Coupling Process-Based Models and Plant Architectural Models: A Key Issue for Simulating Crop Production

Abstract

Process-Based Models (PBMs) can successfully predict the impact of environmental factors (temperature, light, CO2, water and nutrients) on crop growth and yield. These models are used widely for yield prediction and optimization of water and nutrient supplies. Nevertheless, PBMs do not consider plant architecture as a determinant of yield, thus they often lack the flexibility to follow plant plasticity. Leaf area index (LAI), flower and fruit abortion are usually not predicted very well, because PBMs operate at the level of plant compartment (e.g. all leaves together) and unit area of crop instead of the phytomers where the feedback between plant growth and plant architecture operates. Functional Structural Plant Models (FSPMs) use the architecture as the support of functioning, integrate properly the action of the environmental conditions at the phytomer level that gives the plant its full plasticity: feedback between biomass production and biomass partitioning for both development (functioning of meristems) and growth (sink strength and variation, allometry of organs). These kinds of models suffer from the drawbacks that data acquisition is very heavy and model parameter estimations rely on numeric simulation that demands thousands of iterations as no analytical solution is available. Coupling PBM and FSPM, as done in the GreenLab model, is an important step to improve yield prediction not only with respect to its quantity but also to its quality (number and size of organs, branching pattern in ornamental plants). In the GreenLab model biomass production depends on PAR (photosynthetically active radiation) and LAI, a common pool of biomass is assumed and partitioning of biomass among organs is based on their relative sink strength. Plant architecture is simulated dynamically and the organ functioning is modified continuously during growth according to various environmental conditions. Such mathematical models could provide better yield prediction, as crop production depends on the functioning of the plant architecture, and parameter identification is more accurate with inverse methods. Some examples addressing the above-mentioned issues are presented in this paper using the GreenLab model.