Statistical Emulators of Maize, Rice, Soybean and Wheat Yields from Global Gridded Crop Models

TitleStatistical Emulators of Maize, Rice, Soybean and Wheat Yields from Global Gridded Crop Models
Publication TypeJournal Article
Year of Publication2018
JournalAgricultural and Forest Meteorology
Volume236
Pages145-161
Date Published09/2018
Abstract

This study provides statistical emulators of crop yields based on global gridded crop model simulations from the Inter-Sectoral Impact Model Intercomparison Project Fast Track project. The ensemble of simulations is used to build a panel of annual crop yields from five crop models and corresponding monthly summer weather variables for over a century at the grid cell level globally. This dataset is then used to estimate, for each crop and gridded crop model, the statistical relationship between yields, temperature, precipitation and carbon dioxide. This study considers a new functional form to better capture the non-linear response of yields to weather, especially for extreme temperature and precipitation events, and now accounts for the effect of soil type. In- and out-of-sample validations show that the statistical emulators are able to replicate spatial patterns of yields crop levels and changes overtime projected by crop models reasonably well, although the accuracy of the emulators varies by model and by region. This study therefore provides a reliable and accessible alternative to global gridded crop yield models. By emulating crop yields for several models using parsimonious equations, the tools provide a computationally efficient method to account for uncertainty in climate change impact assessments.

URLhttp://dx.doi.org/10.1016/j.agrformet.2016.12.022
DOI10.1016/j.agrformet.2016.12.022
Funding Program: 
Journal: Agricultural and Forest Meteorology
Volume: 236

This study provides statistical emulators of crop yields based on global gridded crop model simulations from the Inter-Sectoral Impact Model Intercomparison Project Fast Track project. The ensemble of simulations is used to build a panel of annual crop yields from five crop models and corresponding monthly summer weather variables for over a century at the grid cell level globally. This dataset is then used to estimate, for each crop and gridded crop model, the statistical relationship between yields, temperature, precipitation and carbon dioxide. This study considers a new functional form to better capture the non-linear response of yields to weather, especially for extreme temperature and precipitation events, and now accounts for the effect of soil type. In- and out-of-sample validations show that the statistical emulators are able to replicate spatial patterns of yields crop levels and changes overtime projected by crop models reasonably well, although the accuracy of the emulators varies by model and by region. This study therefore provides a reliable and accessible alternative to global gridded crop yield models. By emulating crop yields for several models using parsimonious equations, the tools provide a computationally efficient method to account for uncertainty in climate change impact assessments.

DOI: 10.1016/j.agrformet.2016.12.022
Year of Publication: 2018
Citation:
Blanc, E.  2018.  "Statistical Emulators of Maize, Rice, Soybean and Wheat Yields from Global Gridded Crop Models."  Agricultural and Forest Meteorology 236: 145-161, doi:10.1016/j.agrformet.2016.12.022.