Coupling with landscape microclimate models provides not only the vertical inputs used by agricultural models, but also gradients along the landscape. Coupling with hydrological models provides water flow paths such as surface runoff, vertical and lateral groundwater flow, and interactions between shallow soil and groundwater zones and with adjacent surface water bodies . Water quality models provide sediment and solute transport along the landscape controlled by water flows, and other effects such as wind erosion. Integration and upscaling of landscapes into the watershed scale requires three-dimensional coupling of the surface and subsurface water, energy and mass transfers. Condon and Maxwell and Maxwell et al. provide more details on coupled versus integrated models. At this scale, the groundwater aquifer system typically transcends the boundaries of the watershed, necessitating regional scale analysis to evaluate not only the impacts of cropping and animal production systems on water quantity and quality, but also feed backs from the hydrological system into the agricultural system . Further, meso-scale rainfall and evapotranspiration distribution models control the local surface and subsurface flow intensities, pollution and abatement . At this scale, human effects through land use changes, and ecological dynamics and transitions on natural or protected lands are also important components needed to evaluate overall sustainability of agricultural systems .Although some efforts have gone into integrating biophysical models ,mobile vertical rack more is needed to enable comprehensive assessments of agricultural systems across scales and adequately address environmental and economic responses to decisions and policies.
The need to address decisions and policies at scales arises frequently in agricultural system modeling. Resolving the time and space scale differences among model components is often a major issue, particularly when component models are developed independently for different purposes. This problem arises, for example, when one attempts to create a model that combines crop and hydrology models, crop and economic models, or crop and climate models . There have also been efforts, starting in the early 2000s, in which dynamic models have been developed to provide forecasts over aggregated areas . Traditionally, climate model output for a grid cell is down scaled to produce weather data time series for points that are then fed into crop models. However, the land surface also influences climate; processes within the atmosphere and oceans, and on the land, are coupled and dynamically interact over space on timescales from fractions of seconds to thousands of years. Crops are a major component of the land surface of the globe, occupying about a quarter of all land area. Regional climate can be sensitive to large-scale changes in cropped areas that can result from changes in economic or climate conditions . Therefore, another direction for agricultural impact assessments at a large-scale is to dynamically couple crop simulation with models of land and atmospheric processes.Osborne et al., 2009 showed that, in some parts of the world, the impact of changes in cropped area on regional surface temperature can be of the same magnitude as regional human-induced climate change.
This result raises the question of whether or not new fully-coupled climate change impacts studies will revise our previous estimates of food security impacts. It is clear that the full coupling of crop simulations within global climate models is opening up new possibilities for studies of the impact of climate change on agricultural production – studies that capture some of the complex and important feed backs within the Earth system at a large scale. Limitations in the skill of large-area modeling of crop production and yield is dominated by the density of data used in the simulation. More data should equate to better skill. However, the skill of large-scale modeling is determined by the smallest data set, whether this is the grid cell with the shortest run of observed yields, or the data grid with the largest resolution . We have seen recent increases in the resolution of climate input data and global grids of crop management and soil information. In this field of agricultural modeling, any future increase in data resolution should produce more skillful model simulations.We next discuss the state of current agricultural system science relative to its capabilities and limitations in providing information to assist a wide range of decision makers represented by the five Use Cases. Each Use Case contains a set of interactions between systems and users in a particular environment in a systems analysis. The Use Cases are for developing and developed country settings, demonstrating a range of needs for widely different applications at different scales and levels of intensification. Antle et al. indicated that these Use Cases need crop, livestock, and farming system models. The question that we address here is whether current agricultural system models, existing sources of data, and existing decision support systems are adequate for providing information needed for these Use Cases.Can existing crop, livestock, and farming system models, data, and ICT tools provide the information that Sizani needs to advise the small farmer? The short answer is “No”; there are currently no easily accessible and usable applications that would allow her to analyze the particular farmer’s situation. Or apps that can connect with models in the “cloud” to make runs needed for her to advise the farmer. Although there are models that partially meet her needs, and there are well documented examples of using models to develop insights on productivity enhancement strategies in the face of resource constraints and climate risk they have not been integrated or are not packaged for use by this type of non-expert user.
Models can, for example, simulate responses of crops to soil and weather conditions as well as water and nitrogen fertilizer input but do not generally simulate actual yield in production situations where, weeds, pests or diseases are not controlled. Two of the most serious limitations of many crop-soil models are their inabilities to accurately simulate soil infertility and their failure to represent losses associated with the wide range of pest, disease, and weed species that damage crops. In many intensive production systems, soil fertility, weeds, pests, and diseases are controlled so that responses in those situations can be represented by the costs of management inputs and the production responses to climate and water management. Typically, cropping system models simulate yields that are higher than actual yields in farmers’ fields, which are reduced due to poor management. In addition, fields are usually not homogeneous; for example, spacing between plants may vary considerably,whereas the models assume homogeneity. However, if pest and disease data are observed and available, these data can be input to some existing crop models to compute yield loss associated with specific pests and to diagnose the reasons for the gap between potential and actual yield . Keating and McCown have shown, however, that expert application of well adapted models can still lead to useful insights on many of the key constraints to productivity enhancement in small-holder situations. Generally, farming system models now in use have some capabilities needed to analyze this Use Case. However, most farming system models are not developed to be easily implemented by non-expert users nor for farms with characteristics different from those for which they were developed. An exception to this is the TOA-MD farming system model , although that model also needs reliable data from farm surveys to simulate a population of farms in contrast to a particular farm.It is impractical for Sizani to collect information on a particular farm, go back to her office and work with an analyst to evaluate options for the farmer. Instead, data are needed to describe a range of farming systems so that she could select the combination of biophysical, farming system, and household characteristics from available data. This would include information to allow her to tailor inputs to most closely match the conditions of specific farms. This includes climate, soil, management practice, labor and other inputs available for production and marketing of outputs, typical pest and disease pressures, availability and prices for farming inputs, and other farm, economic, and environmental information. Generally, sufficient data on the biophysical, environmental, and socio-economic conditions of each farm or for a range of farm typologies in the regions are not available. Although some data, such as climate and soil data, are available, generally these are not organized nor are they sufficiently site-specific that agricultural systems models can readily access them for analysis of specific farms. Although research has shown that some analyses needed to advise a farmer can be made, the availability of input data for agricultural systems models remains a major limitation.Most existing DSS tools that are available in Apps are focused on relatively narrow issues ,vertical grow rack such as when to apply a fungicide to a particular crop, when to apply the next irrigation, or how much N fertilizer to apply to a particular crop that will be grown on a particular type of soil in a specific setting. There are few DSS tools that make use of more integrated models to help advisors advise farmers in making farming system decisions . We envision a DSS platform that will connect various models, databases, analysis, and information synthesis tools in an easy-to-use interface for Sizani to set up the analyses and outputs to answer questions about the management of that particular farms’ biophysical and socioeconomic situation and the uncertainties in those estimates. Such DSS platforms are possible, but not yet constructed.Models of maize and other crops, livestock, and the farm household are also needed for this Use Case. These models are available for at least partially performing this type of analysis. Starting in the 1980s, several groups began using crop simulation models to evaluate alternative management systems in developing countries .
Models used in those efforts were generally based on CERES and other crop models now in DSSAT and on the ORYZA rice model developed by IRRI. More recently, the Global Futures and Harvest Choice CGIAR research projects led by the International Food Policy Research Institute have used crop and economic models to evaluate the potential benefits of developing new technologies, including new crop varieties . For example, Singh et al. used the DSSAT CROPGRO groundnut model with climate and soil inputs at six locations in India to evaluate different crop traits being targeted by CGIAR plant breeders. They found that the effect of combining various traits was beneficial, with estimated yield gains varying, depending on location and climate change conditions. Rapid advances in biotechnology and molecular plant breeding are helping researchers incorporate molecular markers and genes into models so that ultimately genetic composition of crops can be used to predict performance of future varieties to help target expensive and time consuming plant breeding efforts . The paper by Hwang et al. presents some concepts now being explored for next generation crop models. Similarly, considerable work has been done on farming system models to evaluate options for improving the livelihoods of farmers. These include farm simulation models , optimization models that attempt to select the best combination of enterprises and their management to achieve one or multiple goals of the farmer . Also, the Trade off Analysis model is currently being used as the basis for model-based impact assessments . Furthermore, this approach can incorporate results from crop and livestock models, as well as environmental and social outcome models, and it can be adapted for smallholder or large commercial farming systems. However, there are important limitations in the capabilities of these models, similar to those mentions in Use Case 1 Thus, there may be large yield gaps between actual yields in farmers’ fields and the potential productivity in those fields . When water, nitrogen, and climate are the major limitations in crop productivity, current models are highly useful, assuming that soil, weather, cultivar, and management input data are available for the analyses. In this Use Case, it is likely that other factors, including other soil nutrients, pests, diseases, and weeds, need to be taken into account. The challenge for next generation models includes not only modeling those factors but also collecting data that describe the production situation with all of the important yield-limiting and reducing factors. Another question is whether existing biophysical models can predict performance of the wide range of intensification options that may be used by farmers for this Use Case.