Multivariate analysis of phenotypic traits of indigenous sheep revealed new population in western part of Ethiopia

This study was carried out to phenotypically characterize and differentiate sheep populations of Western part of Ethiopia. Physical observations and body weight and linear measurements were studied. Multivariate canonical and discriminant analysis were employed to differentiate population. Three districts; Dendi, Guduru and Mida Kegn districts were selected purposely based on sheep production potential and previous characterization studies. For example, majority of male sheep in Dendi district had horn while majority are polled in Guduru and Mida Kegn districts. The two canonical variates extracted explain a total of 100% of the total variation (80.6% by Can1 and 19.4% by Can2). The Can1 and Can2 presented high weighing for ear length and pelvic width, respectively. It showed that these traits demonstrating their importance to discriminate and to classify the sheep population in the areas. Sheep population of Dendi, Guduru and Mida Kegn districts were correctly grouped into their own geographical area for about 71.15%, 54.62% and 69.23%, respectively. The largest mahalanobis distance was found between sheep population in Dendi and Guduru (2.27). The smallest distance was found between Guduru and Mida Kegni with mahalanobis distance of 0.649. These results showed that sheep population in Dendi was found to be unique and confi rmed the existence of two populations (Dendi and Guduru-Mida Kegn) in the study area. Research Article Multivariate analysis of phenotypic traits of indigenous sheep revealed new population in western part of Ethiopia Gobena Wagari1*, Tesfaye Getachew2 and Elias Bayou3 1Department of Animal Science, College of Agriculture, Oda Bultum University, Chiro, Ethiopia 2International Center for Agricultural Research in the Dry Areas (ICARDA), Addis Ababa, Ethiopia 3Department of Animal Science, College of Agriculture and Natural Resources, Mizan Tepi University, P. O. Box 260, Mizan Teferi, Ethiopia Received: 15 May, 2020 Accepted: 06 June, 2020 Published: 08 June, 2020 *Corresponding author: Gobena Wagari, Oda Bultum University, College of Agriculture, Department of Animal Science, Chiro, Ethiopia, E-mail:


Introduction
Ethiopia has a diverse sheep population and its distribution is paralleled with its diverse ecology. According to CSA (2016/17) [1] report, 30.7 million sheep are estimated to be found in the country with at least 9 sheep breeds and 14 traditional sheep populations [2]. Despite low level of productivity due to several technical (genotype, feeding and animal health), institutional, environmental and infrastructural constraints [3], indigenous sheep breeds have great potential to contributing more to the livelihoods of the people in low-input, small-holder crop livestock and pastoral production systems [4].
According to Kassahun and Solomon [5], Horro sheep breed is widely distributed in Shewa to East and West Wollega, body weight and body measurement variations which might be evidence of population stratifi cation in Wollo sheep breed [6]. Notable variation among indigenous population was observed within Wollo sheep population. This variation might be due to environment and /or genotype. Phenotypic comparison based on morphological characters can provide to an extent a reasonable representation of genetic differences among populations. Multivariate analyses of morphological traits have been proved to be suitable in assessing genetic variation within and between populations when all morphological variables are considered simultaneously. Discriminant analysis encompasses procedures for classifying observations into groups and describing the relative importance of variables for distinguishing among groups [7]. In this study, an attempt has been made to phenotypically characterize and differentiate sheep populations of western part of Ethiopia i.e. West Shewa and Horro Guduru Wollega zones based on phenotypic traits, using canonical discriminant analysis to see whether they are similar or distinct.
Although large number of sheep found in West Shewa [1], information on the morphological characteristics of sheep is limited. This population was classifi ed under Horro sheep breed and most of the characterization and research interventions are focus in Horro Guduru Wollega Zone. However, there was a hypothesis that the sheep breed in West Shewa particularly in and around Dendi district exhibit different morphological character. Thus, conducting this study is useful to plan different developmental strategies like community-based genetic improvement program, to obtain better knowledge of their present and potential future uses for food and agriculture in defi ned environments, and to recognize indigenous sheep exist in and around Dendi district as morphologically distinct sheep breed from or similar with Horro sheep breed based on their phenotypic traits. Thus, this study was aimed to describe qualitative and quantitative characteristics of indigenous sheep breeds and to investigate phenotypic variation among geographical areas using canonical discriminant analysis in different location of western Ethiopia.

Sampling technique
Dendi, Guduru and Mida Kegn districts were selected purposively for this study. Guduru district was selected to represent Horro Guduru Wollega area, which is beloved to be the main breeding tract and origin of Horro sheep (Horro area) where most of the previous sheep characterization and interventions were focused on [9]. Dendi district was selected to represent Ginchi sheep (central highland sheep) which is well known in Addis Ababa market. Mida Keng district was selected as the area found between Guduru and Dendi and population sampled from this district were helped to assess how the morphological characters behave as we move from Guduru to Dendi district in other way from Horro Guduru Wollega to West Shewa zone. From each district, two rural kebeles were selected purposively considering the sheep fl ock population.
According to FAO [10], physical measurements should be taken only from a representative set of adult animals (as judged by dentition): about 100-300 females and 10-30 males. Based on this, body weight and linear measurements were taken from a total of 780 adult sheep. Each experimental adult animal was classifi ed by sex, district, and dentition. Dentition record was included, as this was the only reliable means to estimate the approximate age of an animal. Adult sheep was classifi ed into two age groups; 1 pair of permanent incisor (1PPI) and ≥ 2 pairs of permanent incisor (2 PPI).

Data collection
Data was collected through fi eld measurement and observations which included morphological characters like: coat color type and pattern, head profi le, ear form, presence of wattle and horn, horn shape and orientation and tail shape were observed and recorded using a format developed for the purpose whereas body measurements: chest girth (CG), body length (BL), height at wither (HW), rump height (RH), pelvic width (PW), ear length (EL), horn length (HL), tail length (TL), tail circumference (TC) and scrotum circumference (SC) were collected using tailors measuring tape while body weight was measured using suspended spring balance having 50kg capacity. Linear body measurements were taken by restraining and holding the animals in a stable condition. All measurements were carried out by the same person in order to avoid inter-individual variations.

Statistical analysis
Both qualitative and quantitative data were analyzed by Statistical Analysis System (SAS, release 9.1, 2008). Qualitative data from individual animal observation was analyzed following the frequency procedures of SAS. GLM procedures of SAS were employed to analyze quantitative variables to determine effects of class variables (sex, district and dentition) and their interaction. For adult sheep; sex, district and age group of the sheep and their interaction were fi tted as independent variables while body weight and linear body measurements were fi tted as dependent variables. Scrotum circumference was analyzed by fi tting age group and district as fi xed factor.
Model for the least square mean analysis in females and males except scrotal circumference was: Y ijk =μ + D i + A j +S k + (AxS) jk + eijk Where: Y ijk = Observed body weight or linear body measurements, μ = Overall mean, D i = the fi xed effect of i th district (Dendi, Guduru, Mida Kegn), A j = the fi xed effect of j th age classes (1PPI, ≥2PPI), S k = the fi xed effects of k th sex (male, female), (AxS) jk =the interaction effects of j th age and k th sex and e ijk = random residual error Model for the least square mean analysis in males for scrotal circumference was: Where: Y ij = Observed/measured scrotal circumference, μ = Overall mean, D i = the fi xed effect of i th district (Dendi, Guduru, Mida Kegn), A j = the fi xed effect of j th age classes (1PPI, ≥2PPI) and e ij = random residual error Stepwise discriminant procedure was applied using PROC STEPDISC to determine which phenotypic traits have more discriminant power than others. The relative importance of the morphometric variables in discriminating of indigenous sheep types in the study area was assessed using the level of signifi cance (P<0.05) and partial R 2 values ≥0.01. The CANDISC procedure was used for univariate and multivariate one-way analysis that calculate the Mahalanobis distances of the phenotypic traits, and derive canonical functions (linear combinations of the continuous variables that summarize variation of sheep types among the areas. The ability of these canonical functions to assign each individual sheep to its sampling area was calculated as the percentage of correct assignment of each sheep types in each geographical area using the DISCRIM procedure. The number of misclassifi ed individuals indicates the degree of similarities between the populations.

Characterization of qualitative traits
Description of the physical characteristics of livestock breeds is very important for developing a breeding strategy in a particular production system [11]. The result indicated that all observed morphological characters were signifi cantly associated (P<0.01) with districts except presence of wattle (

Live body weight and linear measurements
Information on live body weight and linear measurements of the existing breed types has a vital role in the selection Program [6].

District effect
There was signifi cant difference for body weight and all recorded linear body measurements (P<0.01) among the districts. Average least squares means and standard errors of body weight and chest girth were 29.4±0.23, 28.2±0.23 and 29.8±0.24 for sheep in Dendi, Guduru and Mida Kegn, respectively ( Table 2)   shortage problem was more prevalent in Guduru district rather Taye, et al. [11] reported that the overall least square mean body weight, heart girth, body length and height at rump were affected by sex.

Age effect
Body weight and all recorded linear body measurements were signifi cantly (P<0.01) affected by age group of animal.
All measured traits show an increment with age of animal.

Multivariate analysis
Stepwise discriminant analysis: Step-wise discriminant analysis was fi rst applied to select the most important discriminator variables used for differentiation among sheep types under study. The result of the stepwise discriminant analysis is presented in Table 3. All of the nine variables considered in the analysis were found to be signifi cant (P<0.01 and Partial R 2 >0.01 to P<0.0001 and Partial R 2 0.1181). Ear length was found to be the most discriminating trait followed by body length, chest girth, height at wither and pelvic width in decreasing order of discriminating power. This implies that taking these basic measurements (EL, BL, CG, HW and PW) consistently could be more important in differentiating between the three geographical areas than acquiring numerous additional measurements.

Canonical discriminant analysis:
The univariate and multivariate test statistics are presented in Tables 4,5, respectively. The univariate statistic testing shows that all variables are highly signifi cant (P<0.0001) except chest girth (P<0.001). The resulting partial R 2 values ranged from 0.0173 for chest girth to 0.1181 for ear length. The multivariate test for differences among the districts is also highly signifi cant (P<0.0001), in all of the four types of multivariate tests (Wilks' Lambda, Pillia's Trace, Hotelling-Lawley Trace, and Roy's Greatest Root).
The total canonical structure, canonical correlation and total variation explained by each canonical variable for sheep populations are presented in Table 6. In this study, canonical discriminant analysis showed how each trait aligned to the two canonical variations; weighing each original trait according to its contribution on each canonical variable.
The fi rst canonical variable explained 80.6% of the total variation and Can2 explained 19.4% of the total variation, the two canonical varieties extracted explains a total of 100% of the total variation. EL, HW and RH had higher weighing in extracting Can1 while PW, BW and BL loaded high in Can2. The Mahalanobis distance matrix among districts is presented in Table 8. Mahalanobis distance was estimated from the mean values of variables having more discriminant power. All pair