AgMIP is a distributed climate-scenario simulation exercise which includes model intercomparisons using historical data as well as analyses of the effects of future climate change conditions on existing and adapted crop management systems. The project involves the participation of multiple crop and agricultural economics modeling groups around the world with the goals of improving the characterization of risk of hunger and world food security due to climate change and to enhance adaptation capacity in both developing and developed countries.
At the core of AgMIP is the use of multiple models for the purposes of quantifying model uncertainties and improving the performance of all participating models. Figure 1 illustrates the general categories of tools that are envisioned for AgMIP to enable the multiple climate, crop, and economic models to be used for model intercomparisons and assessments.
Data harmonization / data translation
One of the goals of the AgMIP project is a comparison of the simulated results from various crop models for the purposes of quantifying model uncertainties and as a basis for model improvement. Multiple crop models are used to generate simulated estimates of crop yields and other data using identical input data for each model. But the format of the input data varies among the models. Some models use text file inputs, some use XML files, some use Excel spreadsheets, etc. Manual translation of data into the various model formats is not practical for large amounts of data that will be processed by AgMIP researchers.
To solve this problem, AgMIP will store crop experiment data in a harmonized data format, referred to as the ACE (AgMIP Crop Experiment) data. Once in harmonized format, the data can be exported to the formats specific to each model. We are in the process of developing the data translation tools to create the input data specific to each model. The AgMIP IT team is working with the crop model development teams for each crop model to accomplish this goal. A series of development workshops are being organized that focus on programming the applications necessary for exporting data from the AgMIP harmonized crop experiment database to the data formats specific to each of the major crop models associated with AgMIP.
The first IT development workshop was part of the AgMIP South Asia Regional workshop held at ICRISAT in Patancheru, India, February 20-24, 2012. We paired crop model experts with software application developers to produce data translation tools for producing model-ready input files specific to each crop model. Additional tool development workshops are currently being planned.
AgMIP tools and applications are being developed collaboratively using a github repository (http://github.com/agmip). A wiki describing AgMIP tool development protocols is hosted at The AgMIP IT Team Wiki. The wiki shows information about our development environment, coding standards, and the GIT version control system. AgMIP applications developers can sign up for a GitHub account and fork one of the AgMIP repositories and begin collaboration immediately.
Crop model input data from (field experiment sentinel sites) comprise just one category of data that will be used in AgMIP. Other data types, as listed in Figure 1, are climate scenarios, generalized soil properties, GIS raster and vector data, Representative Agricultural Pathways (RAPs) and crop scenarios, crop model output data, regional economic model inputs and outputs, and global economic model outputs. Each of these categories of data will be associated with a harmonized database and the data translation tools appropriate for the associated models.
Automated model integration
It is anticipated that model integration of climate, crop and economic models will progress in stages. The first step is to get multiple models running using the same input data, as described in the previous paragraphs. Outputs from climate models will be stored in a database for use individually by the crop models. The outputs from the crop models will be stored in a database and aggregated for use by the global and regional economic models. Outputs from the global economic models will provide pricing structures and external factors used by the regional economic models. All model output data will be stored for use by aggregation, analysis and visualization to produce final products for display and viewing on the AgMIP web site.
The second phase of integration will involve a more automated process by which the full workflow from climate modeling through crop modeling and economic analysis can proceed based on user-selected scenarios. This will require the coordination of very high speed computers with fast access to terabytes of information. We envision that the AgMIP integrated modeling architecture will be based on Galaxy software (http://galaxy.psu.edu/). The Computation Institute at the University of Chicago is developing a generalized version of the Galaxy platform which integrates several unique features which would be of use to AgMIP, including improved data transfers; authentication and authorization; and parallel job scheduling in High-Performance Computing (HPC) environments on clusters, supercomputers, grid, and cloud resources. This work is beyond the scope of currently funded projects within AgMIP, but we are currently writing proposals to secure funding to allow more automated model integration to be possible for AgMIP researchers.