Impacts of climate change on spatial wheat yield and nutritional values using hybrid machine learning

Summary

This study integrated controlled field experiments with three wheat cultivars and multiple micronutrient treatments across varying soil and climate conditions with national yield statistics to train and test machine learning algorithms. An automated stacked ensemble model robustly predicted grain yield and nutritional concentrations (nitrogen, iron, zinc), which were then used to project spatial impacts of climate change across three global circulation models and two emissions pathways to mid-century. Results suggest climate change will modestly increase wheat yield and protein concentration but decrease iron and zinc concentrations, presenting a paradoxical challenge to food security and nutritional adequacy.

UK applicability

UK wheat production may experience similar yield increases under mid-century climate change, but the projected declines in iron and zinc concentrations are concerning for national nutrition security. The model's transferability to UK conditions depends on whether the diverse soil and climate conditions in the training data adequately represent British growing environments, which is not explicitly addressed in the abstract.

Key measures

Grain yield (GY), nitrogen (N), iron (Fe), and zinc (Zn) concentrations; R² values for model predictions (GY >0.78, N >0.75, Fe >0.71, Zn >0.71); projected percentage changes in yield and nutrient concentrations under two climate scenarios (SSP2-45 and SSP5-85) for mid-century (2020–2050) relative to historical period (1980–2010)

Outcomes reported

The study modelled spatial wheat grain yield, protein, iron, and zinc concentrations under mid-century climate scenarios using machine learning ensemble methods trained on controlled field experiments and national yield statistics. Projections indicate climate change will increase wheat yield and protein concentration but decrease iron and zinc concentrations by mid-century.

Topic tags