In this paper, four strategies were proposed for modeling tree crown volume using as independent variable stem variables, crown variables, combination of stem and crown variables, and stem volume. We used a dataset comprised of 170 trees from 12 temporary plots located in forest stands in southern Brazil. Models composed of stem variables presented weaker predictive ability. The best model contained crown variables, which explained 78.95% of observed variability. However, implementation of such model is bounded by its independent variables, which are not often measured in forest inventories. The model composed by diameter at breast height and crown length proved to be an adequate modeling approach. The predictive capability was kept by model , which is composed by most easily measured variable in a forest -diameter at breast height, also by the most easily acquirable crown variable -crown length. In our suggested model, estimates of and are coefficients that convert volume of a regular geometric solid -RGS is dbh² times crown length) -into crown volume, whilst estimate of is an allometric constant.