Live body weight and linear body measurements of indigenous sheep population in their production system for developing suitable selection criteria in Central Zone of Tigray , Northern Ethiopia

The aim of this study was to characterize phenotype indigenous sheep population in central zone of Tigray. A total of 450 adult sheep were sampled randomly for this purpose. Based on dentition adult sheep were classified into four age categories. Average BW of the sampled sheep in Tanqua-Abergelle, Kola-Tembien and Adwa districts were (20.19±0.19, 22.18±0.22 and 23.68±0.33) kg, respectively. Location had highly significant (p<0.0001) effect on body weight (BW) and most of the linear body measurements (LBM). Adwa sheep were heavier than Tanqua-Abergelle sheep, but comparable with Kola-Tembien sheep. Sex of the animal had significant (p<0.0001) effect on BW and most of the LBM. BW in rams and ewes in the study area were 23.23±0.25 and 20.81±0.13 kg, respectively. Age group had significantly (p<0.0001) affected BW and most of the LBM. The trend of BW and LBM increased with increased dentition class. Highly significant (p<0.0001) correlations were observed between BW and most of the LBM. Chest girth (CG), rump length (RL), tail length (TL) and body length (BL) were found to be the most important traits in the regression model determining male BW accounting for variability of 74%, while for female CG, pelvis width (PW) and RL were the most important traits accounting 73% of the total variability of female BW. Based on the present result one may develop selection criteria and productivity schemes of the local sheep.


INTRODUCTION
Ethiopia is endowed with 29.33 million sheep (CSA, 2015) with diversified genetic pools adapted to a wide range of agro-ecologies.Environmental pressure also maintains a wide range of genotypes, each adapted to a specific set of circumstances (Getachew et al., 2010).At least 9 sheep breeds and 14 traditional sheep populations are found in Ethiopia (Gizaw et al., 2007).
Genetic improvement of the local livestock through appropriate techniques or selection and breeding programme is the need of the day (Yakubu, 2010).The usefulness of breed characterization of the indigenous livestock in general and sheep in particular is never in doubt, because characterization, inventory and monitoring of animal genetic resources (AnGR) are essential to their sustainable management and facilitate effective planning of how and where they can best be used and developed (FAO, 2015).
Even though the study area is potential in sheep production little works was done to characterize or to improve the indigenes sheep population of Tanqua-Abergelle, Kola-Tembien and Adwa districts (Gizaw, 2008;Tajebe et al., 2011) providing some information on some physical body measurements and characteristics, but it was limited.
Phenotype characterization of the existing sheep population, is the base for designing community based breed improvement and genotypic characterization, since genetic resources and production systems are not static and thus routine inventories and thus on-going monitoring is needed (Sölkner et al., 1998).The the objectives of this study were to characterize sheep in Central Zone of Tigray in their production system based on quantitative traits so that suitable selection criteria would be suggested.

Description of study area
The study was conducted in three districts of central zone of Tigray (Tanqua-Abergell, Kola-Tembien and Adwa) (Figure 1).The central zone of Tigray covers about 9741 square km with a total population of about 786,271 cattle, 406,018 sheep, 1, 139, 452 goat, 81,468 colonies of honey bee, and 1,390,782 poultry (CSA, 2015).The elevation of the area ranges from 1332 to 2921 m a.s.l.Annual rainfall is variable within a range of 466-758 mm.Temperature ranges from 14 to 22°C.Most of the lands are cultivated with some patchy grazing bottomlands and degraded hilly sites (CSA, 2015).

Site selection and sampling technique
From among the nine districts of central zone of Tigray, three districts were selected using multi-stage purposive sampling techniques, based on the sheep population density and road accessibility, in consultation with the zonal and districts bureau of agricultural experts.From each selected districts two rural kebeles (Felege-Hiwet and Gera from Tanqua Abergelle district, Werka-Emba and Debre-Tsehay from Kola-Tembien and Debre-Gent and Endamaryam-Shewito from Adwa) were selected purposively based on the sheep flock density and accessibility for transportation.Accordingly, a total of 450 healthy adult sheep (135 males, 315 females in the proportion of 30 males: 70 females) were selected randomly (209 from Tanqua-Abergelle, 143, from Kola-Tembien and 98 from Adwa district).The sheep were identified by sex, districts and four age groups (1PPI-1 pair of permanent incisor), 2PPI (2pair of permanent incisor), 3PPI (3 pair of permanent incisor) and 4PPI (4 pair of permanent incisors) for the base line data collection (characterization) as per Wilson and Durkin (1984).

Quantitative traits
The standard breed descriptor list for sheep developed by FAO (2012) was closely followed in selecting quantitative (body measurement) traits like: body weight (BW), body length (BL), wither height (WH), chest girth (CG), chest width (CW), rump length (RL), pelvis width (PW), ear length (EL), tail length (TL) and scrota circumference (SC).Measurements were made using flexible measuring tape while weight was measured using suspended spring balance having 50kg capacity with 200 g precision.Each experimental animal was identified by sex, districts and age group.Linear body measurements were taken by restraining and holding the animals in a stable condition.

Data management and analysis
General Linear Model (GLM) procedures of the Statistical Analysis System (SAS, release 9.2 2008) were employed to analyze quantitative variables to determine effects of class variables (sex, district and age class).Sex district and age were fitted as fixed effect while body weight and other linear body measurements were fitted as dependent variables.The effects of class variables and their interaction were expressed as Least Square Mean (LSM) ± SE.When analysis of variance declares significance, least square means were separated using adjusted Tukey-Kramer test.

Model used to analyze body weight and other linear body measurements except scrotal circumference were:
Yijk = µ+ Si + AK +Dj+ eijk Where: Yijk = the observed k th body weight / linear body measurements of the i th sex of the j th district µ=overall mean Si= the effect of i th sex (male and female) Dj = the effect of j th district (1 to 3 districts) AK = the effects of k th age (1, 2, 3 and 4pair of permanent incisor) eijk = random residual error

Model used to analyze the scrotal circumference was
Where Yij= the observations of linear body measurements /scrotal circumference the j th age group of the i th district µ=overall mean Ai=the effects of i th age group (i= 1, 2, 3 pair of permanent incisor) Dj= the effects of j th district (1, 2 and 3 districts) (AD)ij = the effect of interaction of j th age and i th district eijk= random error

Multivariate analysis
The quantitative variables or traits from female and male sheep were separately subjected to Discriminant analysis (PROC DISCRIM of SAS) and canonical Discriminant analysis (CAN DISC) program to ascertain the existence of population level phenotypic difference among the sample sheep population in the study area.

Pearson correlation
The relationships between body weight and other linear body measurements were calculated for male and female population separately using Pearson's correlation coefficient.

Multiple linear regression
The stepwise multiple linear regression analysis was done to obtain models for estimation of live body weight from other linear body measurements for males and females within each age group.To determine the best fitted regression equation for the prediction of body weight, step wise regression procedure of SAS was employed.Initially, selection of variables at (P<0.05) was employed by incorporating all variables at the same time to see the order of selected variables and then stepwise regression analysis was made.Best fitted model was selected based on the smaller values of Conceptual Predictive Criterion C(p), Akaike Information Criterion (AIC), Schwarz Bayesian Criteria(SBC), Root Mean Square Error (RMSE) and the higher value of adjusted-R 2 for simplicity of measurements under field condition to determine those traits that contribute much to the response variable (Kaps and Lamberson, 2004).The following models were used for the analysis of multiple linear regressions.

Analysis of live body weight and linear measurements
The least squares means ± Standard errors of body weight (BW) (kg) and linear body measurements (LBM) (cm) and fixed effects of location, sex and age group on body weight (kg) and LBMs (cm) for Tanqua-Abergelle, Kola-Tembien and Adwa sheep are presented in (Table 1).

Effect of location (district)
Location had highly significant (p<0.0001) for body weight (BW) and other linear body measurements except pelvic width (PW) and ear length (EL).Similarly scrotal circumference was not influenced (p>0.05) by location.The reason for the significance difference of body weight Higher body weight value in rams than in ewes indicated that this might be due hormonal differences and growth rate of the two sexes (Sobola, 2007).
The linear measurements of rams recorded in the study area were larger than the linear measurements of ewes except CW and PW.These listed body measurements of both sexes in the study area were higher than with the value of (Peter et al., 2013) reported in Djallonke and Sahel sheep in Northern Ghana for male and female 60.9±0.8,64.3±0.7,73.5±1.0 and 64.4±0.9 cm and 54.4±0.4,58.0±0.3,64.0±0.5 and 57.7±0.4 cm, respectively for BL, HW, CG, and RL.

Effect of age
Body weight and linear measurements of the sampled sheep populations (

Correlation between body weight and body measurements
The correlation coefficient among body weight and other linear body measurements are presented in (Table 3).
Strong positive correlation (P<0.001) between body weight and chest girth, height at wither, rump length, and body length with values of (0.79, 0.68, 0.64 and 0.62) respectively, was found in male sampled sheep population.The highest relationship between body weight and chest girth was observed in male for the pooled data with value of (0.79).Similarly, Pearson matrix correlation of female sampled sheep in pooled data also confirmed that strong positive correlation (P<0.001) between body weight and chest girths, rump length, pelvis width and tail length with the values of (0.81, 0.68, 0.62 and 0.56) in age classes 1-4 was observed.The highest relationship between body weight and chest girth with value of (0.81) observed in female for the pooled data sampled sheep in the study districts.
Similar to this study the strong positive correlation between the dependent variable body weight and the independent variable chest girth to predict body weight were observed in different previous works on sheep breed.For example, in North Wollo zone, Northern Ethiopia, Habru, Gubalafto district and in Selale Area, Central Ethiopia, Debre Libanos and Wuchale district of sampled sheep reported by Abera et al. (2014) and Mohammed et al. (2015) chest girth was the best variable for predicting live body weight than other linear body measurements for both male and female sampled sheep population.

Multiple linear regression analysis
Stepwise multiple linear regression analysis for male and female sampled sheep populations of Tanqua-Abergelle, Kola-Tembien and Adwa districts for predicting live body weight (LBW) from linear body measurements was found to have positive correlation with body weight.(Tables 4  and 5), shown the number of parameters entered in each step to predict the best fitted variables to estimate live body weight and their contribution in terms of adjusted coefficient of determination (R 2-adjusted ), mallows conceptual predictive criterion C(p), Akaike Information Criterion (AIC) and Schwarz Bayesian Criterion (SBIC) at different dentition and sex categories.Smallest Cp value indicates precision and small variance in estimating the population regression coefficients; while the coefficient of determination (adjusted-R 2 ) represents the proportion the total variability explained by the model.Regression model was developed for males and females using the pooled data for all age groups due to the small proportion of animal at each age classes.Ten variable body measurements (body length, height at wither, chest girth, chest width, rump length, pelvic width, ear length, tail length and horn length) were used for females to estimate body weight, while for the estimation of male body weight scrotal circumference also take under consideration.Chest girth, rump length, tail length and body length were included in the model in order of importance and they accounted 74% of the total variability of the male sampled sheep population and chest girth alone accounted for 61% variation in body weight in the study area.In female sampled sheep population however, the three variables with positively contribution to the prediction model which were chest girth, pelvis width and rump length and were fitted first, second and third accounted for 73% of the total variability of the female sampled sheep population and chest girth alone accounted for 66% variation in body weight in the study area.Parameters used in estimation of the multiple linear regression models showed that the male sampled sheep population had higher adjusted R 2 (74%) than the values of female sampled sheep population (73%).This indicated that those linear measurements might predict more accurate in males than in females.In most circumstances chest girth was found to be the most important in accounting sizeable proportion of the changes in the body weight of the sampled sheep population in the study districts.Comparable measurements were reported for Habru, Gubalafto, Debre Libanos and Wuchale sheep in Ethiopia (Abera et al., 2014;Mohammed et al., 2015).Chest girth was more reliable in predict body weight than other linear body measurements at field level when there is no facilitate to take the whole measurements.Using pooled age group the best fitted model with criteria of AIC, Cp, adjusted-R 2 , Root Mean Square Error (RMSE) and SBC criteria for male was Y= -51.59 + 0.43CG + 0.33RL + 0.15TL +

Multivariate analysis
Multivariate analysis was conducted using quantitative traits for male and female mature females and mature males independently at all age classes.among the multivariate analysis canonical and Discriminant analysis.

Canonical discriminant analysis
All squared Mahalanobis' distances obtained among districts populations for male and female were significant (P<0.0001),indicating the existence of measurable differences between male and female districts populations (Table 6).The largest distance was found between district 1 and 3 for both male (3.

Discriminant analysis
Quantitative variables varied between sex groups and correct classification percentages were calculated separately for male and female sheep populations.The overall classification rates (hit rate) of male and female sampled sheep population were 29.2 and 37.6, respectively.For males, most individuals were classified into their source population (76.29% for Tanqua-Abergelle, 66.67% for Kola-Tembien, and 56.9% for Adwa) (Table 7).
As indicated in (Table 8) females also a more or less similar patterns were observed and most Individuals were

Conclusion
Sheep populations in Adwa district were heavier than Tanqua-Abergelle sampled sheep, and comparable with the sheep of Kola-Tembien district.Most of the BW and linear body measurements were higher in rams than those in ewes.BW in the study area for rams and ewes were 23.23 and 20.81 kg, respectively.Body weight and the linear body measurements in both sexes increased with increased age (dentition class) up to the fourth age group.Body weight of male sheep in the study area was positive and highly significantly (p<0.0001)correlated with CG, HW and BL, respectively, while BW of female sheep population was positive and highly significantly (p<0.0001)correlated with CG, RL and PW, respectively.
Step wise regression analysis showed that CG, RL, TL and BL in males and CG, PW and RL in females accounted for 74 and 73% of the total variability in body weight, respectively.

Figure 1 .
Figure 1.Map of the study area.

Table 1 .
Least squares Means ± Standard errors for fixed effects of location and sex group on body weight (kg) and LBMs (cm) for Tanqua-Abergelle, Kola-Tembien and Adwa sheep.

Table 2
dentition class.According to Mekasha(2007) body size and shape of the animal rises until the animal reaches optimal growth.BW, BL, HW, CG, CW,

Table 2 .
Least squares Means ± Standard errors for fixed effects of age group on body weight (kg) and LBMs (cm) for Tanqua-Abergelle, Kola-Tembien and Adwa sheep.

Table 4 .
Multiple regression analysis of live weight on different body measurements for rams at different age groups in the study area.

Table 5 .
Multiple regression analysis of live weight on different body measurements for ewes at different age groups in the study area.Adjusted= coeffiennt of determination, C (P) = Conceptual Predictive Information Criterion, AIC= Akaike Information Criterion, RMSE= Root Mean Square Error, SBC=Bayesian Information Criterion, 1PPI= 1Paire if Permanent Incisor, 2PPI= 2 Paired of Permanent Incisor, 3PPI= 3Paire of Permanent Incisor, 4PPI= 4Paire of Permanent Incisor, CG= Chest Girth, RL= Rump Length, TL= Tail Length, PW= Pelvis Width, HW= Height at Wither, BL=Body Length.

Table 6 .
Squared Mahalanobis' distance among district populations for male and female populations.

Table 7 .
Percent classified into each district (hit rate) for male populations using discriminant analysis.

Table 8 .
Percent classified into each district (hit rate) for female populations using discriminant analysis.