Multi-resolution methods are widely used in scientific visualization, image processing, and computer graphics. While many applications only require an one-time construction of these data-structures which can be done in a preprocess, this pre-process can take a significant amount of time. Considering large datasets, this time consumption can range from several minutes up to several hours, especially if this pre-process is frequently needed. Furthermore, numerous new applications, such as visibility queries, arise which often need a dynamic reconstruction of a scene database. In this paper, we address several problems of the construction or reconstruction of recursive tree hierarchies in parallel. In particular, we focus on parallel dynamic memory allocation and the associated synchronization overhead.