The Global Gridded Crop Model Intercomparison phase 1 simulation dataset

Christoph Müller, Joshua Elliott, David Kelly, Almut Arneth, Juraj Balkovič, Philippe Ciais, Delphine Deryng, Christian Folberth, Steven Hoek, R. C. Izaurralde, Curtis D. Jones, Nikolay Khabarov, Peter Lawrence, Wenfeng Liu, Stefan Olin, Thomas A. M. Pugh, Ashwan Reddy, Cynthia Rosenzweig, Alex C. Ruane, Gen Sakurai, Erwin Schmid, Rastislav Skalský, Xuhui Wang, Allard de Wit, Hong YangScientific 2019 · DOI

TL;DR

This paper introduces the Global Gridded Crop Model Intercomparison (GGCMI) phase 1 dataset, a comprehensive collection of global crop model simulations. This dataset aims to advance our understanding of crop model performance, the relationship between agricultural productivity and environmental impacts, and to guide future improvements in global crop modeling frameworks.

What Problem Does This Paper Address?

Accurately predicting how global agriculture will respond to climate change and other environmental pressures is crucial for food security and sustainable development. While numerous crop models exist, their outputs can vary significantly due to differences in model structure, input data, and parameterization. This makes it challenging to compare results, understand uncertainties, and identify areas for model improvement. The GGCMI phase 1 addresses this by creating a standardized, large-scale dataset of simulations from multiple global gridded crop models (GGCMs) under consistent protocols, allowing for a systematic evaluation of model performance and inter-model variability.

Methods

The core of this work is the creation of a standardized dataset through the GGCMI phase 1 protocol. Fourteen different crop modeling groups participated, each running their GGCMs using up to 11 historical input datasets (weather, soil, etc.) spanning from 1901 to 2012. Simulations were conducted under three management harmonization levels (default, fullharm, harmnon) and for both rainfed and near-perfectly irrigated conditions. This comprehensive approach allows for the analysis of model responses to different climate scenarios, management practices, and the inherent uncertainties associated with these factors. The paper details the setup and calibration narratives for each of the 14 participating models, providing transparency into their methodologies.

Datasets

The primary dataset introduced is the GGCMI phase 1 simulation dataset. This dataset comprises annual data fields at a 0.5 arc-degree spatial resolution, covering the global ice-free land surface. It includes outputs for up to 15 different crops and 14 output variables.

The simulations were driven by up to 11 historical input datasets, including:

  • AgCFSR
  • AgMERRA
  • CFSR
  • ERA-I
  • GRASP
  • GSWP
  • Princeton GF
  • Princeton GF version 2
  • WATCH
  • WFDEI.CRU
  • WFDEI.GPCC

These input datasets were derived from various reanalysis products and were bias-corrected against different observation-based products (e.g., CRU, GPCC) to account for input data uncertainty.

Key Insights and Ideas

  • Quantifying Model Uncertainty is Crucial: The GGCMI phase 1 highlights that the choice of crop model and its parameterization can lead to significant uncertainty in simulated yields. This dataset provides a framework to systematically assess and understand this uncertainty.
  • Harmonization is Key, but Not Perfect: While efforts were made to harmonize input data and management assumptions, the paper acknowledges that not all aspects could be standardized due to model complexity and data availability. This diversity, however, can also reflect real-world variability in agricultural practices.
  • Beyond Yields: A Richer Dataset: The dataset includes more than just crop yields, offering variables like evapotranspiration, biomass, and nutrient application. This allows for a more holistic understanding of crop-environment interactions and model performance beyond just yield prediction.
  • Guiding Future Model Development: By evaluating model performance against historical data and identifying areas of divergence, this dataset serves as a critical benchmark for improving the accuracy and robustness of future global gridded crop models.

Key Results

The paper primarily describes the characteristics and creation of the GGCMI phase 1 dataset. While it doesn't present a single overarching quantitative result in terms of model accuracy across all models and crops, it establishes that:

  • Broad Coverage: The dataset covers the global ice-free land surface, simulating up to 15 crops across various historical climate and management scenarios.
  • Model Diversity: Fourteen different GGCMs contributed to the dataset, showcasing a range of modeling approaches and complexities.
  • Comprehensive Outputs: For each simulation, up to 14 output variables were provided, offering a rich resource for diverse analyses.
  • Initial Evaluation: A preliminary evaluation study (referenced in the paper) indicated that GGCMs generally perform better than statistical models in explaining observed yield variability, but their explanatory power diminishes in regions where yield variability is driven by factors other than weather. No single GGCM consistently outperformed others.

Limitations and Future Work

  • Soil Data Not Harmonized: The paper explicitly states that soil data was not supplied to modelers and they were requested to use their own input data. This lack of harmonization in soil parameters is acknowledged as a limitation, given the importance of soil for crop yield simulations.
  • Inconsistent Growing Season Variables: The protocol for defining growing season weather variables (sumt, gsrsds, gsprec) was not sufficiently clear, leading to inconsistent reporting across models. Users are advised to compile their own indicators using weather input data and sowing/maturity dates.
  • Data Masking and Aggregation: The dataset includes simulations for areas where crops are not currently grown. Users are cautioned to use crop- and irrigation-specific masks or handle data with care, and to analyze results for larger spatial entities rather than individual sites.
  • Future Work: The authors aim to promote further analyses using this dataset to understand crop model performance, productivity-environment relationships, and to identify pathways for improving global gridded crop model frameworks. They also suggest that the dataset can serve as input for other models and for quantifying model uncertainties.

Summary updated

Abstract

The Global Gridded Crop Model Intercomparison (GGCMI) phase 1 dataset of the Agricultural Model Intercomparison and Improvement Project (AgMIP) provides an unprecedentedly large dataset of crop model simulations covering the global ice-free land surface. The dataset consists of annual data fields at a spatial resolution of 0.5 arc-degree longitude and latitude. Fourteen crop modeling groups provided output for up to 11 historical input datasets spanning 1901 to 2012, and for up to three different management harmonization levels. Each group submitted data for up to 15 different crops and for up to 14 output variables. All simulations were conducted for purely rainfed and near-perfectly irrigated conditions on all land areas irrespective of whether the crop or irrigation system is currently used there. With the publication of the GGCMI phase 1 dataset we aim to promote further analyses and understanding of crop model performance, potential relationships between productivity and environmental impacts, and insights on how to further improve global gridded crop model frameworks. We describe dataset characteristics and individual model setup narratives.
BibTeX
@inproceedings{muller2019global,
  author = {Christoph Müller and Joshua Elliott and David Kelly and Almut Arneth and Juraj Balkovič and Philippe Ciais and Delphine Deryng and Christian Folberth and Steven Hoek and R. C. Izaurralde and Curtis D. Jones and Nikolay Khabarov and Peter Lawrence and Wenfeng Liu and Stefan Olin and Thomas A. M. Pugh and Ashwan Reddy and Cynthia Rosenzweig and Alex C. Ruane and Gen Sakurai and Erwin Schmid and Rastislav Skalský and Xuhui Wang and Allard de Wit and Hong Yang},
  title = {{The Global Gridded Crop Model Intercomparison phase 1 simulation dataset}},
  booktitle = {Scientific},
  year = {2019},
  doi = {10.1038/s41597-019-0023-8},
}
Figure