The food system is a complex system with a variety of actors, feedback and interdependence mechanisms, heterogeneity, and spatial and dynamic complexity (IOM and NRC, 2015). The system is composed of large-scale interconnected systems of systems that include ecosystems (both natural and agricultural), climate, food processing and distribution network data and information systems, and socioeconomic systems. Models are used to better understand how various entities interact within and integrate across the system, and to understand relationships among the natural environment, society, economy, and the various biophysical components of the food system. For example, an understanding of those relationships enables agroecology and ecohydrology approaches to be used to simultaneously manage multiple goals (e.g., lowering water use, sequestering carbon, and minimizing environmental impacts). However, modeling complex interactions and feedback loops within and between each of these systems can be improved through (1) a better understanding of the drivers (e.g., physical, behavioral, market drivers) and their interactions, and (2) a greater ability to collect and analyze large amounts of data with high geospatial and temporal variability.
The study of such a complex system will serve the ultimate goal of informing decision making in food and agriculture practices and policies. Such systems models are also needed to evaluate the impacts of a technological breakthrough on the sustainability of the food system, to rapidly respond to perturbations in the systems to avoid adverse outcomes, and
to propose interventions that lead to overall efficiency gains. The socioeconomic elements of the system need to be deeply integrated with the biophysical drivers of the agroecosystem to invoke any meaningful changes. This chapter presents the many challenges in conducting systems-level models and analysis of the food system, and the opportunities to advance this scientific field.
2.1 Challenge 1: Create and Link Efficient Integrated Systems Models to Data Collection Efforts
In describing subsystems of the food system, models have been successfully built that are biophysics based (based on understanding of the behavior of a system’s components) and empirically based (based on observation and measurements). For example, soil and water balance models were developed to understand climate constraints on agriculture, soil models were developed to predict impacts of soil erosion on crop productivity, and insect and disease models were developed for integrated pest management strategies (Jones et al., 2017). In soils, great strides have been made in understanding N2 fixation and uptake and in determining types of perturbations that may impact the N2 cycle (Salvagiotti et al., 2008). Models for predicting weather at hyperregional scales are available, such as the National Center for Atmospheric Research’s Weather Research and Forecasting model and The Weather Company’s Deep Thunder (The Weather Company, 2016; NCAR, 2018). However, these various modeling efforts are generally discipline-specific and the subsystem models are not yet well integrated across all of the different disciplines within the food and agricultural system. There are many challenges to creating a reliable integrated systems model for the food system.
Systems Are Not Well Defined
The systems are not well defined, and in many cases the linkages between the various systems are not delineated well enough to support mechanistic-based physical models. For example, how do changes in the microbiome of soil microorganisms related to N2 fixation affect the resilience of plants against heat or salt stress, to the nutritional value of the food produced, or to feed conversion efficiency in animals? How would a mechanistic understanding of the interactions of the soil microbiome with the plant, root, and leaf microbiomes be combined with genetic tools to create more resilient plants?
In addition to linkages, a significant challenge in defining any particular
subsystem is determining the main drivers of change for consideration in the analysis. For example, to help identify means to address the differences in productivity and other outcomes between and within regions, it would be helpful to measure and analyze the causes (e.g., drivers) for such differences. In another example, although trade-off analysis is a useful tool for integrating biophysical models with socioeconomic models to predict the impacts of different drivers on the oversustainability of the system (Stoorvogel et al., 2004), these models require a priori assumptions about the principal drivers affecting the model outcomes, and consider only a subset of all possible drivers. Life-cycle analysis (LCA) takes a more holistic approach to modeling sustainability indicators, and can be useful for determining, for instance, the range of the magnitude of life-cycle impacts for a given agriculture activity (e.g., milk production) or set of dietary food choices. This can be used to identify system-level opportunities for intervention to improve sustainability (Boehm et al., 2018; Poore and Nemecek, 2018). However, the outcome of the LCA will depend on whether it is being optimized for water use, yield, or nutritional content (Notarnicola et al., 2017; Pourzahedi et al., 2018). The outcome may also depend on the system boundaries that are selected.
Lack of Cohesion Between the Different Modeling Communities
As previously noted, the absence of standardization of data formats and reporting, data privacy concerns, proprietary data, and lack of data repositories are impediments to data sharing. Data standardization and open access are important since the output of one system model will likely be the input to a different system model.
Lack of Methods for Integrating Models
Better methods for integrating subsystem models across different spatial and temporal scales and for integrating physics-based mechanistic models with empirical/statistical models or economic models are needed (Kling et al., 2017). Mechanistic (biophysical) models of the key components of a single system may be possible, but adding only a few additional nodes to a system analysis creates significantly more connections that may each take many years of research to achieve (see Box 8-1). This precludes the ability to create biophysics-based mechanistic models for the food system. The problem of scale is exacerbated by the fact that many different forms of perturbations can affect the systems simultaneously (e.g., society’s values and climate). Each of the system nodes has complicated feedback loops that will need to be considered together in a systems-level analysis. Tackling this challenge will likely require the development of empirical (statistical)
models, rather than purely mechanistic ones. The spatial resolution of many components of the food system may be quite coarse (e.g., soil moisture determined from satellite sensing has kilometer to tens of kilometers scale resolution), which will require averaging, simplifying assumptions, and conducting statistics to downscale the models for more-local predictions, including analysis of differences within and between regions. However, better ground sensing as described in Chapter 3 may improve this situation. Methods to blend these two types (and scales) of models (mechanistic and empirical) and data are needed.
Sensor development and movement toward data-driven empirical models for the food system will require massive amounts of high-quality data to calibrate the models. The current approaches for integrating data into models to update model parameters are slow. Rapid (perhaps real-time) updating of the models with data is necessary, whether the subject is water and nutrient use in crop agriculture, feed efficiency in animal agriculture, or food safety and quality in food processing and distribution.
2.2 Challenge 2: Uncertainty Within and Across Multiple Elements of the System
The high degree of uncertainty in the food system is a significant challenge for developing accurate models. Uncertainty is especially challenging for long-range strategic planning efforts to cope with a growing population. Uncertainty comes in many forms; for example, it is difficult to predict extremes in environmental conditions (water, heat), yields, pricing, and market conditions. While it is also difficult to predict the behavior of growers and consumers related to technology adoption and new inputs, it is important for the models to take behavior into account. Incorporating information on adoption behaviors and market response is required to best quantify response to the effects of technology changes, and useful examples include recent applications to water (Khanna and Zilberman, 2017; Taylor and Zilberman, 2017; Zilberman et al., 2017). There is a lack of systems-level efficiency data (e.g., water-use or nitrogen-use efficiency at the watershed scale) available to calibrate the models. High degrees of uncertainty across multiple system elements can make accurate predictions of overall system behavior inherently indeterminate.
2.3 Challenge 3: Access to Agricultural Systems Models and Better Integration of Management and Society into the GEMS Model
Agricultural systems modeling is a well-established field. Significant progress has been made in understanding the biophysical processes driving agricultural productivity (Jones et al., 2017), and integrating biophysical and economic models for analysis (van Wijk et al., 2014). To date, modeling efforts have focused on understanding the linkages between system components and drivers of the system responses. Challenges remain in making these models more accessible to users, in creating models that can address specific user needs (e.g., including interactions between multiple crops and between crops and livestock), and in developing user-friendly decision tools (Antle et al., 2017). Another challenge in agricultural systems modeling is in data management. There are vast amounts of farm-level data available, and effectively storing, validating, and sharing those data are crucial to developing the next generation of agricultural systems-level models.
The genotype-environment-management-society (GEMS) approach to modeling agricultural systems offers great opportunities to improve system efficiency by acknowledging complex interactions. Expanding the existing systems models to include additional societal drivers would enable better predictions of behavior in the food system, and is necessary to invoke meaningful changes to the system. For example, food choices and behaviors affecting food waste could have significant impacts on the overall sustainability of agriculture. The price of commodities is driven by policies (e.g., biofuel mandates), technology choice, and consumer demands, which in turn affects the mix of crops selected by growers. These do not always align with climate or resource availability if prices and other factors do not fully reflect the opportunity cost of alternative uses.
This section includes three opportunities to advance the systems approach for food and agriculture. Breakthroughs in agricultural research will come from the collaborative efforts of scientists and experts in fundamental and applied science areas, policy, and human behavior. An example of systems thinking and integrated systems modeling for watershed management is noted in Box 8-2.
3.1 Opportunity 1: Complex System Modeling to Determine Methods for Integrating Data, Models, Management Strategies, and Socioeconomic Behaviors
Development of data-driven integrated food system models will inform future research efforts toward the most promising agricultural breakthrough opportunities to be pursued, or the best management practices for efficient resource use. The process of continuously updating and validating models with data will likely lead to more mechanistically based data-driven models and advancements in minimizing and managing uncertainty within and between systems to improve forecast reliability. Matching model complexity to scale and to user needs can result in better decision support tools for producers.
As previously mentioned, the study of the food and agricultural system as a complex system serves the ultimate goal of informing decision making and—specifically relevant to this report—of evaluating the impacts of technological breakthroughs on the sustainability of the food system. The complexity of the food system, the integration of its various systems, including the natural and built environments, and the significant need to increase its resilience and efficiency make the food system and its subsystems an ideal model for developing the capacity to model complex systems in general. Initiatives to study the food system provide the opportunity to determine appropriate boundaries for analysis within and between systems, to identify the key data gaps and missing linkages within systems and at the boundaries of different systems, and to identify and build the required systems analysis and decision support tools for complex agricultural systems. There is an opportunity to identify appropriate “behavior” models to include in agricultural systems models, and to standardize different system models through model intercomparisons using “base-case” scenarios for model comparison. As weather and climate models increase resolution and better represent small-scale processes such as vegetative feedbacks, it will open research into soil–plant–atmosphere–water cycling from field to continental scales. Modeling the system response to climate extreme events can improve the utility of the models to growers. Scaling up models from field scale to landscape scale can also help to manage agroecosystems coupled with natural ecosystems using ecohydrology methods.
3.2 Opportunity 2: Pioneer New Methods for Improving Data-Sharing and Privacy Issues for Public–Private Partnerships
As previously noted in Chapter 7, data-driven approaches to managing agriculture offer unprecedented opportunity for increasing the sustainability of the food system. The availability of open, harmonized data has
already proven to be essential to creating models of the food system (Jones et al., 2017). Continued development and applications of these approaches to the food system provides an opportunity to discover new methods of data collection, provenance, and sharing. It also provides opportunities to integrate data into models in real time, and to develop machine-learning and artificial intelligence approaches for managing inputs and outputs along the food value chain. There is a great opportunity to advance systems modeling by developing methods to integrate both public and private data for use in democratized decision support tools (Antle et al., 2017).
3.3 Opportunity 3: Apply Systems Thinking and Convergence Methods to Agricultural Sustainability Problems
The food system touches on a wide range of issues, including climate change, water resource use, plant and animal genetics, food safety, nutrition, policy, and economic sustainability. The focus on individual subsystems, without consideration of the interconnectedness of the subsystems, has led to unintended consequences to other components of the system (Turner et al., 2016). In other words, approaching agricultural research more holistically from a system-of-systems perspective offers the potential to optimize across multiple components of the system. The basic tenet of “systems thinking” is recognizing the interconnected nature and linkages that cross multiple dimensions (subsystems). There is also tremendous opportunity to promote convergence of knowledge to solve agriculture’s most vexing problems (e.g., water use, soil degradation, and food waste). For example, combining expertise from plant biology, microbiology, soil science, nanotechnology, sensor design, wireless communications, and behavioral sciences may lead to innovative and socially acceptable methods of applying nanotechnology-based sensors into plants and soils to monitor plant productivity without negatively affecting the plants, environment, or the ability to sell the food produced. Combining expertise may help to convert “waste” products to value-added products or to important sources of inputs and energy to other parts of the food system. Programs aimed at tackling all of the elements of these problems will require better integration of fundamental science and engineering with economics, medicine, behavioral sciences, and social sciences. New knowledge and research is needed on how to design technology packages and institutional structures to enhance technology adoption and new processes. Research in systems-thinking approaches to agricultural problems will ultimately trickle down into new educational programs and into practice as those graduates enter the profession.
Barrier 1: Data provenance, security, sharing, and costs of collection and storage. Data-driven approaches to manage the food system offer great potential to improve resilience and sustainability. However, issues around governance and the development of a viable business model for democratizing and sharing the benefits derived from those data are significant barriers to success (Halewood et al., 2018). Institutional and policy challenges are needed to promote the formation of public–private partnerships, determine ownership of resources, and manage sharing of benefits across countries. The lack of appropriate platforms, standards, and incentives for sharing of data and models is also a substantial barrier to success.
Barrier 2: Systems are inherently indeterminate. There need to be realistic expectations about the applicability and accuracy of the food system models. As Box 8-1 showed, as more subsystems are included in the analysis, the number of subsystem interactions increases considerably. The causal relationships between many elements of the food system are not well characterized. Moreover, the high spatial and temporal variability within these systems and between the systems and the environment, as well as the unpredictability of human behaviors, may simply make the system too complex to model with reasonable accuracy. As an example, the predominant methods for irrigation in water-limited western states are flood and furrow irrigation systems. Although these are highly inefficient systems, farmers fear that switching to more efficient sprinkler and drip irrigation systems will erode their stake in and ownership to their water rights (i.e., the “use it or lose it” syndrome), which is likely the reason many farmers hold back from making changes to more efficient irrigation systems (Osborn et al., 2017). Evaluating the effects of policy-induced changes to incentives for switching to the more efficient irrigation systems requires the development of coupled models that account both for the geospatial component and water usage as well as for human behaviors related to adaptation and induced technological change.
Barrier 3: Limited incentives to consider systems-level impacts. Policies and regulations affecting agriculture (e.g., water policy, water quality criteria, and nutrient management) are often made without consideration of their impact on the system (Alfredo and Russo, 2017). The research enterprise has not been structured with the goal of applying systems approaches to problem solving across the span of the food and agricultural system, and there are limited incentives for research cooperation among disciplines.
Barrier 4: Unintended consequences are not immediately identifiable. Unidentified risks in any system or system of systems are unavoidable. The more complex and interrelated the system becomes, the greater the potential for unidentified risks. Although it is impossible to identify all of
the risks of any system, a move toward systems thinking in food and agricultural research will help to reduce any unintended consequences due to interventions—for example, altering photosynthesis routes, manipulation of the soil microbiome, or environmental costs of information and communication technologies associated with widescale deployment of autonomous data-collecting sensors.
Recommendation 1: Identify opportunities to improve the performance and adoption of integrated systems models of the food system and decision support tools. The community of stakeholders should identify current barriers to successfully applying integrated systems models to the food system. The community should also develop a roadmap to overcome key challenges specific for modeling the food system (e.g., integration of subsystems, standardization and interoperability, data storage and sharing, data infrastructure, simplified decision support tools, crop–crop and crop–livestock interactions, and validating the performance of complex systems models using data). The community should clearly delineate where the inability to model the integrated system components is limiting progress toward sustainability and resilience, rather than where other barriers exist such as poor policies or undervaluation of natural resources.
Once the key systems and system of systems to be modeled are determined, the research community needs to improve understanding of the drivers behind the subsystem interactions. This will require improving the ability to collect and analyze large amounts of data with high geospatial and temporal variability (e.g., sensor development). Connecting the sub-models into more integrated systems models will also require consistency in applying system boundaries and assumptions and scale of analysis (Kling et al., 2016), as well as more standardized data reporting and model input/output to make the subsystems interoperable. These integrated system models can then be used to direct research needs and to avoid unwanted consequences of technology interventions or policies.
Recommendation 2: Incorporate elements of systems thinking and sustainability into all aspects of the food system (from education to research to policy). A paradigm shift is needed in the management of the food system to anticipate the effects of environmental or policy-induced changes. This will be a fundamental change in the way the food system is viewed, and the way stakeholders are educated to operate. In research, it will require more transdisciplinary research and team science around approaching and working to solve problems in the food system. It will also require more integrative, systems-level approaches that can assess policy alternatives for systems-level sustainability outcomes. It will also require adequate incen-
tives and funding for systems-level convergent approaches to research and education.
Alfredo, K. A., and T. A. Russo. 2017. Urban, agricultural, and environmental protection practices for sustainable water quality. WIREs: Water 4(5). doi: 10.1002/wat2.1229.
Antle, J. M., B. Basso, R. T. Conant, H. C. J. Godfray, J. W. Jones, M. Herrero, R. E. Howitt, B. A. Keating, R. Munoz-Carpena, C. Rosenzweig, P. Tittonell, and T. R. Wheeler. 2017. Towards a new generation of agricultural system data, models and knowledge products: Design and improvement. Agricultural Systems 155:255-268.
Boehm, R., P. E. Wilde, M. Ver Ploeg, C. Costello, and S. B. Cash. 2018. A comprehensive life cycle assessment of greenhouse gas emissions from U.S. household food choices. Elsevier 79(C):67-76.
Halewood, M., T. Chiurugwi, R. S. Hamilton, B. Kurtz, E. Marden, E. Welch, F. Michiels, J. Mozafari, M. Sabran, N. Patron, P. Kersey, R. Bastow, S. Dorius, S. Dias, S. McCouch, and W. Powell. 2018. Plant genetic resources for food and agriculture: Opportunities and challenges emerging from the science and information technology revolution. New Phytologist 217(4):1407-1419.
IOM and NRC (Institute of Medicine and National Research Council). 2015. A Framework for Assessing Effects of the Food System. Washington, DC: The National Academies Press.
Jones, J. W., J. M. Antle, B. Basso, K. J. Boote, R. T. Conant, I. Foster, H. C. J. Godfray, M. Herrero, R. E. Howitt, S. Janssen, B. Keating, R. Muñoz-Carpena, C. Porter, C. Rosenzweig, and T. R. Wheeler. 2017. Brief history of agricultural systems modeling. Agricultural Systems 155:240-254.
Khanna, M., and D. Zilberman. 2017. Inducing water conservation in agriculture: Institutional and behavioral drivers. Choices 32(4).
Kling, C. L., R. W. Arritt, G. Calhoun, and D. A. Keiser. 2017. Integrated assessment models of the food, energy, and water nexus: A review and an outline of research needs. Annual Review of Resource Economics 9:143-163.
NCAR (National Center for Atmospheric Research). 2018. Weather Research and Forecasting Model. Available at https://www.mmm.ucar.edu/weather-research-and-forecasting-model (accessed June 8, 2018).
Notarnicola, B., S. Sala, A. Anton, S. J. McLaren, E. Saouter, and U. Sonesson. 2017. The role of life cycle assessment in supporting sustainable agri-food systems: A review of the challenges. Journal of Cleaner Production 140:399-409.
Osborn, B., A. S. Orlando, D. L. Hoag, T. K. Gates, and J. C. Valliant. 2017. The Economics of Irrigation in Colorado’s Lower Arkansas River Valley. Available at http://www.cwi.colostate.edu/media/publications/sr/32.pdf (accessed June 26, 2018).
Poore, J., and T. Nemecek. 2018. Reducing food’s environmental impacts through producers and consumers. Science 360(6392):987-998.
Pourzahedi, L., M. Pandorf, D. Ravikumar, J. Zimmerman, T. Seager, T. Theis, P. Westerhoff, L. Gilbertson, and G. V. Lowry. 2018. Life cycle considerations of nano-enabled agrochemicals: Are today’s tools up to the task? Environmental Science: Nano 5:1057-1069.
Rabotyagov, S. S., T. D. Campbell, M. White, J. G. Arnold, J. Atwood, M. L. Norfleet, C. L. Kling, P. W. Gassman, A. Valcu, J. Richardson, R. E. Turner, and N. N. Rabalais. 2014. Cost-effective targeting of conservation investments to reduce the northern Gulf of Mexico hypoxic zone. Proceedings of the National Academy of Sciences of the United States of America 111(52):18530-18535.
Salvagiotti, F., K. G. Cassman, J. E. Specht, D. T. Walters, A. Weiss, and A. Dobermann. 2008. Nitrogen uptake, fixation and response to fertilizer N in soybeans: A review. Field Crops Research 108(1):1-13.
Stoorvogel, J. J., J. M. Antle, C. Crissman, and W. Bowen. 2004. The tradeoff analysis model: Integrated biophysical and economic modeling of agricultural production systems. Agricultural Systems 80(1):43-66.
Taylor, R., and D. Zilberman. 2017. Diffusion of drip irrigation: The case of California. Applied Economic Perspectives and Policy 39(1):16-40.
Turner, B. L., H. M. Menendez, R. Gates, L. O. Tedeschi, and A. S. Atzori. 2016. System dynamics modeling for agricultural and natural resource management issues: Review of some past cases and forecasting future roles. Resources 5(4):40.
van Wijk, M., M. Rufino, D. Enahoro, D. Parsons, S. Silvestri, R. Valdivia, and M. Herrero. 2014. Farm household models to analyse food security in a changing climate: A review. Global Food Security 3:77-84.
The Weather Company. 2016. The Weather Company Launches “Deep Thunder”—the World’s Most Advanced Hyper-Local Weather Forecasting Model for Businesses. Press Release. Available at http://www.theweathercompany.com/DeepThunder (accessed June 26, 2018).
Zilberman, D., R. Taylor, M. E. Shim, and B. Gordon. 2017. How politics and economics affect irrigation and conservation. Choices 32(4).
This page intentionally left blank.