Analyzing and modelling the genetic variability of aerial architecture and light interception of oil palm (Elaeis guineensis jacq.)

The development of new breeding strategies to find more sustainable and productive systems is a major challenge to cope with ceaseless increasing demands for vegetable oils, notably palm oil. Optimizing plant architecture to increase radiation interception efficiency could be an option for enhancing potential oil palm production. Indeed, studies in cereals showed great improvement of yields by altering plant architecture, in combination with agronomic practices. By analogy, we proposed to investigate the influence of oil palm architecture on the capacity of the plant to intercept light, by using 3D reconstructions and model-assisted evaluation of radiation-use eciency. The first objective of this study was to analyse and model oil palm architecture and light interception taking into account genetic variability. A second objective was to explore the potential improvements in light capture and carbon assimilation by manipulating oil palm leaf traits and propose architectural ideotypes.
Data were collected in Sumatra, Indonesia, on five progenies (total of 60 palms), to describe the aerial architecture from leaflet to crown scales. Allometric relationships were applied to model these traits according to ontogenetic gradients and leaf position within the crown. The methodology allowed reconstructing virtual oil palms at different stages over plant development. Additionally, the allometric-based approach was coupled to mixed-effect models in order to integrate inter and intra progeny variability through progeny-specific parameters. The model thus allowed simulating the specificity of plant architecture for a given progeny while including observed inter-individual variability. The architectural model, parameterized for the different progenies, was then implemented in AMAPstudio to generate 3D mock-ups and estimate light interception efficiency, from individual to stand scales.
Model validations were performed at different scales. Firstly at organ scale, the geometry of the stem, the leaves and the leaflets were compared between virtual mock-ups and actual plants measured in field. Secondly, at plant scale with indicators derived from terrestrial laser scanning (TLS) to assess crown dimensions and porosity. These indicators integrated topological and geometrical information related to the amount of light intercepted by an individual. Finally, validations were performed at plot scale using hemispherical photographs (HP) to assess the variability of canopy openness for the five studied progenies.
Significant differences in leaf geometry (petiole length, density of leaflets and rachis curvature) and leaflets morphology (gradients of leaflets length and width) were detected between and within progenies, and were accurately simulated by the modelling approach. The comparison of plant area obtained from TLS and virtual TLS highlighted the capacity of the model to generate realistic 3D mock-ups. The architectural variabilities observed at plot scale between and within progenies were also satisfactory simulated. Finally, light interception estimated from the validated 3D mock-ups showed significant variations among the five progenies.
Sensitivity analyses (Morris's method and meta-modelling approach) were then performed on a subset of architectural parameters to identify the architectural traits impacting on light interception efficiency and potential carbon assimilation over plant development. Daily carbon assimilation was estimated with a photosynthesis model coupled to the radiative balance model, which enabled to integrate the temporal and spatial variations of photosynthetic organ irradiances.
The most sensitive parameters over plant development were those related to leaf area (rachis length, number of leaflets, leaflets morphology), but fine attribute related to leaf geometry showed increasing influence when canopy got closed. In adult stand, optimized carbon assimilation was estimated on plants presenting a leaf area index (LAI) between 3.2 and 5.5 m2.m2, with erected leaves, short rachis and petiole and high number of leaflet on rachis. Four ideotypes were identified in respect to carbon assimilation, exhibiting specific geometrical features that optimize light distribution within plant crown and reduce mutual shading among plants.
In conclusion, this study highlighted how a functional-structural plant model (FSPM) can be used to virtually explore plant biology. In our case of study, the 3D model of oil palm, in its conception and its application, permitted to detect the architectural traits genetically determined and influencing light interception. The limited number of traits revealed in the sensitivity analysis and the combination of traits proposed through ideotypes could guide further breeding programs. Forthcoming work will be dedicated to integrate in the modeling approach other physiological processes such as stomatal conductance and carbon partitioning. The improved FSPM could then be used to test different scenarios, for instance in climate change context with low radiations or frequent drought events. Similarly, the model could be used to investigate different planting patterns and intercropping systems, and proposed new multi-criteria ideotypes of oil palm.

Composition du jury :
M. Gerhard BUCK-SORLIN, Professeur, IRHS AgroCampus Ouest - Rapporteur
M. Christophe PLOMION, Directeur de recherche, INRA - Rapporteur
M. Eric DUFRÊNE, Directeur de recherche, CNRS - Examinateur
M. Alain RIVAL, Professeur, CIRAD - Examinateur
Mme Evelyne COSTES, Directrice de recherche, INRA - Directrice de thèse
M. Jean DAUZAT, Chercheur, CIRAD AMAP - Co-directeur de thèse
M. Jean-Pierre CALIMAN, Directeur de recherche, SMARTRI - Invité