Automatic calibration of a functional-structural wheat model using an adaptive design and a metamodelling approach

Abstract

- Background and Aims Functional-structural plant models are increasingly being used by plant scientists to address a wide variety of questions. However, the calibration of these complex models is often challenging, mainly because of their high computational cost. In this paper, we applied an automatic method to the calibration of WALTer: a functional-structural wheat model that simulates the plasticity of tillering in response to competition for light. - Methods We used a Bayesian calibration method to estimate the values of 5 parameters of the WALTer model by fitting the model outputs to tillering dynamics data. The method presented in this paper is based on the Efficient Global Optimisation algorithm. It involves the use of Gaussian process metamodels to generate fast approximations of the model outputs. To account for the uncertainty associated with the metamodels approximations, an adaptive design was used. The efficacy of the method was first assessed using simulated data. The calibration was then applied to experimental data. - Key Results The method presented here performed well on both simulated and experimental data. In particular, the use of an adaptive design proved to be a very efficient method to improve the quality of the metamodels predictions, especially by reducing the uncertainty in areas of the parameter space that were of interest for the fitting. Moreover, we showed the necessity to have a diversity of field data in order to be able to calibrate the parameters. - Conclusions The method presented in this paper, based on an adaptive design and Gaussian process metamodels, is an efficient approach for the calibration of WALTer and could be of interest for the calibration of other functional-structural plant models .

Functional–Structural Plant Modeling Highlights How Diversity in Leaf Dimensions and Tillering Capability Could Promote the Efficiency of Wheat Cultivar Mixtures

“…However, it is expected that both sowing density and the proportions of the two cultivars in mixture should have an impact on the results (Grace and Tilman, 1990). Moreover, the sowing density has been identified, as expected, as one of the most influential inputs in WALTer (Lecarpentier et al, 2019;Blanc et al, 2021), supporting the interest in exploring variations of this parameter. FSPMs are particularly adapted to explore these agricultural practices (Evers et al, 2019;Gaudio et al, 2019).…”

“…Nevertheless, as for any modeling exercise (Passioura, 1996), our study presents some approximations that must be considered when interpreting the results. Indeed, even though WALTer allows for a realistic simulation of the tillering dynamics in response to competition for light (Lecarpentier et al, 2019;Blanc et al, 2021), some of the underlying hypotheses of the model should be mentioned. First of all, light is the only environmental factor considered in the model: water and nutrients (nitrogen, phosphorus, potassium, sulfur.…”

Towards a Stochastic Model to Simulate Grapevine Architecture: A Case Study on Digitized Riesling Vines Considering Effects of Elevated CO2

“…In a climate impact study of Ben Touhami and Bellocchi [ 139 ] it was found that the inclusion of prior information was beneficial for uncertainty estimates, even if these information came from different climate conditions. Blanc et al [ 140 ] point out that a pitfall of Bayesian model calibration for functional-structural plant models is that with transferring uncertainties in model parameters repetitive simulation runs are necessary for robust estimates of output variables. To address this, they propose a surrogate modeling approach to substitute time consuming FSP model simulations with simpler models, such as Kriging models.…”