Socioeconomic and environmental drivers are important determinants urban plant richness patterns. The scale at which these patterns are observed in different regions, however, has not been explored. In arid regions, where forests are not native, the majority of the urban forest is planted, and trees are presumably chosen for specific attributes. Here, we investigate the role of spatial scales and the relative importance of environmental vs. socioeconomic drivers in determining the community structure of southern California's urban forest. Second, we assess the usefulness of ecosystem service-based traits for understanding patterns of urban biodiversity, compared with species composition data. Third, we test whether resident preferences for specific tree attributes are important for understanding patterns of species composition and diversity. We studied tree communities in 37 neighborhoods in three southern California counties (Los Angeles, Orange, and Riverside). The urban forest in southern California is very diverse with 114 species. Using multiple regression analyses we found socioeconomic drivers were generally more important than environmental and the strength of the relationship between urban forest community structure and socioeconomic drivers depended on whether we were analyzing within or across counties. There was greater tree richness in wealthier neighborhoods compared with less affluent neighborhoods across all counties and Orange County, but not in Los Angeles or Riverside counties alone. We also found a greater proportion of residential shade trees in hotter neighborhoods than in cooler neighborhoods, which corresponds with survey results of residents' preferences for tree attributes. Ultimately our study demonstrates that the species richness and functional traits of urban tree communities are influenced by managers' and residents' preferences and perceptions of urban tree traits.