GGCMI introduces first online tool for crop model evaluation

Figure6GGCMI Global gridded crop model (GGCM) intercomparison using time series correlation coefficients for the top-10 maize producing countries. Rows display the individual countries ordered by production; left-hand labels describe the best performing GGCMs for that country and the correlation coefficients. White boxes indicate that correlations are not statistically significant. See paper for more.
C. Müller et al. 2017

By Christoph Müller

Model improvement is one of the central objectives of The Agricultural Model Intercomparison and Improvement Project (AgMIP.) For gridded large-scale crop models, model improvement has been hampered by the lack of suitable reference data. The Global Gridded Crop Model Intercomparison of AgMIP has now established a common benchmark for the evaluation of global gridded crop models (GGCMs) for wheat, maize, rice and soybean. In the recently published paper “Global gridded crop model evaluation: benchmarking, skills, deficiencies and implications”, open access in Geoscientific Model Development, this benchmark is used to test the performance of 14 participating GGCMs at global, national and pixel scale.

In setting new standards for model evaluation, GGCMI not only defines a benchmark against which models can be tested, it also supplies an online tool where all modelers can evaluate their models, including new models and subsequently improved model versions. The tool reproduces the analysis of the paper and was developed with the support of the Geoshare team. It is hosted on mygeohub.org at mygeohub.org/tools/ggcmevaluation where simulation results can be uploaded, processed online and results and figure plots can be downloaded after processing.

“Global crop modelers can now evaluate and demonstrate their models’ skills in a standard procedure, which greatly improves intercomparability between studies and facilitates targeted model improvement”, lead-author Christoph Müller commented.

“We also try to set new standards in the agricultural modeling community with respect to open knowledge by publishing in an open-access journal, making all input and all output data publicly available, as well as the processing scripts used in the analysis, and by providing an online tool”, added co-author Joshua Elliott, who led the publication of the modeling protocol for this phase of simulations in GGCMI.

Links:
Evaluation publication: C. Müller et al. 2017
Protocol publication: Elliott et al. 2015
Online tool for global crop model evaluation: mygeohub.org
github repository of GGCMI data processing and analysis scripts: github
data access to input and output data: RDCEP.org


About the Author:
Christoph Müller works at the Potsdam Institute for Climate Impact Research (PIK) in Potsdam, Germany
and co-leads the AgMIP gridded crop model initiative (Ag-GRID) and GGCMI.