SUNLAB: A functional-structural model for genotypic and phenotypic characterization of the sunflower crop

Fenni Kang, Veronique Letort, Hugo Magaldi, Paul-Henry Cournede, Jeremie Lecoeur
2012 2012 IEEE 4th International Symposium on Plant Growth Modeling, Simulation, Visualization and Applications  
8 A new functional-structural model SUNLAB for the crop sunflower (Helianthus annuus L.) was developed. It is dedicated to simulate the sunflower organogenesis, morphogenesis, biomass accumulation and biomass partitioning to organs. It is adapted to model phenotypic responses of different genotypic variants to diverse environmental factors including temperature stress and water deficit. A sensitivity analysis was conducted to quantify the relative parameter influences on the main trait of
more » ... main trait of interest, the grain yield. The model was calibrated for four genotypes on two experimental datasets collected on plants grown under standard non-limiting conditions and moderate water stress. Its predictive ability was then tested on an additional dataset. The four considered genotypes -"Albena", "Melody", "Heliasol" and "Prodisol" -are the products of more than 30 years of breeding effort. Comparing the values found for the four parameter sets associated to each variant, allows to identify genotype-specific parameters. The model also provides a novel way of investigating genotype performances under different environmental conditions. These promising results are a first step towards the potential use of the model as a support tool to design sunflower ideotypes adapted to the current worldwide ecological and economical challenges and to assist the breeding procedure. 9 grated traits to more gene-related traits that should reveal more stable under 31 varying environmental conditions (Yin et al., 2004; Hammer et al., 2006). 32 Consequently, an important question to examine is how to design models 33 that can be used in that context. The models should simulate the phenotypic 34 traits of interest (e.g. yield) with good robustness and predictive capacity. 35 The models should also present a trade-off between mechanistic aspect and 36 complexity: Chapman et al. (2003) state that, for such use, a growth model 37 should include 'principles of response and feedbacks' to 'handle perturbations 38 to any process and self-correct, as do plants under hormonal control when 39 growing in the field' and to 'express complex behavior even given simple op-40 erational rules at a functional crop physiological level'. Casadebaig et al. 41 (2011) discuss that question in the case of their model SUNFLO (Lecoeur 42 et al., 2011). SUNFLO is a biophysical plant model that describes organo-43 genesis, morphogenesis, and metabolism of sunflower (Helianthus annuus L.). 44 It has shown good performances to identify, quantify, and model phenotypic 45 variability of sunflower at the individual level in response to the main abi-46 otic stresses occurring at field level but also in the expression of genotypic 47 variability (Casadebaig et al., 2011). The authors mixed mechanistic and sta-48 tistical approaches to deal with highly integrative variables such as harvest 49 index (HI). This HI factor is determined by a simple statistical relation-50 ship dependent on covariables previously simulated by the mechanistic part 51 of the crop model throughout the growing season. Although this statistical 52 solution and the large datasets used for its parameterization conferred good 53 robustness to the prediction of HI and thereby crop harvest, biomass parti-54 3 tioning to other plant organs and trophic competition between organs were 55 not taken into account. Fenni: Here, in last review, the third reviewer posed 56 a question " The idea of formalizing trophic competition between organs (p 4 57 l 27-33) is at the core of this new model. You write that such a formalization 58 should help representing feedback effects of biomass partitioning on "other 59 processes" (p 4 l 13-15). What do you mean?", I simply deleted the sentence 60 that saying "feedback effects of biomass partitioning on other processes can-61 not be taken into account", but I still tried to mention it in the discussion 62 saying "The plants are considered only at the minimal level of organ com-63 partments. PBM model ignores plant architecture and its plasticity. The 64 lack of individual organ's simulation can influence the simulation of plant 65 functioning. For example, PBMs models normally use the relative values of 66 the sink strength of organs to simulate biomass partitioning. These sink val-67 ues are assessed directly from experiments and the sources and sinks have no 68 significant direct interaction in these models (de Reffye et al., 2008). How-69 ever, the lack of trophic competition simulation may hinder the simulation of 70 feedback effects of biomass partitioning on other processes. As Pallas et al. 71 (2008) state, trophic competition influenced the organogenesis of grapevine 72 in their research". That paragraph had been deleted by you in last revision. 73 Do you think we should still mention somewhere this idea "mechanistic sink-74 source solver helps the simulation of biomass partitioning, and also the future 75 possibility of simulating feedback effects"?. I added one sentence in the dis-76 cussion to add about considering feedback effects:"'Introducing a mechanism 77 of trophic competition at organ level in a PBM, as done in this study, opens 78 the possibility to model feedbacks effects of biomass partitioning on other 79 4 processes such as photosynthesis or organogenesis (Mathieu et al., 2009). 80 "'. Do you agree? Moreover, it was shown in Lecoeur et al. (2011) that 81 HI is the parameter that contributes the most (14.3%) to the coefficient of 82 variation of the potential grain yield. It was also shown that when ranking 83 the processes in terms of their impact on yield variability, the first one was 84 biomass allocation (before light interception according to plant architecture, 85 plant phenology and photosynthesis). Therefore, Lecoeur et al. (2011) sug-86 gested that a better formalisation of the trophic competition between organs 87 could be a way to improve our understanding of genotypic variation for the 88 harvest index Fenni: here the third reviewer posed another question last 89 time " Please elaborate and give appropriate references to clarify the idea of 90 formalizing trophic competition (direct formalization? It could also be indi-91 rect) and to construct a sound argument (e.g. which outlines to satisfy the 92 need of feedback effects of biomass partitioning)." I actually didn't answer 93 his question about direct formalization or indirectIn fact I don't understand 94 the reviewer's question; what do you think he means by "'direct or indi-95 rect formalization"' ?. In order to take up this challenge, a new sunflower 96 model, named SUNLAB, was derived from SUNFLO. The representation 97 of plant topological development and allocation process at individual organ 98 scale were inspired by the functional-structural plant model GREENLAB, 99 which has been designed as a "source-sink solver" (Christophe et al., 2008) 100 and which is accompanied with the appropriate mathematical tools for its 101 identification (Cournede et al., 2011). SUNLAB thus inherits from GREEN-102 LAB the flexible rules of sink competition for biomass partitioning at organ 103 scale Fenni:the deleted phrases "(organ type includes blade, petiole, intern-104 5 ode and capitulum; a leaf consists of a blade and a petiole; for the modeling 105 of trophic competition, blade and petiole are considered as two organ types)" 106 were actually trying to answer the second reviewer's question in the previous 107 review " you are using the word "blade" to designate the leaves all over the 108 manuscript. This is somewhere confusing. The manuscript would benefit to 109 clarify this". Maybe somewhere a small explanation of blade could be added 110 if it is deleted here Yes, I had taken care of that and I had added it in the MM 111 part, in plantstructure paragraph as "' The different organ types, denoted as 112 o, include leaves (decomposed into blades and petioles),"' Do you think that 113 it will be ok ?, together with the more detailed representation of ecophysio-114 logical processes and environmental stress effects on biomass production and 115 yield from SUNFLO. 116 This paper presents in detail the mechanisms of SUNLAB and the pa-117 rameter estimation procedure based on field experimental data. A sensitivity 118 analysis is performed on the model parameters, using the Sobol method, to 119 investigate the relative contribution of each parameter and their interac-120 tions to the model output uncertainty. The output that we consider is the 121 main trait of interest in most breeding procedures, that is the final grain 122 yield. The potentials of SUNLAB for genotypic characterization are illus-123 trated by comparing the parameters obtained after the estimation process 124 for four genotypes, namely "Albena", "Heliasol", "Melody" and "Prodisol". 125
doi:10.1109/pma.2012.6524833 fatcat:7ksyj27vgrc2zbkvor5mlbwzgm