An objective approach to model reduction: Application to the Sirius wheat model.

Area of Science:

• Agricultural Science
• Computational Biology
• Crop Modeling

Background:

• The Sirius model is a widely used simulation tool for wheat growth and development.
• Model evaluation is crucial for ensuring accuracy and efficiency in agricultural research.
• Systematic model reduction can identify areas for improvement in complex simulation systems.

Purpose of the Study:

• To systematically evaluate and reduce the complexity of the Sirius wheat growth model.
• To identify redundant variables within the model that do not significantly contribute to prediction accuracy.
• To enhance the efficiency and rigor of crop model evaluation through automated procedures.

Main Methods:

• A systematic model reduction procedure was applied using software control to replace model variables with constants.
• Predictions from reduced models were compared against growth analysis observations (biomass, yield, leaf area) from 9 trials.
• Model performance was assessed under optimal, nitrogen-limiting, and drought conditions in the UK and New Zealand.

Main Results:

• Out of 111 variables, 16 were identified as potentially redundant.
• Redundancy was observed in areas including carbon translocation, nitrogen physiology, temperature adjustments, vernalization, and soil processes.
• The study demonstrated that simplified model versions could maintain predictive power for key growth parameters.

Conclusions:

• The representation of certain processes within the Sirius model may not be justified by their contribution to prediction accuracy under the tested conditions.
• The automated model reduction approach provides an efficient and systematic method for rigorous model evaluation.
• This 'model intra-comparison' enhances the understanding of variable importance in crop simulation.