A total of 357 adult cows (age ranging from 4 to 16 years old) and 29 sires (from 4 to 5 years old) belonging to the Niger Zebu Bororo cattle population were assessed for 13 body measurements and 11 qualitative traits. Assessment was carried out in 12 different sites of the provinces of Tahoua (5 sites, 75 cows and 3 sires) and Maradi (7 sites, 282 cows and 26 sires) of the Republic of Niger (Fig. 1). Although the northern parts of these two provinces belong to the Sahara, sampled sites were located within the Sahel area. Sites sampled in the Tahoua province belonged to the municipalities of Boundou (14∘50′ N, 6∘56′ E), Goulbi (13∘27′ N, 7∘02′ E), Bangui (13∘40′ N, 06∘16′ E) and Djangangari (14∘55′ N, 06∘34′ E) while sites sampled in the Maradi province belonged to the municipalities of Chigrenne (15∘03′ N, 06∘36′ E), Bermo (15∘16′ N, 06∘44′ E) and Birni Lalley (14∘26′ N, 06∘45′ E). Each sampling location was georeferenced using GPS Garmin-50 devices (Garmin Ltd., Olathe, KS, USA). Assessments were carried out in 2015, during the rainy season (August–September).

The original dataset included records of 21 quantitative traits. However, only 15 of them coincided with those obtained by Traoré et al. (2015, 2016) in nine West African cattle populations. Moreover, definition of two of the traits recorded (tail length and facial width) differed between the current research and the project of Traoré et al. (2015, 2016). Therefore, after editing, only 13 quantitative traits were used for analyses (Tables 1 and 2): facial length (from orbital fossa to upper lip), muzzle circumference, horn length (greater curvature), ear length, height at withers, heart girth, height at hips (tuber coxae), body length (from lateral tuberosity of the humerus to tuber ischii), thorax depth, shoulder width (between lateral tuberosities of the humerus), pelvic width (between tuber ischii), ischium width (between tuber ischii) and rump length (from tuber coxae to tuber ischii).

Eleven qualitative traits were scored with the same within-trait levels, codes and definitions used in Traoré et al. (2015, 2016): cephalic profile, ear shape, muzzle pigmentation, eyelid pigmentation, hoof pigmentation, horn colour, dewlap size, backline profile, horn shape, spotting pattern and coat colour pattern (Table 3).

Age of the individuals was approximated by examining dentition and direct enquiries to owners. For statistical purposes, the age of the individuals was grouped as follows: 4 years old (5 cows and 6 sires), from 5 to 10 years old (252 cows and 23 sires) and older than 10 years (100 cows). Body measurements were carried out, with the animals standing stationary on a flat floor, using Lydthin stick, tape measure and Vernier caliper. No ethics statement was required for data collection. Body measurements and trait scores were obtained from different technicians visiting farms with the permission of the owners. Animals were managed by the owners.

Data were jointly analysed with 375 individuals belonging to the 4 West African zebu cattle populations previously analysed by Traoré et al. (2015, 2016): Zebu Azawak (29 cows and 8 sires), Zebu Mbororo (64 cows and 14 sires), Zebu Peul of Burkina Faso (266 cows and 24 sires) and Zebu Peul of Benin (128 cows and 17 sires). Zebu Mbororo of Burkina Faso and Zebu Bororo of Niger can be considered local representatives of the long-horned West African zebu cattle group (Rege and Tawah, 1999). However, in the current analysis both samples were considered as belonging to different populations. Consequently, the name given by Traoré et al. (2015, 2016) to the long-horned zebu of Burkina Faso (Zebu Mbororo) was used throughout the text to point out the different geographical origin of data. For consistency, Zebu Peul of Burkina Faso and Benin were treated as belonging to different populations. Traoré et al. (2015, 2016) did not find clear differentiation at the qualitative traits level among West African zebu populations. Therefore, when necessary for descriptive purposes, frequencies of qualitative traits were pooled for West African zebu cattle other than Zebu Bororo.

Body traits were analysed separately for cows and sires to avoid bias due to sexual dimorphism (see Traoré et al., 2016, for a review on this issue). Preliminary analyses showed that neither country of origin (Benin, Burkina Faso and Niger) nor type of West African zebu (long-horned: Zebu Bororo and Zebu Mbororo; short-horned: Zebu Peul and Zebu Azawak) had statistically significant influence on data. Therefore, a very simple model, including the effect of the breed (five levels) and the age of the individual (three levels for cows and two levels for sires), was fitted using PROC GLM of the SAS/STATTM package (SAS Institute Inc., Cary, NC) to estimate least squares means, and their corresponding standard errors, for each level of the breed effect. Additionally, Duncan's multiple-range test was performed on the breed effect means. Furthermore, the between-breeds Mahalanobis distance matrix was computed on body measurements using the CANDISC procedure of SAS/STAT.

Relationships among body measurements were summarised via principal component analysis (PCA), using the PROC FACTOR of SAS/STAT, to determine the number of independent traits that account for most of the phenotypic variation in body measurements. This analysis was computed from the correlation matrix among measurements to ensure that all traits were treated as equally important, giving the same weight to the variables regardless their own variance. A VARIMAX rotation was applied to the retained components in order to obtain factor pattern coefficients considerably less correlated than the original body measurements. Only factors accounting for more variation than any individual type trait (eigenvalue > 1) were retained.

Frequencies of each level of the qualitative traits analysed were computed using the PROC FREQ of SAS/STATTM. Statistical significance of the differences in the frequencies observed among the Zebu Bororo breed and the other West African zebu cattle was assessed pooling both sexes via chi-squared Mantel–Haenszel test. Relationships between qualitative traits were assessed via correspondence analysis using the PROC CORRESP of SAS/STATTM. Two canonical dimensions and their eigenvectors were computed to account for association between the levels of the traits scored. Following Parés-Casanova and Jordana (1999) and Grema et al. (2017) each score of the qualitative traits assessed was consider arbitrarily as polymorphism of the trait and used to compute the between-breeds Reynolds distance matrix using the program MolKin (Gutiérrez et al., 2005). Statistical confidence on the distance values was assessed via bootstrapping using 1000 replicates.

Eigenvectors computed for each individual via PCA and correspondence analyses were used to construct contour plots illustrating 75 % confidence region (per breed) of the relationships among individuals using the library ggplot2 of R (http://CRAN.R-project.org/). Eigenvectors computed for each individual via PCA were regressed on latitude (in decimal format) of the sampling site using the Proc REG of SAS/STAT to assess variation related to geography.

Least-squares means for the body measurements assessed by breed are given in Table 1 (cows) and Table 2 (sires). Data corresponding to the Zebu Azawak, Zebu Mbororo and Zebu Peul populations were previously analysed in Traoré et al. (2015, 2016). However, least-squares estimates may vary due to differences in both the datasets analysed (including the split of Zebu Peul into two different geographical populations) and the models fitted for analyses. For both males and females, Zebu Bororo cattle tended to have the higher estimates of the dataset and the Zebu Peul of Benin the lower. In any case, differences in estimates for body measurements between Zebu Bororo and the other long-horned zebu population analysed (Zebu Mbororo) were low. Even in the case of facial length, Zebu Bororo cows had lower estimates than Zebu Mbororo (51.0 ± 0.3 vs. 52.5 ± 0.5, respectively). Furthermore, these two breeds had very similar values for key traits such as height at hips and ischium width for both cows and sires. Actually, the higher differences between the Niger Zebu Bororo and the Burkina Faso Zebu Mbororo populations were assessed in measurement traits such as ear length and horn length closely related with key qualitative features such as ear shape and horn shape.

Both in males and in females, PCA identified two factors with eigenvalues higher than 1. In the females dataset factor 1 (eigenvalue = 7.34) accounted for 56.43 % of the total variation and factor 2 (eigenvalue = 1.21) explained 9.32 %. In the sires dataset factor 1 (eigenvalue = 8.04) accounted for 61.88 % of the total variation and factor 2 (eigenvalue = 1.07) explained 8.21 %. In females (Table 1) factor 1 clearly summarised the general size of the individuals (height and length) while factor 2 summarised the body width. In sires (Table 2), with a relatively limited sample size, factor 2 was less informative on body trait variation. The eigenvalues computed via PCA for each individual were plotted in a two-dimensional space to illustrate the between-breeds relationships for body measurements (Fig. 2). No differentiation among West African zebu populations was found. The 75 % confidence regions computed for Zebu Bororo, Zebu Azawak, Zebu Mbororo, Zebu Peul of Burkina Faso and Zebu Peul of Benin were intermingled, particularly in the case of females in which sample size was not a limitation (Fig. 2a). The eigenvalues computed for each individual corresponding to the factor 1 identified either the cows' or sires' dataset were also regressed on latitude to ascertain the existence of a geographical pattern of variation on body measurements in West African zebu cattle. Both in females (R2 = 0.598) and in males (R2 = 0.370) positive and significant regression coefficients were computed (0.34467 in cows and 0.24049 in sires) showing that overall body size increased with latitude).

The between-breeds Mahalanobis distance matrix is given in Table 3. All pairwise distances were statistically significant (p < 0.001). The hypothesis that the breeds' means are equal in the populations analysed was also tested using Wilks' lambda. This parameter took a significant value (p < 0.0001) for both the cows (λ = 0.04662812; F = 75.37; degrees of freedom = 52) and the sires (λ = 0.02895334; F = 8.64; degrees of freedom = 52) datasets. Therefore, differences found were statistically different from zero. For both sires and cows, the largest distances were found between the Zebu Bororo cattle and the other breeds. In both sexes the higher differentiation was found for the pair Zebu Bororo–Zebu Azawak (41.65 for cows and 47.49 for sires). Zebu Bororo had the lowest differentiation with Zebu Mbororo.

The frequencies (in percentage) of each level of the 11 qualitative traits recorded for Zebu Bororo cows and sires are given in Table 4. Data corresponding to the Zebu Azawak, Zebu Mbororo, Zebu Peul of Burkina Faso and Zebu Peul of Benin populations were pooled with no sex separation to be used as a reference for Zebu Bororo. A chi-squared Mantel–Haenszel test showed that incidence of all the analysed traits varied significantly between Zebu Bororo cows and sires and the other West African zebu cattle for p < 0.001 except for cephalic profile and backline profile (Table 4). For these two traits zebu cattle were majority straight, regardless of the population to which the individuals belonged. However, Zebu Bororo cattle had large differences for most qualitative traits with the other West African zebu cattle. The “most frequent” Zebu Bororo individuals had dropped ears, non-pigmented muzzle, grey-coloured lyre-shaped horns and red-pied coat colour pattern. This general appearance substantially departs from that of the other West African zebu cattle populations analysed in which horizontal ears and pigmented muzzle are the rule and it is possible to find a high variation in horn shape and coat colour (Table 4).

A correspondence analysis was carried out on the 11 qualitative traits recorded. Two correspondence dimensions identified accounted for 36.77 and 21.60 %, respectively, in the cows' dataset and 32.43 and 26.57 % of the total variation, respectively, in the sires' dataset. The solutions provided for each individual by the correspondence analysis were plotted in a two-dimensional space. Figure 3 shows that the 75 % confidence region computed for the Zebu Bororo breed is clearly separated from those of the other zebu populations at both the cows and the sires level which. In turn, the non-Niger West African zebu populations showed highly intermingled 75 % confidence regions.

Between-breeds Reynolds' distance matrices were computed for each sex to quantify differentiation due to qualitative type traits (Table 5). The pairs involving Zebu Bororo cattle had the higher distance values.