Team Digiplante

Members
Overall Objectives
Scientific Foundations
Software
New Results
Partnerships and Cooperations
Bibliography
PDF e-pub XML


Section: Overall Objectives

Research fields

Plants play a major role in our society for human food, biodiversity, as environmental regulator, materials or energy source. The human management of plants is therefore a crucial issue for the future of our societies, and the way this management is conducted today is far from satisfactory: pollution by intensive farming, agricultural shortage in some parts of the world, environmental catastrophe caused by extreme deforestation, loss of biodiversity. And yet, plants are organisms that are well known by biologists. This important knowledge has only poorly allowed the development of management tools, probably from the lack of a systemic approach. Mostly empirism and Fisher's statistics have been used to help define farming, forestry or breeding strategies.

The first models of plant biophysics only appeared in the seventies (Monteith, 72,77; De Wit, 78). At this stage, energetic budgets are considered at field scale. In parallel, mostly for computer graphics applications, individual-based models of plant geometry were developed, based on formal grammars (Smith, 84; Prusinkiewicz, 88,89) or automata (De Reffye et al., 88). The need for applied mathematics remained quite limited at this stage (numerical simulation, computer graphics). The biophysical and geometrical approaches would finally meet in the late 90's and see the functional-structural plant models (FSPM) community emerge: interactions between plant structure and biophysical functioning is henceforth seen as the key to understand plant growth and offer a framework for the integration of existing biological knowledge. The mathematical issues raised by this new type of models are spectacular. Plants are now described as full systems, both structurally and functionally, and appear as typical complex systems: multi-scale issues, multiple and interacting biophysical processes, regulation and retroaction loops, that can no longer be fully studied with only simulations.

The Digiplante team was born in this context around Philippe de Reffye in 2002, first incubated in the Metalau group at INRIA Rocquancourt (M. Goursat, J.P. Quadrat) and as an independent joint team between INRIA, Centrale and CIRAD in 2004, with P. de Reffye and P.H. Cournède as permanent researchers. The main objective of the team at this time was the development of the GreenLab model [De Reffye and Hu, 2003], its mathematical formalism and the fitting of real plant data to study the genericity and applicability of the model. New concepts are introduced, like the interactions between structure and functioning, the effects of competition for light or a new model of secondary growth. The Digiplante software developed by Centrale propose a versatile fitting tool, that is used to study a wide variety of real plants (sugar beet, maize, rice, sunflower, Arabidopsis, rapeseed, cucumber, sweet pepper, tomato, cotton, grapevine, beech tree, pine tree, cecropia) with different biological partners. In 2007, Marc Jaeger joined Digiplante after 5 years of LIAMA's direction. He brought with him the computer graphics and plant visualization themes in Digiplante. In 2007, the team was also transferred to INRIA Saclay, it became an 'Equipe-Projet' INRIA in 2008 and was evaluated positively in 2009.

When confronting the GreenLab model to real plant data, the chosen approach was mostly horizontal, that is to say that the general idea was to see if the concepts implemented in the GreenLab model could apply to a wide variety of plants. This strategy allowed the implementation of new features in the model, making it suitable for plants as diverse as Arabidopsis and Beech tree. This step can be seen as successful. However, in only very rare cases the robustness of the model and its predictive capacity have been studied. When implemented, this vertical approach was not fully satisfactory, which makes the model still unproper for real applications: modeling the interactions with the environment should be improved, no measure and control of model uncertainty is provided. The reason for this is still the lack of a proper mathematical methodology for the analysis of models and their full evaluation.

Digiplante is now a joint team between INRIA Saclay and Centrale, under the scientific responsibility of P.-H. Cournède. The objective of Digiplante is now clearly orientated towards the development of a proper mathematical methodology for the analysis of models and their full evaluation, around 3 axes: