Combining crop growth modelling and statistical genetic modelling to evaluate phenotyping strategies

Genomic prediction of complex traits, say yield, benefits from including information on correlated component traits. Statistical criteria to decide which yield components to consider in the prediction model include the heritability of the component traits and their genetic correlation with yield. Not all component traits are easy to observe or measure. Therefore, it may be attractive to include proxies to yield components, where these proxies are measured in (high throughput) phenotyping platforms during the growing season. Using the APSIM (Agricultural Production Systems Simulator) cropping systems model, we simulated phenotypes for a wheat diversity panel segregating for a set of model parameters regulating phenology, biomass partitioning, and the ability to capture environmental resources. The distribution of the additive QTL effects regulating these parameters followed the same shape and rate as observed for a Gamma distribution of QTL effects on real phenotypic data. In the multi-trait prediction models for final yield we included proxies to biomass and canopy cover as observed over the growing season. Various drought stress scenarios were evaluated. Each scenario triggers different adaptive mechanisms and the importance of component traits differs between drought scenarios. The combined use of crop growth models and multi-trait genomic prediction models provides a procedure to assess the efficiency of phenotyping strategies. It also allows to quantify the impact of yield components under different environment types on yield prediction accuracy. Heritability for the parameters of P-splines or parametric models to characterize the dynamics of secondary traits was larger than that of individual time points. This increases the potential of secondary traits to obtain a larger prediction accuracy for the target trait. Yield prediction accuracy benefitted from including biomass and green canopy cover parameters in scenarios with no- or limited water stress. The advantage of the multi-trait model was smaller for the early-drought scenario, due to the reduced correlation between the secondary and the target trait. Therefore, multi-trait genomic prediction models for yield require scenario-specific correlated traits.