[Defended thesis] Mario Serouart: Intraspecific competition in maize : Observation through in situ phenotyping, in silico modelling, light regime and application to sowing structure optimization

Intraspecific competition in maize : Observation through in situ phenotyping, in silico modelling, light regime and application to sowing structure optimization

• Starting date : 3rd May 2021
• University : A2E – ED 536 Agrosciences & Sciences
• PhD school: University of Avignon
• Scientific field: Agronomy, IT, Statistics
• Thesis management: Frédéric Baret, Inrae EMMAH Avignon
• Thesis supervisors: Raul Lopez-Lozano, Inrae EMMAH | Jean-Charles Deswarte & Benoît de Solan, Arvalis institut du végétal
• Funding: CIFRE Arvalis | ACTA
• #DigitAg : Cofunded thesis – Axe 6 : Modélisation et simulation (systèmes de production agricole), Axe 2 : Innovations en agriculture numérique, Axe 3 : Capteurs, acquisition et gestion de données, Challenge 1 : Le challenge agroécologique, Challenge 2 : Le phénotypage rapide

Keywords: Deep Learning, LiDAR, Feature Extraction, High Speed Phenotyping, Agricultural Data, Artificial Intelligence, Light Detection, Plant Phenotyping

Abstract: Agroecology integrates ecological principles into agriculture to enhance crop resilience by limiting inputs and maximizing efficiency of natural resources or mechanisms. This latter point is the focus of this thesis. Indeed, one option to enhance the performance of agricultural systems is to optimize plant architecture increasing radiation interception efficiency. This improvement could be achieved through a natural mechanism called phenotypic plasticity. In a competitive environment, with high plant density, maize plants optimize their position by adjusting the arrangement of their photosynthetic organs (leaves) based on neighboring plants, through biophysical processes. Thus, a shade reduction dynamic could limit the illuminated surface area lost at the canopy level. Following this hypothesis, this study aims to understand the influence of plant canopy on its ability to intercept light. This involves two levels of study: maize architecture at the individual plant level and the overall canopy architecture through sowing patterns, defined by density and row spacing. Our analyses rely on both true field trials, to be as close to real conditions (in situ) as possible, and 3D reconstructions of canopy, establishing a radiative balance on these virtual scenes (in silico). Concerning in situ data, they were collected over the three years of thesis. An unique panel of five hybrid was studied under various sowing patterns configurations, creating more or less competitive environments depending on plant density and spacing between them (from 6 to 12 plts.m−2 and from 0.4 à 0.8m, respectively). A set of aerial characteristics was precisely described for each treatment and trial. This was done through manual acquisitions or indirect estimations using high-throughput phenotyping. Special emphasis was placed on genetic and environmental differences in leaf orientations. This focus, on this specific trait, is due not only to its importance in light interception but also to a lack of knowledge identified in the state of the art. For this purpose, we developed an image analysis algorithm to describe the interactions G×E of induced plasticity by modifying row spacing. Subsequently, thanks to this focus and using data from other aerial parameters collected, allometric relationships were used to model (in silico) virtual canopies. These relationships are considered realistic as they explicitly derive from what was observed in the field. In these cases, they are referred as 3D numerical representations. This reconstruction method allowed simulating several architectural specificities both at the individual and canopy levels. Thus, estimating their influence on the light regime to extract ideotypes, i.e., the best theoretical configurations for optimizing yield. Finally, these conclusions were compared to the actual yield measurements from the three years of experiments to study the reliability of the integrative approach on the yield maximization issue. The results of these analyses revealed significant differences both within hybrids and sowing patterns. Particularly in leaf positioning, showing a strong predominance of leaves oriented perpendicularly to the row direction with increasing intra-row competition. However, we found that this behavior ultimately had little influence on light interception thanks to sensitivity analysis (max. 25% importance). However, it turned out that vertical leaf inclination is the main architectural variable that regulates light intensity and its distribution. In this regard, plants with more erectophile leaves will result in increased canopy photosynthesis. Finally, the configuration of the sowing pattern played a crucial role. According to this thesis work, reducing row spacing systematically had a positive impact on yield, regardless of the density considered. The magnitudes of yield gains involved in narrowing rows are approximately +1.0 to 2.0 t.ha−1 when sowing at high density (>10 plts.m−2) and considering relatively short row distances (<0.6m). A hypothesis is that there may be better light interception and distribution in squared patterns, allowing for up to 20% more radiation capture compared to rectangular ones.

Social networks: LinkedIn – Site – Github

Contact : mario.serouart [AT] inrae.fr