Development of next-generation crop models can be divided into several categories: significant improvements in simulation of important crop processes and responses to stress; extension from simplified crop models to cropping systems models that address complexity in space, time and the number of processes considered; and scaling up from field-based models to landscape, national, continental, and global scales.Several crop processes require major advances in understanding and simulation capability in order to narrow uncertainties around how crops will respond to changing atmospheric conditions, both changes in mean variables and changes in extremes. Experimentalists and modelers need to work together from the outset to ensure that the right research questions are posed as experiments are planned, critical field data are gathered at appropriate times, and process-based understanding is captured so as to transfer new insights from the field to the crop models directly and expeditiously.Extreme weather and climate events are responsible for significant economic and social costs in agricultural regions around the world and are expected to increase in duration and intensity under most climate change scenarios.This includes more precise understanding of what thresholds qualify a weather or climate event as “extreme” for different crops in different regions and what simulation processes need to be improved to describe crop responses and their variability to such extreme events. Sequential periods of yield-reducing weather conditions can be especially damaging,growing blueberries such as two or more consecutive dry years as often experienced in sub-Saharan Africa . Other extreme events could be extended periods of record-high temperatures or flooding during a growing season.

Developing predictive capacity that scales from genotype to phenotype is challenging due to biological complexities associated with genetic controls, environmental and management effects, and interactions among plant growth and development processes. Crop model improvements are needed to link complex traits at gene network, organ, and whole plant levels . Phenotypes are linked to changes in genomic regions via associations with model coefficients . Hwang et al. discuss how genetic information can be incorporated into next generation crop models.Crops are already experiencing higher levels of carbon dioxide and temperature in agricultural regions around the world. Understanding of how accelerated rates of CO2 and temperature rise will interact to affect crop growth and productivity is growing , but this improved understanding needs to be incorporated into crop models . Water relations of soils and crops are also important, and optimizing carbon and nutrient cycling, as well as multiple nutrient interactions – beyond the current focus exclusively on nitrogen x water interactions – plays a crucial role in sustainable intensification. The simulation of all of these processes and their interactions and management, especially under conditions of stress and considering soil biology mediated processes, needs to be radically improved.Crop modelers, breeders, physiologists, and human health and nutrition researchers need to broaden the scope of modeling to include nutritional quality of food as well as behavioral aspects of food utilization so as to enable more fully developed projections of future risk of hunger due to climate change and development pathways. For crop modelers, this requires moving from a yield-only perspective to one that includes processes that affect nutritional quality, such as carbon dioxide concentration, drought, and insect pressure . Simulations of non-staple crops, for many of which crop models have not been developed, are needed to understand nutritional effects . For example, as people move out of hunger, one of the primary correlates of health is fruit and vegetable consumption and better models of how these crops may be affected by climate change and other processes will become increasingly important. At cropping system level, there is mounting evidence on the strong positive correlation that exists between crop diversity, nutritional functional diversity and balanced diets in developing regions .

The field of crop modeling was built on a single crop-by-crop approach. A new paradigm is now emerging that moves beyond ‘crops’ to ‘farming systems.’ These new farming system simulation tools incorporate the complexity that comes with many interacting biophysical and socioeconomic components . Such farm system models, however, have tended to place emphasis on farm-scale interactions between system components while reducing the detail with which cropping systems are simulated, to reduce uncertainty and numerical dispersion, by means of so-called ‘summary models’ or more parsimonious or ‘minimum-data’ approaches . Such trade offs between model complexity and modeling capabilities can be now better overcome through the increasing computational power and data available for NextGen models. From the crop modeling perspective, progress on the number and types of simulated crops , uncertainty propagation related to model parameters and structure, ex ante testing of adaptations, and scaling are needed .Many models now incorporate the ability to simulate multi-year crop rotations . Some models allow for more than one crop in the same field. For example, Corre-Hellou et al. developed a model for pea–barley intercropping based on STICS that allows an inversion of dominance in height between species during the crop cycle and a trophic link between crop growth rate and the potential for N2 fixation. Next steps in regard to rotations and intercrops are to advance technology so that modelers can rapidly incorporate multiple crops within fields, and multiple crops over time as the usual practice. Then the response of these more complex cropping systems can be tested under different sustainable intensification management strategies utilizing the updated simulation environments. Similarly, inversion studies can be performed to determine optimal cropping systems and management strategies for particular desired outcomes.Most smallholder farming in the world involves integrated crop-livestock systems that cannot be represented by crop modeling alone. Thus, next-generation farming system models include key linkages to livestock. Progress towards these linkages has been made in the NUANCES-FARMSIM that allocates limited resources across the farm and simulates the way organic matter is recycled or redistributed within the farm in both crops and livestock ; these decisions determine the long term production capacity of the system .

Valdivia et al. developed and applied a biophysical and econometric simulation model that includes dynamic interactions between crops and livestock through nutrient cycling. As discussed below, livestock linkages that need to be fully incorporated include growth and productivity models for perennial grasslands and range lands as well as the usual annual crops used as fodder. Modeling of grassland and range land systems requires also considering the grazing/browsing behavior of herbivores and their interaction with grass/range species, which typically leads to spatial heterogeneity in productivity and other ecosystem services. Information from local experiments will be required to develop and test the grassland and range land models in a wide range of environments. These models will then be capable of deployment with livestock models, regional farm data, and inputs related to management and climate. On the management side, the effects of animal labor need to be included as well.New farming system simulation tools are incorporating the complexity that comes with many interacting biophysical and socioeconomic components, especially in smallholder farming systems in developing countries. A key issue is how to represent heterogeneity in these complex systems; some modeling frameworks use “typologies” , while others take a distributional, point-based or gridded approach that may represent the range of conditions more fully ,square plant pots if data needed to characterize those variations are available. The question of appropriate model detail is clearly important . For the cropping portions of the more complex farming systems models, future research should focus on improving crop models for larger-scale applications. To date, large-area crop models have not been developed to capture the relationships important at an aggregated regional scale and long time-horizons ; the AgMIP Coordinated Global and Regional Assessment is undertaking this task . Other areas for research aimed at better understanding of scaling-up of crop models for large-area assessments include: inclusion of spatial variability in soils and in management, particularly for N fertilization, sowing dates, and crop varieties. A particular challenge is to understand the impact of methods to scale-up crop rotations . Cropping system models need to be able to simulate easily a diverse set of farms rather than just one or several representative farms. There are several approaches for scaling up, including use of gridded models and development of simpler quasi-empirical models for landscape scale analysis . Large-scale computation can allow for much more extensive use of gridded models than in the past.

Soils and climate input datasets become important as simulation goes from field to landscape scale. There are several types of dynamic process gridded crop models: those developed from the site-based models such as DSSAT and APSIM; ecosystem-based models; and dynamic land-surface models. An example of a more statistical model is the agroecological zone approach developed by IIASA and the FAO tended to have exposure to, and in-depth knowledge of, a single crop model . Crop models allow useful extrapolation and prediction for prescriptive management, but most current crop models lack the ability to handle spatially connected processes within a field or landscape. Use of the models with real time, remotely sensed data is not currently available to most farmers or farm advisors . AgMIP aims to increase efficiency of model improvement and application by sharing information between different models and encouraging the use of multiple models in impact assessment . Ideally, parameters from one crop model can be uploaded into databases and then downloaded, reformatted for use in another model. However, AgMIP has found that this sharing of parameter values between models is not necessarily straightforward. AgMIP is bringing different modeling groups together to compare and improve their models. The aims are to develop a better understanding of different crop models across the agricultural modeling community; improve both individual crop models and the entire group of models for a particular crop; and improve the efficiency and effectiveness of multi-model applications in agriculture.In addition to the linkages with crop models discussed above, including the need for modularization and inter-operability, there are a number of areas in which advances in livestock modeling could improve the information needed to support a variety of Use Cases, for both farm level and landscape-scale decisions. More comprehensive livestock models are needed, covering a wide diversity of ruminant and other species, adequately pre-parameterized for most common situations and with default values for users to parameterize models to their conditions. Summary or meta-models from comprehensive, dynamic models could be developed as on-farm decision aids. These tools could include summary models for intake, production and greenhouse gas emissions calculations. Some of these summary models could be developed as mobile phone technologies. Other improvements could include development of extensive, standardized feed libraries linked to a GEO-WIKI for improving mapping of feeds globally. These libraries could also be used for deriving functions of feed quality for different agroecological conditions. One way this could be accomplished would be to expand existing household data collection protocols to include suitable data for livestock. As a step towards addressing heterogeneity, more detailed crop and livestock production systems typologies would be useful. These typologies could be derived from existing farm household, agro-ecology, farm, rangeland, population, markets and other spatial data. NextGen production systems mapping needs to include intensification, gender dimensions of family labor and control over assets and income, and operation size indicators. Better spatial data are needed, including spatially explicit standardized feeds and productivity data. Ideally these data would be linked to crowd sourcing and large data rescue initiatives. Improved spatially explicit farm and regional data on production costs for different livestock technologies are also needed. This information is seldom available and is crucial for both regional and global analyses. These data would enable bio-spatial analysis of livestock yield gaps to guide investments and to identify opportunities to use livestock as a vehicle for agricultural development, poverty reduction, and environmental protection.Pastures and range lands are integral to all livestock production systems and are often closely integrated with crop production systems . The biophysical components of these systems and driving data required to model them are largely similar to those of crop production systems, but several features of these components of agricultural systems need to be addressed in next generation models.