Evaluating the sensitivity of agricultural model performance to different climate inputs

Area of Science:

• Agricultural Science
• Climate Modeling
• Crop Simulation

Background:

• Accurate agricultural models and climate inputs are essential for validating future food production projections.
• Previous studies have not fully distinguished errors from climate inputs versus crop models due to a lack of observed yield data.

Purpose of the Study:

• To systematically evaluate the reliability of a widely-used crop model (pDSSAT) for simulating U.S. maize yields.
• To compare model performance using various observational climate data products versus uncorrected reanalysis data.

Main Methods:

• The parallelized Decision Support System for Agrotechnology Transfer (pDSSAT) model was used.
• pDSSAT was driven with climate inputs from multiple sources: reanalysis, bias-corrected reanalysis, and a control dataset.
• Model outputs were compared against observed historical U.S. maize yields.

Main Results:

• Model simulations showed higher accuracy when driven by observation-based precipitation products compared to uncorrected reanalysis data.
• Biased precipitation distribution significantly impacted yields only in arid regions, contrary to previous findings.
• Crop yields demonstrated oversensitivity to precipitation fluctuations and undersensitivity to floods and heat waves across all climate input scenarios.

Conclusions:

• Limitations within the crop models themselves, rather than climate input data, appear to be the primary factor affecting the accuracy of agricultural projections.
• Further research should focus on improving the structural integrity and realism of crop models to enhance predictive capabilities.