Fruit trees present developmental characteristics which are similar to other perennials, such as the existence of ontogenetic gradients and dependencies between consecutive growth. However, because of their agronomical use, fruit tree species also raise specific issues. For instance, the practice of grafting on dwarfing rootstock considerably reduces the juvenile period during which trees do not flower. During the mature phase, flower formation is often instable from one year to the next, leading to an irregular fruit bearing. This represents a major disability for fruit production that is usually prevented through a large number of manipulations such as grafting, pruning, chemical applications for fruit thinning, water and nitrogen supply. Presently, fruit production is facing a number of new challenges such as climatic changes or the necessary reduction of incomes in orchards, which may endanger its performance. A deep renewal of practices has to be engaged, involving a large set of competences. Indeed, numerous factors (genetic, climatic, abiotic and biotic stresses) interacting in a complex network must be optimised to find innovative answers to the current economical and environmental challenges. The role of modelling will thus be crucial in the formalisation of this complexity and the exploration of innovative scenarios. In this talk, we will attempt to highlight key ideas and concepts that underline our research and that aim to face the different levels of complexity. We will also submit our current open questions for discussion. Our approach consists in the exploration of plant structure from cellular to whole plant scale, taking into account one or several scales depending on the target question. For the exploration of tree perennial development over several years and dependencies between consecutive growth, macroscopic scales of description, such as annual shoots or growth units, are usually considered. Mixed probabilistic/structural models have been built in which a decomposition approach was applied to separate the growth components due to ontogeny or environmental factors in the measured plant development. Their application in apple tree has led us to identity unexpected patterns, in which two successive phases corresponding to different patterns of alternations between flowering and vegetative growth units were identified. The current challenge concerns the analysis of the influence of genetic and environmental factors on the development of perennial plant structures over several years. Such mixed probabilistic/structural models have also been combined with different physical and eco-physiological models (carbon acquisition, transpiration, carbon and water transport, effect of gravity, etc.) in fruit tree simulation systems. However, the integration of different sub-models, either mechanistic or stochastic, into a global simulation model has generated new issues due to the high complexity of the whole system which needs to be carefully addressed. Moreover, due to the large number of potential sources of variation (parameters, environmental variables, sub-model choices, ...), exploring the behaviour of such models is not possible analytically, and in silico experiments combined with sensitivity analyses are required. However, the strategy and intermediate objectives that must be defined to progress towards the build of efficient and integrated perennial plant models at macroscopic scales remain open questions. At the microscopic scale, the understanding of the relative contribution of mechanisms at different scales, from molecules, cells and organs to the construction of individual plant structures, is a major challenge for biologists. Following pioneer studies on model plants, projects have recently emerged for agronomic species. After having demonstrated the genetic control of architectural traits in segregating populations of apple, we have engaged in the description of the cellular patterning of apple tree organs, internodes and leaves, in different allelic combinations. The relative contribution of the number of cells and cell size to the final internode shape was shown to be genotype-depend and affected by a period of soil water deficit. Similarly, the relative contribution of the number of cells and cell volumes in apple leaf histogenesis is currently under study. The large databases generated by these studies will be presented and are likely to lead to new 3D modelling developments with the objective to link cellular to organ scale.