Engineering Meteorological Features to Select Stress Tolerant Hybrids in Maize

Area of Science:

• Agricultural Science
• Environmental Science
• Data Science

Background:

• Crop yield is significantly impacted by environmental stresses like drought and heat.
• Accurate prediction and quantification of these stresses are crucial for agricultural resilience.
• Existing indices may not fully capture the complex interactions affecting crop performance.

Purpose of the Study:

• To develop novel stress metrics integrating meteorological and soil data for drought and heat stress.
• To predict maize yield using a random forest regression model and explainable AI.
• To rank environments based on stress levels and identify stress-resilient maize hybrids.

Main Methods:

• Meteorological and soil data were used to calculate drought and heat stress indices across four maize growth stages.
• A random forest regression model was employed for yield prediction, incorporating feature importance analysis.
• Environmental stress levels were quantified using model-derived feature contributions, creating a new environmental index.

Main Results:

• A new environmental index was developed to quantify combined drought and heat stress.
• Environments were ranked based on their stress levels, aiding in targeted agricultural management.
• Over 2400 maize hybrids were evaluated and classified for tolerance to heat, drought, or combined stresses.

Conclusions:

• The developed stress metrics provide a novel approach to assessing environmental stress in agriculture.
• Model explanation techniques enhanced the understanding of feature contributions to yield reduction.
• The study successfully identified maize hybrids with improved tolerance to combined heat and drought stress, contributing to the Syngenta Crop Challenge 2019.