Genetic variability in light-related parameters of maize models revealed by linking 3D-reconstruction of plant architecture with light model and phenotyping platform

International audienceRadiation interception efficiency (RIE) and radiation use efficiency (RUE) are the main driving forces of dry mass accumulation, so parameters related to RIE and RUE, e.g. light extinction coefficient (k) and photosynthetic parameters, have strong impacts on yields. Since these parameters are strongly related to the canopy structure, architectural traits of plants are important for characterization of the genetic variability of these parameters, a critical work for modelling the interactions between yield and environmental conditions and predicting future genetic gain. In this work, we propose a new method to estimate the RIE- and RUE-related parameters of maize for the crop models by a high-throughput phenotyping platform, PHENOARCH (https://goo.gl /x3C6oN). In autumn 2014, 330 maize lines were grown under non-stressed condition with five repetitions. For each plant, one top image and 12 side images of the plant were taken every 2-3 days. These images were used to reconstruct the 3D-architecture of the plants. To estimate RIE of a maize line on day t, the 3D plants of this line were used for constructing a virtual canopy consisting of 15 plants in three rows. Based on the architecture of this virtual canopy, RIE was calculated by the RATP light model and leaf area index (LAI) was estimated by the images analyses and the reconstructed 3D-architecture. k was obtained by computing the slope between natural logarithm of light transmittance through the canopy and LAI. Relationship between RIE and plant developmental stage was fitted to a sigmoidal function with three parameters: maximum RIE (RIEmax), maximum change of RIE (smax) and time taken to reach smax (ts). Maize lines were genotyped with 354K polymorphic SNPs and genome wide association analysis (GWAS) was performed over model parameters. Between genotypes, significant differences in k, RIEmax, smax and ts were found, ranging from 0.31-0.75 (unitless), 0.43-0.98 (unitless), 0.12-0.27 (d-1) and 18.3-39.5 (d), respectively. GWAS revealed 16 QTL for k, 77 for RIEmax, 1 for smax and 7 for ts. This study is an example of model-assisted phenotyping and we conclude that using the 3D architecture of plants reconstructed by phenotyping platform has strong potential to discover the genetic variability of light-related traits. Further possibilities include the estimation of radiation use efficiency and relative canopy photosynthetic capacity