Study areas

The study was conducted in two regions, Oromia and Sidama. Three districts – Diksis district of Arsi zone and Goba district of Bale zone in Oromia, and Hula district in Sidama (Figure 1) – were covered. Some parameters of the sampled districts including weather conditions and agroecology are presented in Table 1.

Site and animal selection

Representative samples of Arsi, Bale and Jemjem cattle breeds were selected from their respective breeding tract. Information on their breeding tract and distribution areas were identified from previous studies (Assefa & Hailu, 2018; Legesse & Zeleke, 2021; Rege & Tawa, 1999; Yimamu, 2014). Accordingly, Diksis district was randomly selected from the highland districts of the Arsi zone to represent Arsi cattle, Goba district was randomly selected from the highland districts of the Bale zone to represent Bale cattle, while Hula district was randomly selected from the highland districts of the Sidama region to represent Jemjem cattle. Two sampling sites (Kebeles) were randomly selected from each district. Forty households that reared cattle were randomly selected from each sampling site (kebele). Within each household, the adult cattle aged four years and above were first separated from the young ones to avoid age bias. Genetically unrelated animals were separated to make the sampling representative. Then, two animals were selected randomly for the morphometric and morphological recording to avoid sampling bias. Selected animals were controlled carefully by their owners and trained labourers. Aggressive animals that could not properly stand on the flat ground were not recorded to avoid measurement bias.

Data collection

Data on morphometric (quantitative linear body measurements) and morphological (qualitative characteristics) traits were collected based on the data collection procedures described in the UN’s Food and Agriculture Organization (FAO) guidelines (FAO, 2012). Data collection was performed in the morning to avoid errors regarding feeding and watering. Five researchers were involved in the data collection procedure: three handled the quantitative data while the remaining two took care of the qualitative data decision-making and recording. To reduce bias, morphometric data recording was performed by the same researcher throughout the study. Animals were measured using a centimeter-unit textile measurement tape. A total of 441 cattle (342 females and 99 males) were subjected to nine morphometric measurements (Table 2) and 16 morphological/qualitative traits (Table 3).

Univariate analysis

UNIVARIATE procedure of SAS (SAS, 2002) was used for data normality test, the frequency procedure (Chi-square test) was used for morphological (qualitative) data analysis, and the general linear model (GLM) procedure was used for morphometric (quantitative) data analysis. The following statistical analysis model was used to analyze the morphological data:

Y_ij = μ + S_i + B_j + e_ij

where Y_ij is an observation, μ is the overall mean, S_i is the fixed effect of sex (i = male, female), B_j is the fixed effect of breed (j = Arsi, Bale, Jemjem) and e_ij is the random error. Quantitative data were analyzed separately for each sex by fitting breed as a class variable. Means (LSM) were separated using the adjusted Tukey-Kramer test (Kramer, 1956; Tukey, 1953).

Multivariate analysis

Stepwise discriminant analysis (SAS, 2016) was used to detect morphometric traits that better discriminate the cattle breeds, while discriminant analysis was applied to allocate individuals to known breeds and assess possibilities of misclassifications. Canonical discriminant analysis was also employed to deliver maximal separations between breeds (SAS, 2002). Graphic interpretation of breed differences was plotted using the scored canonical variables. Pairwise Mahalanobis distances between breeds were computed as $D^2\left(i|j\right)={\left(x_i-x_j\right)}^{\text{'}}{cov}^{-1} \left(x_i-x_j\right)$ . Where $D^2\left(i|j\right)$ is the distance between breeds $i$ and $j$ , ${cov}^{-1}$ is the inverse of the covariance matrix of measured variables, $x_i$ and $x_j$ are the means of variables in the $i$ ^th and $j$ ^th breeds.

Qualitative characteristics

Figure 2 shows the coat colour distribution across the two sexes and three breeds studied. While coat colour was not significantly influenced by the animals' sex (chi-square value 4.5, p = 0.3480), it did vary by breed (chi-square value 37.9, p < 0.0001). Black coat colour was predominant across all three breeds, while black + white coat animals were observed more frequently in Jemjem cattle. Black + white coloured animals are those with predominantly black coat colour with some white patches, spots or shades. Their coat colour pattern can be also indicated as pied, spotty or shaded. The same applied to the red + white coloured animals.

The effects of breed and sex on the qualitative characteristics of Arsi, Bale and Jemjem breeds are presented in Table 3 along with the respective chi-square values and levels of significance. Sex affected 6 out of the 13 traits while breed significantly affected 7 out of the 15 traits. The majority of the studied cattle had forward-oriented (65.8%), widely spaced (71.4%) and curved horns (76%). They also mainly had small (66.7%), erected (96.2%) humps located at the cervicothoracic position (77.8%). All (100%) of the studied cattle had a straight face and a back profile as well as a sloppy rump. Medium (44.4%) or long (45.6%) tails were equally common and medium dewlap width were observed on most (72.1%) of the cattle. On the other hand, 57.6% of the males had medium perpetual sheath while naval flap was not observed on 53.2% of their female counterparts. A uniform coat colour pattern was observed on most (78%) of the cattle while all of them (100%) had straight-edged ears (Figure 3).

Comparing sexes, laterally oriented straight horns were more frequently observed in males than females. The majority of females had an erect hump while some males had a dropping hump. Males also had larger humps located at the thoracic position while females possessed small humps located at the cervicothoracic position. Comparing breeds, a higher proportion of narrow horn spacing was observed in Jemjem cattle. However, no significant differences were observed among the cattle breeds in terms of most of the qualitative characteristics.

Morphometric traits

Means (least squares), standard errors and pairwise comparisons showing the effect of breed on the morphometric traits of the studied male and female cattle populations are presented in Table 4. Relative differences among breeds were observed more in females than males. Within females, Bale cows had the largest body length (101.6cm), heart girth (139.4cm) and hock circumference (28.8cm). The Arsi cows had the smallest body length (97.1cm), pelvic width (29.5cm) and muzzle circumference (36.1cm) while their horns were the longest (22.6cm). The Jemjem cows had relatively intermediate measurements for most of the traits including body length (99.7cm), pelvic width (30.5cm) and hock circumference (27.6cm). Similarly, the Bale males had the largest heart girth (150.5cm) and hock circumference (30.7cm).

Multivariate analysis

Canonical discriminant analysis

Multivariate statistics including eigenvalues using the first and the second canonical structures (Can 1 and Can 2) are shown in Table 7. In classifying the cattle breeds, Can 1 had a higher proportion for females (0.6066) and males (0.7876) than Can 2. However, the lowest eigenvalues were observed for both canonical structures under both sexes.

Table 7: Multivariate statistics outputs from the two canonical structures. Can, canonical structure.

	Females		Males
Multivariate Statistics	Can 1	Can 2	Can 1	Can 2
Canonical correlation	0.6138	0.5307	0.6814	0.4353
Proportion	0.6066	0.3934	0.7876	0.2124
Eigenvalue	0.6047	0.3921	0.8670	0.2338

Pairwise Mahalanobis distances between the breeds studied are presented in Table 8. The shortest and the longest distances were observed among males. The shortest distance (1.77) was observed between Arsi and Bale males while Bale and Jemjem oxen were related distantly (7.31). The overall results showed the lowest and non-significant distances among Arsi, Bale and Jemjem cattle breeds.

Table 8: Pairwise squared distances between breeds. Females above diagonal, males below diagonal.

From breed	Arsi	Bale	Jemjem
Arsi	0	2.88	3.32
Bale	1.77	0	2.72
Jemjem	4.21	7.31	0

A plot of Can 1 and Can 2 showing the maximum separation among the cattle breeds is presented in Figure 4. In line with the result of the Mahalanobis distances, females were separated less than males. Accordingly, Arsi, Bale and Jemjem cows were inseparable and categorized in the same group while relative separation was observed between Arsi and Jemjem cows. Similarly, a relative separation between Bale and Jemjem oxen was also observed.

Qualitative characteristics

Due to their easily observable nature, unique qualitative characteristics can be used for breed differentiation. Alongside other morphometric and morphological traits, similarities in coat colour and coat colour pattern among breeds may indicated genetic similarity (Getachew, Abegaz, Misganaw, & Fekansa, 2014; Mustefa, Aseged, Kenfo, & Hunde, 2024). According to Getachew et al. (2014), the majority (73.62%) of Ogaden cattle exhibited a uniform body colour pattern, with most (69.33%) having a grey coat colour. Similarly, Mustefa et al. (2024) suggested that the Guraghe and Jimma cattle populations might belong to the same breed based on their phenotypic similarities. They reported that 66% of Guraghe and 77% of Jimma cattle populations had a uniform coat colour pattern with 55% of Guraghe and 65% of Jimma cattle populations having a red coat (Mustefa et al., 2024). In line with these results, the cattle breeds examined in the current study – Arsi, Bale and Jemjem – shared similarities in both coat colour and coat colour patterns. These phenotypic similarities suggest a potential genetic link between these breeds. However, contrasting reports fromMustefa et al. (2023) on Harar cattle, which displayed a diverse range of coat colours and patterns, highlight the need for further molecular characterization to confirm the results of the phenotypic study.

The dominantly black coat colour and uniform body colour pattern observed in this study are in line with the results ofYimamu (2014) on Arsi cattle. The dominance of dark colours over light colours might be associated with the highland environment (Titto et al., 2016), since animals with darker coats are better adapted to cold conditions by absorbing more heat than lighter colour coats (Titto et al., 2016). Moreover, the dominantly observed black coat colour might also be associated with farmers’ preferences and selection criteria as black was favoured in the studied areas.

Beyond coat colour, similarities in other qualitative characteristics were also observed. The resemblance in horn, hump, tail, dewlap, naval flap and perpetual sheath besides their perfect match in the face, back and rump profiles among Arsi, Bale and Jemjem challenge their classifications as different breeds. The slight differences noted can be taken as within-breed differences. Such differences in cattle sampled from different locations were reported byTerefe, Dessie, Haile, Mulatu, and Mwai (2015) in Mursi cattle andMustefa et al. (2021) in Raya cattle.

Morphometric traits

Results of morphometric traits, alongside qualitative traits, can provide reliable information for quantifying the degree of relationships among breeds. In this study, the observed qualitative similarities among Arsi, Bale and Jemjem were also supported by quantitative measurements. Significant differences in morphometric measurements that would indicate distinct breeds were not observed. This was also in line with Mustefa et al. (2024), who reported comparable morphometric measurements between Guraghe and Jimma cattle populations suggesting they belong to the same breed.

As noted in the qualitative analysis, the differences observed among the three cattle breeds might be due to within-breed variation (Mustefa et al., 2024). Bale cows seem to be the largest, with higher measurement values for body length, heart girth and hock circumference. Intermediate measurement values were observed in Jemjem cows, while Arsi cows were the smallest, with lower values for body length, pelvic width and muzzle circumference although they possessed the longest horns. However,Yimamu (2014) reported relatively higher measurements for body length, heart girth and height at withers for Arsi cattle in the same study area.

When compared to other Ethiopian breeds, the morphometric values of Arsi, Bale and Jemjem cattle were lower than Afar cattle (Tadesse, Ayalew, & Hegde, 2008), Begait cattle (Ftiwi, 2015), Begaria cattle (Getachew et al., 2020), Fogera cattle (Girma, Alemayehu, Abegaze, & Kebede, 2016), Gojjam Highland cattle (Getachew & Ayalew, 2014), Harar cattle (Mustefa, 2023), Kereyu cattle (Nigatu & Tadesse, 2020), Nuer cattle (Minuye, Abebe, Dessie, & T, 2018), Ogaden cattle (Mustefa et al., 2023) and Raya cattle (Mustefa et al., 2021). On the other hand, the morphometric values of Abergelle and Irob cattle (Zegeye, Belay, & Hanotte, 2021) were lower than the Arsi, Bale and Jemjem. Comparable morphometric values were also reported in Arado cattle (Genzebu et al., 2012), Gofa cattle (Kebede et al., 2017), Horro cattle (Bekele, 2015) and Mursi cattle (Terefe et al., 2015).

Effect of sex

In most morphometric traits, males were observed to be larger than females. Such differences might be attributed to the secretion of testosterone in males, which promotes skeletal development and muscle mass growth (Baneh & Hafezian, 2009). The endocrine system plays a significant role in differentiating the two sexes, with the growth-limiting effects of estrogen being more prominent in females (Baneh & Hafezian, 2009; Chriha & Ghadri, 2001). The findings of this study are in line with the reports ofMustefa et al. (2023) on Harar and Ogaden cattle,Mustefa et al. (2021) on Raya cattle, andTerefe et al. (2015) on Mursi cattle.

Multivariate analysis

Morphometric traits were identified and ranked based on their ability to differentiate between the cattle breeds. In line with the results ofMustefa et al. (2024) on Guraghe cattle, lower partial R-Square and F-values (Table 5) were observed in the stepwise analysis, showing that morphometric traits have limited potential to discriminate the breeds into different categories. The higher the R-Square and F-values the higher the potential of the traits in differentiating the cattle breeds (Mustefa et al., 2023).

The higher error rate (Table 6) suggests greater shared similarities among the breeds, which reduces the chances of clearly categorizing the breeds into different clusters. On the other hand, the lower the error rate, the lower the similarities shared among the breeds. This highlights the uniqueness of each breed. The moderate classification with a considerably higher error rate (34.17%) observed in the current study, showed the presence of shared similarities among the breeds. An error rate of 1% was reported in classifying the phenotypically unrelated Harar and Ogaden cattle breeds (Mustefa et al., 2023).

The low eigenvalues reported for both canonical structures in both sexes (Table 7) do not support the classification of the animals into different breeds. An eigenvalue higher than 1 is accepted to approve the discrimination analysis. If the value is lower than 1, the discrimination of the studied animals into different breeds is not significant. In this study, the observed low eigenvalue disproved the presence of three breeds in the study area.

Similarly, the higher the Mahalanobis distances between breeds (Table 8) the higher the possibility of classification into different clusters. However, the Mahalanobis distances in the current study were low although Jemjem males showed relatively higher distances. This could be due to the small sample size of Jemjem oxen. The accuracy of the analysis increases with larger sample sizes. Due to the low eigenvalue (< 1) in the multivariate analysis, the distances observed were not significant, supporting the conclusion that the studied cattle breeds are phenotypically inseparable.