Crop science: A foundation for advancing predictive agriculture

The pathway to impact from proof of concept to adoption for improved maize yield productivity in the water-limited environments of the US corn-belt was achieved through a long-term research effort involving private-public partnerships.

“The pathway to impact, from proof of concept, through reduction to practice, to adoption for improved maize yield productivity in the water-limited environments of the US corn-belt, was achieved through a long-term research effort involving private–public partnerships (Messina et al. 2020a).”

Alternative approaches based on advancing prediction methods to explore and exploit the full potential of the G × E × M state-space are emerging and deserve greater consideration.

“Alternative approaches based on advancing prediction methods to explore and exploit the full potential of the G × E × M state-space are emerging and deserve greater consideration (Chapman et al. 2012; Peng et al. 2020; Messina et al. 2020b).”

A general principle guiding development of future crop productivity prediction technologies is to develop the capability to identify crop improvement strategies that can avoid problems before they arise rather than solving those problems once they have occurred.

“Beyond the specifics of predicting G × E × M interactions and selecting for G–M technology combinations, a general principle guiding development of future crop productivity prediction technologies is to develop the capability to identify crop improvement strategies that can avoid problems, before they arise, rather than solving those problems once they have occurred (Messina et al. 2020b).”

Prediction of G × E × M interactions will require an integration of progress in genomic prediction of traits for breeding applications with advances in understanding crop responses to environmental variation for agronomic prediction.

“Prediction of G × E × M interactions will require an integration of the progress that has been made to enable genomic prediction of traits for breeding applications, together with advances in understanding crop responses to environmental variation for applications in agronomic prediction (Hammer et al. 2019; Voss-Fels et al. 2019; Messina et al. 2020b).”

We can use our collective understanding from scientific disciplines contributing to crop improvement to predict the interplay of genetics and management with projected environmental changes associated with climate change to provide a foundation to understand the impact of climate change on the future of food production.

“We can use our collective understanding from the scientific disciplines contributing to crop improvement to predict the interplay of genetics and management with the projected environmental changes associated with climate change to provide a foundation to understand the impact of climate change on the future of food production (Chapman et al. 2012; Messina et al. 2020b; Hammer et al. 2020; Snowdon”

Messina et al. (2020a, b) proposed an iterative empirical-modelling approach based on a fusion of mechanistic biophysical models, statistical and machine learning approaches to explore the genetic, environment and management dimensions simultaneously.

“To apply this foundation for G × E × M prediction, Messina et al. (2020a, b) proposed an iterative empirical-modelling approach based on a fusion of mechanistic biophysical models, statistical and machine learning approaches to explore the genetic, environment and management dimensions simultaneously.”

Messina et al. (2020a, b) proposed an iterative empirical-modelling approach based on a fusion of mechanistic biophysical models, statistical and machine learning approaches to explore the genetic, environment and management dimensions simultaneously.

“To apply this foundation for G × E × M prediction, Messina et al. (2020a, b) proposed an iterative empirical-modelling approach based on a fusion of mechanistic biophysical models, statistical and machine learning approaches to explore the genetic, environment and management dimensions simultaneously.”

The focus would shift from describing G × E × M interactions to predicting G × M interactions for different G-M technology combinations and quantifying their performance characteristics for the environment-types of the TPE, defined as predicting (G-M) × E interactions.

“Thus, instead of describing G × E × M interactions the focus would shift to predicting G × M interactions for different G–M technology combinations and quantifying their performance characteristics for the environment-types of the TPE, herein defined as predicting (G–M) × E interactions, following Messina et al. (2020b).”

An alternative strategy is to focus on research methodologies to identify G-M technology combinations.

“An alternative strategy is to focus on research methodologies to identify G–M technology combinations (Hammer et al. 2014; Messina et al. 2020b).”

Such advanced knowledge provides feedback to breeders to guide design and refinement of MET stages of breeding programs.

“Such advanced knowledge provides feedback to breeders to guide design and refinement of the MET stages of breeding programs (Messina et al. 2020a).”

Prediction-based crop improvement strategies provide a viable approach to tackle the high dimensionality of the G × E × M factorial.

“Firstly, they provide a viable approach to tackle the high dimensionality of the G × E × M factorial (Messina et al. 2020b).”

For complex traits that are targets of crop improvement programs, prediction methods based on a fusion of mechanistic crop models and machine learning, which both include trait genetics, will be required.

“For the complex traits that are the targets of crop improvement programs we can anticipate that prediction methods based on a fusion of mechanistic crop models and machine learning, which both include trait genetics, will be required (Messina et al. 2020b; Ersoz et al. 2020; Washburn et al. 2020; Diepenbrock et al. 2021).”

Large commercial breeding organisations have the resources to undertake comparisons of alternative crop improvement strategies and to resource their practical implementation.

“Large commercial breeding organisations have the resources to undertake such comparisons of alternative crop improvement strategies and to resource their practical implementation (e.g. Messina et al. 2020a).”