Modeling plant growth and development underwent considerable development with strong incentives from various consortia. It emerged as an efficient tool in ecology and genetics to face new challenges raised by competition for resources and to benefit from breakthroughs in biotechnology. In this presentation, we propose a classification of approaches used in modeling plants based on our experience at the Laboratoire d'Ecophysiologie des Plantes sous Stress Environnementaux (Montpellier, France). We present 5 types of models and discuss how they participate in a general workflow aimed at designing plant ideotypes for future climates. Type 1 are frameworks for spatiotemporal analysis of growth and development. They consist in quantitative description of when and where growth occurs in a specific organ, or how growth is synchronized between organs at whole plant level or, more generally, how development occurs. They are developed almost systematically when a novel species or organ are to be studied in the lab, as they are critical to set up experimental designs (what to measure, when, ...) or to reason the sampling of tissues for omics studies (which part of the organ should be taken, when). Type 2 are data-driven (statistical) representations of the responses to the environment and of determinant of growth. Often based on specialized statistical methods (multivariate analysis, structural equation models, mixed models,...), they allow us to formalize responses, reveal linkages and hierarchies between processes or multiple factors affecting growth, and demonstrate causalities between them. They are used as such to produce synthetic representations of growth processes or decipher between genetic and environmental effects. They are also used as bases for building other types of models. Type 3 are empirical models of response to environmental factors specially designed for parameterizations in high throughput phenotyping. This entails parsimony in the parameterization (hence in input variables), high heritability of parameters (traits) and compatibility with high throughput measurements. Such models have proved to be very efficient in genetic studies, for characterizing allelic diversity or finding QTLs. They also are ready to use and genetically well parameterized components for more integrative models. Type 4 are computer-assisted simulation models, used to make predictions a priori. They are based on hypothetical frames of integration of biological processes within the plant, and on physical models for the interactions with soil and atmospheric compartments. They allow simulation of complex system (crop models). They are used in the lab to simulate the behavior of plants in different climatic scenarios and thus to assess the relevance of a given set of traits for a given location. They are also produced to help reasoning agricultural practices, and especially irrigation. Type 5 are computer simulation models, used to reconstruct or interpret experiments a posteriori. Such models are designed to be fitted to experimental data, either to compute hidden variables (like sink strengths, radiation interception and efficiency), interpolate variables of interest from fragmental inputs in time or space (e.g. LAI time course) or interpret a macroscopic response by extracting meaningful parameters (e.g. hydraulic characteristics from difference in growth kinetics). During this presentation, new challenges are proposed to modelers in each type of models. They consist in bottlenecks which currently hamper plant scientists from achieving multiple goals in biology, ecology, agronomy or breeding. Work done in collaboration with Christian Fournier, Francois Tardieu, Denis Vile, Angelique Christophe, Eric Lebon, Christine Granier and Bertrand Muller.