Livestock Science 141 (2011) 232–241

Contents lists available at ScienceDirect

Livestock Science

j ourna l homepage: www.e lsev ie r.com/ locate / l ivsc i

Eco-geographical structuring and morphological diversities in Ethiopiandonkey populations

E. Kefena a,e,⁎, A. Beja-Pereira b, J.L. Han c, A. Haile f, Y.K. Mohammed d, T. Dessie e

a Ethiopian Institute of Agricultural Research, Holetta Agricultural Research Center, P. O. Box, 2003, Addis Ababa, Ethiopiab Research Center in Biodiversity and Genetic Resources (CIBIO), University of Porto, Campus Agrario de vairao, Rua Padre Armando, Quintas, 7, Porto, Portugalc CAAS-ILRI Joint Laboratory on Livestock and Forage Genetic Resources (JLLFGR), Institute of Animal Science, Chinese Academy of Agricultural Sciences (CAAS),Beijing 100193, P.R. Chinad Haramaya University, P. O. Box, 138, Dire Dawa, Ethiopiae International Livestock Research Institute (ILRI), P. O. Box 5689, Addis Ababa, Ethiopiaf International Center for Agriculture in Dry Areas (ICARDA), P. O. Box 5466, Aleppo, Syria

a r t i c l e i n f o

⁎ Corresponding author at: Ethiopian Institute ofHoletta Agricultural Research Center, P. O. Box, 2003,Tel.: +251 911 74 83 81; fax: +251 112 37 03 77.

E-mail addresses: kefenaol@yahoo.com (E. Kefena(A. Beja-Pereira), h.jianlin@cgiar.org (J.L. Han), a.hailem.yusufkurtu@yahoo.com (Y.K. Mohammed), t.dessie

1871-1413/$ – see front matter © 2011 Elsevier B.V.doi:10.1016/j.livsci.2011.06.011

a b s t r a c t

Article history:Received 19 February 2011Received in revised form 20 May 2011Accepted 16 June 2011

We carried out a nationwide survey to morphologically characterize and identify thegeographic distribution of Ethiopian donkey populations. Here we re-evaluated some of thedonkey populations previously identified and explored new ones fromwhichwe confirmed theexistence of six distinct donkey populations and rejected one donkey population that wasmisidentified previously. We used twelve selected morphometric variables that recorded on atotal of 569 adult donkeys (289 jacks and 280 jennets) spanning over wide arrays of ecologiesin Ethiopia. Mean pairwise multiple comparisons (MPMC) between traits and multivariateanalysis were carried out separately for jacks, jennets and aggregated gender. The MPMCshowed significant differences between donkey populations for most of the traits studied.Principal component analysis showed that three body heights (height at wither, back andrump) and body length alone account for about 87.5% the total variations (eigenvalue=10.49)of the 12 measured morphometric variables. Mahalanobis as well as Euclidean distances werealso found to be highly significant (Pb0.01) for pairwise comparisons of all populations.Stepwise discriminant analysis showed that, except in the order of their occurrence, the firsttwo traits that separate between jennets and jacks were similar showing that there are noexplicit anatomical dissimilarities or sexual dimorphism between genders in donkeys. Percentassignment of discriminant function analysis showed that, on average 64.83% of individualswere classified within their respective populations implying that populations were distinct.This study also explicitly showed that morphological character system (body size and coatcolor patterns) varies according to ecological patterns and altitudinal gradients. Suchmorphological divergences in contrasting ecologies may reflect response and adaptation of agiven population to the prevailing ecological variables and biophysical resources.

© 2011 Elsevier B.V. All rights reserved.

Keywords:African wild assesEthiopian donkey populationsAltitudinal gradientsAdaptive divergenceBiophysical resources

Agricultural Research,Addis Ababa, Ethiopia.

), albanobp@fc.up.pt@cgiar.org (A. Haile),@cgiar.org (T. Dessie).

All rights reserved.

1. Introduction

Until recently, the origin and domestication events of thedomestic donkeys have been controversial and poorly under-stood. Though donkeys are among the widely distributed largedomestic mammals of the world, they haven't been a focus ofzooarchaeological and developmental research in Africa andelsewhere in the past because: 1) Archaeological remainswerenearly absent and unlike other livestock species, rock art

233E. Kefena et al. / Livestock Science 141 (2011) 232–241

drawings, wall-painting and other iconographies are limited inthe continent except with few rock arts that were restricted tothe Nile Valley in Egypt and Nubia; 2) often, donkeys are notconventional sources of meat and their uses for packing andtraction do not fit within the stereotyped perspective oflivestock agencies; 3) donkeys are considered as animals ofthe poor and women and therefore, highly stigmatized,marginalized and were not a focus of research (Blench, 2004;Epstein, 1971; Marshall, 2007; Rossel et al., 2008).

Donkeys are anAfricandesert-like adaptedheritage animals(Rossel et al., 2008) and genetic research by Beja-Pereira et al.(2004) confirmed that they were descended from two Africanwild ass subspecies: the Nubian wild ass (Equus africanusafricanus) and the Somaliwild ass (Equus africanus somaliensis).Nevertheless, they were not certain where this first occurred,but suggested two domestication events probably in north-eastern Africa.

Genetic studies (Hanotte et al., 2002) unraveled thatpastoralism and mobile way of life were discovered to firsttake place in northeast Africa and African pastoralists weresuggested to domesticate the donkey for the first time (Beja-Pereira et al., 2004; Marshall, 2007). Kimura et al. (2011) andMarshall (2007) and references therein unanimously conclud-ed that the Nubian wild ass dominates the central Sahara withbroad distribution from northern Eritrea, across the Sudanto parts of the Sahara and north to Egypt dated about6000 years BP. On the other hand, the Somali wild ass wasonly observed farther north than Eritrea at about 11.6°N.Moreover, early domestic stocks have been identified at about3500 BC in northeastern Ethiopia and if early herders in theHorn of Africa domesticated Somali wild ass, it is likely to haveoccurred nearly 4000 years BP (Marshall, 2007). From theseevidences, it can be argued that domestication of the historicNubian wild ass predates its Somali counterpart by about2000 years. However, recent mitochondrial DNA studies(Kimura et al., 2011) excluded the Somali wild ass as theancestry of the domestic Somali donkey raising new questionthat research on the origin and time of domestic donkeys is yetan on-going debate and is far from complete. Several otherstudies (Beja-Pereira et al., 2004; Kimura et al., 2011; Rosselet al., 2008) suggested that donkeys played vital roles in

Table 1Summary of sampled domestic donkey populations and their distinct characteristic

No. Breeds/populations Characteristic features and uses

1 Abyssinian Predominantly brown, hairy and dwarf. Compact, hardheavily working animals

2 Afar Grey, grayish red and light rose. Leg stripes are commoFree ranging and have no or less work loads

3 Hararghe Grey and brown coat colors are common. Leg strips arecommon. Heavily working animals

4 Ogaden Mainly grey and grayish red. Heavily built animals. Legstripes are not uncommon. Relatively working animals

5 Omo/Hamer Coat color is light grey with shinny skin. Head coat colmainly rosy. Fatty and heavily built donkeys. Free rangwith no or less work loads. Used to serve as source of fsome tribal groups in the past. Now occasionally used

6 Sinnar Tall with variable coat colors. White and leopard coat ccommon. Leopard coat colored donkeys have ridge onHeavily working animals in the western lowlands andabsent beyond

transforming ancient societies, land-based transport systemsand trade routes in Africa and Eurasia. Today, donkeys continueto serve their destiny in the epoch of motorized vehiclesand advanced transport systems in the poorer regions of theworld, particularly in Africa where transport infrastructure isunderdeveloped.

Next to China, Ethiopia possesses the biggest donkey pop-ulations in the world (Gebrewold et al., 2004). Moreover,Ethiopia is one of themost strategically significant places in theHorn of Africa for domestic donkey diversities. For instances,Epstein (1971) indicated that donkey populations in Ethiopiaexhibit considerable conformational differences, have dif-ferent coat color patterns andbody sizes.However, thepatternsof morphological diversities and geographical distributionof Ethiopian donkeys have not been fully investigated andremained unknown. Moreover, though they are key draft,transport and pack animal sources for themajority of the poor,donkeys in Ethiopia are highly neglected, stigmatized and havepassed unnoticed as trivial animals in the Ethiopian historicalchronicles.

Gubitz et al. (2000) demonstrated an outstanding modelthat showed morphological character system might reflectecological selection regimes (coat color patterns), history (bodydimensions) or both (scalation). Gizaw et al. (2007) alsoshowed that morphological characters and patterns of mor-phological variations are highly associated with ecology inEthiopian sheep populations. Therefore, we hypothesized thatin addition to differences in the origin of their wild ancestors,eco-geographical variations and associated biophysical re-sources might have resulted in the current domestic donkeydiversities in Ethiopia. This paper is, therefore, aimed at tomorphologically characterize and identify the geographicaldistributions and morphological relationships among theindigenous donkey populations of Ethiopia.

2. Materials and methods

2.1. Sampling strategy

The sampling structure and sites are shown in Table 1 andFig. 1, respectively. At the initial stage, we considered reports

s with respect to coat color patterns, uses and locations in Ethiopia.

Regions/locations

y and Widely distributed following mountain chains on both sidesof the Great Rift Valley

n. Afar plain interfacing with the Somali wild ass

Oromia/Hararghe highlands in eastern Ethiopia

Somali/Ogaden plain. Widely distributed in all pastoral andagro-pastoral areas in Somali Regional State

or isingood foras food

South Omo pastoral lowlands around Lake Turkana andHammer, Gnyangatom and Dasenech districts in SouthernEthiopia. Very common at site called “Chew Bahir” means“salt sea”

olors aretheir ribs.rare or

Sparsely distributed in the northwestern lowlands along theEthio-Sudanese boarder from Humera in north-west to Assosain the west

Fig. 1. Map of Ethiopia showing sampling locations and agro-ecological classification based on altitude.

234 E. Kefena et al. / Livestock Science 141 (2011) 232–241

that classified Ethiopian donkey populations based on ethnicand/or geographic nomenclatures. Secondly, we reviewed acountry report of the domestic donkey populations of Ethiopiathat have been officially reported in the Domestic AnimalDiversity-Information System (DAD-IS) of the Food andAgricultural Organization of the United Nations at (http://www.dad.fao.org/). Thirdly, we considered donkeys as a breedif they qualify the definition given in the World Watch List(2000) and Breeds of Livestock (1995) Oklahoma StateUniversity at (http://www.ani.okstate.edu/populations/). Inaddition to these standard criteria, we set additional criteriathat a givenpopulation is considered asunique if that particularenvironment/s is a true breeding tract for that particularpopulation/s. Any donkey populations that failed to meet thestandard criteria were not considered as distinct populationsand were rejected. Using these criteria as benchmarks, we

Fig. 2. Schematic diagram showing major morph

carried out both a confirmatory and exploratory nationwidesurvey to identify distinct donkey populations as well as tolocate their eco-geographical distributions. At this stage, it isimportant to note that we rejected the Jimma donkeypopulation previously reported in http://www.dad.fao.org/ asit failed to qualify the set criteria.

2.2. Data collection

Twelve quantitativemorphometric variables (AppendixA)were measured on a total of 569 full-mouthed adult donkeys(289 jacks and 280 jennets) that randomly selected from sixdistinct donkey populations (Table 2). Donkeyswere carefullyhandled by trained laborers and positioned to stand properlyon flat and hard groundswith parallel legs. Theywere allowedto get calm before body measurements were taken. Body

ological variables measured on donkeys.

Table 2Summary of number of jacks and jennets and total number of donkeyssampled per breed/population.

Breeds/populations Jacks Jennets Total Proportion

Abyssinian 41 62 103 0.18Afar 40 42 82 0.14Hararghe 46 51 97 0.17Ogaden 53 47 100 0.18Hammer/Omo 40 56 96 0.17Sinnar 69 22 91 0.16Total 289 280 569 1.00

235E. Kefena et al. / Livestock Science 141 (2011) 232–241

measurements were not taken on aggressive and improperlystanding donkeys. For the majority of the donkeys, the lengthand circumferences of hind limb cannon bonewere difficult tomeasure and were excluded from the final dataset.

Five major reference points were carefully marked (Fig. 2)before measurements were taken. These points were: 1) thepoint of wither; 2) the point of back; 3) the point of rump;4) the point of shoulder joint and 5) the point of pin bone. Atotal of twelve linear and circular morphometric variableswere carefully selected from standard procedures developedby Oulehla (1996) for horses and were measured.

Body measurements were regularly taken from the rightside except when measurements are repeated in cases whenwe were doubtful that donkeys were not properly standingon leveled and hard grounds. A single person recorded all themeasurements throughout this study to avoid measurementerrors.

2.3. Statistical analysis

Population means were estimated separately for jacks,jennets and aggregated gender. Tests for significance differ-ences between population means were separated by t-testusing Bonferroni adjustment procedure in Holm (1979) withstep-down method to account for multiple comparisons.

The data were standardized following the procedure ofManly (1986) before Euclidian distances were computedbetween populations. The distance matrix was used via clusterprocedure to build a dendrogramusing unweighted pair-groupmethod using arithmetic mean (UPGMA). Percentage assign-ments of individuals to clusterswere evaluated by discriminantfunction analysis. Squared Mahalanobis distances werealso computed between populations as: D2

ij=(Yi− Yj)‘COV−1(Yi− Yj); where D2

ij is the distance between populationi and j and COV−1 the inverse of the covariance matrix ofmeasuredvariables x and Yi and Yj are themeans of variable x inith and jth populations, respectively. Except the differences inthe magnitude of distances values, both Euclidian andMahalanobis distance procedures produced similar topologyof the dendrogram and thus distance based on Mahalanobisdistance procedure (Mahalanobis, 1936) was preferably usedthroughout cluster analysis.

Principal component analysis was performed on standard-ized datasets to summarize the variables into few meaningfulones that account formost of the variations in the population. Astepwise discriminant function analysis with forward selectionprocedure was performed to find out subsets of quantitativevariable that best reveals differences among populations.Canonical discriminant function analysis was also performed

to find out linear combination of quantitative variables thatprovidemaximal separations between populations. The scoredcanonical variableswere used to plot pairs of canonical variableto aid visual interpretation of group differences. Data was an-alyzed using SAS (2004).

3. Results

3.1. Breed means

Mean values of morphological variables and their co-efficients of variation (CV) are depicted in Tables 3–5 for jacks,jennets and aggregated gender, respectively. Pairwise meancomparisons for jacks showed significant differences formanytraits. Overall body size evaluation that took into account bodyheights (height at wither, back and rump) and body lengthshowed that Abyssinian jack population was the shortestwhile Sinnar jack population was the tallest. For instances,differences in height atwither between Abyssinian and Sinnarjacks populations was to the extent of 15.13 cm. Moreover,differences in heightmeasurements between four donkey jackpopulations viz. Afar, Omo, Hararghe and Ogaden with Sinnarwere highly significant (Pb0.05). Though shorter than Sinnarjack populations, Ogaden jack was significantly longer bodied(Pb0.05) as compared to all jack populations except with theSinnar (PN0.05). Except with minor differences, Afar, Omo,Hararghe and Ogaden jack populations were generally similarin their overall body dimensions and placed between theAbyssinian and Sinnar jack populations.

Nearly similar results were obtained for jennets (Table 4)and aggregated gender (Table 5). Pairwise mean compar-isons for aggregated gender for body heights revealed thatAbyssinian donkeypopulation remained inferior in the overallbody dimension while Sinnar donkey population was signif-icantly the tallest of all (pb0.05). Width of hip (WH) andchest width (CW) were significantly higher (Pb0.05) forAfar, Hararghe, Ogaden and Omo donkeys than Sinnar andAbyssinian donkeys populations. However, chest circumfer-ence (CC) was significantly higher (Pb0.05) for Omo, Ogadenand Sinnar donkey populations with lowest chest circumfer-ence obtained for Abyssinian donkey population. GenerallyCC, CW and WH describe the shape and body conditions ofdonkeys. This showed that donkey populations sampled fromarid semi-arid lowlands are fatter and blocky. Though Sinnardonkey was significantly the tallest of all donkey populations,they are thinner and lean when compared, particularly withAfar, Ogaden and Omo donkey populations. On the otherhand, our results could indicate that height is probablyinversely proportion to body condition or fatness in donkeys.

3.2. Multivariate analysis

Results of the first four principal components are shown inTable 6. The first four principal components explain about87.5% (eigenvalue=10.49) of the total variability of the orig-inal 12 measured variables. Results of discriminant functionanalysis also indicated that (data not shown), on average 64.83% of sampled individuals were classified into their respec-tive clusters/populations.

Squared Mahalanobis distances between jacks and jennetsand aggregated gender are shown in Tables 7 and 8,

Table 3Means and pairwise comparison of body measurements with coefficients of variations in each breed/population: 1) jakes.

TraitsΨ Breeds/populations

Abyssinian CV Afar CV Hararghe CV Ogaden CV Omo CV Sinnar CV

HW 94.99c 4.4 103.60b 3.4 102.88b 3.3 102.72b 3.4 103.85b 3.2 110.12a 3.2HB 96.68c 4.6 105.13b 3.3 104.30b 3.2 104.04b 3.4 105.25b 2.9 112.05a 3.1HR 97.12c 4.7 106.32b 3.4 105.01b 3.0 105.53b 3.7 105.90b 2.6 112.97a 3.0BOL 101.34c 4.5 111.11b 3.0 110.24b 2.3 111.01b 2.7 115.63a 3.3 114.86a 3.0BAL 52.66c 5.8 57.94b 5.5 57.51b 5.2 58.37b 6.2 58.36b 4.7 60.49a 5.0NL 47.13c 6.0 51.03b 4.7 50.85b 4.9 52.20b 5.3 54.28a 5.8 54.21a 4.7HL 43.33c 3.1 45.39b 2.3 45.08b 2.7 45.97b 2.5 47.21a 2.4 47.46a 3.3LCB 19.35d 4.8 22.92bc 6.1 22.62c 5.2 23.07bc 6.0 23.59b 5.5 24.47a 4.8LFL 47.05d 5.0 52.99bc 3.8 51.98c 4.1 52.76bc 4.8 53.95b 4.0 57.62a 4.1WH 30.85c 9.5 34.56a 9.1 33.84ab 7.0 34.49ab 7.2 35.14a 6.0 32.86b 5.8CW 21.97c 9.4 26.11a 7.0 25.88a 8.3 25.24ab 9.7 26.55a 6.4 24.60b 7.2CC 104.98c 4.6 113.68b 4.3 113.36b 4.6 115.37ab 3.7 116.09ab 3.6 117.69a 3.9

Means with the same superscripts in each row are not significantly different (PN0.05). Ψ HW = Height at wither; HB = Height at back; HR = Height at rump;BOL = Body length; BAL = Back length; NL = Neck length; HL = Head length; LCB = Forelimb length of cannon bone; LFL = Length of foreleg; WH = Width ofhip; CW = Chest width and CC = Chest circumference.

236 E. Kefena et al. / Livestock Science 141 (2011) 232–241

respectively. All Mahalanobis distances between jack popula-tions were highly significant (Pb0.001) except between Afarand Hararghe jacks (PN0.05). The largest distance wasobserved between Abyssinian and Sinnar jacks (Table 7).Abyssinian jack population was also significantly distant(Pb0.001) from any other jack populations. Relatively closerMahalanobis distances were obtained among Afar, Hararghe,Ogaden and Omo jack populations.

Results of squared Mahalanobis distances between jen-net donkey populations are also depicted in Table 7. AllMahalanobis distances were highly significant (Pb0.001) be-tween all jennet populations. Unlike in the case of jack pop-ulations, the longest Mahalanobis distance was obtainedbetween Abyssinian and Omo jennet populations followedby between Abyssinian and Sinnar jennets (Table 7). Like inthe case of jacks, though all Mahalanobis distances werehighly significant between all jennet populations, relativelyshorter distances were observed between Afar, Hararghe,Ogaden and Omo jennet populations.

Squared Mahalanobis distances for aggregated gender aredepicted in Table 8. All Mahalanobis distances were highly

Table 4Means and pairwise comparison of body measurements with coefficients of variati

TraitsΨ Breeds/populations

Abyssinian CV Afar CV Hararghe C

HW 93.14c 3.9 101.04b 3.6 99.58b 3HB 94.59c 3.9 102.80b 3.8 100.85b 3HR 95.36d 4.9 104.39bc 3.5 102.11c 3BOL 101c 4.3 110.69b 3.0 109.59b 2BAL 53.02c 5.4 56.79b 5.9 56.49b 5NL 46.59c 5.0 50.35b 4.4 50.24b 4HL 42.66c 3.7 45.03b 2.3 45.25b 2LCB 19.17c 4.9 21.99b 7.2 21.49b 6LFL 46.60d 5.0 51.80bc 4.1 50.54c 4WH 31.15c 7.9 34.72ad 8.1 34.25ab 8CW 21.37c 8.9 25.71ab 8.8 25.06ab 8CC 103.27c 4.9 112.96b 3.4 110.98b 3

Means with the same superscripts in each row are not significantly different (PN0.BOL = Body length; BAL = Back length; NL = Neck length; HL = Head length; LCBhip; CW = Chest width and CC = Chest circumference.

significant (Pb0.005) for aggregated gender. Relatively shorterMahalanobis distances were consistently obtained betweenAfar, Hararghe, Ogaden and Omo donkey populations. Abys-sinian and Sinnar donkey populations were also consistentlydistant from each other as well as from the rest of donkeypopulations considered in this study.

Cluster analysis of squaredMahalanobis distances (Table 8)based on twelve morphometric variables grouped the overallEthiopian donkey populations into four distinct clusters(Fig. 3): 1) the dwarf and hairy Abyssinian donkey populationthat inhabit highland ecology on both sides of the Great RiftValley except theHararghe highland; 2) The tallest, elegant andunique coat colored Sinnar donkey population that inhabit thewestern lowlands along the Ethio-Sudanese border; 3) lightgrey coat colored, shinny skinned and roseheadedOmodonkeypopulation that inhabit the southern arid and semi-aridlowlands; and 4) Ogaden, Afar and Hararghe donkeys thatinhabit the northeaster, eastern arid and semi-arid lowlandsand Hararghe highlands, respectively.

Results of stepwise discriminant function analysis basedon the contribution of the traits in separating among jacks

ons in each breed/population: 2) jennets.

V Ogaden CV Omo CV Sinnar CV

.7 101.00b 3.8 101.61b 3.0 108.71a 3.3

.7 102.36b 3.8 103.22b 3.3 110.54a 3.7

.6 103.62bc 4.0 104.35b 3.3 111.70a 3.6

.2 109.82b 2.4 114.25a 3.6 114.16a 2.6

.8 56.92b 4.8 57.93ab 4.2 59.57a 5.5

.1 51.42b 4.5 52.98a 5.0 53.23a 4.8

.2 45.53b 2.2 46.73a 2.8 46.55a 3.5

.4 22.09b 7.7 22.82b 5.4 24.34a 6.3

.3 51.64bc 3.8 52.91b 3.9 57.41a 4.3

.7 35.38a 9.7 35.24a 6.6 32.91b 6.8

.0 25.35ab 7.8 25.90a 6.8 24.44b 5.9

.9 116.54a 3.7 116.64a 4.4 115.80ab 3.7

05). Ψ HW = Height at wither; HB = Height at back; HR = Height at rump= Forelimb length of cannon bone; LFL = Length of foreleg; WH = Width o

;f

Table 5Means and pairwise comparison of body measurements with coefficients of variations in each breed/population: 3) aggregated gender.

TraitsΨ Breeds/populations

Abyssinia CV Afar CV Hararghe CV Ogaden CV Omo CV Sinnar CV

HW 93.87c 4.2 102.29b 3.7 101.15b 3.8 101.92b 3.7 102.54b 3.2 109.78a 3.2HB 95.42c 4.3 103.94b 3.7 102.48b 3.8 103.25b 3.7 104.06b 3.2 111.68a 3.6HR 96.06d 4.5 105.33b 3.6 103.48c 3.6 104.63bc 4.0 104.99bc 3.1 112.66a 3.2BOL 101.14c 4.3 110.89b 3.0 109.90b 2.3 110.45b 2.6 114.82a 3.5 114.70a 2.9BAL 52.88c 5.5 57.35b 5.8 56.97b 5.6 57.69b 5.7 58.11b 4.4 60.26a 5.1NL 46.80d 5.4 50.68c 4.6 50.53c 4.5 51.84b 5.0 53.52a 5.4 53.97a 4.8HL 42.93d 3.5 45.21bc 2.3 45.16c 2.4 45.77b 2.4 46.93a 2.6 47.24a 3.4LCB 19.24d 4.9 22.44bc 6.9 22.02c 6.4 22.61b 7.1 23.14b 5.6 24.44a 5.2LFL 46.78d 5.0 52.38b 4.1 51.22c 4.4 52.23b 4.5 53.34b 4.0 57.57a 4.1WH 31.03d 8.6 34.64ab 8.6 34.05b 7.9 34.91ab 8.6 35.20a 6.3 32.87c 6.0CW 21.61c 9.1 25.91a 7.9 25.45a 8.3 25.29ab 8.8 26.17a 6.7 24.56b 6.9CC 103.95c 4.8 113.31b 3.9 112.11b 4.3 115.92a 3.7 116.41a 4.0 117.23a 3.9

Means with the same superscripts in each row are not significantly different (PN0.05). Ψ HW = Height at wither; HB = Height at back; HR = Height at rump;BOL = Body length; BAL = Back length; NL = Neck length; HL = Head length; LCB = Forelimb length of cannon bone; LFL = Length of foreleg; WH = Width ofhip; CW = Chest width and CC = Chest circumference.

237E. Kefena et al. / Livestock Science 141 (2011) 232–241

and jennet populations are shown in Table 9. The levels ofsignificance and squared partial correlation criteria were usedto select for variables in a given order that contribute most indiscriminating between populations. The length of the forelegwas the most important variable used in separating betweenjacks populations followed by body length (Table 9).

Unlike between jack populations, the most important traitsfor discriminating between jennet populations were bodylength followed by length of foreleg. Chest circumference thatappeared in the tenth position in jack populations appeared inthe top five most important traits discriminating betweenjennet populations. The discriminatory power of chest width ismore important in jennet than in jack populations. Generally,the first two morphological variables that used in discriminat-ing between jack and jennet populations were the same exceptin the order of their appearances.

For aggregated gender, the first three most important mor-phometric variables used for discriminating between donkeypopulations were the length of foreleg, body length andchest width (Table 10). Literally, length of foreleg is a subsetanddescriptor of three bodyheights (height atwither, backandrump) as it measures height from the ground. Body length wasfound to be the secondmost important factor in discriminatingbetween donkey populations.

Table 6Eigenvectors, eigenvalues and cumulative values of the first four principal compon

TraitsΨ Components and eigenvectors

PC1 PC2 PC3 PC4

HW 0.329 −0.193 −0.188 −0HB 0.325 −0.209 −0.202 −0HR 0.326 −0.186 −0.201 −0BoL 0.310 0.044 0.083 0BaL 0.284 −0.077 0.487 −0NL 0.284 −0.042 0.556 0HL 0.289 0.009 0.251 0LCB 0.307 −0.086 −0.257 0LFL 0.320 −0.171 −0.157 −0WH 0.154 0.664 0.187 −0CW 0.187 0.557 −0.359 0CC 0.288 0.294 −0.114 −0

ΨHW = Height at wither; HB = Height at back; HR = Height at rump; BOL = BodForelimb length of cannon bone; LFL = Length of foreleg; WH = Width of hip; CW

Result of canonical discriminant analysis from two canon-ical variates for the overall donkey populations is shown inFig. 4. The first canonical variate separates Afar, Hararghe andOgaden from two groups: Abyssinian donkey population andOmo and Sinnar donkey populations. The second canonicalvariate roughly separates between Omo and Sinnar donkeypopulations. However, Afar, Hararghe and Ogaden donkeypopulations were hardly separable based on morphologicalgrounds.

4. Discussion

4.1. Morphological diversities in Ethiopian donkey populations

In our study, we exclusively depended on morphometricmeasurements, eco-geographical gradients and to some extentcoat colors patterns to characterize and identify the geograph-ical distributions of Ethiopian donkey populations. This comesfrom the assertions of Gubitz et al. (2000) who showed thatmorphological character systems reflect ecological selectionregimes (color patterns), history (body dimensions) or both(scalation).

In Ethiopia, the most peculiar donkey population withunique morphological characteristics is the Sinnar. Its

ents.

Eigenvalue Proportion Cumulative

.144 8.132 0.678 0.678

.168 1.264 0.105 0.783

.169 0.560 0.047 0.830

.084 0.541 0.045 0.875

.043 0.386 0.032 0.907

.330 0.266 0.022 0.929

.223 0.252 0.021 0.950

.032 0.233 0.019 0.970

.032 0.217 0.018 0.988

.600 0.108 0.009 0.997

.615 0.025 0.002 0.999

.128 0.015 0.001 1.000

y length; BAL = Back length; NL = Neck length; HL = Head length; LCB == Chest width and CC = Chest circumference.

Breeds/populations

Abyssinian

Omo

Afar

R-Squared1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0

Sinnar

Ogaden

Harar

Fig. 3. A dendrogram constructed using Mahalanobis distances betweenEthiopian donkey populations derived from morphological variables.

Table 7Squared Mahalanobis distances between jacks (below diagonal) and jennets(above diagonal) of Ethiopian donkey breeds/populations.

Breeds/populations

Abyssinian Afar Hararghe Ogaden Omo Sinnar

Abyssinian 0 29.73 25.13 34.51 66.59 44.28Afar 24.26 0 1.94 3.54 9.26 12.56Hararghe 21.37 0.96 0 4.24 10.83 16.69Ogaden 27.22 2.96 3.55 0 7.35 14.66Omo 47.85 11.26 12.98 9.42 0 15.13Sinnar 60.77 16.56 22.56 20.08 24.90 0

238 E. Kefena et al. / Livestock Science 141 (2011) 232–241

geographical niche is limited to the western lowlands alongEthio-Sudanese border and it is the tallest of all donkeypopulations found in Ethiopia. It has variable coat colors suchas white, leopard, black and brown. In addition to its use asworking animals, Sinnar is an excellent desert-adapted animalsused for riding. Sinnar jacks (Shokile/allele inOromifa language)are the primary choice for mule breeders. They rarely foundbeyond the western lowlands and if any, they are exclusivelyused to produce a mule that command premium price inEthiopia. Based on these unique and peculiar phenotypiccharacteristics, we speculate that its ancestral trunk might bedifferent from the rest of donkey population, but needs to beverified further by genetic studies.

According to previous studies (Beja-Pereira et al., 2004), wesuppose that themorphologically diverse andwidely distributeddomestic donkeys with characteristic leg strips might bedescendants of the Somali wild ass (E. africanus somaliensis).However, a study by Kimura et al. (2011) revealed that there isless support of molecular data for the Somali wild ass to be theancestor of Somali domestic donkey suggesting that Somalidomestic donkeys might be from unknown ancestral origin thatprobably became extinct. Kimura et al. (2011) argued that wildass in northeast Africa may have additional, yet unrecognizedgenetic substructure invoking new thoughts that domesticdonkeys in theHornofAfricahave still enigmatic ancestral origin.

However, suggesting that all the current Ethiopian donkeypopulations were descended from a common ancestral origin,phenotypic divergence in body size, coat color patterns or bothmight be created by their adaptive divergence to ecologicalvariables and biophysical resources. It may also underline thesignificance of landscape ecology and habitat suitability indetermining overall body dimensions, coat color patternsor both in donkeys, as it was adequately explained for otherspecies (Demetrius, 2000; Gizaw et al., 2007; Gubitz et al.,2000; Storz, 2002). Such variations are also suggested to havegenetic underpinning (Forsman et al., 2008; McKechnie et al.,2010; Paaby et al., 2010). Other explanations for variation in

Table 8Squared Mahalanobis distance between Ethiopian donkey breeds/popula-tions: aggregated gender.

Breed Abyssinian Afar Hararghe Ogaden Omo Sinnar

Abyssinian 0 – – – – –

Afar 51.65 0 – – – –

Hararghe 43.48 2.45 0 – – –

Ogaden 59.20 5.76 5.84 0 – –

Omo 109.52 19.58 21.84 15.22 0 –

Sinnar 120.55 31.64 43.84 35.53 42.25 0

body dimensions are associated with isolation by distance(Wright, 1943) and habitat fragmentation (Bennett andSaunders, 2010; Keyghobadi et al., 2005).

In the present study, morphometric evidences explic-itly showed that body size and coat color patterns follow eco-geographical patterns and altitudinal gradients in domesticdonkeys. For instances, Abyssinian donkey population thatmainly inhabit the highland ecology are phenotypically distinct(small in body size and predominantly brown coat colored)than arid and semi-arid domestic donkey populations, whichare bigger in size and grayish in coat color. This can be suf-ficiently explained in terms of the fact that such apparent dif-ferences are likely to be mediated by spatially varyingenvironmental factors and biophysical differences that even-tually led to divergence amongpopulations (Gubitz et al., 2000;Storz, 2002) and also has links with the concept of landscapeecology (Bennett and Saunders, 2010). Fairly bigger body sizesand similarities in coat color patterns among donkey popula-tions of arid and semi-arid lowlands also show that donkeys areactually desert-adapted heritage animals as reported in severalliteratures (e. g. Beja-Pereira et al., 2004; Blench, 2000; Epstein,1971; Marshall, 2007; Rossel et al., 2008).

4.2. Multivariate analysis

Large distances among Ethiopian donkey populations ingeneral and between Abyssinian and Sinnar donkey popula-tions in particular show noticeable diversities in Ethiopiandonkey populations. Nevertheless, relatively shorter distancesamong Afar, Hararghe, Ogaden and Omo donkey populationsappear to substantiate the assertion that in morphologicalcharacter systems are governed by similarities in ecologicalvariables and/or altitudinal gradients. Due to their geographicalproximity to each other, Hararghe highland donkey populationshare phenotypic similaritywithbothAfar andOgaden lowlanddonkey populations, but obviously distinct based on morpho-logical grounds.

Principal component analysis showed that among thetwelve measured morphometric variables, the three bodyheights (height at wither, back and rump) and body lengthalone accounted for about 87.5% of the variations in the overalldonkey populations. This might be due to high correlationcoefficients, particularly among major linear measurements

Table 9Order of traits used in discriminating between jack and jennet donkey populations in the stepwise discriminant analysis.

Jacks Jennets

Step Variable enteredΨ Partial R-square F value PrNF Variable entered Partial R-square F value PrNF

1 LFL 0.67 115.4 b0.0001 BOL 0.65 103.03 b0.00012 BOL 0.32 26.76 b0.0001 LFL 0.34 28.56 b0.00013 HR 0.21 15.25 b0.0001 CW 0.23 15.94 b0.00014 CW 0.21 14.71 b0.0001 CC 0.17 10.77 b0.00015 LCB 0.15 9.60 b0.0001 HR 0.15 9.79 b0.00016 HL 0.12 7.61 b0.0001 HL 0.14 8.95 b0.00017 BAL 0.07 4.23 b0.0010 WH 0.07 3.98 b0.00178 NL 0.06 3.60 b0.0036 HW 0.04 2.19 b0.05619 HB 0.06 3.33 b0.0062 NL 0.04 2.43 b0.035310 CC 0.05 2.77 b0.0183 BAL 0.03 1.74 b0.125011 WH 0.06 3.36 b0.0058

ΨHW = Height at wither; HB = Height at back; HR = Height at rump; BOL = Body length; BAL = Back length; NL = Neck length; HL = Head length; LCB =Forelimb length of cannon bone; LFL = Length of foreleg; WH = Width of hip; CW = Chest width and CC = Chest circumference.

239E. Kefena et al. / Livestock Science 141 (2011) 232–241

(data not shown). Manly (1986) underlined that high correla-tions between measured variables are precondition to carryout principal component analysis. Body height and bodylength generally describe the overall body dimension or linearmeasurements of animals. A sufficient interpretation for this isthat differences among the domestic donkey populations are amatter of body dimensions (body heights and body length)than other morphometric variables that describe body shape(widthofhip, chestwidth and chest circumference) in donkeys.

Stepwise discriminant functions analysis showed that thefirst two morphometric variables that discriminate betweenjack and jennet donkey populations were the same except inthe order of their appearances. This implies that there are noexplicit anatomical dissimilarities and/or sexual dimorphismbetween genders in donkey populations. Folch and Jordana(1997) also reported little sexual dimorphismbetween gendersin Catalonian donkey breed. Nevertheless, it has been shownthat jennets are longer bodied but shorter and jacks are tallerbut shorter bodied. Like chest circumference, chest width wasalso more important in discriminating between jennet donkeypopulations implying that jennets are fatter and blocky than

Ab= Abyssinian (Blue)

Af= Afar (Red)

Ha= Hararghe (Green)

Og= Ogaden (Cyan)

Om= Omo (Majenta)

Si= Sinnar (Orange)

15

10

5

0

-5

-7.5 -5.0 -2.5 0.0 2.5 5.0Can1

Can

2

Fig. 4. Plot of canonical discriminant analysis showing the first against thesecond canonical variant for the overall populations.

jacks. For aggregated gender, however, a variable that indicatebody height (length of foreleg) and body length were moreimportant than any other variables considered in this study.

The first two canonical variables can sufficiently separateamong the overall donkeypopulations (eigenvalue=91.85%).For instances, Abyssinian donkey populations are separatedwell by the first canonical variate from Afar, Hararghe andOgaden as well as from Omo and Sinnar population. This ismore or less consistent with the pairwise mean comparisonscomputed for aggregated gender. On the other hand, Afar,Hararghe and Ogaden donkey populations are hardly separa-ble by either of the canonical variables implying that they aresimilar basedonmeasuredmorphological variables regardlessof visible phenotypic differences among them.

5. Conclusion

Nationwide confirmatory and exploratory characteriza-tions of Ethiopian donkey populations showed that there isexplicit morphological diversity in the populations. Most of

Table 10Order of traits used in discriminating between donkey breeds/populations inthe stepwise discriminant analysis: aggregated gender.

Step Variable enteredΨ Partial R-square F-value PrNF

1 LFL 0.67 226.58 b0.00012 BoL 0.34 58.55 b0.00013 CW 0.21 30.72 b0.00014 HR 0.17 23.12 b0.00015 HL 0.13 16.06 b0.00016 CC 0.09 10.37 b0.00017 WH 0.07 8.33 b0.00018 LCB 0.05 6.44 b0.00019 NL 0.04 4.26 b0.000810 BaL 0.04 4.96 b0.000211 HW 0.04 4.38 b0.000612 HB 0.02 2.26 b0.0477

ΨHW=Height at wither; HB=Height at back; HR=Height at rump; BOL=Body length; BAL = Back length; NL = Neck length; HL = Head length;LCB = Forelimb length of cannon bone; LFL = Length of foreleg; WH =Width of hip; CW = Chest width and CC = Chest circumference.

240 E. Kefena et al. / Livestock Science 141 (2011) 232–241

the variations in the morphological character systems varywith eco-geographical patterns and biophysical resources.However, whether the variations in these morphologicaltraits are caused by adaptive or non-adaptive sources need tobe further verified by comparing between relative levels ofpopulation divergence in quantitative traits and neutral DNAmarkers.

Morphometric variables that were selected and used asmorphometric character descriptors in this study may beused as a priori to explain diversities in Ethiopian donkeypopulations. However, other morphometric variables andcorporal indices need to be further incorporated and used tofully characterize and describe donkey populations in Ethiopia.Furthermore, like as for other livestock species, a standardizedand uniform morphological descriptors and guidelines need tobe developed and used at wider scale in equines in general andin donkeys in particular.

Acknowledgements

This piece of work is a part of PhD project given to the firstauthor by the GermanAcademic Exchange Services (DAAD) incollaboration with the International Livestock ResearchInstitute (ILRI) in their regional program.We thank all farmersand pastoralists for sparing their animals for free and allregional agricultural offices for their inexorable cooperationduring data collection.

Appendix A. Description of measuring devices anddefinitions of the twelve morphological measurementstaken were as follows

No. Definition of quantitative morphologicalcharacters

Measuringdevice

1 Height at wither: distance from the highestpoint of the processus spinalis of the vertebrathoracic to the floor

Measuring stick

2 Height at back: Distance from the highestpoint of the back to the floor

Measuring stick

3 Height at rump: Distance from the rum (ilium)to the floor

Measuring stick

4 Body length: Distance from the most cranialpoint of the shoulder joint to the most caudalpoint of the pin bone

Measuring tape

5 Back length: Distance from the highest pointat wither to the highest point at rump

Measuring tape

6 Neck length: Distance from the highest pointat wither to nape by normal posture of thehead

Measuring tapeMeasuring tape

7. Head length: Distance from the nape to thebeginning of the nostril

Slide caliper

8 Length of cannon bone (forearm): Distancefrom the lateral toberculum of the osmetacarpale IV to the fetlock joint

Slide caliper

9 Length of foreleg: Distance from the pasternto the point of shoulder:

Measuring stick

10 Width of hip: Distance from the left to theright point of hip

Measuring tape

11 Chest width: Distance from the left to theright upper arm

Measuring tape

12 Chest circumference: Measured in place ofthe girth

Measuring tape

References

Beja-Pereira, A., England, P.R., Ferrand, N., Jordan, S., Bakhiet, A.O., Abdalla,M.A., Mashkour, M., Jordana, J., Taberlet, P., Luikart, G., 2004. African originof domestic donkey. Science 304, 1781.

Bennett, A.F., Saunders, A.S., 2010. Habitat fragmentation and landscapechange. In: Navjot, S.S., Paul, R.E. (Eds.), Conservation Biology for All.Oxford University Press, New York, pp. 87–104.

Blench, R.M., 2000. A history of donkeys, wild asses and mules in Africa. In:Blench, Roger M., MacDonald, Kevin C. (Eds.), The origin and develop-ment of African livestock: archaeology, genetics, linguistics andethnography. UCL Press, pp. 339–354.

Blench, R.M., 2004. The history and spread of donkeys in Africa. In: Fielding,D., Starkey, P. (Eds.), Donkeys, People and Dvelopment. A resource bookof the Animal Traction Network for Eastern and Southern Africa(ATNESA). Technical Center for Agricultural and Rural Cooperation(CTA), Wageningen, The Netherlands, pp. 24–32.

Breeds of Livestock, 1995. Horse Breeds. Oklahoma State University,Stillwater, Oklahoma. http://www.ansi.okstate.edu/breeds/horses/.

Demetrius, L., 2000. Directionality theory and the evolution of body size.Proceedings of the Royal Society of London 267, 2385–2391.

Epstein, H., 1971. The Origin of Domestic Animals of Africa. AfricanaPublication Corporation, New York, p. 719.

FAO (Food and Agricultural Organization of the United Nations), ). DomesticAnimal Diversity-Information System (DAD-IS)http://www.dad.fao.org/.

FAO (Food and Agricultural Organization of the United Nations), 2000. In:Scherf, B. (Ed.), World Watch List for domestic animal diversity. FAO,Rome, Italy.

Folch, P., Jordana, J., 1997. Characterization, reference ranges and theinfluence of gender on morphological parameters of the endangeredCatalonian donkey breed. Journal of Equine Veterinary Science 17 (2),102–111.

Forsman, A., Ahnesjo, J., Caesar, S., Karlsson, M., 2008. A model of ecologicaland evolutionary consequences of color polymorphism. Ecology 89 (1),34–40.

Gebrewold, A., Tegenge, A., Yami, A., 2004. Research need of donkeyutilization in Africa. In: Fielding, D., Starkey, P. (Eds.), Donkeys, peopleand development. A resource book of the Animal Traction Network forEastern and Southern Africa (ATNESA). Technical Center for Agriculturaland Rural Cooperation (CTA), Wageningen, The Netherlands, pp. 77–81.

Gizaw, S., Van Arendock, J.A.M., Komen, H., Windig, J.J., Hanotte, O., 2007.Population structure, genetic variation and morphological diversity inindigenous sheep of Ethiopia. Animal Genetics 38, 621–628.

Gubitz, T., Thorpe, R.S., Malhotra, A., 2000. Phylogeographic and naturalselection in the Tenerife gecko Tarentoal delalandii: testing historical andadaptive hypothesis. Molecular Ecology 9, 1213–1221.

Hanotte, O., Bradley, D.G., Ochieng, J.W., Vergee, Y., Hill, E.W., Rege, J.E.O.,2002. African pastoralism: genetic imprints of origins and migrations.Science 296, 336–339.

Holm, S., 1979. A simple sequentially rejective multiple test procedure.Scandinavian Journal of Statistics 6, 65–70.

Keyghobadi, N., Ronald, J., Matter, F.S., Strobeck, C., 2005. Among- andwithin-patch components of genetic diversity respond at different ratesto habitat fragmentation: an empirical demonstration. Proceedings ofthe Royal Society Series B 272, 553–560.

Kimura, B., Marshall, F., Chen, S., Rosenbom, S., Moehlman, P.D., Tuross, N.,Sabin, R.C., Peters, J., Barich, B., Yohannes, H., Kebede, F., Teclai, R., Beja-Pereira, A., Mulligan, C.J., 2011. Ancient DNA from Nubian and Somaliwild ass provides insights into donkey ancestry and domestication.Proceedings of the Royal Society Series B 278, 50–57.

Mahalanobis, P.C., 1936. On the generalized distance in statistics. Proceedingof the National Institute of India 2 (1), 49–55.

Manly, B.F.J., 1986. Multivariate Statistical Methods. Chapman and Hall,London, p. 159.

Marshall, F., 2007. African pastoral perspectives on domestication of thedonkey: a first synthesis. In: Denham, T., Iriarte, J., Vrydaghs, L. (Eds.),Rethinking Agriculture: Archaeological and Ethnoarchaeological Per-spectives. Left Coast Press, Walnut Creek, California, USA.

McKechnie, S.W., Blacket, M.J., Song, S.V., 2010. A clinally varying promoterpolymorphism associated with adaptive variation in wing size inDrosophila. Molecular Ecology 19, 775–784.

Oulehla, J., 1996. Zuechterische Standards in der Lipizzanerpferde-Popula-tion. Habilitationsarbeit, Brno-Piber.

Paaby, A.B., Blacket, M.J., Hoffmann, A.A., Schmidt, P.S., 2010. Identificationof a candidate adaptive polymorphism for Drosophila life history byparallel independent clines on two continents. Molecular Ecology 19,760–774.

Rossel, S., Marshall, F., Peters, J., Pilgram, T., Adams, M.D., O'Connor, D., 2008.Domestication of the donkey: timing, processes and indicators.

241E. Kefena et al. / Livestock Science 141 (2011) 232–241

Proceedings of the National Academy of Sciences of the United States ofAmerica 105, 3715–3720.

SAS (Statistical Analysis System Institute), 2004. SAS guide for personalcomputers, version 6. SAS Institute, Kary, NC, USA.

Storz, J.F., 2002. Contrasting patterns of divergence in quantitative traits andneutral DNA markers: analysis of clinal variation. Molecular Ecology 11,2537–2551.

Wright, S., 1943. Isolation by distance. Genetics 28, 139–156.