION OFTHESIS AND COPYRIGHT 
Student's Full Name / 
ýLý1L ý lnJl 
H 
Academic Session / 
Research Title / 
(\, x1 (.! \ \\ýýj t Matric No. / ý67 
CýPýý'IMFkYK >c C", 
vt 'ýý 
`ýý)ýey YýpiaS)ýýýý")1) Setiýrýný avýCý ýoý 
I hereby declare that the work in this thesis/ pro ect paper is my own except for quotations 
and summaries which have been duly acknowledged / 
I acknowledged that Universiti Sams Islam Malaysia reserves the right as follows / 
i_ lU1 3, iLt v; JLII <ýLwyl ýpL. )i wwlý 
I The thcs'is/ postgraduate project paper is the property ofIUnivcrsiti Sains Islam Malaysia / 
t j)llläýýl ýlý IwcaLl IIýý)I1a1 7ý J!. 
The library of tiniversiti Saws Islam Malaysia has the right to publish my thesis/ 
postgraduate project paper as online open access (fullteat) and make copies for the 
purpose of research or teaching and learning only / 
ý` 
ýs-°-t"1i ý'ý,. 
li 9i : ý1a1i 
ý-ý-1i i. ý ,.:. 
Ji y. ý vrJlli : L,,, ýl.. yi ýyl, Ji : u,. v- ýý I 
(J°IS,. Ii ýýJý) 
ý-ýý: -...:. 
(Student's Signature) 
0 14 L) 
............................  ............ (IC No. /Passport No. ) 
I- q- 2cý... 5 
(I)atc) 
(Supeivisbr's Signature) 
rý'ýýý 
rYiol Yp 
.:..................... (Name) 
.................. .... ............... DR. MOHD ýUkýI BIN HASSAN 
Director 
Ins0lute of Nalal Reearoh & ManapernerM pHR11Aq 
urtiveraib seins hlam Malaysia 
i 
iU 
11 
AUTHOR DECLARATION 
I hereby declare that the work in this dissertation is my own except for quotations and 
summaries which have been duly acknowledged. 
Date: 15"Mav, 2015 
S ignaturq 
Name: Hana Saleh Alhadi Abdarhman 
Matric No: 3130043 
Address: P6-1 IA-05 Plaza Indh 
Scpakat Indahl. Kajang 
43000, Selangor 
111 
BIODATA OF AUTHOR 
Hana Saleh Alhadi (Matric No. 3130043) was born on 8"' February, 1988. She is a 
Libyan by Nationality. She is currently living at P6-11A-05 Plasa Indah Apartment 
Sepkat Indah 1, Kajang, Selangor, Malaysia. She obtained her Bachelor of Science 
(Botany) from 7''' October University, Bani Walid, Libya. Being an enthusiastic person, 
with the purpose of exploring knowledge and skill of the field she enrolled in 
September, 2013 for her Master Degree in Food Biotechnology at Universiti Sains 
Islam Malaysia, Bandar Baru Nilai, Malaysia. She can be contacted by email via 
(hs. zbida(i yahoo. com). 
IN' 
ACKNOWLEDGEMENTS 
My first and foremost thanks is to Allah (SWT) who has made all things possible. I 
would like to express my gratitude to my supervisor Emeritus Professor Dr. Jalani 
Sukaimi for the useful comments, remarks and engagement through the learning 
process of this master dissertation. Your advice on both research as well as on my 
career have been priceless. Furthermore, I would like to thank my co-supervisor. Dr. 
Mohd Sukri Hassan for his supervisor on chemometrics for Illy thesis. 
Also, I would like to thank staff of the Faculty of Science and Technology of the 
university. the Dean of the faculty, Prof. Bachok Taib, staff of Food Biotechnology, 
laboratory assistants and technicians of the faculty. I am thankful for their aspiring 
guidance, invaluably constructive criticism and friendly advice during my study. I am 
sincerely gratetül to them for sharing their truthful and illuminating views on a 
number of issues related to the research and career. 
A special thanks to my family. Words cannot express how grateful I am to my mother, 
father, brothers and sisters for all of the sacrifices that you've made on my behalf. 
Your prayer for me was what sustained me thus far. At the end I would like express 
appreciation to the Government of Libya for financial support of my studies. Without 
you, my appreciation would not he completed. 
V 
ABSTRAK 
Anggaran kepelbagaian genetik dan menentukan hubungan antara koleksi adalah 
strategi yang penting untuk memastikan pengumpulan dan penggunaan germplasma 
yang cekap. Bahan germplasma sawit yang dikutip dari Senegal dan Gambia yang 
ditanam di Stesen Kluang MPOB telah dicirikan mengikut kepelbagaian genetik. 
Sebanyak 44 ciri - ciri agronomi germplasma sawit ini telah dinilai melalui statistik 
yang mudah untuk menilai kepelbagaian genetik; dan dua teknik kemometrik (PCA 
dan analisis Muster) untuk mengenal pasti sifat - sifat yang menyumbang kepada 
perubahan keseluruhan dan mengklasifikasikan bahan-bahan itu berdasarkan 
persamaan. Keputusan profil kepelbagaian itu menunjukkan bahawa germplasma 
kelapa sawit Senegal dan Gambia mempunyai kebole hubahan yang rendah kepada 
tinggi untuk pelbagai ciri. Sembilan komponen utama dengan nilai eigen >1 
bersamaan dengan 88% daripada jumlah variasi dengan majoriti variasi itu dikuasai 
oleh PC1. Kebanyakan ciri-ciri, khususnya, BTS, ABW, BWT, MFW, P/B, M/F, 
OY, TEP, TDM, BDM, e, S/F, K/F, 0/B dan K/B menyumbang kepada 
kepelbagaian di antara dan dalam germplasma, yang mana menunjukkan bahawa 
variasi yang luas wujud dalarn bahan germplasma yang telah dikaji. Analisis 
kumpulan Ward adalah berdasarkan keputusan - keputusan PCA yang telah 
dikumpulkan untuk aksesi 42 minyak kelapa sawit kepada enam kelompok, dan 
kelompok-VI mempunyai bilangan tertinggi ahli. Tambahan pula, ia tiada kaitan 
antara kepelbagaian genetik dan asal geografi. Purata ciri-ciri agronomi setiap 
kelompok menunjukkan bahawa kelompok-III mempunyai nilai purata tertinggi pada 
sifat hasilnya. Selain itu, kumpulan - kumpulan kelompok yang mempunyai nilai 
purata yang tinggi bagi ciri-ciri yang diingini boleh dipilih untuk ciri-ciri setiap satu. 
Pengaksesan ini boleh digunakan untuk menghasilkan bahan-bahan kelapa sawit yang 
berhasil tinggi. 
Kata kunci: Kelapa sawit, germplasma, kepelbagaian genetik, kimometrik, analisis 
komponen utama, analisis kelompok. 
vi 
ABSTRACT 
Estimation of genetic diversity and determination of the relationships between 
collections are useful strategies for ensuring efficient germplasm collection and 
utilization. Oil palm germplasm materials collected from Senegal and Gambia 
maintained at the MPOB Kluang Station were characterized for genetic diversity. A 
total of 44 agronomic traits of these oil palm materials were subjected to simple 
statistics to evaluate the genetic variability; and two chemometric techniques (PCA 
and Cluster analysis) to identify the characters contributing to the overall variation and 
classify the materials based on similarity. Results of the variability profile showed that 
the Senegal and Gambia oil palm germplasm exhibited low to high variability for the 
various traits. Nine principal components with eigenvalue >1 accounted for 88 % of 
the total variation with PCI capturing majority of the variation. Most of the traits 
especially, FFB, ABW, BWT, MFW, P/B, M/F, OY, TEP, TDM, BDM, e, S/F, K/F, 
O/B and K/B contributed to divergence between and within the germplasm, indicating 
that wide variation exists in the germplasm materials studied. Ward's cluster analysis 
based on the PCA results grouped the 42 oil palm accessions into six clusters with 
cluster-VI having the highest number of members. Furthermore, there was no 
association between genetic diversity and geographical origin. Means of the 
agronomic traits of each cluster showed that cluster-III had the highest mean value of 
yield traits. Also, the cluster groups having high mean values for desired traits could 
be selected for the traits per se. These accessions could be used to produce high 
yielding oil palm materials. 
Keywords: Oil palm, germplasm, genetic diversity, chemometrics, principal 
component analysis, cluster analysis. 
vii 
ýi UL4 
öýlsi.: YýS äA)911 ýýLa11 ý' jLo. e1 ý , °" ýLuil)iwl ý Lcyýuýl }, : ýL9, Jsýl 11iS 
s9 
MPOB K1Ua11(y AÜ'ý 
44 sLlý 
(CA) 
ýSýysuJý _ ` ) 
i-IIS (PCA) 
, 1; i ýýý911 Js-a'» ( PCA) - 
)I ýsSaJi }; Li , L-- 
ý1< %ti I Jl äa, äJl ýä,. ; ', ) yaLc ä: ýls dll . al'; c 
l >"S . ýyLsi ýs. c)c Lualcý ýjL=1. 
J1 
`j o ,_ 
jll 
MI-W, BW1'. ABW, FIB i olý "14. a11 ý, ýaL.. . 
PC1 ý ý, lýyý ý., lýis 4Lý 88 
K/ B9 O/B, K/F, S/F, BDM, TDM, TE, P OY, MFP/B, 
(CA) 
y, si ýyý, y 5 (PCA) 
. ltyj Jý 
3 äLSli ýi 
, 4ýilL , 
ý,; Iyu . "', Iý ýýI 
ýCl1tlIIOllICiCICS ýýI jq. 
ý I, jý, ll üajý gs l)ýI , j9ý1 : ý', ý, ti 
0. siliboll ülalSil 
ýýJi, Ji ýi 
VIII 
CONTENT PAGE 
Contents 
AUTHOR DECLARATION 
BIODATA OF AUTHOR 
ACKNOWLEDGEMENTS 
ABSTRACT 
ABSTRACT 
ABSTRACT 
CONTENT PAGE 
LIST OF TABLES 
LIST OF FIGURES 
LIST OF APPENDICES 
ABBREVIATION 
CHAPTER 1: INTRODUCTION 
1.1 Problem Statement 
1.2 Aim 
1.3 Research Questions 
1.4 Objectives 
CHAPTER II: LITERATURE REVIEW 
2.1 HISTORY OF THE GENUS ELAEIS 
2.1.1 Evolution of the Genus Elaeis 
2.1 .2 
Morphology and Biology of E. guinec'nsis 
2.1.3 Infra Specific Classification of E. guinc'cnsis 
2.1.4 Ecology and Morphology of E. guineensi. s" 
2.2 PLANT GENETIC RESOURCES 
Page 
I 
11 
111 
X, 
vi 
V11 
Vlll 
X1 
X11 
x iii 
xiv 
I 
I 
3 
3 
3 
4 
4 
5 
6 
7 
9 
13 
ix 
2.2.1 Genetic Diversity 13 
2.2.2 Characterization of Genetic Diversity 14 
2.3 CHEMOMETRICS 15 
2.3.1 Principal Components Analysis (PCA) 16 
2.3.1 Clustering Analysis 18 
2.3.1.1 Types of Clustering 20 
2.4 PREVIOUS STUDIES 2I 
CHAPTER III: MATERIALS AND METHODS 27 
3.1 Breeding Materials and Site Location 27 
3.2 Data Collection 27 
3.2.1 Yield and Yield Components 27 
3.2.2 Bunch Analysis 28 
32.3 Vegetative Measurements and Physiological Characters 30 
3.2.4 Fatty Acid Traits 31 
3.3 Statistical Analyses 32 
3.3.1 Variability Profile 32 
3.32 Principal Components Analysis (PCA) 33 
3.3.3 Cluster Analysis 34 
CHAPTER IV: RESULTS 36 
4.1 Variability Profile 36 
4.2 Principal Components Analysis (PCA) 36 
4.2.1 Scores Plot 38 
4.2.2 Loadings 38 
4.2.3 Scores 42 
4.2.4 Bi-Plot 42 
4.3 Cluster Analysis 45 
4.3.1 Genetic Distance 49 
CHAPTER V: DISCUSSION 50 
5.1 PCA 50 
x 
5.1.1 Scores Plot 
5.1.2 Loadings 
5.1.3 Scores 
5.1.4 Bi-Plot 
5.2 Cluster Analysis 
5.2.1 Genetic Distance 
CHAPTER VI: CONCLUSIONS AND RECOMMENDATIONS 
REFERENCES 
APPENDIX 
51 
51 
53 
54 
54 
56 
57 
59 
65 
X1 
LIST OF TABLES 
Table 1: Extent of Variation 
Table 2: Variability profile of Senegal and Gambian oil palm germplasm 
as analyzed by descriptive statistics 
Table 3: Principal components analysis for Senegal and Gambia oil palm 
germplasm based on 46 agronomic traits 
Table 4: Scores of the 42 Senegal-Gambian oil palm germplasm on the 
extracted PCs 
Table 5: Characteristics means of six clusters generated by Ward's 
cluster analysis based on 46 agronomic traits 
Table 6: Inter cluster distance as analyzed by proximity matrix of squared 
Page 
36 
37 
39 
43 
48 
Euclidean distance 49 
X11 
LIST OF FIGURES 
Page 
Figure 1: Scores plot where PCs I and 2 are orthogonal to each other 18 
Figure 2: Clustering pattern 19 
Figure 3: Scores plot of principal component analysis between percentage 
variance and number of principal components 38 
Figure 4: Scattered diagram of 46 oil palm germplasm traits for first two 
components contributing almost half of the total variability 41 
Figure 5: Scattered diagram of 46 oil palm germplasm traits showing 
correlation to the first two components 41 
Figure 6: Two dimensional ordinations of 42 Senegal-Gambian oil palm 
accessions on principal axes 1 and 2 44 
Figure 7: Bi-plot of 46 oil palm agronomic traits and 42 oil palm accessions 
on PC I and PC 2 44 
Figure 8: The relationship among the oil palm germplasm reflected by 
cluster analysis 46 
x ill 
LIST OF APPENDICES 
Page 
Appendix 1: Proximity Matrix of Squared Euclidean Distance 66 
x1V 
ABBREVIATION 
AB W average bunch weight 
BNO bunch number 
BWT bunch weight 
BDM bunch dry matter 
BI hunch index 
C12: 0 lauric acid 
C14: 0 myristic acid 
016: 0 palmitic acid 
C 16: 1 palmitoleic acid 
C 18: 0 stearic acid 
C 18: 1 oleic acid 
C18: 2 linolcic acid 
C 18: 1 linolenic acid 
C20: 0 arichidic acid 
DIAM diameter of trunk 
radiation conversion efficiency 
/ fractional interception of radiation 
Fr B tTult/hunch 
FFB fresh fruit bunch 
FP frond production 
GAM Gambia 
HT height 
IV iodine value 
KB kernel /hunch 
K/F kernel/fruit 
KY kernel yield 
LA leaf area 
xv 
LAI leaf area index 
LL leaflet length 
LN number of leaflet 
LW leaf width 
MARDI Malaysian agricultural research and development institute 
M/F mesocarp/fruit 
MFW mean fruit weight 
MN W mean nut weight 
MPOB Malaysian palm oil board 
NAR net area ratio 
NIFOR Nigerian institute for oil palm research 
O/B oil/bunch 
O/DM oil/dry mass 
O/WM oil/wet mass 
OY oil yield 
P/B parthenocarpic fruit/bunch 
PCA principal component analysis 
PC(s) principal component(s) 
PCS petiole cross section 
PORIM palm oil research institute of Malaysia 
RL rachis length 
RRS reciprocal recurrent selection 
SEN Senegal 
S/F shell/fruit 
TEP total economic product 
TDM total dry mass 
UPGMA unpair weighted group method using arithmetic averages 
VDM vegetative dry mass 
WHCA Ward's hierarchical clustering analysis 
CHAPTER I 
INTRODUCTION 
Oil palm (Eluic'. s guincen. sis Jacq. ) is a diploid monocotyledon belonging to the family 
Arecaceae. The crop produces two types of oil, namely palm oil and palm kernel oil. 
Palm oil is one of the world's most traded vegetable oils in the international market 
and most widely consumed edible oil (Corley & Tinker, 2003; Rajanaidu, 1994). It 
has been predicted that by the year 2020, the world production of oils and fats will 
increase to 174 million tonnes and palm oil production to 35 million tonnes and that 
by then, palm oil will be the dominant vegetable oil in the world (Jalani, 1998). 
The majority of oil palm plantation and palm oil production is provided by Indonesia 
and Malaysia, as both contribute about 44 % and 41.5 % respectively to the world 
palm oil production (Oil World, 2008). However, the oil palm breeding populations in 
both countries are derived from a narrow genetic pool. Most of the commercial 
planting materials utilized are derived from the Deli duru, which was first introduced 
in Indonesia in 1848 and afterwards in Malaysia (Corley & Tinker, 2003). The narrow 
genetic pool of oil palm resulted into quite a lot of expeditions being mounted by 
researchers in Malaysia to collect oil palm germplasm in Africa and south-central 
America (Rajanaidu et al., 1999). 
Evaluation of oil palm genetic variability is needed for various purposes such as for 
the selection of superior palms and it also serves as a first step towards effective utility 
in breeding programmes (Mandal, 2008). With the dawn of advanced computer 
technologies, it has become possible to study the complex relationship among 
genotypes through chemornetric or multivariate analyses which offers enhanced 
understanding of the structure, predominantly of large germplasm collections 
(Martinez-Calvo et al., 2008). In this view, chemometrics or multivariate statistical 
methods particularly the principal component and cluster analyses have gained wide 
recognition in the evaluation of germplasm materials of many species (Muhammad et 
al., 2013; Ajmal et al., 2013; Doumbia, 2013; Lohani et al., 2012; Deyong, 2011; 
Lacis et al., 2010 and Bozokalfa et al., 2009). 
Chemometric is the science of relating measurements made on a chemical system or 
process to the complex state of the system via multivariate statistical methods 
(Martens & Naes, 1993). One of the main advantages of chemometric method is that it 
is possible to explore complex co-linear multivariate information in a graphic display. 
Principal component analysis (PCA) is the fundamental chemometric method based 
on vector algebra (Martens & Naes, 1993). The main purpose of the method is to 
reduce dimensions of complex multivariate data and to simplify data interpretation by 
finding new orthogonal variables, principal components (PCs), 
describing the variance 
in data. Cluster analysis is used in the categorization of germplasm materials into 
groups based on similarity or dissimilarity (Ariyo, 
1993). 
1.1 Statement Problem 
There is a need to introduce oil palm germplasm as the genetic 
base of oil palm 
industry in Malaysia. The oil palm in Malaysia mostly originating from the four palms 
planted in Bogor in 1848. The germplasm 
introduced needs to be evaluated and 
characterized. 
1.2 Aini 
To study variation in oil palm germplasm by using chemometric analysis 
[principal component analysis (PCA) and clustering]. 
1.3 Research Questions 
i. Is there variation in the oil palm germplasm? 
ii. What are the traits contributing to the variation? 
iii. What is the relationship among accessions in the germplasm? 
1.4 Objectives 
The objectives of the study were: 
i. To determine the level of variation in the oil palm germplasm; and 
ii. To identify and classify the groups of accessions with different genetic diversity 
based on quantitative traits. 
CHAPTER II 
LITERATURE REVIEW 
2.1 HISTORY OF THE GENUS Elaeis 
The genus Elacis is derived from a Greek word elaion, which means oil. The oil palm 
is today grown commercially in South-east Asia, Equatorial America, Africa and 
South Pacific. The high and increasing yields of the oil palm have led to a rapidly 
expanding world industry, now based in the tropical areas of Asia, Africa and 
America (Rajanaidu & Jalani, 1994). Its origin is believed to have been in Africa, but 
the most productive parts of the industry at present are in Malaysia and Indonesia, 
which provide most of the oil entering international trade (Oil World, 2012). As a 
commercial crop. its history is rather short, dating back to 1807 on the West African 
coast where its cultivation was commenced. It came to the East via the island of 
Mauritius in 1848, to Indonesia where four seedlings were planted and from there to 
Singapore Botanical Gardens about 1870 and only then into Malaya in 1875 (Jalani, 
2012). The Family Palmae or Arecaceaae, where the genus Elacis belongs is 
considered as old as any other family of flowering plants with fossils discovered in 
Cretaceous rocks dating back some 120 million years (Purseglove, 1972). Many plant 
taxonomists believe it to he the first monocotyledon to have branched out from 
primitive dicotyledonous stock and as such, the progenitor of all monocotyledons. 
The world production of oil palm products has always been impossible to assess 
accurately owing to the quantities of produce that are not recorded, because they are 
produced in groves, smallholder plots and farms, and used for the farmer's domestic 
5 
purposes or sold locally (Oil World, 2012). Estimates suggest that world-wide 
production rose from 2.2 million tonnes of palm oil and 1.2 million tonnes of kernels 
in 1972 to 21 million tonnes of oil, 6 million tonnes of kernels and 2.6 million tonnes 
of kernel oil in 2000. Most of this increase can be attributed to Malaysia and 
Indonesia, and to some smaller Asian producers. The production of paten oil has now 
overtaken that of other vegetable oils, apart from soybean oil. The oil extracted from 
the oil palm is of two types which are palm oil and palm kernel, of which 90 O /o is used 
for food purposes and 10 % for non-food purposes (Jalani, 2012). 
2.1.1 Evolution of Genus Elaeis 
The genus Elacis which belongs to subfamily Cocoideac is believed to have 
originated either in Africa or America and is one of the 240 genera of the family 
Arccaccac. It cannot be viewed in isolation from the other genera because of high 
degree of homogeneity among the chromosomes of the palms (Dransfield et al., 
2005). The first species of the genus, E. guincensis, known as the African oil palm 
was described by Jacquin in 1763. The second species is E. oleifera known as the 
South American oil palm initially described as E. inclanococca and applied by 
Gaertner in 1788 to be a form of E. guineensis. This species is distinguished from E. 
guinccnsis by its trunk. Purseglove (1972) however doubted the classification as E. 
olcifcra hybridizes easily with E. guinecnsis and the degree of divergence between the 
two species rather justifies separation on a specific level rather than generic level. The 
third species was previously known as Barcella odoru, but was named Elaeis odoru 
by Wessels. Boer (1965); it is not cultivated, and little is known about it. 
However, molecular markers indicated that inclusion of E. odoru within the genus 
Elacis is justified (Barcelos ct al., 2002): the genetic distance between E. odoru and 
6 
the other o species of Elueis was similar to the distance between the latter, and less 
than the distance from Cocos nucifcru, another member of the Cocosoideae subfamily. 
The fourth species, E. inuduguscuriensis, was described by Beccari (1941) but is still 
controversial. Some taxonomists believe that these species is a variant of E. guineensis 
which was introduced around the 10°i century when Africa influence entered 
Madagascar (Purseglovc, 1972). The species is distinguished from E. guineensis by its 
male flower in which the fused filaments of the staminal tube are shorter and the 
anthers erect, instead of spreading at anthesis while the fruits are smaller, rounded and 
surrounded by larger bracts. Uhl & Dransfield (1987) recognized only two species, 
namely E. guineensis and E. oleiTeru, the third species as belonging to Burcellu odoru 
(Trail) Trail cx Drude and the fourth species as a variant of E. guineensis to 
Madagascar as mentioned above. For the purpose of this research, E. giiinecnsis will 
only be extensively focused on. 
2.1.2 Morphology and Biology of E. guineensis (African Palm oil) 
Eiucic guineensi. s is a large, pinnate-leaved palm having a solitary columnar stem with 
short internodes. There are short spines on the leaf petiole and within the fruit bunch. 
The separate upper and lower ranks of leaflets on the rachis give the palm a 
characteristic untidy appearance. The palm is normally monoecious with male or 
female, but sometimes mixed, inflorescences developing in the axils of the leaves. The 
fi-uits are borne on a large, compact bunch. The fruit pulp, which provides paten oil, 
surrounds a nut, the shell of which encloses the palm kernel. 
The early descriptions of the oil palm are listed in Hartley (1988). The only one of 
more than historical interest is the botanical description by Jacquin (1763). He 
described the palm from material from Martinique (to where it must have been 
7 
introduced), his description was detailed, but he described the flowers as either female 
or hcrnutpphrc, diti steriles and seemed unaware that flowers of the two sexes were in 
separate inflorescences. The production of male and female inflorescences was first 
recorded by Miller in his Gardener's dictionary (London, 1768). 
Before the end of the eighteenth century Gaertner (De f iictihus et sc'minihus 
planturiun. Stuttgart, 1788) gave a more detailed description of the flower parts, 
recording that the male and female flowers are on separate inflorescences. Most of the 
early attempts at classification of varieties were unsatisfactory, as they were based on 
very limited acquaintance with the palm, and no knowledge of the inheritance of the 
characters described. However, it is the first description by Preuss (1902) of the 
li. somhc' palm, a name used in Congo, Cameroon and Nigeria for the thin-shelled 
tencru fruit form and still employed in quite recent times. 
Janssens (1927) and Smith (1935) provided the first simple classifications which, in 
their essentials, have stood the test of time. Although nothing was known of the 
inheritance of the characters described, Janssens recognised that the fruit forms duru 
and teneru, distinguished by the thickness of shell, could be found in fruit types of 
different external appearance. Thus, both the common fruit type nigresccns and the 
green-fruited virescen. c were divided by Janssens into three forms, dim, tenera and 
/nsi/L'ru. 
2.1.3 Infraspecifrc Classification of E. guineensis 
The taxonomy of E. guincc'nsis" variant forms has not yet been clarified. Several 
authors have described them as cultivars, varieties, subspecies or races. Purseglove 
(1972) believed that cultivars in the real sense do not occur in the species. He 
8 
therefore produced an infi-aspecific classification based on fruit characters, viz., dura 
(DD), tenera (Dd) and Cpisilera (dd), without committing them to specific taxonomic 
categories. Other author such as Whitcmore (1979) debated the various forms 
idalatrica, tenera, deli Jura and pis"i/era as varieties. As categories such as subspecies, 
variety or form may be inappropriate in the infraspecitic classification of the species, 
the term race is proposed. Race is defined as a permanent variety or a mierospecies. 
This term may prove appropriate because, in the genetic sense, races of a species are 
capable of exchanging genes and this has been amply demonstrated by hybridization 
between the races of E. guineensis. 
Brief Descriptions of the various common races of E. guineensis (Willy, 2010). 
" Race thira (DD): endocarp 2mm-8mm thick, comprising 25 %- 55 % of fruit 
weight: medium mesocarp content of 35 %- 55 % by weight: kernel tends to 
be large comprising 7%- 20 % of fruit weight. The Deli din"u is an ecological 
variant of duru, differing from latter in only minor differences such as kernel 
content (4 %- 10 %), mesocarp content (up to 65 %) and as such larger fruit 
size. The mucrncanvu form with endocarp 6mm-8mm thickness is an extreme 
rani race. 
" Race teneru (Dd): endocarp 0.5mm-3mm thick, comprising I%- 32 % of fruit 
weight. medium to high mesocarp content of 60 %- 95 %: fibre ring darker in 
colour. The dividing line between duru and teneru is sometimes rather distinct 
as shown by the presence of a fibre ring in the mesocarp of the latter which is 
diagnostic: kernel usually smaller, comprising 3%- 15 `Y° of fruit. The oil 
content is higher at 24 %- 32 %. 
9 
" Race jpisifý'ra (dd): no endocarp, with small pea-like kernel in fertile fruits, but 
predominantly sterile with fruits rotting prematurely. It has a much higher sex 
ratio than clura and tenera and a significantly higher number of spikelcts, even 
when fertile, the fruit to bunch ratio is low; infertile palms show strong 
vegetative growth. 
2.1.4 Ecology and Morphology of E. guineensis 
In its natural habitat in West Africa, E. guincensis often occur in association with 
Raphia or, if alone, in fresh water swamps (Chevalier, 1943). It is absent from 
undistributed tropical rain forests, where there is competition from the forest flora, 
absence of light due to the forest canopy and excessive moisture. It cannot survive or 
regenerate in high secondary forests for the same reasons. It may tolerate temporary 
flooding, if the water is not stagnant. Under semi-wild conditions, E. guincc'nsis 
groves owe their presence to alteration of the natural vegetation by man for their 
development. In southern Nigeria, with its high population pressure, oil palm reaches 
its highest development and has given rise to paten groves. In both wild and semi-wild 
conditions, the yields are very low compared to plantation conditions, where 
individual plants are properly spaced and adequately cared for (Corley &Tinker, 
2003). 
The biggest area of semi-wild E. guineensis groves in the lowland regions of Western 
and Central Afi-ica lies between 10° N and 1 0°S, which have a marked dry season 
lasting up to tour to six months. The genus is ecologically suited to such conditions, 
although physiologically, it lowers yields. The present knowledge indicates that under 
plantation conditions, Elac'is cultivation can be extended from its natural habitat of 
10"N and 10"S of Equator to 23"N and 23"S, i. e. between the tropics of Cancer and 
10 
Capricon. This has been brought about by man's deliberate planting of oil palm as it is 
an important component of world's vegetable oil supply (Latiff, 2000; Willy 2010). 
It is generally known that Eluc'is thrives well between a mean minimum temperature 
of 20°C - 23°C and a mean maximum temperature of 28°C - 32°C which normally 
occur in tropical countries. If the temperature falls below this, particularly at night 
temperature of below 19°C, bunch development is affected and the yield reduced 
(Latiff, 2002). There is also evidence to show that a temperature below 15°C stops 
growth in young seedlings. In Malaysia, oil palm can be grown on a wide range of 
soils. It is found that adequate soil moisture is more important than nutrient supply, 
which can be artificially supplied. In Peninsular Malaysia, the best areas are the 
coastal alluvial clay; Sabah's the riverine and coastal alluviums, and the soils of 
volcanic origin. 
Aforpholo,, -f 
0 The oil palm has a typical adventitious root system. When the seed germinates, 
the radicle emerges through the germ-pore. The radicle continues to grow and 
is joined when the seed becomes detached by the primary roots. The short- 
lived radicle is replaced by adventitious roots emerging from the periphery of 
the radicle-hypocotyl junction. Primary roots emerge from the base of the 
swollen part. The primary roots descend deeply from the base of the trunk and 
spreads horizontally at varying depths in soil as the palm develops. They are 
up to 6mm-10nun in diameter and up to 20m in length. The secondary roots 
are produced from the pericycles of the primary roots amounting to 2mm-4mm 
in diameter ascending to the surface of the soil (Purseglove, 1972; Corley 
&Tinker, 2003). 
II 
"A trunk of an oil palm is not visible until the palm is three years or so old 
when the apex reaches its full diameter inform of an inverted cone. The base of 
the trunk is about 60 cm in diameter and the trunk is about 40 cm in diameter. 
The rate of trunk extension is about 25 cm - 50 cm per year. All mature palms 
have a solitary columnar stem with persistent frond bases. The stem of E. 
guinc'c'nsis is stout and despite the crowded steeply ascending frond petioles 
bases, and then ultimately smooth. Frond bases persistenly adhere to the stem 
for at least 12 years and then fall away except for a few near the crown. The 
stem measures between 22 cn1 to 75 cm in diameter (Latiff, 2002). 
" The frond consists of leaflets, each with a lamina and midrib, a central rachis 
to which the leaflets are attached, a petiole and a frond sheath. Only remnants 
of the frond sheath are visible externally. In a developing frond, the sheath is 
tubular and completely encloses but as it extends after growth of the sheath has 
stopped, the fibrous sheath splits and breaks up, leaving a row of spines along 
both edges of the petiole which are bases of fibrous sheath. The leaflets are 
long. ranging from 55 cm to 65 cm and often 100 cm. it is narrow with a width 
from 2.5 cm to 4.0 cm and a midrib with a number of parallel veins on the 
lamina. The cuticle is thick and has a very high resistance to diffusion of water 
vapour. Stomata are situated on the abaxial surface of the leaflets only. The 
leaflets are arranged in two planes (Latiff, 2002; Willy, 2010). 
" The oil palm produces either a male or female or at certain stages a 
hermaphrodite inflorescence in each of the frond axil during the mature stage. 
An average female intlorescence may have more than 100 spikelets with over 
4000 floral buds. The spikelets develop acropetally on the inflorescence. They 
are about 30 cm long with 12 - 30 flowers on central spikelets and less than 12 
12 
flowers on the lower and upper spikclets. Averagely, there are 700 - 1200 
flowers per spikelet. Therefore, the total number of flowers is about 126000 
and the number of pollen grains has been estimated to be in excess of 900 
million with weight of 40 g per inflorescence. The presence of empty frond 
axils is an indication of abortion. The hermaphrodite inflorescence in E. 
guinccnsis is similar to that of E. olci/c'ru. It has about 200 spikelets of varying 
lengths from 5 cm - 15 cm (Willy, 2010). 
" The oil palm gives the highest yield of oil per unit area of any crop and 
produces two distinct oils, palm oil and palm kernel oil, both of which are 
extremely important in world trade. The time from flowering to harvesting of 
ripe fruits is five to six months. The fruits are borne on spikelets which are 
spirally arranged to form a compact bunch. The fruits are drupes of about 10 
kg - 90 kg in weight. The pericarp comprises of three layers, exocarp (external 
layer), mesocarp (outer layer) and endocarp (internal layer). The exocarp 
protects the fruit, the mesocarp contains the oil and the endocarp is a hard shell 
containing the kernel which contains the kernel oil and the endosperm (Latiff, 
2002, Willy, 2010). 
" The seed is a nut which remains after the soft oily rnesocarp has been removed 
usually by retting. It consists of an endocarp and possibly two or three kernels. 
The nut size varies greatly and is dependent both on the thickness of shell and 
size of kernel. It may be 2 cm -3 cin in length and average 4g in weight. Deli 
cluru and African nuts are larger, weighing up to 12 g, and the African tencra 
nuts are usually 2 cm or less in length and average 2g (Corley &Tinker, 
2003). 
13 
2.2 PLANT GENETIC RESOURCES 
Plant genetic resources include the variation in crop plant genetic material that is 
accessible for present and potential future exploitation (FAO, 1996). The variation 
contains diversity which can be found at the nucleotide sequences levels, genotypes 
and alleles. and is required for the improvement of new cultivated varieties along with 
contributing to the flexibility of existing varieties (Hammer et al.. 2003). Plant genetic 
resources are significant basis of genetic diversity and contribute to food security. 
Plant breeding is founded on the exploitation of genetic diversity within and between 
crop species and varieties (FAO, 1996). 
2.2.1 Genetic Diversity 
Genetic diversity has been defined as the germplasm of plants, animals or other 
organisms containing useful characters of actual or potential values, especially when 
these characters provide the variation in genes and genotypes between and within 
species or populations (Cromwell et al., 1999). It is the diversity that enables species 
to adapt to changing environments and provides an insurance against unknown future 
needs or conditions, thereby contributing to stability of farming systems at the local, 
national and global level. Genetic diversity comprises the total genetic variation 
present in a population or species. It is the differences within and between species or 
varieties in genes, alleles and genotypes, caused by mutation and recombination. 
Genetic diversity is the basis of selection in crop plants, it is essential for the 
development of new varieties by using novel combinations and traits (Preston, 2011). 
In the field, genetic diversity among and between individuals and varieties is essential 
fior resistance to pests and diseases, as well as tolerance and adaptation to climatic 
14 
conditions and changing climate (Preston, 2011). According to Newbury and Ford- 
Lloyd (1997), maintenance of the range and magnitude of genetic diversity present 
within a taxon is a primary aim of plant conservation. Therefore, to facilitate the 
conservation of genetic variation for present and future use, and to establish a 
baseline, the diversity of plant genetic resources, such as those held at MPOB needs to 
he characterized. 
2.2.2 Characterization of Genetic Diversity 
The level of genetic diversity present in populations is influenced by many factors 
including life form, breeding system, seed dispersal and geographic range. The overall 
result of this is a higher level of genetic diversity within populations and lower level 
of differentiation between populations in outbreeding species; a lower level of genetic 
diversity within populations and higher level of genetic diversity between population 
differentiations in inbreeding species (Hamrick & Godt, 1996). Genetic patterns of 
diversity may be variable across time such as in species due to changes in agriculture, 
and such as in ex-situ collections (due to genetic drift, cross pollination) and selective 
effects during regeneration, which can result in changes in genetic diversity and allele 
frequencies (Negra & Tiranti, 2009). 
Genetic diversity has been explored using, predominantly, three marker types; 
morphology, proteins and DNA based methods. Before the advent of modern genetic 
technology, morphological markers were the classical method for characterization or 
estimation of genetic diversity. These comprised a diversity of traits and 
measurements that were recorded at all stages of plant development. Using 
morphological markers confer many advantages and morphological studies are often 
the first step in species studies and plant genetic resource activities serving to inform 
15 
molecular studies as to where diversity may exist (Karp et al., 1997). Advantages 
comprise the low cost and level of technology required, and the relation of markers to 
traits of agronomic importance (Newbury & Ford-Lloyd, 1997). 
However, there remain limits to the usefulness or applicability of morphological 
markers. malls of which are met by molecular methods. These include: the limited 
number of informative characters available, some of which may show little variation 
over material. The quantitative nature of their inheritance (being jointly influenced by 
genetics and environmental conditions of growth, because of this, some traits cannot 
be reliably isolated), related to this is the effect of environment that may mask the 
genetic coordinate and therefore influence genetic diversity estimates based on 
morphology (Spooner et al., 2005). 
In the past, determination of the genetic structure of germplasm collections has mainly 
been done using passport data (van Hintem, 2000) or multivariate statistical methods 
such as cluster analysis, principal component analysis, multidimensional scaling; 
usually based on agronomic data (Franco et al., 2001). 
2.3 CHEN1ON1ETRICS 
The development of the discipline chemometrics is strongly related to the use of 
computers in chemistry. Chemometrics is the chemical discipline that uses 
mathematical and statistical methods to design or select optimal measurement 
procedures and experiments, and to provide maximum chemical information by 
analyzing chemical data (Matthias, 2007). Chemometrics is the use of mathematical 
and statistical methods tior selecting optimal experiments and extracting maximum 
amount of information when analyzing multivariate data. Chemometrics can also he 
16 
called multivariate analysis (Matthias, 2007). This is because the statistical method of 
classification is usually by multivariate methods which include Principal Component 
Analysis (PCA), Cluster analysis and Discriminate analysis (Oyelola, 2004). They are 
all under exploratory statistical analysis. Wold and Sjostrom (1998) have reviewed 
that one of the successful area in chemometrics was pattern recognition method which 
is applied in both industry and academia. 
Chemical pattern recognition method often used in chemometrics to seek the pattern 
from the multivariate data. In addition, Gonzalez (2012) has discussed that pattern 
recognition was part of Artificial Intelligence field that focused on finding the 
similarities and variation from large data sets in which brought about natural 
classification and grouping. In most cases, the raw data from the chemistry research 
were in form of chromatographic or spectroscopic. Hence, the chemist has dealt with 
chemometrics analysis to extract the maximum of useful information, thus, 
highlighted the differences in the chromatographic or spectroscopic results of different 
variables. As a result, exploratory data analysis (EDA) gave simple earlier visual idea 
of the main relationship between the objects and the variables (Brereton, 2003). 
2.3.1 Principal Components Analysis (PCA) 
PCA is a widely used mathematical tool for high dimension data analysis; that 
provides a guideline for how to reduce a complex dataset to one of lower 
dimensionality, to reveal any hidden, simplified structures that may underlie it 
(Jolliffe, 2002). Although PCA is a powerful tool capable of reducing dimensions and 
revealing relationships among data items, it has been traditionally viewed as a 
"blackbox" approach that is difficult to grasp for many of its users because of its 
coordinate transtormation from original data space into Eigen space. PCA is a method 
17 
that projects a dataset to a new coordinate system by determining the eigenvectors and 
eigenvalues of a matrix. It involves a calculation of a covariance matrix of a dataset to 
minimize the redundancy and maximize the variance. Mathematically, PCA is defined 
as an orthogonal linear transformation and assumes all basis vectors are an orthogomal 
matrix (Jolliffe. 2002). PCA is concerned with finding the variances and coefficients 
of a dataset by finding the eigenvalues and eigenvectors. 
PCA is one of the most useful statistical tools for screening multivariate data with 
significantly high correlations. Information from PCA may assist the plant breeder to 
identify limited traits for using in hybridization and selection programs (Doumbia et 
al., 201 3). The central idea of principal component analysis (PCA) is to reduce the 
dimensionality of a data set consisting of a large number of interrelated variables, 
while retaining as much as possible of the variation present in the data set. This is 
achieved by transforming to a new set of variables, the principal components (PCs), 
which are uncorrelated, and which are ordered so that the first few retain most of the 
variation present in all of the original variables. PCA tool is used to interpret the 
variables that described the differences between samples, thus, identifies which 
variables contributed most to an observed difference or which variables were 
correlated. In short, PCA is the most basic workhorse of multivariate data analysis that 
acts as fundamental technique for other methods in "The Unscrambler" (CAMO, 
2014). 
Basically, the principle of PCA aims to find the direction in multidimensional space 
along which the distance between the data points is the largest thereby indicating the 
maximum variance in data set. In other words, it constructed as linear combinations of 
the initial variables that contributed most in making the samples different from each 
18 
other, thus minimizing the squared distance from the line to each point of data set. 
Data point was the representation of each sample in a data table, thus, the location was 
determined by the cell values of the corresponding row in the table. Thus, each 
variable act as coordinate axis in multidimensional space. PCs are computed 
iteratively where the first PC contained largest information or most explained variance 
while the subsequence PC contained less explained variance than previous one. Based 
on the geometric concept, a new set of coordinate axis were formed as shown in 
Figure 1 below where PCs were orthogonal to each other. Thus, the user can focus on 
the first few PCs in the interpretation which contained larger information than the 
subsequence ones (Esbensen cl al., 2002; CAMO, 2014). 
Figure 1: Score plots where PCs I and 2 are orthogonal to each other (CAMO, 2014) 
Variable 3 
Variable 1 
2.3.2 Clustering Analysis 
Clustering deals with finding a structure in a collection of unlabeled data. A loose 
definition of clustering could be "the process of organizing objects into groups whose 
members are similar in some way". A cluster is therefore a collection of objects, 
which are "similar" between then and are "dissimilar" to the objects belonging to 
19 
other clusters (Manpreet et al., 2008). Thereby, clustering is the grouping of data 
items based on their similarity. Cluster analysis is classified as unsupervised pattern 
recognition due to the discovery of the data structures without providing any 
explanation as well as the pre-information about the data not being used in the 
classification (CAMO, 2014). 
The goal of clustering is to determine the intrinsic grouping in a set of unlabeled data. 
But how to decide what constitutes a good clustering'? It can be shown that there is no 
absolute "best" criterion, which would be independent of the final aim of the 
clustering. Consequently, it is the user, which must supply this criterion, in such a way 
that the result of the clustering will suit their needs. For instance, we could be 
interested in finding representatives for homogeneous groups (data reduction), in 
finding "natural clusters" and describe their unknown properties ("natural" data 
types). in finding useful and suitable groupings ("useful" data classes) or in finding 
unusual data objects (outlier detection) (Manpreet et al., 2008). 
Figure 2: Clustering pattern 
4 
\\l`X 
\ \l\ý` 
X `l ` .ý/ 
Clustering can be in the türnl of distance-based clustering and conceptual clustering. 
Distance - based cluster is when two or more objects that belong to the same cluster 
20 
are "close" according to a given distance. On the other hand, conceptual clustering 
occurs when two or more objects that belong to the same cluster are defined by 
common concept. In other words, objects are grouped according to their fit to 
descriptive concepts not according to simple similarity measures. 
2.3.2.1 Types of Clustering 
The Unscrambler software offers two main groups of cluster analysis which are 
classified as non-hierarchical clustering (k-means and k-medians) and hierarchical 
cluster analysis (HCA). 
K-means Clustering 
The target of the clustering is to minimize the variability of the samples within the 
clusters along with maximizing the variability between clusters formed starting with k 
random clusters (StatSoft, 2013). The theory of k-means clustering algorithm provided 
by the Unscrambler starts with the randomized set of samples distribution between the 
defined k random clusters. Every defined clusters centroid is calculated using averages 
(k-means) method, then, the distance between each of the samples to its centroid is 
measured within the respective clusters. Consequently, the number of clusters k with 
the shortest distance between the samples to the centroid is selected by moving a set 
of samples to that particular k cluster. This nonhierarchical algorithm is repeated until 
there are no changes occurring and once the centroids have identified the natural 
clustering of points (Tan et al., 2006). 
K- median clustering 
In statistics and data mining, k-medians clustering is a cluster analysis algorithm. It is 
a variation of k-means clustering where instead of calculating the mean for each 
21 
cluster to determine its centroid, one instead calculates the median. This has the effect 
of minimizing, error over all clusters with respect to the 1-norn distance metric, as 
opposed to the square of the 2-norn distance metric (which k-means does. ). K median 
is used when one wishes to minimize the total 1-norn distance from each point to its 
nearest cluster center (Jain & Dubes, 1988, Bradley et al., 1997). 
Hierarchical Clustering Analysis (HCA) 
HCA used different linkage methods along with specific distance measure to create 
clusters. The chosen of both factors need to be appropriately selected based on the 
application domain as well as base on the real-world interpretation because they can 
affect the grouping results. Consequently, a dendrogram is created to portray the 
results of clusters arrangement produced yet formed based on the meet of triangle 
inequality of metric determined from the distance between samples (CAMO, 2014). 
Basically. HCA begins by treating each sample within its class, then, proceeded by 
combining the samples into another cluster based on their similarity until the clusters 
emerged to form one larger cluster (CAMO, 2014). 
As discussed earlier, the distance measure between samples is important to be 
determined for the first part in relating the samples into the same group based on its 
closest distance. Hence, the Unscrambler offers several options for the distance 
measures between samples such as Euclidean distance, squared Euclidean distance, 
city-block distance and Chebyshev distance. Among these methods, Euclidean 
distance is the most usually chosen in measurement of the distance between two 
samples of the data sets that are suitably normalized by taken into consideration the 
different between two samples directly based on the magnitude of changes in the 
sample levels. The squared Euclidean distance intended to normalize the data by 
22 
measuring the similarity between clusters where some variable may control the 
distance between groups (Hoon et al., 2013 and CAMO, 2014). 
2.4 PREVIOUS STUDIES 
Principal component analysis and cluster analysis have been widely used in various 
studies for delineating the variability in large group of genotypes of many crop 
species, some of which are outlined below: 
Shivani and Sreelakshrni (2014) assessed the genetic diversity in indigenous 
germplasm lines of safflower using multivariate tools. Principal component analysis 
revealed that 99% of the total diversity was explained on the basis of the first three 
principal components. The distribution of accessions within eight well defined clusters 
was visible as well with no apparent relationship with the geographical origin. 
In evaluating the variation in seed vigour characters of West African rice (Ory_a 
sativa L. ) genotypes, seeds of 24 West African rice genotypes were selected based on 
seed vigour traits in the laboratory and field in two cropping seasons at the research 
farm of Federal University of Agriculture, Abeokuta, Nigeria. The analysis was 
subjected to principal component analysis and clustering analysis. The first three axes 
of the PCs across the two seasons captured 86.34% of the total variation. Cluster 
analysis classified these genotypes into four distinct groups based on germination and 
emergence percentages. Those characters identified by PCA could be included in the 
crop improvement programme for improved seed quality within West African low 
land rice germplasm (Adebisi et al., 2013). 
Genetic diversity based on cluster and principal component analyses for yield and 
quality attributes in ginger was studied by Ravishanker et at. (2013). Through cluster 
23 
analysis, 25 genotypes were grouped into five main clusters while first six PCs with 
eigenvalues greater than one accounted for 76.19% of the total variability. 
The comparative study of cowpea germplasms diversity from Ghana and Mali using 
morphological characteristics was carried out by Doumbia et al. (2013). Results from 
PCA indicated that first PC explained a total variation of 72.05% while UPGMA 
showed that at 0.97 level of similarity, almost all the 94 accessions studied were 
distinct from each other while at 0.95 levels, they were similar to each other. 
Muhammad et al. (2013) utilized PCA and cluster analysis to study the genetic 
divergence in indigenous spinach. Results obtained shows that the cumulative 
contribution of the first three PCs reflected by PCA was 58.4 % while cluster analysis 
grouped the spinach genotypes into four major clusters regardless of collection origin. 
In the research of WeldeMicheal et at. (2013), forty nine coffee germplasm (Collca 
arabica L. ) accessions were collected from Gomma Wereda in other to estimate 
information on genetic diversity. The experiment was conducted in simple lattice 
design with two replications during cropping season by superimposing on six years 
old coffee trees and grown under uniform coffee shade tree (Sc'shania seshan) 
conditions. Data from 26 quantitative characters were recorded. The analysis of 
variance showed a significant variation among the accessions for all morphological 
traits. This shows the existence of variability among the tested materials. Cluster 
analysis grouped the 49 germplasm accessions into five clusters which make the 
accessions to be moderately divergent. The distances between these clusters are highly 
significant (P<0.01) which suggested a suitable hybridization program is possible. 
Principal component analysis showed a variation in the first two principal components 
which is explained by the larger share of the observed variation. 
24 
Ajmal et al. (2013) used multivariate analysis to study the genetic relationships among 
wheat germplasm comprising of 50 genotypes. The first three PCs with eigenvalues 
greater than one contributed to 70.59% of the variability amongst genotypes while 
cluster analysis set apart the 50 genotypes into 5 clusters based on Ward's method. 
Principal component analysis was done by Lohani et al. (2012) to estimate the 
variability in germplasm of potato. The PCA of 54 potato genotypes based on 
correlation matrix of agronomic and quality traits showed that first 11 components 
explained 96.25% variation. Non-hierarchical Euclidean Cluster analysis based on 
PCA grouped the 54 potato genotypes into seven non-overlapping clusters. 
Darvishzadeh et al. (2012) studied on the genetic diversity in population of Iranian 
Ajowan (curiun copticum L. ) based on agronomical and morphological characteristics. 
Ten populations were collected from different regions and evaluated in a completely 
randomized design with eight to ten replications. Among the characteristics studied, a 
high coefficient of variation was observed for the number of seeds (197.58), plant 
yield (57.56) and shoot dry matter (56.28). Cluster analysis using Ward's method 
classified ten populations of Ajowan into four groups. PCA using 18 agronomical and 
morphological characteristics indicated that the first two principal components with 
cigenvalue of more than one accounted for 74.5% of the total variance. Both PCA and 
cluster analysis result were consistent which displayed a considerable diversity of 
agronomical and morphological is useful in gcrmplasm management. 
In order to reveal the underlying association between agronomic traits of spring 
wheat in eastern region of Qinghai Tibet Plateau, Dcyong (2011) used partial 
correlation analysis, PCA and cluster analysis to study the relationship between yield 
25 
grain and agronomic traits. PCA was able to indicate genotypes with high-yield 
potential and the varieties were grouped into 14 groups through cluster analysis. 
Sanni et al. (2010) studied on the variability of 434 accessions of rice germplasm from 
Cote d'lvoire and evaluated 14 agro morphological traits in upland condition using 
experimental design and analyzed with multivariate methods. The result suggested 
that traits such as plant height, leaf length, tillering ability, number of days to heading 
and maturity, panicle length and grain size were the principal discriminatory 
characteristics. The result of PCA indicated that first three PCs explained about 
58.41% of the total variation among the 14 characters studied. Seven cluster groups 
were obtained using the unweighted variable pair group method of the average linkage 
cluster analysis. 
Multivariate statistical analysis was used by Lacis et al. (2010) to determine 
phonotypical variability and genetic diversity among 40 sour cherry accessions. Both 
cluster analysis and PCA showed an adequate grouping of accessions according to 
phenotypical data and known pedigree data. Nine components with eigenvalues larger 
than 1.0 were extracted which described 80% of the variability of the original traits 
while cluster analysis identified four main clusters. 
Phenotypic diversity for quantitative and qualitative traits in a collection of pepper 
gcrmplasm from different areas of Turkey was assessed by Bozokalfa et al. (2009). 
All the accessions were characterized for 67 agro-morphological traits and the 
morphological traits were subjected to PCA followed by hierarchical agglomerative 
clustering. The first six PC axes accounted for 54.29% of the variance among the 48 
accessions and their lines while seven groups were created based on morphological 
and agronomic properties. 
26 
Pecrasak et al. (2009) evaluated genetic variation in cultivated mungbean germplasm 
to assess the extent and pattern of their diversity. He evaluated 9 qualitative and 21 
quantitative traits in 340 diverse cultivated mungbean accessions collected at the 
world vegetable center, Taiwan. The germplasm displayed a wide range of diversity 
for most of the traits evaluated. Cluster analysis grouped the germplasm into 5 major 
and 1 minor cluster. Principal component analysis revealed that the first three PCs 
explained 74.9% of the total variation. It is recommended that germplasm from west 
Asia he exploited more in cultivar programs. The results can help mungbean breeders 
choose the right combinations of parental genotypes carrying the desirable characters. 
In the study of genetic variability of some quality traits in Luihvrus spp. germpalsm, 
66 accessions representing eighteen species of the genus lathyrus were collected from 
different regions evaluated for variation of quality traits. Cluster analysis of 
coefficient of variance values for each accession identified the 66 accessions into 8 
groups. High variability was attributed both genetic and environmental factors 
exhibited at both inter-specific and intra-specific levels. The most promising accession 
for breeding programs was L. sutivus from Tunisia (Sammour et al., 2007). 
CHAPTER III 
MATERIALS AND METHODS 
3.1 Breeding Materials and Site Location 
The germplasm used in this study originated from Senegal and Gambia. A random 
sample of five families from 13 sites originating from Senegal and also Gambia 
(population; designated as SEN 01, SEN 02, SEN 03, SEN 04, SEN 05, SEN 05, SEN 
06, SEN 07, SEN 08, SEN 09, SEN 10, SEN 11, SEN 12, SEN 13 GAM05.02 and 
GAM05.08) were collected in July-August 1993 by researchers from Malaysian Palm 
Oil Board (MPOB) from the two countries. The germplasm materials were planted at 
the MPOB Research Station Kluang, Johor in 1996. The palms were derived from the 
Independent Completely Randomized Design, where a total number of 415 open- 
pollinated palms were planted in Trial 0.352 (Senegal materials) and Trial 0.357 
(Gambia materials) with two replicates each (41 progenies in replication I and 15 
progenies in replication 2). For the purpose of this study, available quantitative data 
on the Senegal and Gambia germplasm were collected at the MPOB headquarter. 
3.2 Data Collection 
The pertönnance of progenies for yield and oil yield components was assessed based 
on data on fresh fruit hunch (FFB) yield records, bunch number (BNO) and average 
hunch weight (ABW), bunch components, vegetative and physiological characters and 
fatty acid characters. 
28 
3.2.1 Yield and Yield Components 
Harvesting of oil palm usually begins at 36 months after field planting with 
subsequent operations carried out at regular intervals of seven to ten days, i. e. three 
rounds in a month. Data on yield and yield components were evaluated from year 
2000 - 2007. 
Procedure for Yield Recording 
Recordings of bunch weight (BWT) and bunch number (BNO) on individual are 
carried out during the harvesting rounds in oil palm breeding. Fresh bunch weight 
(FFB) is the sum of the bunch weight (BWT) while BNO is the total of all the bunch 
counts and average hunch weight (ABWT) is the quotient between FFB and BNO. 
The yield components were derived as follows: 
FFB (kg/p/yr) = 
r; '=1 BjVT, 
BNO (bunches/p/yr) = 
r; '=1 BNO; 
ABWT (kg/p/yr) = FFB/BNO 
Where n is the number of harvesting rounds. 
3.2.2 Bunch Analysis 
Quantitative data on the bunch components were evaluated in the 2001 until 2006. 
Bunch and fruit components were determined using the bunch analysis technique 
developed by Blaak et al. (1963). 
29 
Procedure. for Bunch Analysis 
Samples of three to five bunches from each palm were analysed. Each bunch was 
weighed and fruit hearing spikelets chopped off the stalk. The spikelets were 
randomly segmented for analysis of the fruit to bunch (F/B) and fruit compositions 
(FC). To ease picking, the F/B portions were kept for two days and the weight of 
empty spikelets. fertile and parthenocarpic fruits were recorded. Analysis of the fruit 
compositions was continued following spikelet sampling on the same day. Fruits were 
detached and randomly sampled, counted, weighed and scraped. The nuts were crack 
opened and kernel weighed. A 5g sample of minced mesocarp was oil extracted in 
hexane for 18-24 hours using soxhlet apparatus. The amount of oil in the sample was 
calculated on weight difference before and after extraction. The components of the 
hunch were calculated using the below formulae: 
FB = Fruit to Bunch (%) = [(FFWT + PFWT) / SWT] x[ (BWT -STKWT) / BWT] x 100 
P/F = Parthenocarpic to Fruit (%) = PFWT / (FFWT + PFWT) x 100 
M/F = Mesocarp to Fruit (%) _ [(FSWT - FNWT) / FSWT] x 100 
MC = Moisture Content of mesocarp(%) =100 -[{(TDMWT-TWT)/(FSWT-FNWT); x 100] 
O/DM = Oil to Dry Mesocarp (%) 
O/WM = Oil to Wet Mesocarp (%) 
O/B = Oil to Bunch (%) 
O'Fi = Oil to Fibre (%) 
K/F = Kernel to Fruit (%) 
MNW = Mean Nut Weight (g) 
MFW =Mean Fruit Weight (g) 
P/B = Parthenocarpic to Bunch (%) 
OPY = Oil per Palm per Year (kg) 
_ [(ETMWT - ETFWT) / (ETMWT - ETWT)] x 100 
=[(100-MC)xO/DM]/100 
_ (F/B x M/F x O/WM) / 10,000 
_ [(ETMWT -ETFWT) / (ETFWT -ETWT)] x 100 
= (KWT / FSWT) x 100 
= FNWT /NOFNUT 
= FSWT / NOFNUT 
= (P/F x F/B) / 100 
= (O/B x FFB) / 100 
30 
KPY = Kernel per Palm per Year (kg) 
Where; 
BWT = Bunch weight 
SWT = Spikelet weight 
PFWT = Parthenocarpic fruit weight 
FSWT = Fruit sub sample weight 
NOFNUT = Number of fresh nut 
TDMWT = Tin + dry mesocarp weight 
ETWT = Extraction thimble weight 
= (K/B x FFB) / 100 
STKWT = Stalk weight 
FFWT = Fertile fruit weight 
ESPKWT = Empty spikelet weight 
FNWT = Fresh nut weight 
KWT = Kernel weight 
TWT = Tin weight 
ETFWT = Extraction thimble + fibre weight 
ETMWT = Extraction thimble + mesocarp weight FFB = Fresh fruit bunch 
3.2.3 Vegetative Measurements and Physiological Characters 
Vegetative characters of the oil palm germplasm were pooled eight years after field 
planting. Frond production was first calculated after the 7th year before other 
parameters were taken a year after, i. e. year 2004. The physiological characters on the 
other hand were assessed based on measurements on collective data of bunch yields 
and hunch quality components, following the methods developed by Squire (1986). 
Data on the physiological characters were assessed in the year 2007. The various 
components of the vegetative and physiological parameters were calculated using the 
below formulae: 
FP = Frond production (no/p/yr) 
PCS = Petiole Cross Sectional Area (em 
RI. = Rachis length (m) 
LL = Leaflet Length 
LW = Lcatlet Width 
LN = Leaflet number (no/p/yr) 
D= Trunk diameter (cm) 
11T = Trunk height (m) 
IITI = Ilcight increment 
LA Leaf Area (m ) 
= number of fronds produced in one year 
= petiole depth x petiole width 
= length from tip of rachis to the first ligule 
=(LLI +LL2+...... LL6)/6 
=(LWI +LW2+.... LW6)/6 
= (number of fronds on one side of rachis)x2 
= diameter of trunk at one meter fromground 
= height of trunk of ground to base of frond 41 
_ (HT at year t) / (age at year t- 2) 
_ [(2; (LL; x LW; ) x LN x 0.57) / 6] / 10,000 
31 
LAR = LeafArea Ratio 
LAI = Leaf Index Ratio 
LDW = Leaf Dry Weight (kg) 
TDW = Trunk Dry weight (kg) 
FDW = Frond Dry weight (kg) 
FI = Frond Index 
I = Fractional interception 
C= Conversion Efficiency (g/MJ) 
VDM = Vegetative Dry Matter (kg/p/yr) 
BDM = Bunch Dry Matter (kg/p/yr) 
TDM = Total Dry Matter 
BI = Busich Index 
NAR = Net Assimilation Rate 
TOIL = Total Oil (k(, /p/yr) 
TEP = Total Economic Product (kg/p/yr) 
Where: 
_ (FP x LA) / VDM 
= (40 fronds x LA x PD )/ 10,000 
= (0.1023 x PCS) + 0.2062 
= 3.142 x (D/2)2 x (I-1T / Age in year) x 1,000 x 0.17 
= FP x LDW 
= LA/ LDW 
=1- exp (_0.47) - (LM -0 
3) 
=TDM/(31 x f) 
_ (FDW + TDW) 
= 0.53 x FFB 
_ [(VDM + BDM) x PD'] / 1,000 
= BDM / (BDM + VDM) 
= TDM / (0.52 x LAI) 
= OPY + (KPY x 0.5*) 
= OPY + (KPY x 0.6") 
PD = Planting Density 
*0.5 = Kernel Oil Extraction Rate 
"0.6 = Relative Price of Palm Kernel Oil to Price of Palm Oil 
FFB = Fresh Fruit Bunch (kg/p/yr) 
OPY = Oil per Palm Year (kg) 
KPY = Kernel per Palm per Year (kg) 
3.2.4 Fatty Acid Traits 
Fatty acid traits were pooled between years 2000 - 2007. The fatty acid composition 
was evaluated using the method proposed by Timms (1978) for routine analysis using 
gas chromatography. 
Procedure. for Fatty Acid Analysis 
The tatty acid analysis was carried out using MPOB method Timms (1978). Fresh 
bunches were collected from the palm, weighed and chopped. Spikelets samples were 
selected one each from apical, middle and basal portion of the bunch. The samples 
32 
were sterilized and fi-uits from the sterilized spikelets separated. The mesocarps were 
separated and oven dried at 105 °C for a minimum of 3 hours. The dried mesocarp 
were minced and dried. After mincing, the dried mesocarp was diluted with 300m1 
hexane and filtered with a spoonful of sodium sulphate. The filtrate was sealed and 
kept in dry place. Oil and hexane mixture was later separated into oil and hexane, 
using rotary evaporator by distilling the hexane vapors under vacuum. The extracted 
oil is then poured in vials, labeled and stored for FAC and carotene analysis. 
For the fatty acid methyl determination, 0.05g of the extracted oil was measured and 
1.9m1 of solvent hexane added and homogenized. Sodium methoxide of 0.11111 was 
also added until a cloudy solution is formed. After 10 minutes, the clear portion of the 
methyl ester was pipetted into gas chromatograph vials and covered with Teflon vial 
caps for analysis. The separation of the FAME by the GC machine was under 
programmed conditions. The fatty acids components were represented by the peaks on 
the chromatograpll. 
3.3 Statistical Analyses 
3.3.1 Variability Profile 
The quantitative morphological data collected was arranged in Excel Microsoft word. 
The oil palm accessions were organized according to their family codes. Simple 
descriptive statistics such as mean, standard deviation, standard error, minimum, 
maximum and variance for each collected traits were calculated using SPSS statistical 
tool. This was done in order to know the extent of variation in the germplasm 
accessions. The average of the quantitative data was also standardized to give equal 
33 
weight to all measurements using the following formula (Microsoft Office Excel, 
2007): 
Z=X-, u/Q (1) 
Where, X= Value to standardize 
p= Arithmetic nlcan 
6= standard deviation of the distribution 
3.3.2 Principal Components Analysis (PCA) 
PCA simplifies the complex data by transforming number of correlated variables into 
a smaller number of variables called principal components. The first principal 
component accounts for maximum variability in the data as compared to each 
succeeding component. PCA was analyzed using "THE UNSCRAMBLERRX" 
software (CAMO software version 10.1). Mathematically, PCA involved in the 
decomposition of original data matrix, X, into a structure part and noise part. In matrix 
representation, the model with a given number of components as follows: 
X= 7P1 +E (2) 
where T is the scores matrix, P the loadings matrix (transposed) and E the error matrix. 
The structured part of data is the combination of scores and loadings in which focused 
by user in interpretation of PCA results while the remaining part is called error or 
residual matrix. The uth column of Tand ath row of P is represented by vectors of 1 
and p,, respectively and both are the vector representations of the uth PC. The number 
of PCs is denoted by A. while a is the number of PC such as 1,2,3 up to A. The 
maximum number of PCs (: 1) determined is either 1- I (number of objects - 1) or J 
34 
(number of variables) depending on which give smaller value. Thus, the first scores 
vector and the first loadings vector are called eigenvectors of the first principal 
component. Therefore, each successive component is characterized by a pair of 
eigenvectors for both the scores and loadings (CAMO, 2014). 
In contrast, the residual matrix, E, represent the fraction of variation that cannot be 
modeled well. It other words, it cannot be explained by available PCs, yet, useful in 
the lack-of-fit measurement of model to the original data where small value of E 
indicated the good model and vice-versa (CAMO, 2014). Usually after PCA, the size 
of each component can be measured and represented by eigenvalue. In PCA, the more 
significant the components indicated the larger of their size, thus have larger 
eigenvalue. Therefore, the eigenvalue of a PC is the sum of squares of the scores and 
represented as follows: 
g'I tij2 (3) 
where g<< is the ath eigenvalue. The sum of all nonzero eigenvalues for a data matrix 
equals the sum of squares of the entire data-matrix, so that 
za=1 9" zi=i zj=1 xij, 
where K is the smaller of 1 or J (Brereton, 2003). 
3.3.3 Cluster Analysis 
(4) 
Cluster analysis identities variable which were further clustered into main groups and 
subgroups using Ward's method through the "THE UNSCRAMBLERK'X" software 
(CAMO software version 10.1). The general purpose of cluster analysis is to group 
similar objects or samples into respective classes based on their specified 
;ý 
characteristics or variables and it includes different type of algorithms such as joining 
(tree clustering), two-way joining (block clustering) and k-means clustering. The aim 
of the tree clustering algorithm was to join the objects or samples in each class of 
themselves together using specified distance measures and linkage rules, thus, lornung 
larger cluster by connecting all the objects at the last step known as hierarchical tree. 
The Ward's method (Ward, 1963) used in this study optimizes an objective function; 
that is, it minimizes the sum of squares within groups and maximizes the sum 01' 
squares between groups. Ward's method is similar to the linkage methods in that it 
begins with N clusters, each containing one object, it differs in that it does not use 
cluster distances to group objects. Instead, the total within-cluster suns of' squares 
(SSE) is computed to determine the next two groups merged at each step of' the 
algorithm. The error sum of squares (SSE) is defined (for multivariate data) as: 
I1 
SSE 
ý=i ; -i 
Výý - V)I (5) 
where vii is the jh' object in the i°i cluster and #i is the number of objects in the i'l' 
cluster. 
CHAPTER IV 
RESULTS 
4.1 Variability Profile 
The oil palm germplasm from Senegal and Gambia exhibited low coefficient of 
variation to high coefficient of variation for the various plant attributes investigated 
(Table 1). A high variance for fresh fruit bunch (FEB), carotene, leaf number (LN), 
mesocarp-to-fruit ratio (M/F), shell-to-fruit ratio (S/F) and oil yield (OY) was 
observed, whereas for the rest of the other traits a low to medium variation was 
observed. 
Table 1: Extent of Variation. 
Characters Coefficient of variation (%) 
Yield traits 24.36 - 197.68 
Bunch quality traits 0.69-85.06 
Vegetative traits 0.50 - 107.20 
Physiological traits 0.76 - 34.30 
Fatty acid composition 0- 59.09 
4.2 Principal Components Analysis (PCA) 
In order to determine with which combination type of the various agronomic traits the 
Senegal and Gambian oil palm germplasm would achieve high yield, PCA was 
performed. The result of the PCA therein explained the genetic diversity of the oil 
palm germplasm. 
37 
Table 2: Variability profile of Senegal and Gambian oil palm germplasm as analyzed 
by descriptive statistics 
Std. 
Parameter N Minimum Maximum Mean Deviation Variance 
FFB 44 24.36 197.68 84.59 33.40 1115.64 
BNO 44 8.82 25.16 18.45 3.40 11.58 
ABW 44 2.75 13.26 4.62 2.06 4.23 
BWT 44 0.90 13.81 4.12 2.24 5.03 
MFW 44 1.93 12.50 3.03 1.95 3.80 
MNW 44 1.24 2.50 1.61 0.23 0.05 
PB 44 0.00 5.10 0.46 0.81 0.66 
MF 44 28.09 85.06 38.02 11.20 125.36 
KF 44 6.06 19.15 15.00 2.75 7.56 
SF 44 8.88 58.62 46.98 9.27 85.87 
ODM 44 61.73 77.77 70.18 3.55 12.58 
OWM 44 32.13 49.68 41.16 4.00 15.99 
I=B 44 47.63 65.63 57.74 3.35 11.23 
OB 44 4.58 23.33 9.31 3.57 12.76 
KB 44 3.49 11.53 8.65 1.70 2.90 
OY 44 0.69 45.76 8.82 8.87 78.71 
KY 44 0.94 10.45 6.83 2.13 4.52 
TEP 44 1.25 50.05 12.92 9.20 84.62 
FP 44 24.00 34.40 31.03 2.34 5.45 
PCS 44 8.98 32.85 17.27 4.51 20.36 
RL 44 3.20 5.33 3.93 0.38 0.15 
LL 44 74.55 107.20 86.27 5.86 34.31 
LW 44 3.46 5.16 4.05 0.39 0.15 
LN 44 105.00 169.00 125.81 11.68 136.52 
HT 44 2.14 3.50 2.81 0.35 0.12 
LA 44 3.84 8.49 5.05 0.97 0.94 
I_Al 44 2.27 5.03 2.99 0.57 0.33 
DIAM 44 0.50 0.86 0.68 0.06 0.00 
F 44 0.60 0.89 0.70 0.06 0.00 
LAR 44 10.09 18.56 13.01 1.65 2.73 
BDM 44 0.67 15.81 6.59 2.76 7.63 
VDM 44 7.27 19.52 12.26 2.35 5.51 
TDM 44 12.26 34.30 18.85 4.17 17.40 
BI 44 0.05 0.49 0.34 0.09 0.01 
E 44 0.56 1.25 0.87 0.13 0.02 
NAR 44 8.13 16.18 12.37 1.63 2.65 
C14: 0 42 0.21 1.13 0.50 0.15 0.02 
C16: 0 42 32.21 44.76 39.22 2.33 5.43 
C 16: 1 42 0.07 0.58 0.15 0.10 0.01 
C18: 0 42 2.20 7.25 5.15 0.81 0.65 
CI 8: 1 42 38.63 53.93 44.77 3.05 9.28 
C18: 2 42 6.43 13.29 9.88 1.25 1.56 
C18: 3 42 0.10 0.60 0.19 0.09 0.01 
C20: 0 42 0.00 0.29 0.12 0.07 0.00 
IV 42 53.27 59.09 56.26 1.50 2.26 
Carotene 42 391.67 2405.88 1650.86 398.73 158987.16 
38 
4.2.1 Scores Plot 
The score plot explained the percentage variance associated with each principal 
component obtained by drawing a graph between eigenvalues and principal 
component numbers. Nine principal components (PCs) exhibited eigenvalue of more 
than one and accounted for 88% of variability and as a result, these nine PCs were 
given due importance for further explanation (Figure 3). PC 1 showed 28% of 
variability with eigenvalue 11.20 in germplasm which then decreased gradually. 
Elbow line is obtained after which after 4th PC tended to straight. After that, little 
variance could be observed in each PC and it ended at 9th PC with variability of 3% 
and eigenvalue of 1.21. 
90 
as 
80 
75 
70 
65 
60 
55 
50 
45 
40 
35 
30 
25 
20 
15 
10 
5 
0 
PC-O PC-1 PC-2 PC-3 PC-4 PC-5 PC-6 PC-' PC-8 PC-l) 
PCs 
Figure 3: Score plot of principal component analysis between percentage variance and 
number of principal components 
4.2.2 Loadings 
The result of the PCA shown in Table 2 indicates the contribution of the various 
agronomic traits across the extracted PCs. 
39 
Table 3: Principal component analysis for Senegal and Gambia oil palm gerrnplasm 
based on 46 agronomic traits 
Traits 
FFB 
BNO 
ABW 
BWT 
MFW 
MNW 
P/B 
M/F 
K/F 
S/F 
O/DM 
O/WM 
F/B 
O/ B 
K/B 
OY 
KY 
TEP 
FP 
PCS 
RI_ 
LL 
LW 
LN 
HT 
LA 
LAI 
DIAM 
F 
LAR 
BDM 
VDM 
TDM 
BI 
E 
NAR 
C14: 0 
C16: 0 
C 16: 1 
C18: 0 
C18: 1 
C18: 2 
C18: 3 
C20: 0 
IV 
Carotene 
Eigenvalue 
Variance (%) 
Cumulative (`%, ) 
Principal Component Axes 
123456789 
-0.89 -0.29 -0.22 -0.09 -0.09 0.04 -0.13 0.13 -0.09 
-0.22 -0.60 -0.65 -0.15 -0.08 -0.05 -0.10 0.14 -0.10 
-0.96 -0.07 0.13 0.00 -0.05 0.10 -0.10 0.08 -0.01 
-0.96 -0.10 0.14 -0.10 0.07 0.05 -0.02 0.03 -0.03 
-0.88 0.03 0.35 0.06 -0.04 0.27 -0.04 -0.04 -0.06 
-0.21 -0.14 -0.01 0.23 -0.26 0.74 -0.02 0.20 0.04 
-0.83 0.06 0.24 -0.15 0.01 0.11 -0.18 -0.19 -0.15 
-0.84 -0.04 0.47 0.03 0.07 0.13 -0.03 -0.15 -0.03 
0.66 -0.06 -0.25 -0.22 -0.14 0.08 -0.32 -0.07 0.49 
0.78 0.07 -0.48 0.03 -0.04 -0.18 0.15 0.20 -0.13 
-0.39 -0.18 0.51 -0.10 -0.01 -0.55 0.14 -0.09 0.09 
-0.39 -0.20 0.51 -0.09 0.05 -0.53 0.08 -0.21 0.11 
0.24 -0.27 0.44 0.13 -0.34 0.49 0.01 -0.16 0.19 
-0.75 -0.16 0.56 0.04 0.05 0.12 0.00 -0.20 0.07 
0.65 -0.16 -0.07 -0.13 -0.25 0.26 -0.28 -0.12 0.52 
-0.94 -0.14 0.25 -0.05 0.01 0.11 -0.05 -0.04 -0.02 
-0.30 -0.61 -0.53 -0.31 -0.07 0.14 -0.17 0.06 0.25 
-0.94 -0.24 0.15 -0.10 0.00 0.13 -0.07 -0.03 0.03 
-0.09 -0.64 -0.27 -0.15 0.00 -0.12 0.39 -0.21 0.22 
-0.45 0.78 -0.26 -0.11 -0.11 -0.03 0.07 -0.06 0.12 
-0.49 0.74 -0.10 -0.20 0.07 -0.10 -0.11 -0.06 0.05 
-0.26 0.43 -0.50 0.17 0.36 -0.08 -0.14 -0.06 0.17 
-0.58 0.33 0.12 0.38 -0.38 0.08 -0.08 0.24 -0.13 
-0.47 0.51 0.03 -0.26 0.13 -0.06 0.14 -0.20 0.29 
0.08 -0.02 -0.26 0.06 -0.31 0.32 0.65 -0.14 -0.22 
-0.74 0.60 -0.15 0.17 0.04 0.03 -0.02 -0.01 0.12 
-0.75 0.59 -0.15 0.17 0.04 0.03 -0.02 -0.01 0.12 
-0.04 0.72 0.07 0.02 -0.52 -0.11 -0.10 -0.08 0.06 
-0.63 0.67 -0.16 0.15 -0.02 -0.06 -0.03 0.03 0.17 
-0.39 -0.57 0.11 0.31 0.50 0.07 -0.15 0.07 0.19 
-0.88 -0.32 -0.26 -0.16 0.00 0.02 -0.05 0.09 -0.06 
-0.38 0.69 -0.35 -0.12 -0.31 -0.03 0.26 -0.17 0.16 
-0.85 0.18 -0.39 -0.19 -0.18 0.00 0.12 -0.04 0.05 
-0.59 -0.68 -0.26 -0.16 0.19 0.03 -0.12 0.15 -0.03 
-0.69 -0.17 -0.46 -0.40 -0.26 -0.03 0.19 -0.07 -0.04 
-0.25 -0.50 -0.43 -0.55 -0.31 -0.09 0.20 -0.08 -0.11 
-0.29 -0.49 0.42 0.39 -0.01 -0.17 0.22 0.30 
0.06 
-0.17 -0.30 -0.41 0.76 -0.10 -0.09 0.02 -0.28 
0.03 
-0.09 0.17 -0.55 0.45 0.30 -0.09 -0.04 0.02 -0.12 
0.30 0.00 0.41 -0.29 -0.30 -0.03 -0.44 -0.25 -0.35 
0.16 0.43 0.11 -0.73 0.33 0.20 0.05 0.24 -0.04 
-0.22 -0.44 0.27 0.49 -0.41 -0.28 0.12 0.04 0.25 
-0.12 0.19 -0.63 0.31 -0.02 -0.06 -0.04 0.22 0.02 
-0.05 -0.34 -0.29 -0.24 -0.30 -0.33 -0.44 -0.10 -0.04 
-0.06 0.16 0.46 -0.50 0.00 -0.06 0.26 0.52 0.29 
0.06 -0.13 -0.09 -0.12 0.56 0.34 0.19 -0.51 0.01 
11.20 6.67 5.01 3.47 2.36 2.13 1.65 1.38 1.21 
28 17 13 965433 
28 45 58 67 73 78 82 85 88 
40 
The most contributing agronomic traits to PC 1 with their corresponding eigenvector 
in bracket were: FFB (-0.89), ABW (-0.96), BWT (-0.96), MFW (-0.88), P/B (-0.53), 
M/F (-0.84), S/F (0.78), K/F (0.66), O/B (-0.75), K/B (0.65), OY (0.66), TEP (-0.94), 
BDM (-0.89), TDM (-0.85), and e (-0.69). On PC 2, the most contributing traits with 
their corresponding eigenvector in bracket were BNO (-0.60), KY (-0.61), FP (-0.64), 
PCS (0.78), RL (0.74), LA (0.60), DIAM (0.72). 
For PC 3 BNO (-0.65), O/DM (-0.51), O/WM (0.57), O/B (0.56), KY (-0.53), LL (- 
0.50), C16: 1 (-0.55), and C18: 3 (-0.63) were the most contributing traits. The most 
contributing traits to PC 4, are PC 5, PC 6, PC 7, PC 8 and PC 9 with their 
corresponding eigenvectors in bracket were: NAR (-0.55), C16: 0 (0.76), C18: 1 (- 
0.73), IV (-0.50): DIAM (-0.52), carotene (0.56): O/DM (-0.55), O/WM (-0.53): HT 
(0.65): IV (0.52), carotene (-0.51): and K/B (0.52) respectively. 
Figure 4 is a PC loading plot and has two dimensional scatter plots showing the 
distribution of the various traits in space. It could be observed from the graph that 
each variable has a loading on the plot. Variables along the vertical axis are the ones 
with the highest contribution to the PC 1 while those along the horizontal axis are 
correlated to the PC 2. It is also visible on the graph that some of the variables are 
superimposed on one another while some of the variables on the other hand are further 
apart on two opposite ends. 
The correlation loadings plot of the various traits in two dimensional axes is shown in 
Figure 5. The variables on the inner circle indicate 50% variance while those found on 
the outer circle indicate 100% variance and they exhibit high contribution to PC 1. 
41 
0.3 , 
F 
Loadings 
PCs nI 
NL. VDM 
uZ i. ii 
LW 
LIN 
0.1 1 
TnM 
eftfv 
ý1ý 
L/4 
ý I __. _. nniG 0 N 
0 
a 
-0.1 
TEP 
BDfVI 
FIB 
-o. 2 
nz 
-0.2 -0.3 
BI 
MNW 
C160 
C182 
C111QR. 
K; " 
BNO LAR 
-0.1 PC-1 (28%) 
C181 
HT 
barotene 
0 
C180 
F/B 
0.1 
S/F 
K/F " 
wB 
Figure 4: Scattered diagram of 46 oil palm germplasm traits for first two components 
contributing almost half of the total variability 
Correlation Loadings (X) 
1 -0.8 -0.6 -0.4 -0.2 0 0.2 
PC-1 (28%) 
C200 
FP 
0.4 0.6 0.8 1 
Figure 5: Scattered diagram of 46 oil palm germplasm traits showing correlation to the 
first two components 
42 
4.2.3 Scores 
On the basis of the extracted PCs, the Senegal and Gambian oil palm germplasm were 
given scores (Table 3). Based on the first three PCs, the oil palm germplasm with the 
highest score on PC 1 was SSC 3 (-16.53). On the same PC, SEN 09.04, SEN 13.07, 
SEN 02.08 and SEN 01.02 had scores -7.87,4.83,4.25 and -3.11 respectively. Lowest 
score was recorded for SEN 07.03 (-0.03). As seen on PC 2, SEN 01.02 had the 
highest score (5.96). GAM 05.08, SEN 02.08 and SEN 13.04 had scores of -5.64, 
5.04, and 4.57 respectively. PC 3 on the other hand, revealed that highest score was 
recorded by SEN 01.02 (-8.83) while the least was by SEN 10.05 (-0.02) and SEN 
05.02 (-0.02). 
The scatter plot as displayed in Figure 6 shows the distribution of the Senegal and 
Gambian oil palm germplasrn on two dimensional plots. As observed from the plot, 
the oil palm germplasm can be seen in groups. Some of the oil palm accessions are 
closely packed while some are further apart. Accessions like SSC 3, SEN 09.04, SEN 
01.02 and GAM 05.08 are singly without any other accessions close to them. It can 
also be observed that SSC 3 is the most dispersed accessions out of all the other 
accessions. Noticeable from the graph as well is the center point of both the vertical 
and horizontal axes where dense distribution of some of the accessions can be 
observed. The accessions found at the point are laid on top of one another. 
4.2.4 Bi-plot 
A PC bi-plot shows the simultaneous distribution of the traits as well as the accessions 
with the high scores for the traits (Figure 7). On the graph, it can be seen that the 
43 
Senegal and Gambian the oil palm accessions are positioned close to the traits that 
best portray them. 
Table 4: Scores of the 42 Senegal-Gambian oil palm germplasm on the extracted PCs 
Principal Components 
Accessions PC-1 PC-2 PC-3 PC-4 PC-5 PC-6 PC-7 PC-8 PC-9 
GAM 05.02 1.29 -0.20 1.23 -1.77 1.03 -0.78 1.67 -2.32 0.97 GAM 05.08 1.79 -5.64 2.19 1.67 2.79 4.07 -1.39 -1.04 1.49 SEN 01.02 -3.11 5.96 -8.83 4.30 1.25 1.33 1.73 1.29 0.23 
SEN 01.03 -0.57 0.09 0.49 -0.45 -1.02 0.60 1.53 0.25 -0.44 SEN 02.01 0.12 0.72 -0.75 -0.57 0.45 -0.58 -0.38 -0.29 0.67 SEN 02.04 0.17 0.02 -0.21 -1.44 -1.10 -0.57 0.63 1.56 0.29 SEN 02.05 -0.12 -3.47 -1.48 -1.93 0.10 -0.75 -0.52 -0.28 0.43 SEN 02.06 0.37 -0.63 -2.06 -1.37 -1.41 -0.31 0.22 0.70 0.25 SEN 02.08 4.25 5.04 3.20 -0.54 1.30 1.89 0.73 -0.05 -1.42 SEN 02.09 0.70 -0.73 -1.73 -0.47 0.54 0.40 -0.07 -0.43 -1.10 SEN 03.03 0.43 0.42 -1.36 0.94 0.33 -0.23 -0.60 -0.12 0.29 SEN 03.06 0.44 -0.49 -1.32 3.93 2.82 -1.35 -0.58 -1.80 -0.70 SEN 03.07 2.63 -1.70 0.20 0.93 2.04 1.45 -1.58 -0.51 -0.62 SEN 04.01 1.48 0.73 0.84 0.39 1.12 -1.74 -1.13 0.50 0.76 SEN 04.02 0.94 -1.58 -1.06 1.46 -1.38 -2.47 -0.79 -0.65 -1.03 SEN 04.03 -0.79 -1.66 -1.64 1.17 0.80 -1.44 -1.34 1.54 1.31 SEN 05.01 0.53 -1.27 1.25 0.60 -0.51 -1.27 -2.31 0.78 0.71 SEN 05.02 -0.06 -3.63 -0.02 0.96 -1.71 3.49 0.43 1.13 1.63 
SEN 05.03 -0.83 -1.83 -1.79 -0.14 -0.49 -1.28 -0.68 0.65 0.48 SEN 05.04 0.74 -0.33 -0.37 -1.04 2.51 0.72 -0.94 0.37 -0.18 SEN 05.05 1.55 -0.94 -1.79 -0.99 -0.01 2.34 0.37 0.36 -0.38 SEN 05.08 0.61 -2.88 -0.57 -2.47 -0.37 -0.72 0.45 0.59 -2.26 SEN 06.01 0.41 -1.67 -0.71 -1.55 -1.68 -1.54 -0.52 0.23 1.33 SEN 06.08 -0.12 -0.69 0.47 -1.07 0.31 1.80 1.47 0.77 0.08 
SEN 07.03 -0.03 -0.68 0.91 -1.62 -0.44 0.01 0.59 0.40 1.08 
SEN 07.04 -0.49 -1.73 -1.59 -1.06 -1.38 -0.12 0.51 -1.37 -1.07 SEN 07.05 2.21 -1.77 1.80 3.19 0.00 -1.11 1.15 -1.94 -0.16 SEN 07.08 0.98 -2.15 -1.96 -1.56 -0.24 -0.02 1.59 0.13 -0.80 
SEN 08.02 -0.45 -1.92 -2.83 1.13 -1.76 0.57 -1.87 0.19 -1.45 SEN 08.03 2.87 0.33 3.28 0.21 1.68 0.59 1.25 2.74 -1.17 SEN 08.04 2.81 -1.19 2.32 1.94 0.33 -0.82 3.11 0.27 -1.78 SEN 09.04 -7.87 -1.73 3.43 2.45 0.09 -1.23 1.83 2.33 1.08 SEN 10.03 1.21 -0.77 1.51 2.25 -0.69 -1.26 1.01 -0.50 1.26 SEN 10.05 1.13 -2.22 -0.02 -1.48 -1.16 -0.80 0.78 -0.53 -1.23 SEN 12.01 2.07 2.37 0.61 -0.19 -0.63 -0.91 0.03 -0.60 0.59 SEN 12.02 2.42 4.12 -0.27 -3.60 3.02 -1.15 -2.41 1.38 -1.04 SEN 12.03 0.92 3.50 0.38 0.44 1.31 -1.83 1.00 -2.07 1.28 SEN 12.05 2.77 3.90 1.24 -2.21 -0.40 0.30 -0.09 0.22 2.56 SEN 13.01 2.45 2.12 -2.19 -0.95 -3.10 2.52 -0.12 -2.35 0.44 SEN 13.04 1.30 4.57 0.52 -1.31 -0.79 0.21 0.87 -0.14 0.99 SEN 13.07 4.83 4.23 3.91 4.31 -4.40 0.65 -2.49 0.92 -1.09 SSC 3 -16.53 3.01 3.46 -1.12 -0.23 1.23 -1.32 -1.60 -1.22 
44 
6 
5 
4- 
SSC 3 
3 
2 
ö1 
r 
Z, 
N 
CL -1 
-2 3 
-4 ý 
-5 - 
1 
-Tý 
-18 -16 -14 
Scores 
SEN b1.02 SEN 02.08 
D3. 
ý3 4F0N 
08 03 
th ___ Rol Ph 
08.04 
7 
-8 -6 -4 -2 
PC-1 (28%) 
Iý 
-12 -10 
Figure 6: Two dimensional ordinations of 42 Senegal-Gambian oil palm accessions on 
principal axes 1 and 2 
1 
0.8 
o. s SSC 3 
o. a 
0.2 
ý 
-- 0 N 
Ü 
°- 
-0.2 
-0.4 
-0.6 
-0.8 
-1 
Bi-plot 
SEN 09.04 
SP OR 
LAI ýý 
TDM 
rtýý 
,r 
TeP ; 
Rfln 
9 
11 
DIA SEN 02.08 
SEN 13.04 
ýSýýa513.07 
4 
24 
EA1V. 03 ". 
0 
6 
1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 
PC-1 (28%) 
Figure 7: Bi-plot of 46 oil palm agronomic traits and 42 oil palm accessions on PC I 
andPC2 
45 
4.3 Cluster Analysis 
The dendrogram generated through Ward's hierarchical method divided the Senegal- 
Gambian accessions into six clusters (Figure 8). Cluster-I, cluster-II and cluster-III 
comprised of only one accession each which include SEN 01.02, SEN 09.04 and SSC 
3 respectively. Cluster-IV had eight accessions which comprised of SEN 02.08, SEN 
12.01, SEN 12.02, SEN 12.03, SEN 12.05, SEN 13.01, SEN 13.04, and SEN 13.07. 
Cluster-V had six accessions which consisted of GAM 05.08, SEN 05.02, SEN 07.05, 
SEN 08.03, SEN 08.04, and SEN 10.03. Cluster-VI was the largest group and 
consisted of GAM 05.02, SEN 01.03, SEN 02.01, SEN 02.04, SEN 02.05, SEN 02.06, 
SEN 02.09, SEN 03.03, SEN 03.06, SEN 03.07, SEN 04.01, SEN 04.02, SEN 04.03, 
SEN 05.01, SEN 05.03, SEN 05.04, SEN 05.05, SEN 05.08, SEN 06.01, SEN 06.08, 
SEN 07.03, SEN 07.04, SEN 07.08, SEN 08.02, and SEN 10.05. 
Means of the agronomic traits that characterize individual clusters are presented in 
Table 4. The germplasm grouped in cluster-I was reflected by high fresh fruit bunch, 
highest mean values for bunch number, nut weight, shell to fruit ratio, petiole cross 
section, frond length, height, vegetative dry mass, palmitic acid, palmitoleic acid, and 
linoleic acid. 
Cluster-II had the second highest mean value of fresh fruit bunch, bunch weight, mean 
fruit weight, mesocarp- to-fi-uit ratio, oil to dry mass, oil to wet mass, oil to bunch 
ratio, oil yield, total economic product, bunch dry mass and bunch index. This cluster 
also recorded highest mean value for frond production and leaf area ration while it had 
lowest height value and carotene content. 
46 
Ward's mev od using Squared Eu idean ds 
33C3 
IN EN 
IN C 
IN 301 
IN 12. ý 
N1? 03- 
. 
IN 3.0i- 
IN 12 01 - 
N 12.02 
F"i'? I- 
SEN01m- 
INI N 
IN 10-- 
IN 10.03 
IN01ý-_ 
1NÜvC2 
0wuýý 
INC111 
IN 
ýN 03 0+ 
cN ; 3.1v 
qN1: 01 
FNCi}: 
IN Ol? 
GlY a02 
IN x! 31 
IN 'J 03 
XNý02 
INCa02 
,. - 
sNvrC - 
NiC)1 
xti02ýb--- 
ýN9? D'; -- 
x'N0? 03- 
RN 02ib 
"s 
Ralamv, s' I, a 
Figure 8: The relationship among the oil palm germplasm accessions reflected by 
cluster analysis 
47 
Cluster-III was noted for having the highest mean value for most of the yield traits as 
well as morpho-physiological traits. This cluster had highest mean value of FFB, 
BWT, MFW, P/B, M/F, O/DM, O/WM, O/B, OY, TEP, LW, LA, LAI,. r BDM, 
TDM, BI, c, NAR and moderate values for fatty acid traits. 
Cluster-IV was characterized by lowest mean values for most of the yield traits and 
highest mean value of kernel to fruit ratio and trunk diameter. Also noted in this 
cluster was high shell to fruit ratio, petiole cross section, leaflet number and moderate 
leaflet length. 
Cluster-V had moderate to high values of most traits under study. The accessions in 
therein also had highest mean value of fruit to bunch ratio and carotene content. 
Cluster-VI had medium to high values for most traits under study. They however had, 
highest mean value for kernel yield and arichidic acid and high carotene content. 
48 
Table 5: Characteristics means of six clusters generated by Ward's cluster analysis 
based on 46 agronomic traits 
Cluster-I Cluster-41 Cluster-III Cluster-IV Cluster-V Cluster-VI 
FFB 106.04 141.68 197.68 49.15 62.51 88.51 
BNO 21.00 18.88 15.86 14.22 16.33 20.19 
ABW 5.05 8.83 12.55 3.37 3.88 4.36 
BWT 3.93 8.49 12.57 2.63 3.41 3.88 
MFW 2.65 6.97 12.50 2.31 3.11 2.51 
MN W 1.91 1.65 1.89 1.53 1.79 1.58 
P/B 0.00 0.55 5.10 0.10 0.24 0.37 
M/F 28.09 57.40 84.31 33.80 38.67 35.12 
K/F 13.29 9.03 6.41 16.70 14.96 15.54 
S/F 58.62 33.58 9.28 49.50 46.37 49.34 
O/DM 61.73 75.15 77.37 68.49 70.43 70.33 
O/WM 32.13 46.11 49.68 39.58 41.66 41.42 
F/B 50.76 57.35 55.60 58.20 60.22 57.33 
O/B 4.58 15.50 23.12 7.79 10.21 8.51 
K/3 6.78 5.29 3.62 9.63 9.02 8.89 
OY 4.86 27.17 45.76 3.36 7.12 7.57 
KY 7.19 5.67 7.16 4.28 6.03 7.79 
TEP 9.17 30.57 50.05 5.93 10.73 12.25 
FP 30.00 33.50 28.14 28.80 31.76 31.74 
PCS 29.52 17.88 27.53 20.13 12.87 16.01 
RL 4.45 3.94 5.00 4.13 3.47 3.85 
LL 107.20 88.54 88.86 86.60 80.08 86.44 
LW 4.57 4.98 5.16 4.10 3.91 3.93 
LN 136.00 125.50 158.07 130.73 121.34 122.35 
HT 3.39 2.60 2.58 2.77 3.01 2.81 
LA 7.59 6.40 8.28 5.24 4.36 4.74 
LAI 4.49 3.79 4.90 3.10 2.58 2.81 
DIAM 0.70 0.69 0.74 0.76 0.61 0.66 
F 0.86 0.78 0.88 0.72 0.65 0.68 
LAR 12.36 17.03 14.60 11.12 14.17 13.06 
BDM 8.32 11.11 15.81 3.46 4.95 6.96 
VDM 18.43 12.77 16.15 13.74 10.00 1 1.74 
TDM 26.75 23.89 31.96 17.20 14.95 18.70 
BI 0.31 0.45 0.49 0.19 0.32 0.37 
F 1.00 0.96 1.18 0.77 0.74 0.89 
NAR 11.46 11.76 12.88 10.81 11.25 13.15 
C140 0.30 1.13 0.52 0.36 0.65 0.49 
C160 44.76 39.74 38.69 37.18 40.27 39.40 
C161 0.53 0.10 0.13 0.12 0.12 0.16 
C180 2.20 4.03 5.32 5.70 4.89 5.20 
C181 43.45 41.52 45.07 47.42 42.85 44.56 
C182 8.16 13.29 9.94 8.93 10.91 9.86 
C183 0.60 0.10 0.18 0.17 0.18 0.19 
C200 0.00 0.07 0.12 0.10 0.08 0.14 
IV 53.58 59.09 56.58 56.82 56.34 56.05 
Carotene 1575.49 1421.42 1607.71 1624.35 1793.83 1638.94 
Figures in hold are maximum values 
49 
4.3.1 Genetic Distance 
The genetic distance of all the oil palm germplasm as analyzed by the proximity 
matrix of squared Euclidean distance (see Appendix 1), revealed that the largest 
genetic distance was between SEN 13.07 and SSC 3 (520.302) followed by SEN 
02.08 and SSC 3 (454.034). The least distance was found between SEN 02.06 and 
SEN 07.08 (10.366). The inter cluster distance as shown in Table 5 also revealed that 
the largest genetic distance was between cluster-III and cluster-V (158.629) while the 
least was between cluster-V and cluster-VI. 
Table 6: Inter cluster distance as analyzed by proximity matrix of squared Euclidean 
distance 
Case 
Squared Euclidean Distance 
1: Cluster-I 2: Cluster-11 3: Cluster-III 4: Cluster-IV 5: Cluster-V 6: Cluster-VI 
1: Cluster-I 0.000 118.586 149.386 99.442 107.001 86.122 
2: Cluster-11 118.586 0.000 74.408 91.789 63.147 56.828 
3: Cluster-III 149.386 74.408 0.000 151.442 158.627 126.870 
4: Cluster-IV 99.442 91.789 151.442 0.000 38.770 36.825 
5: Cluster-V 107.001 63.147 158.627 38.770 0.000 20.757 
6: Cluster-VI 86.122 56.828 126.870 36.825 20.757 0.000 
CHAPTER V 
DISCUSSION 
A clear relationship between agronomic traits as stated by Deyong (2011) will not 
only address the fundamental principle of plant breeding but will also facilitate the 
plant breeding practice. Most of the characters such as the agronomic traits 
investigated in the current study contribute to crop yield. The high variability unveiled 
in some of the yield, morpho-physiological and fatty acid traits are essential and 
important for efficient selection of elite oil palm accessions with high yield and yield 
component characters. This finding supports the result reported by Li-Hammed et al. 
(2014) in variation and correlation studies of MPOB-Nigerian oil palm germplasm. 
The existence of variation is nonetheless vital for diversity and adaptability in 
breeding population (Marhalil et al., 2013). 
5.1 Principal Component Analysis (PCA) 
Deyong (2011) stated that when agronomic character contributed mostly to yield, it 
complicates the designing of an ideal crop architecture and that finding out several 
typical agronomic traits of a crop is of assistance in architecture designing and / or in 
designing high crop yield. This conception according to Deyong (2011) could be 
materialücd through multivariate statistical analysis. PCA through its dimension 
reduction method is of immense help in knowing the traits contributing most to 
variation. In present study, nine PCs with Eigen values greater than one and total 
cumulative variance of 88% were extracted from the numerous variables through PCA 
with ABW and BWT, PCS, BNO, C16: 0, carotene, MNW, HT, IV and K/B 
51 
contributing mostly to PC 1, PC 2, PC 3, PC 4, PC 5, PC 6, PC 7, PC 8 and PC 9 
respectively. This translates that oil palm accession with higher scores for these traits 
seems probably will attain high yield more easily and as a result, these traits should be 
given utmost importance in breeding program of oil paten germplasm under study. 
This result follows similar trends to that of Deyong (2011) in analysis among main 
agronomic traits of spring wheat. Furthermore, it can also be observed from the PCA 
results that most of the yield contributing traits was poor on other PCs except on PC 1. 
From the findings of this study, it is obvious that a good hybridization breeding 
program can be initiated by the selection of genotypes from PC I and PC 2. This 
result also agrees with the findings of Magbool et al. (2010) on morphological 
diversity and traits association in bread wheat. 
5.1.1 Score Plot 
The score plot is a visual aid for determining an appropriate number of PCs. It shows 
the eigenvaluc against the component number. The eigenvalues measure the amount 
of variation explained by each PC and will be largest for the first PC and smaller for 
the subsequent PCs. An eigenvalue of greater than one indicates PCs accounted for 
more variance than accounted by one of the original variables in standardized data and 
it is commonly used as a benchmark for which PCs are retained (Fernadez, 2002). The 
score plot in this study showed that maximum variation was present in the first PC and 
according to Maqbool et al. (2010), selection of genotypes from PC 1 will be useful. 
5.1.2 Loadings 
PC loadings arc correlation coefficients between the PC scores and the original 
variables. PC loadings measure the importance of each variable in accounting for the 
52 
variability in the PC (Fernandez, 2010). In this study, variables on the left quadrant 
with high loadings on PC 1 include TDM, P/B, MFW, ABW, BWT, OY, M/F, OY, 
TEP, FFB, BDM, O/B and e while those on the right quadrant with high loadings 
include S/F, K/F and K/B; these set of variables can be said to be anti-correlated. That 
is to say, those on the left quadrant are traits conferring high yield to palms while 
those on the right quadrant are not yield contributing traits and oil palm with high 
percentage of S%F. K/F and K/B will be low in yield. This is evident from the strong 
negative correlation that was unveiled between variables and FFB and other yield 
traits in the research carried out by Li-Haimned et al. (2014) on association between 
oil palm traits. 
Variables with high contribution to PC 2 include BNO, PCS, RL, FP, DIAM and BI. 
Also from the loadings plot, variables that lie close together along the same PC can be 
said to be highly correlated (CAMO, 2014). As stated by Fernandez (2010), high 
correlation between PC 1 and a variable indicates that the variable is associated with 
the direction of the maximum amount of variation in the dataset al. so, more than one 
variable might have a high correlation between with PC 1. A strong correlation 
between a variable and PC 2 indicates that the variable is responsible for the next 
largest variation in the data perpendicular to PC 1 and so on. 
Furthermore on the PC, some variables particularly amongst the fatty acid traits had 
no significant loadings on any of the extracted PCs. This suggests that the variables 
have little or no contribution to the variation in the Senegal and Gambia oil palm 
germplasm. Theretbre, PCA may often indicate which variables in a dataset are 
important and which ones maybe of little importance (Fernandez, 2010). 
53 
5.1.3 Scores 
PC scores are the derived composite scores computed for each observation based on 
the eigenvectors for each PC (Fernandez, 2010). From the scores of the oil palm 
gcrmplasm summarized in Table 3, oil palm genotypes with high scores on PC 1 can 
be said to be the most diverse. Those with high negative scores, i. e., SSC 3, SEN 
09.04 and SEN 01.02 can be said to compliment variables with high negative loadings 
on PC I while those with high positive scores, i. e., SEN 13.07 and SEN 02.08 are 
complementary to variables with high positive loadings on PC 1 (CAMO, 2014). The 
oil palm genotypes with high negative scores are of the high yield type as yield traits 
as observed from the loading plots were loaded negatively on PC 1 (38 traits) while 
the non-yield traits were also positively loaded on PC I (for 11 traits) and genotypes 
with high positive scores can be said to be non-yield type. Therefore, from the scores 
given to the Senegal and Gambia oil palm germplasm, breeders can select genotypes 
with highest score having desirable characters for further breeding programs. 
Furthermore, score plot which is a dimensional scatter plot signifies how well the data 
is distributed and gives information about the samples. In the scores plot of the present 
study, oil palm genotypes that are closer to one another have close values for the 
corresponding variables while those that are far away from one another are quite 
different in values for corresponding variables (CAMO, 2014). The scores plot of the 
Senegal and Gambia materials portrayed that genotypes that are close together are 
sensed as being similar when rated on all the variables studied while genotypes which 
are further apart are more diverse from other accessions (Smiullah, 2013). 
In the present study, SSC 3 was the most distant oil palm genotype from other oil 
º, aºm accessions. The reason for this is glaring distance as SSC 3 is a standard cross; 
54 
i. e. hybrid between a deu-a and a pisi/era and standard crops also known as tenera are 
known to be f high commercial value as they are high yielding genotypes as compared 
to their aura counterparts (pisifc'ra is normally trail) used in this study (Teoh, 2002; 
Corley & Tinker, 2003). Besides fi-om SSC 3, genotypes which are also diverse as can 
be seen from the scores plot include SEN 01.02, SEN 09.04 and GAM 05.08. The 
Gambia material GAM 05.08 was expected to be different because it is not of the 
same origin with the "SEN" genotypes. 
5.1.4 Bi-Plot 
Bi-plot display is a visualization technique for investigating the inter-relationships 
between the observations and variables in multivariate data (Fernendez, 2010). From 
the bi-plot (Figure 5), genotypes SSC 3, SEN 09.04 and SEN 01.02 will be good 
choice for genetic improvement. GAM 05.08 will be the best choice for high carotene 
palm. The result of this is in conformity with that of Doumbia et al. (2013) who used 
the hi-plot graph to suggest good candidates of cowpea accessions to be used in 
genetic improvement of the crop. 
5.2 Cluster Analysis 
Estimating genetic diversity and determining the relationships between collections are 
very useful for ensuring for ensuring efficient germplasrn collections and different 
markers are available for studying variability among accessions (Rabbani et al., 1998). 
Several techniques have also been used to classify and measure the patterns of 
phenotypic diversity in the relationships of species and germplasm collections for a 
variety of crops. However, morphological characterization constitutes the first step in 
the description and classification of germplasm (Mostafa & Harried, 2011). 
55 
Doumbia et al. (2013) stated that PCA alone may not give a clear character 
representation in terms of their contribution to genetic diversity and hence, the need 
for PCA to be complemented with other techniques such as cluster analysis which 
provides more information about the relative positions of the accessions. Cluster 
analysis of the germplasm resources is helpful for parental selection in the plant 
breeding program (Deyong, 2011). In this study, all the Senegal and Gambia oil palm 
germplasm were clustered into six types with each group having its own peculiar 
characters. Such information would be effective in selecting parental materials to 
breed new expected oil palm varieties. The MPOB-Nigerian oil palm germplasm also 
grouped into eight types by cluster analysis (Li-Hammed et al. 2014). 
Though cluster analysis was able to group accessions with greater morphological 
similarity together, the grouping did not necessarily group accessions from the same 
origin together. This can be observed in the grouping of Gambia materials which were 
not in the same cluster group but were found grouped together with some of the 
Senegal materials. This shows that there is no consistency between geographical 
origin and genetic distance. Seymus and Brulent (2010), Lacis et al. (2010), Talebi 
and Rockzadi (2013) and Ajmal et al. (2013) also reported lack of relationship 
between geographical origin and distance. Sonnante and Pignone (2007) were of the 
opinion that the association between genetic similarity and geographic distance among 
genotypes is not always clear. This may be due to migration of the oil palm materials 
from one region to another in collection site. 
Based on the means value for each cluster, cluster-I, cluster-II and cluster-III 
contained genotypes with high yield characters and as stated by Ajmal et al. (2013), 
increased yield potential is a stated goal for plant breeders. Hence, these genotypes 
56 
could be exploited for their release as high yielding accessions after testing them on a 
wide range of environments. Furthermore, these genotypes can also be utilized as 
parents in hybridization programs to develop high yielding oil palm varieties. This 
finding is in conformity with research of Ajmal et al. (2013). Also cluster-II had the 
lowest height and hence, the genotype in this group could be used for breeding of 
short palms as it is a preferred trait in oil palm breeding for easy harvesting of fi-esh 
fruit hunch (Sapey et al., 2012). Groups with desired traits can also be exploited 
directly for such traits. 
5.2.1 Genetic Distance 
According to squared Euclidean distances (D) among genotypes, the largest genetic 
distance was between SEN 13.07 and SSC 3; this was followed by SEN 02.08 and 
SSC 3. Hence, crosses between morphologically distant genotypes will result in 
maximum hetcrosis. The importance of genetic diversity to maximum heterosis has 
been reported by many researchers. WeldeMicheal et al. (2013) reported maximum 
heterosis in some germplasm accessions of some Ethiopian specialty coffee based on 
morphological traits. 
Genotypes with the largest genetic distance between them can be hybridized as 
hybrids of maximum distance result in high yield. On the other hand, the least genetic 
distance was recorded in SEN 02.06 and SEN 07.08. Therefore, crosses between 
genotypes of close proximity should be avoided. However, Rahim et al. (2010) 
suggested that crosses between close genotypes could be useful for backcross 
breeding programs. Similar to the findings of Khodadadi et al. (2010), who reported 
that cluster analysis can be used for finding high yielding genotypes in wheat. 
CHAPTER VI 
CONCLUSIONS AND RECOMMENDATIONS 
Morphological diversity among the Senegal and Gambia oil palm germplasm was well 
defined by both principal component and clustering analyses. Considering the 
different morpho-bio-agronomic descriptors, it has been possible to observe a 
noteworthy inter and intra-group diversity. The characters that were dominants in the 
first components are closely related to yield and yield components; while vegetative 
and physiological traits like the radiation conversion efficiency (e), fi-actional 
interception of radiation (/) and leaf characters were associated with second 
components and other extracted components. The fatty acid traits were not that 
dominant on the extracted PCs. This recommends the likelihood of attaining, through 
selection. suitable genotypes combining high yield with desirable traits for direct 
release as cultivars in the Malaysian Palm Oil Board (MPOB). 
Cluster analysis also aided in distinguishing accessions on the basis of their different 
levels of similarity. Six groups were identified with precise differences according to 
the extracted principal components. Compared with other groups, cluster-III showed 
highest mean values for yield traits. With the exception of cluster-IV which had lowest 
mean value for yield traits, all other groups had moderate to high values of yield and 
yield related traits. The cluster analysis however, provides valuable information in 
order to utilize directly the most promising accessions for production (for example 
SSC I, SEN 09.04) or for future usage in selection programs. Furthermore, there was 
no consistency in assigning of members to groups, as accessions from different 
geographical origin were placed together. 
58 
However, caution must be taken as regard the accessions because they are an 
expression of linked genetic and environmental effects, i. e. findings were based on 
morphological traits which are usually influenced by environment. Therefore, further 
studies should be undertaken to ascertain the claims of this study. For example, 
molecular techniques which are more precise and are not affected by the environment 
can be conducted to further verify the findings from this study. The findings of this 
research will however be of utmost help to researchers, plant breeders and policy 
makers. 
59 
REFERENCES 
Adebisi, M. A., F. S. Okelola, M. O. Ajala, TO. Kehinde, I. O. Daniel & O. O. Ajani. 
2013. "Evaluation of Variations In Seed Vigour Characters of WestAfrican Rice 
(orýv: a sativa L. ) Genotypes Using Multivariate Technique". American Journal 
of'Plant Sciences, Vol. 4 (2), 2013. p. 356-363. 
Ajmal, S., N. M. Minhas, A. Hamdani, A. Shakir, M. Zubair & M. Ahmad. 2013. 
"Multivariate Analysis of Genetic Divergence in Wheat (Triticum acstivum) 
Germplasm". Pakistan Journal ofBotanv. Vol. 45. (5): p. 1643-1648. 
Ariyo, O. J. 1993. "Genetic Diversity in West African Okra (Abelmoschus cuillei L. 
Chev. ) Multivariate Analysis of Morphological and Agronomical 
Characteristics". Genetic Res. Crop Evol. Vo1.40. p. 25-32. 
Barcelos, E., P. Amblard, , J. Berthaud & M. Seguin. 2002. "Genetic Diversity and 
Relationship in American and African Oil Palm as Revealed by RFLP and AFLP 
Molecular Markers". Pesq. Agropec. Bras., Brasilia. Vol. 37. (8): p. 1105-1114. 
Blaak, G. 1967. Oil Palm Prospection Four in the Bamenda Highlands of West 
Cameroon. London: Internal report, Unilever. 
Bozokalfa, M. K., D. Esiyok & K. Turhan. 2009. "Patterns of Phenotypic Variation 
in a Germplasm Collection of Pepper (Cal)s"icum annum L. ) from Turkey". 
Spanish Journal o/ Agricultural Research. Vol. 7. (1): p. 83-95. 
Brereton, R. G. 2009. Chemometrics. f r Pattern Recognition. UK: John Wiley & Sons 
Ltd Publisher. 
Bradley, P. S, O. L. Mangasarian & W. N. Street. 1998. "Clustering via Concave 
Minimization, " in Advances in Neural Information Processing Systems, vol. 9, 
M. C. Mozer, M. I. Jordan, and T. Petsche, Eds. Cambridge, MA: MIT Press, 
1997, pp. 368-374 
Brereton. R. G. 2003. Chemometrics: Data analysis for the laboratory and chemical 
plant. England: John Wiley & Sons Ltd. 
CAMO. 2011.1 he Unscrumbler® X 2009-2011. Version 10.1 (32-bit). CAMO 
software AS. 
CAMO. 2011. What is Mn, ultii'uriute Anulvsis? USA: CAMO Software AS. 
CAMO. 2014. http: //www. camo. com/resources/methods. html. (Assessed on 6°i 
January 2014). 
Chavclicr, A. 1943. "Taxonomic, Biogeographie et Selection des Palmiers au Genre 
Elueis". Bot. App)[ Agric. Trojp. Vol. 23. (295). 
Chong, C. L. 1994. "Chemical and Physical Properties of Palm Oil and Palm Kernel 
Oil" in Selected Readings on Palm Oil And Its Uses. Kuala Lumpur, Malaysia: 
Palm Oil Research Institute of Malaysia. 
60 
Corley, R. H. V. & P. B. Tinker. 2003. The Oil Palm. 4th Edition. UK: Wiley-Blackwell. 
ISBN: 978-0-632-05212-7. 
Darvishzadeh R., M. Dalkani & A. Hassani. 2012. "Determination of the genetic 
Variation in Ajowan (Carum Copticum L. ) Populations Using Multivariate 
Statistical Techniques". Revista Ciencia Agronomica, Vol. 43 (4): p. 698-705. 
December, 2013. 
Deyong, Z. 2011. "Analysis among Main Agronomic Traits of Spring Wheat (Triticum 
aestivun1) in Qinghai Tibet Plateau". Bulgarian Journal ofAgricultural Science. 
Vol. 17. (5): p. 615-622. 
Doumbia, I. Z., R. Akromali & J. Y. Asibuo. 2013. "Comparative Study of Cowpea 
Germplasms Diversity from Ghana and Mali using Morphological 
Characteristics". Journal of'Plant Breeding and Genetics. Vol. 1. (3): p. 139-147. 
Dransfield, J., N. W. Uhl, C. B. Asmussen, W. J. Baker, M. M. Harley & C. E. Lewis. 
2005. "A New Phylogenic Classification of the Palm Family, Arecaceae". Kew 
Bulletin. Vol. 60. (40): p. 559-569. 
Esbensen, K. H., D. Guyot, F. Westad & L. P. Houmoller. 2002. Multivariate Data 
Analysis in Practice . an Introduction to Multivariate Data Analysis and 
Experimental Design (Fifth Ed. ). Norway: CAMO Process Publisher. 
FAO. (1996) Report on the state ofthe world's plant genetic resources Tor food and 
agriculture. Rome: Food and Agriculture Organisation of the United Nations. 
Fernandez G. 2010. Principal component analysis. RS701 D. 
http: //www. cabnr. unr. edu/saito/classes/ers701/pca2. pdf. (Assessed on 30°i 
October 2014). 
Franco, J., J. Crossa, M. L. Warburton, S. Taba & S. A. Eberhart. 2006. "Sampling 
strategies for conserving maize diversity when forming core subsets using 
genetic markers". Crop Science. Vol. 46. p. 854-864. 
Gonzalez, A. G. 2012. "Critical Aspects of Supervised Pattern Recognition Methods 
for Interpreting Compositional Data" in Chemometrics in Practical 
. 4pplicutions". 
Varmuza, K. (Ed. ). China: Intech Publisher. 
Hall R. I. P. R. Leavitt, R. Quinlan, A. S. Dixit & J. P. Smol. 1999. "Effects of 
Agriculture, Urbanization, and Climate on Water Quality in the Northern Great 
Plains". Limnol. Oceunogr. Vol. 44. (3-2): p. 739-756. 
Hammer, K., N. Arrowsmith & T. Gladis. 2003. "Agrobiodiversity with emphasis on 
plant genetic resources". Naturwis. censchuftcn. Vol. 90. p. 241-250. 
Hamrick. J. L. & M. J. W. Godt. 1996. "Effects of life history traits on genetic diversity 
in plant species". Philosophical Transactions of the Roval Society B: Biological 
Sciences. Vol. 351. (1345): p. 1291-1298. 
Hatcher. L. & E. Stepanski. 1994. A step-by-step approach to using the SAS system 
for univariate and multivariate statistics. Cary, NC: SAS Institute Inc. 
61 
Haydar, A., M. B. Ahmed, M. M. Hannan, M. A. Razvy, M. A. Manal, M. Salahin, 
R. Karim & M. Hossain. 2007. "Analysis of Genetic Diversily in Some Potato 
Varieties Grown in Bangladesh". Middle East Journal of Scientific Research. 
Vol. 2. (3-4): p. 143-145. 
Hoon, M., S. Imoto & S. Miyano. 2013. The Clustering Library for cDNA Microarrav 
Data. Japan. The University of Tokyo: Institute of Medical Science, Human 
Gerome Center. 
Hornovoka, 0., M. Zavodna, M. Zakova, J. Kraic & F. Debre. 2003. "Diversity of 
Common Bean Landraces Collected in The Western And Eastern Carpatien". 
Check Journal of'Genetic Plant Breeding. Vol. 39. p. 73-83. 
Jain, A. K. & R. C. Dubes, Algorithms for Clustering Data. Prentice-Hall, 1988. 
Jalani, B. S. 1998. "Research and development of oil paten towards the next 
millennium". 1998 International Oil Palm Con/erence. IOPRI and GAPKI, Bali, 
Indonesia. 
Jalani, B. S. 2012. Malaysian Oil Palm Industy, Contribution, Challenges und Future 
Prospects. 71800, Bandar Baru, Nilai, Negeri Sembilan, Malaysia: Universiti 
Sains Islam Malaysia. 
Jolliffe. I. T. 1986. Principal Component Analysis. Berlin: Springer-Verlag. 
Jolliffe, I. T. 2002. Principal Component Analysis. (2nd Ed. ). UK: Springer Publisher. 
Karp, A., S. Kresovich, K. V. Bhat, W. G. Ayad & T. Hodgkin. 1997. Molecular 
100/ i/1 /)leant genetic resources conservation: a guide to the technologies. 
IPGR1 Technical Bulletin No. 2. Rome, Italy: International Plant Genetic 
Resources Institute. 
Khodadadi, M., M. H. Fotokian & M. Miransari. 2011. "Genetic Diversity of wheat 
(7riticum acstivum L. ) Genotypes Based on Cluster and Principal Component 
Analyses for Breeding Strategies". Australian Journal of Crop Science. Vol. 5. 
(1): p. 17-24. 
Lacis, G., V. Trajkovski & I. Rashid. 2010. "Phenotypical Variability and Genetic 
Diversity within Accessions of the Swedish Sour Cherry (Prunus cerasus L. ) 
Genetic Resources Collection". Biologija. Vol. 56. (1-4): p. 1-8. 
Latif A. 2000. "The biology of the Genus Elaeis" in Advances in Oil Palm Reseurch. 
B. Yusof, B. S. Jalani, & K. W. Chan (ed. ). Bangi: Malaysian Palm Oil Board, 
Bangi. p. 19-38. 
Li-Hammed, M. A., A. Kushairi, N. Rajanaidu, H. Mohd Sukri, Che Wan Zanariah, 
C. W. Ngah & B. S. Jalani. 2014. Correlation Studies in the MPOB-Nigerian Oil 
Palm (Elac'is guinc'ensis Jacq. ) Germplasm. 13th Symposium of the Malaysian 
Society Of Applied Biologi;. Cherating, Pahang, Malaysia. 8-10 June 2014. 
Li-Hammed, M. A. 2014. Principal Component Analysis and Clustering of Ex Situ 
oil Palm (Elacis guineensis Jacq. ) Germplusm. (MSc Thesis). Nilai, Universiti 
62 
Sains Islam Malaysia. 
Lohani, M., D. Singh & J. P. Singh. 2012. "Genetic Diversity Assessment Through 
Principal Component Analysis in Potato (Solununn tuherosum L. )". Vegetable 
Science. Vol. 39. (2): p. 207-209. 
Mandal, P. K. 2008. Bunch Analysis of Oil Palm. Padavegi: National Research Centre 
for Oil Palm (NRCOP). 
Manprect S., K. Keerat & S. Bhavdeep. 2008. "Cluster Algorithm For Genetic 
Diversity". World Academy of Science, Engineering and Technology. Vol. 18. 
(2008). 
Maqbool, R, M. Sajjad & I. Khaliq. 2010. "Morphological diversity and traits 
association in bread wheat (Triticum aestii'um L. )". American-Eurasian Journal 
of'Agriculture and Environmental Science. Vol. 8. (2): p. 216-224. 
Martens, H. & T. Naes. 1993. Multivariate Calibration. New York: Wiley. 
Martinez-Calco, J., A. D. Gisbert, M. C. Alarnar, R. Hernandorena, C. Romero, G. 
Ilacer & M. L Badenes. 2005. "Study of a germplasm collection of Loquat 
(Eriohotnur 
, 
japonica Lindi) by multivariate alaysis". Genetic Resources and 
Crop Evolution. Vol. 55. (5): p. 697-703. 
Matthias, 0.2007. Chemornetrics. Weinheim: WILEY-VCH Verlag GmbH & Co. 
Mostafa, A. & H. Felenji. 2011. "Evaluating Diversity among Potato Cultivars Using 
Agro-Morphological and Yield Components in Fall Cultivation of Jiroft Area". 
American-Eurasian Journal of Agricultural and Ennvironrnc'ntul Science. 11(5): 
655-662. 
Muhammad, A.. S. A. Jatoi, T. Rafique & Abdul Ghafoor. 2013. "Genetic Divergence 
in Indigenous Spinach Genetic Resources for Agronomic Performance and 
Implication of Multivariate Analyses for Future Selection". Science, Technologv 
and DeVelohn7ent. Vol. 32. (2): p. 7-15. 
Negri, V. & B. Tiranti. 2010. "Effectiveness of in situ and ex situ conservation of crop 
diversity. What a Phaseolits vulgar-is L. landrace case study can tell us". 
Genetica. Vol. 138. p. 985-998. 
Novembre. J. & S. Matthew. 2008. "Interpreting Principal Component Analysis of 
Spatial Population Genetic Variation". Nature Genetics. p. 646-649. 
Oil world. 2008. Oil world 2007 annual report. Mielke, Hamburg. 
Oil World 2012. Oil world. Hamburg, Germany. 
Oyelola, B. A. 2004. (Paper). The Nigerian Statistical Association Preconference 
Workshop. Conference centre, University of Ibadan, Nigeria. 20-21 September. 
Peerak S., T. Yimram & P. Somta. 2009. "Genetic Variation in Cultivated Mungbean 
germpalsm and its Implication in Breeding for High Yield". Field Crops 
research. Vol. 1. (12): p. 260-266. 
63 
Preston, J. 2011. The characterization of heritable vegetable. (PhD Thesis). 
University of Birmingham. 
Purseglove, J. W. 1972. Tropical Crops. Monocotyledons. London: Longman Group 
Ltd., London. p. 479-510. 
Rabbani M. A.. A. Iwabuchi, Y. Murakami, T. Suzuki & K. Takayanagi. 1998. 
"Genetic Diversity of Mustard (Brussicu, juncea L. ) Gennplasm from Pakistan 
as Determined by RAPDs". Euphvticu. Vol. 103. (2): p. 235-242. 
Rahim M. A., A. A Mia, F. Mahmud, N. Zeba & K. Afrin. 2010. "Genetic Variability, 
Character Association and Genetic Divergence in Mungbean (Vignu radiate L. 
Wilczek)". Plant Omic. Vol. 3. p. 1-6. 
Rajanaidu, N. & B. S. Jalani. 1994. "Oil Palm Genetic Resources, Collection, 
Evaluation, Utilization and Conservation". (Paper). PORIM Colloquium on Oil 
Palm Genetic Resources. PORIM, Bangi. 13 September. 
Rajanaidu, N. 1994. "Oil palm cultivation and FFB production" in Selected Readings 
on Palm Oil and Its Uscs. Kuala Lumpur, Malaysia: Palm Oil Research Institute 
of Malaysia. p. 11. 
Rajanaidu, N., B. S. Jalani, A. Kushairi & V. Rao. 1999. "Oil palm genetic resources- 
collection, evaluation, utilization and conservation" in Proceeding of'the 
symposium on the science of oil palm breeding. N. Rajanaidu & B. S. Jalani 
(cd. ). Bangi, Malaysia: PORIM. 
Ravishanker, S. D. K. Kumar, Baranwal, A. Chatterjee & S. S. Solankey. 2013. 
"Genetic Diversity Based on Cluster and Principal Component Analysis for 
Yield and Quality Attributes in Ginger (Zingiher officinale Roscoe)". 
International Journal of Plant breeding and Genetics. Vol. 7. (3): p. 159-168. 
Raychaudhuri, S.. J. M. Stuart & R. B. Altman. 2000. "Principal components analysis 
to summarize microarray experiments: application to sporulation time series". 
(Paper). Pacific Svmposium on Biocomputing. 
Rokach, L. & 0. Main-ion. 2010. "Clustering Methods" in Data Mining and 
Knolrlc'dgc Discovery Hand Book. London: Springer. p. 321-357. 
Sanimour R., Mustatä, S. Badr, & W. Tahr. 2007. "Genetic Variability of Some 
Quality Traits in Lathyrus Spp. Germplasm". Acta Agriculture Slovencia, 
Genctika. Vol 43 (1), p. 129-140. 
Sarni, K. A., I. Fawole, A. Ogunbayo, D. Tia, E. A. Sornado, K. Futakuchi, M. Sie, F. E. 
Nwilene & R. G. Guei. 2010. "Multivariate Analysis of Diversity of Landrace 
Rice Germplasm". (Paper). Second African Rice Congress. Innovation and 
Partnerships to Realize Africa's Rice Potential. 22-26 March. 
Shivani, D., & Ch. Srelakshmi. 2014. "Assessment of Genetic Diversity in Indigenous 
Germplasm Lines Safflower (Curthumus tinctorius L. )". Canadian Journal of 
Plant Breeding. Vol. 2. (1): p. 1-4. 
64 
Smiullah, F. A., A. Khan, Afzal, Abdullah, U. Ijaz & R. Iftikher. 2013. "Genetic 
diversity assessment in sugarcane using principal component analysis (PCA)". 
International Journal of Modern Agriculture. Vol. 2. (1): p. 34-38. 
Sonnante, G. & D. Pignone. 2007. "The Major Italian landraces of Lentil (Lens 
culinaris Medik): Their Molecular Diversity and Possible Origin". Genetic 
Resources of Crop Evolution. Vol 54. p. 1023-1031. 
Spooner, D., R. van Treuren & M. C. de Vincente. 2005. Molecular Markers for 
Genchunk Mangainent. IPGRI Technical Bulletin No. 10. Rome, Italy: 
International Plant Genetic Resources Institute. 
Squire, G. R. 1986. "A Physiological Analysis of Oil Palm Trials". PORIM Buletin. 
Vol. 12. p. 12-31. 
StatSoft. 2013. Electronic Statistics Textbook. USA. 
http: //www. statsoft. com/Textbook/Cluster-Analysis#general. (Assessed on 3"' 
January 2014). 
Tan, P. N., M. Steinbach & V. Kumar. 2006. Introduction to Data Mining. USA: 
Pearson Addison-Wesley Publisher. 
Teoh, C. H. 2002.7hc palm oil industry in Mulavsia: from seed to 
. 
frying pan. 
Malaysia: Wild World Fund (WWF). 
Timms, R. E. 1978. "Artefact Peaks in The Preparation And Gas Chromatographic 
Determination of Methyl Esters". Australian Journal of Dairy Tcchnologv. 
March. p. 4-5. 
Uhl, N. W. & J. Dransfield. 1987. Genera Palmarum. A Classification ?f Palms Based 
on the Work of'Harold E. Moore Jr. Kansas: The L. H. Bailey Hortorium and the 
International Palm Society. Allen Press, Lawrence. p. 514-516. 
Ward, J. H. 1963. "Hierarchical Grouping to Optimize an Objective Function". Journul 
of lmerican Statistical A. tisociution. Vol. 58. p. 236-244. 
WeldeMichcal G., S. Alamerew, T. Kufa & T. Benti. 2013. "Genetic Diversity 
Analysis of Some Ethopian Specialty Coffee (Coffea Arabica L. ) Germplasm 
Accessions Based on Morphological Traits". Time Journals o/Agriculture and 
l'cterinuw Sciences. Vol 1 (4): p. 47-54. 
Wessels Boer, J. G. (1965). The Indigenous Palms of Suriname. palm. Vol. 18, No, 2. 
P 230-233. 
Whitemorc, T. C. 1973. The Palms o/'Malaya. Malaysia: Longmans. 
Willy, V (2010). "Growth and Production of Oil Palm" in Soils, Plant Growth and 
Crop Production VoIII. Encyclopedia of Life Support System. 
Wold, S. & M. Sjostrom, 1998. " Chemometrics, Present and Future Success". Journal 
of Chc'momc'tries and Intelligent Laboratory Systems. Vol. 44. p. 3-14. 
ä F' e "ý' pmý, ^ '-ýý oRe., Seä tC S "ý aSý, : s: PR m> 5ve.,. = Aý m.., aa5eýx S$ am eý`3 ý , ': x`' erR 7i ° .. 
ýs ss 
_^° e e° 
ea eýý e e^ 
rýs smraa äýy e aý ss e° 
x: 
°ýýýas ° 
-nmR -°°- 
. .. 
s -- xe 12 2em2e°; ý 
---------- -- -------- 
. .. o; 
s3 "` w _. 
"-., .,. 
ýaz.; s: °ý 
.. ý=ý"aü, ... 
a n; s" 
.- 
_ _ - - ý -ý ý -- ý --e >e- -ý- -ý --ý ý--° _---- - - 
,.. 'xaa Is ýa aý Yaa oa acaaraaY s° ra ss a° ýa aý amaý 
a= gi ý ,., 
. j: aaaa.. ^, aaasRRexaea 8{ eý aas a° v, s; r m__'e' a rv ,:; s .,.. - 
- -- -------------- 
... X: 
a°-_^k: x" ^a . -, 
ý to RöýR xs s° se ^°.., 
o. 
s; ýa (C :, 
ý^äaaa ; 1: ý, 
----- 
.. 
°äsý-äR ý' ýä ýý_ 7E ^- xx 7E ý_ rv e° EZ a%3 Ra F° }, y ,g £R IC Yi Nn aa%: 
>x 
LL 
narY '_' J, 
F"I rRc. R IC . ý, !Z... ai aaaa 
ti e=K° Ri S iý ö 3. K' ni "A b: ýýý-. 9 ýl 
---------------------- 
.., a$ 73 e %ý s. p CI S6.: , ss R° P' 
ý Fý, XY&RpRý`a >r qi ýi E. x `ss `"'vo v uni 
Fi -ý FA iý ýaN" 
eaee , -. . -, _. eem.. .,., e ._.., ,. 
?A 9i m"am 
, s', 
d E- a$ ýf Nmm2, 
.. >ü sý a6 ö Y° eve eý a ss öý mý xý asýx aý a aý .ý ý 
ý t, ,,. sý- 
Y Fi rmwx... ti ö Xi R . s; wRaY, rv IR ;2 : -: e -. .. 
"ý 
Cý P° RV Ký% 3pmxda 'a" amS 1Ö «_"u. 
.>. - ._e-a.., 
:., ^u-ae. °- ^° °'"---. mýae.. .., e , -. ..... 
1 
.., ° IE a .. .%ý 
ýýnL, 
. p° 7C a 7E ! L' ^ -: R !CY.., R %; _' F3 f R' 3%1 St öa iC Pt 7C . °. gý aaS . -, 
--------------- 
... " .-"a '_" .,: qa_. ^^$ , -, en rt; iC °"-i iC -_ '' a iC BýaM . ^f R .caRRY 
Yi =' ý 
:Y , -., Yi RR Gý'e8 Ü° ä k° _x 
IN '", "i^ . _. e R_ äS Id' °m . ^.. aNxý. ° °° Sý« m 
.. ':. 
S YA '^ äY ýR säma$rNýF. ý F, R 
ill ARr LY Re . 'E1 
ffim^ a '9 ^ý p pý . ^, 
x e2 gGnRw ö^ G 
.-- ---`-- 
ý 
... { ` 
.1 
... ,,.. ... .... .. . -. ., e,.. .....:. ... ..... .. ._.., e .. , -, ..... ,,, em.,, ,,, 
.... .ý .- . -. ,,. ... .. _ý