Next Steps for AgMIP in North America

Contributed by Carolyn Mutter, Jerry Hatfield, Senthold Asseng, Bruno Basso, Jean Steiner, Sanford Eigenbrode, Cheryl Porter and the AgMIP Open Session Participants.


The Open Session of the Agricultural Model Intercomparison and Improvement Project (AgMIP) at the Fall 2017 meeting of the American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America brought together about 50 US and international agricultural systems researchers to discuss priorities and future directions for AgMIP in North America. Moderated by Dr. Jerry Hatfield of USDA ARS, the session included perspectives of five panelists on how to better synthesize North America agricultural systems interactions and outputs at present and in the future, before opening the floor to discussion. Panelists included Drs. Senthold Asseng (U Florida), Bruno Basso (Michigan State U), Jean Steiner (USDA-ARS), Sanford Eigenbrode (U Idaho), and Cheryl Porter (U Florida). The session was organized by the AgMIP Coordination Unit at Columbia University.

Session Introduction: A key motivation of the AgMIP community of scientists is to add value to ongoing experiments and modeling activities. This includes advancements to interoperable data and models for model intercomparison and improvement, agricultural system assessments at global and regional scales, and next generation approaches (see also “The US community of AgMIP is very active in the global-scale studies”, Hatfield noted. “However, given the significance of US and North American agricultural contributions to the global food system, it is important to advance collaborative studies here [in North America] also”.

Crop Modeling: There are many areas for systematic advancement in the understanding of current and future crop growth in North America that are within the range of current technology but not yet represented in models. These include high resolution impact assessments for all staple crops inclusive of soil, management, and cultivar information; the impact of genetic adaptation and/or land use changes (e.g., shifts North); impact assessments (including land use) for fruits and vegetables; harvest quality assessments; shifts in pest, weed, and disease dynamics; impacts on soil, including movement of carbon, nutrients, and pesticides (with links back to plant and human health); and, expanding to linkages with life cycle assessments.

Corn-in-Iowa There is critical need for improved characterization of soil processes in the models. “In a recent meta-analysis of 300 peer-reviewed agricultural research papers, only 5 considered soil in a robust manner,” Basso said. “Many account for feedbacks, but it is easy to get that right for the wrong reasons,” he added. The US is well positioned with high-resolution data, including mapping of within field variability. Electrical resistivity maps are increasingly used to more efficiently assess soil depth across regions. Adaptation planning should include consideration of soil nutrients, temperature, and so forth. We have the tools and the thinking. We need strengthened agency support.

AgMIP has international experience in impact assessment in mixed systems and we need to apply it to regions of North America. “AgMIP can help by establishing standards and protocols to enable comparability across scales and with other regions in the world, facilitating also strong linkages with field experimentation and data management; climate science advances and economics assessments; and the training of next generation scientists”, said Asseng. Research Partnerships focused on climate change assessment issues will benefit by forming research-question-driven teams with realistic goals and timelines. Partnerships need to embrace AgMIP principles – that is, transparent, open and inclusive; utilizing the best available approaches, with cross-disciplinary teams, co-leadership, stakeholder engagement, and so forth.

Crop-Livestock Systems: The Grazinglands Research Laboratory (GRL) has been the host of the Southern Plains LTAR since 2012 – one of 18 USDA/ARS LTARs conducting a common experiment to better understand how to sustain or enhance productivity, profitability, and ecosystem services in agro-ecosystems and agricultural landscapes. The common LTAR experiment considers the future for two scenarios; ‘business as usual’, and ‘aspirational’.

There are many critical things to tackle. We need to model plant quality as well as quantity, and relate that to livestock gain and performance. “We also need to model physiological responses of plants to defoliation and trampling – cows are not lawn mowers, though models sometimes reduce them to this”, Steiner emphasized. “Also, we need to better characterize soil physical responses and nutrient cycling associated with grazing”.

AgMIP is already helping through its support to collaborative networks of crop modelers working on grazing systems, and its interest to improve interoperability of multiple models and data sets. A partnership to establish a common, interoperable modeling strategy for the LTAR network is needed, as are ways for effectively and usefully handling large sets of observational data, keeping in mind the diversity of data types acquired.

Pests, Weeds & Disease: There is a general agreement that pests, weeds and disease are inadequately represented in the models. However, just as there is need to move away from mono-crop models and increasingly consider the agricultural system, there is need also to keep in mind that pests, weeds, and disease operate as systems also. “Pests can foster or reduce other pest interactions, weeds interact with the soils and change nutrient profiles, and so forth”, said Eigenbrode. “There are many levels of symbiosis and interaction, and it is not yet clear how to best represent the hindering systems without trying to capture each and every dynamic interaction in the models”. With a better understanding of what models can and cannot do, entomologists and biologists may be better able to help identify the key characteristics that models need to emulate. Similarly, with better understanding of the interactions among and between the crop loss agents, modelers may better describe how the system simulation might be better organized and advanced.

maize_unl Southern Rust disease on a Maize leaf
Photo from University of Nebraska-Lincoln

There is need for much greater interaction between these communities, and for continued support to collaborative networks of modelers, especially in support of work to model alternative systems and rotations. In addition to sustaining partnerships, we need to encourage strengthened involvement of extension faculty and others to establish long-term data records. Support for proofs of concept in one or a few of the USDA Long Term Agro-Ecosystem Research Stations (LTARs) would be very helpful.

Data, Models and Uses: AgMIP and others have already identified as a priority the facilitation of data as a resource for research, extension and education. “Creation of data as a resource is aided by standards for data collection, and by standards for data vocabularies”, said Porter. “The latter is especially important when it comes to adding new domains such as pest, weeds and disease; agricultural socio-economics; livestock and grazing; or, nutrition”. Critical partners include USDA (LTARs, NAL, others), CGIAR (Big Data Platform, Crop Ontology, Agronomy Ontology, Ag Trials, CCAFS modeling tools), GODAN, and Industry.

From project funding, AgMIP has created collaborative networks for data, model, and software sharing, and this has enabled harmonization of agricultural research data sets. There are tools for generating model-ready inputs from experimental data, and for incorporating data from other domains. A sustained support mechanism is needed to systematically update and improve these tools in an interactive manner, responsive to needs of the community.

Discussion, Q&A

Q: In some parts of the world, there is concern that supply chains are collapsing – for example, there are some major agribusinesses in North America that are talking about producing fresh food in the store. Can AgMIP incorporate major agri-business and technology trends into its work?

A: The assessment of sustainability of the supply chain involves many things. In addition to what you mention, there are tradeoffs between leaching, production, soil carbon, and other factors. There is need to incorporate economic analyses and even life cycle analysis. AgMIP has already demonstrated capacity to anticipate shifts in management and technology into its regional integrated assessment methodology.

Q: What about impacts to fruits and vegetables and issues of water scarcity?

A: To understand trade-offs for difference management practices we need to simulate the changing systems and improve multiple model methods, including their uncertainty. It means being able to synthesize and incorporate greenhouse and glasshouse management practices also. Production decisions will be driven in part by resource scarcity – for example, the depletion of aquifers in California may cause relocation of fruits and vegetable production in the future. But decisions are also driven by market preferences – for example, fruit and vegetable breeders presently select for properties that facilitate mechanized production while retaining visual appeal on the shelf, rather than nutrition.

AgMIP has already introduced multiple model methods to analyze systems and strategies at present as well as preferences for the future, methods that have been shown to provide valuable insights to guide decision making with uncertainty. But these methods have not yet been applied to fruits and vegetables. In addition, there is need to better incorporate crop production factors that are known but as yet not modeled (e.g., pest, weeds & disease).

Q: How are priorities in model development established?

A: There are several factors. Models are driven by data, and available data are needed to develop and test them. There are many data sets, but to be useful, data need to be well organized and described. There is critical need for partnerships between researchers and major producers, as well as partnerships between industry, farmer, and researchers. We need to think of new ways for data to be collected, and to establish how best to use remotely sensed information. Drones have also been very useful in helping us to understand what data are missing and most needed. We need to simulate, see gaps, and establish the next experiments.

Session Wrap Up

Models, data, and AgMIP North America: A big emphasis of the international AgMIP work has been around the data. It opens up so many areas for intercomparison and learning. There are many opportunities for advancement in the US and North America – we could learn so much through engagement with the LTARs, the pest, weed, and disease community, and communities concerned with the quality of food. We need to identify the models and research needed to answer the key questions, we need to encourage the practices that allow harmonization between large data sets, whether rangeland, soil, or other. Models and data go together.

USDA-ARS has a lot of people doing experiments and creating data. We want to have the principles of AgMIP applied to the priorities for North America. We need to return to the underlying motivation for this Open Session – to establish a body of AgMIP work in North America.

The US Climate Science community got together to advance the early model intercomparisons and they still lead the world in this endeavor. Motivated by their principles and progress, a group of US scientists created AgMIP. But, to date, though AgMIP has many US participants, it has worked more internationally than domestically. The European partners to AgMIP have keenly engaged their nations to advance the systems approaches. Through much effort, they have amassed data and these nations now have new science programs to advance critical agricultural system research objectives with major resource streams.

The US effort remains relatively disaggregated with many small initiatives. Our agricultural research community needs to come together and agree on priorities, establish the critical areas for advancement, define the interoperable systems and harmonization strategies to facilitate data and model sharing at a level that will deliver the answers decision makers seek, and also capture the interest of our government to address the critical questions about the future of agriculture on this continent.

Next Steps: There is need for realistic simulation of agricultural systems, and need to test the sustainability of aspirational management for US-based production systems. “The LTAR common experiment framework appears very well suited for consideration and advancement of AgMIP methods for North America”, said Hatfield. “The advances will include data and model discovery and development innovations for simulation of complex systems of systems”. The next step is to establish a process for interested partners to come together and create a preliminary plan for AgMIP North America, to be openly shared and developed with the community. “There will be opportunities for input at the upcoming 7th international workshop of AgMIP” Mutter noted, “to be held April 24-26, 2018 at the IICA Headquarters in San Jose, Costa Rica. We will also work with USDA and other partners to establish an AgMIP North America Workshop in 2018”.