By Carolyn Cox
AgMIP Approach to Data and Model Harmonization
AgMIP and USDA held a workshop on January 10, 2014 on “Harmonizing Agricultural Data and Models.” The workshop was organized by Jim Jones, Jerry Hatfield, Cynthia Rosenzweig, and Carolyn Cox (Coordinator of the Florida Climate Institute, which hosted the event). The purpose was to explore concepts for harmonizing agricultural databases and models being developed and used by various USDA research projects and AgMIP. Harmonization would lead to web-based data access for evaluating and improving agricultural models and using them for analyses of climate change impacts, adaptation strategies, and potential food security issues. The USDA-ARS and NIFA-funded projects are increasingly being mandated to ensure their data are open for broad use, with the recognition that data are highly valuable for use beyond the immediate goals of specific projects and represent a substantial public investment. However, even though similar data may be collected across projects, the fact that different databases and formats are used by different projects and labs makes the discovery and access to valuable datasets impractical.
There is increasing use of crop, livestock, and forestry models in research on climate change, food security, and other major issues. Crop models, for example, use similar input data on soils, weather, and management; however, those crop models have been developed by different research groups and use different data formats and units. AgMIP has already made major progress toward harmonizing data inputs and outputs across different crop models because of its emphasis on using multiple models. This work was done by the AgMIP IT Team, led by Cheryl Porter (University of Florida) and Sander Janssen (Wageningen University), with direct contributions by crop modeling groups worldwide (see www.agmip.org/it-team). This AgMIP crop model harmonization approach, which does not require code to be changed, has been implemented for six different crop model families and work is continuing for adding additional models. According to the AgMIP IT Team, a similar approach could be developed to allow web users to access databases on crops, livestock, or other domains without requiring the databases to use the same data structures or the same database software.
Representatives from different USDA projects and labs presented summaries of the data that they are collecting, the approaches that they are using to store data, and their thoughts about making their data openly available. In addition, Simon Liu of the USDA National Agricultural Library (NAL) presented concepts that he has been working on to help harmonize agricultural data (click here to see a list of representatives). Breakout sessions allowed the attendees to address key questions concerning approaches for harmonization, the benefits, and ways forward.
The following conclusions were presented in the closing plenary session:The various NIFA-funded and ARS projects and AgMIP can gain mutual benefits through cooperation on data and agricultural model activities.
The approach developed by AgMIP to harmonize among crop models could be adapted to harmonize data from different USDA and NIFA databases with inputs from NIFA/USDA ARS colleagues and the NAL.
Collaboration can also lead to similar development for other systems, such as livestock, such that an approach will evolve that facilitates model intercomparisons and improvement and access to data across platforms and database systems.
There are low hanging fruit that can be harvested via collaboration among NIFA projects, ARS projects, the NAL, and AgMIP to demonstrate and evaluate the harmonization of data.
Multiple models within AgMIP can be used to assist in analyses of data collected in the NIFA-funded projects.
A follow-up workshop should be organized to coordinate the efforts of different USDA projects and labs toward database harmonization.
Common goals and potential collaborations between AgMIP and the ARS/NIFA projects on evaluation, improvement, and use of agricultural models (crop and livestock) were also identified. The collaboration will ultimately help to answer the overarching policy question: What is the vulnerability of the U.S. food and fiber system to climate extremes and change?