Embracing multiculturalism has become an important initiative for enhancing cultural identity, social cohesion, and harmony in many migrant-receiving countries. While a large body of literature has examined ethnic diversity in relation to migration, less attention has been given to tracking changes in ethnic diversity over time and from a spatial perspective. Our study aims to contribute to this gap by examining nationwide ethnic diversity in Australia from 2001 to 2021 at various spatial scales and exploring its relationship with the neighbourhood features of ethnic settlement to provide practical implications for urban planning and design. To measure ethnic diversity, we draw on the five census datasets from 2001, 2006, 2011, 2016, and 2021 and apply Shannon's diversity index to the smallest census unit, as well as to the levels of capital cities, rural areas, and states. We then employ a machine learning method to model the relationship between ethnic diversity and neighbourhood features of migrants' settlement, including demographic and socioeconomic features and housing characteristics, and to reveal changes in this relationship over time. Our key findings suggest that the level of ethnic diversity dropped during 2006–2011, possibly due to changes in migration policy in Australia, and dropped again during 2016–2021, possibly due to the outbreak of COVID-19. Neighbourhoods with high ethnic diversity are associated with populations that speak English fluently and hold high education degrees, as well as easy access to public transit and affordable housing. Our study provides practical suggestions and implications for migration and settlement policy-making and promoting better social cohesion through urban planning, neighbourhood design, and housing regulation.