In recent years, object-based methods have become a well-recognised tool for data fusion. Commercially available software using this concept rely heavily on user inputs as a basis for the knowledge rules. A common criticism of this knowledge-derived approach is that the user needs to have a significant knowledge of the scene in order to choose the best parameters for segmentation and subsequent fusion of the multisource data. However, in the data-driven fusion model the user plays no part until the computational aspects are completed. In this article, a new data-driven fusion model is presented for modelling individual trees in a forest by applying watershed segmentation and subsequent fusion, using tree heights and crown signatures derived from airborne lidar data and four-band multispectral imagery. The study area is part of the Moira State Forest, New South Wales where the dominant tree species are native eucalypts. A tree crown model (TCM) was computed as the difference between lidar first and last returns. A watershed segmentation algorithm was used to extract individual tree crowns from the TCM. The attributes of the tree objects are then generated from multispectral imagery and lidar data. Finally, tree objects are fused to delineate individual tree species. The application of the data-derived fusion technique led to an 80 percent fusion accuracy for forest species identification at tree level. The results indicate that the data-derived fusion model may prove suitable for estimating and mapping the crown area, height and species of forest at tree level.