Such information can be used to derive inputs for crop models in conjunction with yield mapping analysis to identify areas in the field that are stable over space and time. Crop models can be executed on those areas to provide insights on the reason of variability as well as estimates of potential economic return of variable-rate input prescriptions. Thus, the limitation of availability of spatially-variable data is being overcome through new sensors, communication technologies, and algorithms for producing spatial inputs for use in precision farming as well as statistical and model-based analyses. This is in stark contrast to limitations associated with Use Cases in data-poor regions as noted in Section 3.2. The assessment of spatial soil water availability is crucial for understanding the interaction of water stress and crop yield variability in agricultural fields, especially now with increased climate variability and extended drought periods. Spatial variation in soil water is often the cause of crop yield spatial variability due to its influence on the uniformity of the plant stand at emergence and in-season water stress. Soil water content is highly variable within a field due to spatial variation in rainfall, topography, soil properties, and vegetation. The ability to simulate spatial soil water content over time is important for models used for agricultural and hydrological systems assessment . Process-based crop models have proven to be effective in simulating the water balance of soils when the drainage is assumed to be vertical. However, this assumption is incorrect in many fields. For instance, runoff computed by one-dimensional models is not distributed over space, and thus results in inaccurate predictions of surface soil water balance in neighboring areas within a field. The automation of terrain analysis and the use of Digital Terrain Models have made it possible to quantify the topographic attributes of the landscape and to use topography as one of the major driving variables for many hydrological models . Basso et al. developed a spatial soil water balance model that simulates three-dimensional surface and subsurface water flow. The model requires a digital elevation model for partitioning the landscape into a series of interconnected irregular elements,mobile vertical farm daily weather data, and spatial soil information for the soil water balance simulation.

These aspects are considered a serious limitation in crop models and despite their importance have hitherto received limited attention, thus warranting additional improvements and testing. An example that combines strategic and tactical application of a crop model in a spatial context is described by Basso et al. . A dualcriteria optimization through a tested model could determine the nitrogen rate that minimizes nitrate leaching and increases net revenues for the farmer for three zones within the same field characterized by different yield potentials.Jennifer, an economic analyst in a corporate sustainability group, embraces sustainability as the core of their mission: marketing food while conserving resources. She needs to help the corporation’s contract farmers with decisions regarding when to plant, when to irrigate and when and how much fertilizer to apply to conserve energy, save water, minimize waste and reduce greenhouse gas emissions in efforts to make these products more sustainable from farm to fork. An example of the application of crop models to illustrate how reduced N fertilizer rates result in reductions of greenhouse gas emissions at the field scale are described in Basso et al. who used the IPCC emission factor approach to model N2O emissions. Shcherbak et al. showed that IPCC emission factor approach, despite its simplistic approach, is able to closely reproduced measured data of N2O emissions. Also, two important aspects to consider in simulating the crop nitrogen uptake and soil nitrogen balance is the initialization of the soil carbon pool , and to run models in a continuous mode without annually reinitializing soil conditions like soil water and nitrogen content in order to properly simulate soil carbon and nitrogen dynamics and their impacts on production and environmental outcomes. Models have also been used to evaluate the energy efficiency in agronomic management as reported in Bertocco et al. . The next generation of crop models with capability of using real time weather and historical climate conditions will be able to identify strategies to optimize the amount of fertilizer used at a particular location and time, soil and weather conditions with the goal of increasing yield and reducing greenhouse gas emissions. Crop models can evaluate the effects of unknown weather conditions and help decide the optimal nitrogen to apply to crops using different amounts within the field using precision agriculture prescription maps. Communication companies have partnered with different high-tech companies to deliver solutions for the meteorological, geo-spatial and operational challenges facing the agriculture industry. Remote monitoring solutions, as an integral part of the Next-Gen model platform, along with advanced cloud services, will help farmers with decisions regarding when to plant, when to irrigate and when and how much N fertilizer to apply.

Some of the large corporate supply chain companies have recently set a goal to improve fertilizer-application efficiency of U.S. row crop farmers in its food supply chain by 30% by 2020. System models can further help these companies by setting emission reduction protocols, benchmarks and baselines to compare emissions among different management strategies, and by incorporating sustainable agricultural criteria into their future plans validating mechanism, including certification to verify that the farmers are meeting sustainability criteria. There is a need to provide information on the total greenhouse gas, water and other footprints of food systems that are being considered to improve sustainability of future supply from field to fork . There are existing life cycle analysis models that are being used for this purpose, but there are various challenges in providing robust LCA results for complex food systems . Integration of LCA with biophysical and economic models would provide enable more comprehensive food system sustainability analyses. We are not aware of integrated models or knowledge systems that combine the power of agricultural system models with LCA analyses that would provide strategic foresight indicators of sustainability for use by the food corporation that Jennifer represents, although there have been studies that demonstrate this approach .Various farm-level data and decision tools are in use, and are evolving rapidly along with innovations in computer power, software, mobile information technologies and technologies for site-specific management . A key feature of these tools is that they use both public and private data to generate detailed information and outcome-based data that are useful for farm-level management decisions. This information and data can be used to monitor the economic and environmental performance of a farm operation over time and space. The value of these data for improved farm management performance should motivate producers to collect accurate information. In addition, producers increasingly need detailed management data for purposes of quality certification, e.g., for organic or sustainable certification, or to meet regulatory standards. Various issues need to be addressed to advance the use of these tools for management, certification and related purposes. One issue is how to make data acquisition and analytical tools appropriate for and easy to use by farm-level decision makers . Another is how to facilitate the use of data and management tools through effective outreach programs that communicate the value of the tools and the importance of the data for private and public uses. The confidentiality, security and appropriate use of private data when it is shared is another critical issue. Privacy concerns have been the subject of recent discussions among farmers and commodity organizations as they explore the use of new technologies and big-data analytics.This review of agricultural systems modeling shows that major contributions have been made by various disciplines, addressing different production systems from field to farms, landscapes, and beyond. There are good examples of component models from different disciplines being combined to produce more comprehensive system models that consider biophysical, socioeconomic, and environmental situations to produce a wide range of system responses.

For example, crop, livestock, and economic models have been combined to study farming systems as well as to analyze national and global impacts of climate change, policies, or alternative technologies for different purposes. There is a wide range of approaches used in agricultural systems modeling and in the application of those models to scientific and policy questions. Approaches vary according to objectives of developers, their intended uses and data availability. Developers of agricultural systems models have made good use of theories and concepts from a wide range of disciplines, including agricultural and environmental sciences, ecology, engineering, physics, economics, and statistics. The development of agricultural system models continues to evolve through efforts of many organizations worldwide.The Use Cases examined in this special issue demonstrate that a minimum set of component models are needed to develop useful agricultural system models. These component models include, in particular, crop models, livestock models,vertical farming racks and farming system models. Crop models combine weather, soil, genetic, and management components to simulate yield, resource use, and outputs of nutrients and chemicals to surrounding water, air, and ecological systems, taking into account weed, pest and disease pressures. They integrate information to predict performance for a range of inputs and practices that apply from subsistence situations to those systems using highly controlled, intensive production technologies and modern varieties. Similarly, livestock models take into account climate, herd management, feed sources, and breeds. Farming system models integrate various livestock and cropping systems, including their interactions, with economic models that represent the behavior of farm decision makers. These models are needed at the level of the individual operation as well as at the population level so that the bio-physical and socio-economic heterogeneity of systems and their economic, environmental and social impacts can be evaluated by individual farmers as well as policy makers from farm to global scales. Based on this review, we conclude that different platforms for combining models and data for specific purposes are necessary, and that the design of next generation models and data should take into account this need for a range of platforms. The Use Cases studied included relevant examples across the spectrum of users from small-holder agricultural systems in developing countries to intensive production systems, and from systems supported by agricultural industries to those with little support from the private sector. They include examples that need models and associated data to evaluate technologies at a field or farm scale and others requiring the integration of component models to address socioeconomic, food security, and environmental issues at different scales. Although the adequacy of available models varied among Use Cases, one limitation was common across all of them, namely the scarcity of data.

Data are the foundation for all agricultural systems analyses. This limitation restricts the capabilities of existing models to include factors of importance in most Use Cases, limiting researchers’ abilities to evaluate models across wide ranges of conditions and limiting information that can be used as inputs to apply models . Data limitations are more important than gaps in conceptual theories and approaches. We argue that limitations of current agricultural system models and tools are more strongly rooted in inadequate data than in knowledge gaps. This limitation restricts users’ confidence in model abilities to provide reliable results and thus their use for decisions or policies. This lack of data is especially severe in less developed regions. This is true for production models of crops and animals as well as economic models across the Use Cases that addressed issues in data-poor areas in Sub Saharan Africa. But it is also clear that many data-rich regions also suffer from lack of accessible and usable data. For example, thousands of agricultural researchers conduct studies each year, comparing technologies and management under specific conditions. Even though those data could be highly useful for developing, evaluating, and using models, they are generally not available except to those involved in specific studies. There is perhaps one exception to the statement about a common limitation across Use Cases. The capabilities and limitations for management support for precision agriculture are different from the others. Although data that characterize spatial variability at a high resolution in individual fields has been limited, this situation is rapidly changing as new sensors and observation platforms are being developed by the private sector in response to clear business opportunities. This interest by the private sector is likely to rapidly increase the use of agricultural system models to help farmers like Greg use precision farming to increase resource use efficiencies, reduce environmental risks, and increase profits.