While smallholders are not bound to PAD’s recommendations, PAD promises higher yields and profits, holds a monopoly on knowledge, and is using its knowledge to ensure obedience. This external control, a dynamic of urbanized agrarian knowledge, is problematic in its own right: while often very specific about climate and soil conditions, PAD appears to take little heed of local social, cultural, and political context, and determines its users’ best interests for them. The of-site decision-making, though, becomes more questionable when we learn that PAD has recently partnered with Bayer. Bayer funds PAD’s work in Bangladesh and provides PAD with contact information for its former customers . PAD then advises these and other farmers on how much of which inputs to use, such as Bayer’s fertilizer. PAD reported in 2019 that “farmers . . . were 18% more likely to report using a Bayer product, while trust and satisfaction did not change. Farmers also recommended Bayer products to 8 other farmers, on average” . Effectively, PAD is a marketing arm for agribusiness as it enlists smallholders into global commodity production. In pursuit of “long-term financial sustainability,” and given the success of the Bayer pilot, PAD has begun to consider if “it is worth exploring whether incorporating brand promotions can help PAD and other partners develop commercial advisory services that can be sold to for-profit agribusinesses and offered free of charge to farmers.” Here PAD shifts from being a service for farmers to being a service for for-profit agribusiness. PAD continues to offer free advice to farmers, but its client has changed. The agribusiness company has replaced the farmer as PAD’s primary relationship. E-Sagu represents an earlier example, beginning in 2004, of an e-extension company that connected farmers with urban agricultural experts and that ultimately turned to partnerships with input vendors to stay afloat . Given the wealth extraction already wrought on agrarian and post-colonial zones and the exacerbation of that extraction via digital agriculture,blueberry grow pot such an e-extension service is unlikely to soon exist without similar privatization, unless supported by the state.

We can expect e-extension to continue to be a means, like digital agriculture more broadly, of corporate and urban influence. Second, digital agriculture not only perpetuates certain agricultural practices— namely export-oriented and input-dependent—it also, through partnerships such as with PAD and Bayer, and more straightforwardly through privately-owned PA companies, privileges larger farms and furthers corporate control of independently owned agricultural land. In the US, the number of farms between ten and 1,999 acres has fallen since 2007, while the percentage of land in farms larger than 2,000 acres has increased from 40% to 47% in 2017 .This accompanies a general trend of increasing farm sizes , especially in high-income countries , and the much discussed global land grab . While the reasons for this pattern are various, digital agriculture continues the privileging of larger farms. Digital agriculture favors wealthier farms that grow according to methods conducive to data collection and which produce profits sufficient to afford the technology. This privileging begins with the installation of these technologies. Implementation requires capital investment in sensors to acquire data, connectivity infrastructure to connect the data, and advanced machinery to use the data. The cost of this equipment limits much of its application to wealthier or more financialized farms .These significant investments give uneven returns, which further privilege larger-scale commodity-oriented farms.Digital agriculture adds to the litany. By dictating decision-making, firms achieve control of farms’ inputs and outputs without the risk of fixed assets or of production, and without any obligations to labor welfare. This risk minimization parallels the strategy of contract farming, in which firms set prices and conditions with farms at the beginning of the season. Firms, here too, dictate inputs, even providing loans for them. Under this arrangement farmers carry the risk of production—firms instruct what to grow but bear no liability for a bad harvest—all while production is organized into a form that caters to the interest of investment.

Digital agriculture, though, does not just parallel contract farming; it has also become a tool for contract farming. Smallholder farmer management platforms streamline the contracting process by facilitating communication from firm to farm and allowing firms to have more oversight of farms; by making contract farming easier and cheaper, these platforms then spread the model. Farm force is a particularly notable example of such a platform, and through its Syngenta-provenance indicates contract farming’s appeal to agribusiness corporations . Digital agriculture further supports contractors by increasing their ability to forecast prices and thereby minimizing their price risk. Though this risk is minimal for contractors, primarily resting on growers , firms still bear some degree of the price risk. While some platforms have also emerged to better inform farmers of market prices, firms remain better positioned, with greater computational capital, to forecast global production and demand, allowing them to set prices more in their interest. These digital agriculture models not only minimize economic risk, they also minimize political risk. By allowing family ownership of farms, contract farming and e-extension give the appearances of independence and a distributed means of production and are therefore less provocative of land reform; agribusiness does not need to fear land seizure. In places where land reform has already occurred, such as Zimbabwe, these mechanisms represent a way forward for corporate control.Rather than a land grab, digital agriculture in the Global South facilitates a data and production grab. The appearance of smallholder ownership makes these new grabs more palatable and may demobilize rural classes. Finally, digitalization disrupts agricultural labor. As an intensification of industrialized and automated agriculture more broadly, digital agriculture is anticipated to eliminate the need for farm labor , but its effects on labor are broader. Digital agriculture is likely to deskill workers, further bind their fortunes to the global commodity market and potentially turn them into urban migrants. Digital agriculture’s land consequences described above shape urbanization at its sites of both explosion and implosion.

As it reconfigures land ownership in the operational landscapes of extended urbanization by privileging large estates and by making smallholding more amenable to capital’s interests, it simultaneously denies the autonomy of the farmers on these smaller plots. Both of these are likely to incorporate more growers into the global commodity market to sustain non-agrarian production. Especially in the Global South, where subsistence farming is more common, digital agriculture’s orientation toward larger farms may eventually displace smallholders and convert them into wage workers, as they leave their own plots and work for the commercial outfits. Meanwhile, farmers that retain ownership are also further incorporated into the global commodity market because of PA and e-extensions recommendations. As such, their food security is undermined . They lose the means of subsistence, even as they maintain the means of production—they own land but increasingly do not own their time or behavior—and become more vulnerable to the “vagaries of world market prices” .This threatens smallholder farmers’ very ability to survive and pushes them toward wage work and cities for imagined greater stability . Contract farming and extension, even more so under their digital exacerbation, could lead to dispossession and displacement, and ultimately de-ruralization, sending peasants to cities to become informal urban surplus labor. Digital agriculture also contributes to deskilling. As described above, digital agriculture changes how agrarian knowledge is produced and disseminated. As with urbanization at large, this change is important not only for how it concentrates, but also for how this concentration folds back onto the countryside. The disruption of agricultural learning deskills rural workers, ultimately undermining the farmer welfare digital agriculture allegedly pursues. Originally observed in manufacturing contexts, deskilling is the degradation of labor through the separation of mental from manual work; laborers are “more expensive and less controllable” than machines, and thus require replacement . Stone takes this theory and partially applies it to agricultural production in the GMO era. He finds that deskilling appears differently in an agrarian context as, among other differences, farming is “much more dynamic” and the farmer needs to make many more decisions than does a factory worker. He therefore finds that with agriculture, deskilling is primarily useful as a metaphor rather than a theoretical model.

A decade later, though, digital agriculture may make agricultural deskilling much more literal, by moving the decision-making of-site. With GMOs, farmers’ learning process and ability to make decisions are disrupted by a rapid pace of new technologies they do not understand; they still, however, must make decisions. With digital agriculture, which informs farmers about what to do––whether through sensors or extension––this is no longer the case: farmers no longer need to make decisions as these decisions are made for them,hydroponic bucket from a distance. More data is needed to understand the effects of deskilling from digital agriculture’s various technologies, but the bio-engineering feats of the late twentieth century give an indication of what is to come. As deskilling is not only the disruption of particular knowledges but the “disruption of the process of experimentation and development of management skill” , deskilling and the potential obsolescence of on-site decision-making has urban implications, especially should digital agriculture contribute to continued deruralization. This decapacitation of management skills not only disempowers the farmer as farmer, but also potentially renders them less qualified for the urban labor market, and potentially contributes to a less equipped urban reserve of labor. Within agrarian zones, deskilling could also have destructive ecological effects: Vandeman observed that deskilling alienated farmers’ knowledge of their own land. Agricultural systems science as we know it today has evolved over the last 50 or more years with contributions from a wide range of disciplines . Generally during this same time period, appreciation for and acceptance of agricultural systems science has increased as more scientists, engineers, and economists graduate from universities with training in systems modeling, analytical approaches, and information technology tools. Over this time period, there has also been a corresponding increase in demands for agricultural systems science to address questions faced by society that transcend agriculture. Relevant questions range from how to better manage systems for higher and more efficient production, what changes are needed in a farming system for higher profitability without harming the environment, what policies are needed to help farming systems evolve to meet broader societal goals, and what systems are needed to adapt to the continual changes that agriculture faces, including climate change, changes in demand for agricultural products, volatile energy prices, and limitations of land, water, and other natural resources.There is a strong agenda of new Sustainable Development Goals , which will require models of nutritional quality of food beyond bulk yields and multi-functional landscape models for policy analyses.

Sustainable solutions that address multiple goals will likely benefit from a convergence of science and technologies that make use of information and cognitive sciences . In order to analyze these different dimensions of agriculture and food systems, ideally we would have a virtual laboratory containing models, data, analytical tools and IT tools to conduct studies that evaluate outcomes and trade offs among alternative technologies, policies, or scenarios. The virtual laboratory would allow users to define scenarios, specify analyses covering different social, political, and resource situations and different spatial and temporal scales, and produce outputs suitable for interpretation and use by decision makers. Clearly, that virtual laboratory does not exist. But where are we currently relative to this ideal situation? The purpose of this paper is to address that question by reviewing the state of agricultural systems science and its capabilities for the Use Cases described by Antle, Jones and Rosenzweig that represent two important areas of agricultural systems model applications: for smallholder agriculture in developing countries and for commercial agriculture in industrialized countries. This paper builds on earlier reviews of specific components. In the concluding article of this Special Issue, Antle, Jones and Rosenzweig discuss the implications of NextGen for global change research, another major area of agricultural systems model applications.Several crop modeling review papers have recently been published , summarizing model capabilities and uses. For example, Rivington and Koo surveyed crop model developers and users to assess the state of crop models for use in research and decision making related to climate change. They emphasized the need for additional model development as well as the need for more and better quality data. Ewert et al. reviewed crop models relative to their adequacy in performing integrated assessments of climate change impacts, and pointed out important limitations in most crop models.