Historically, the Merino has been the dominant sheep breed in Australia having first been imported to the country over 200 years ago. In addition, starting in the late 1990s, sheep producers began importing Dohne Merino embryos from South Africa to improve on attributes such as reproduction and carcase composition. Since then, this breed has continued to expand in Australia but the number of genotyped and phenotyped purebred individuals remains low, calling into question the feasibility of genomic selection in this breed. The Australian Merino on the other hand has a very large reference population in a separate genomic evaluation. Many studies have shown that combined genomic evaluation of several populations can be advantageous in terms of accuracy when it involves genetically close populations. In particular, combined evaluations can be very advantageous for small breeds that can benefit from the large reference population of another breed. This study was based on 27,632 Australian Merino (M), Dohne Merino (D) and crossbred (C) genotyped individuals and a pedigree of more than 4 million animals and more than 5.2 million phenotypes for two wool traits (fibre diameter and greasy fleece weight) and one weight trait (yearling liveweight). The first objective of this study was to characterize the genomic structure and relationships between these populations. A Principal Component Analysis of the genomic relationship matrix as well as computations of Hudson's fixation index (Fst), that were below 0.08, revealed a low genetic differentiation between M, D and C populations. This suggests that crossbred or combined predictions may be feasible. In the following step of this work, we will investigate the accuracy of genomic evaluation in a Dohne validation population based on purebred D, purebred M, crossbred C or the combination of these three reference populations.