Using standard survey approaches, we carried out a nationwide survey to morphologically characterize and identify ecozones of Ethiopian horse populations. Accordingly, we explored one new feral horse population previously unreported and seven other distinct horse populations. A total of 17 selected morphological variables were recorded on 503 horses (293 stallions and 210 mares) that belong to five out of eight identified horse populations. Pair-wise multiple mean comparisons (PMMC) using one-way analysis of variance and multivariate analyses were performed separately for each sex and least-squares means was used in the case of aggregated sexes. Results of PMMC showed that there were significant differences (P < 0.05) between the means for most of the variables recorded. Principal component analysis showed that height at wither, height at back, height at rump, body length, back length and barrel length jointly account for about 80 percent of the variations. All squared Mahalanobis distance between populations were significant (P < 0.01). The greatest phenotypic divergence was observed between Bale and Selale horse populations and the least phenotypic divergence was between Horro and Kafa populations. Canonical discriminant function analysis showed that 77.05 percent of individuals were correctly categorized into their respective populations. Moreover, cluster analysis based on squared Mahalanobis distances grouped the five measured Ethiopian horse populations into three major breed groups and five distinct horse populations.