The world population is on track to increase by approximately one-third—to more than 9 billion people—in the next 40 years. Feeding a population that large, which may require doubling agricultural productivity, will only be feasible with significant, even revolutionary, improvements in agricultural processes and equipment. In the past, agriculture has benefited from, and often driven, improvements in technology. For example, in this country, increased mechanization on modern farms led to the change from an agrarian to an urban demographic (NAE, 2000).
In this article we explain how information technology (IT), a relatively recent field that began a little more than 60 years ago (Shannon, 1948), could have at least as big an impact on agriculture in the next half-century as mechanization had in the previous century. We present a systems-level perspective on the challenges and opportunities afforded by the integration of agriculture and IT.
In this section we outline five challenges associated with agriculture that must be overcome to achieve the desired increases in productivity: inherent heterogeneity; unanticipated disturbances; large geospatial dispersion; security and safety requirements; and constraints.
By its very nature, agriculture works with biological systems that are inherently heterogeneous in composition and processes. Fields can vary in soil type and moisture content down to a resolution of a square meter. Weather patterns can vary spatially and temporally in terms of sunlight and rain. Raw materials themselves have basic genetic variations from plant to plant and animal to animal. Indeed, genetic variation is often biologically useful for increasing resistance to diseases and pests. But compare this heterogeneity with the homogeneity of other industrial processes, such as Henry Ford’s assembly line, and the challenges associated with maximizing product yield using a minimum of resources become apparent.
Agricultural processes are much more vulnerable to unanticipated disturbances than many other industrial processes. The weather can cause floods or bring hail storms that can devastate crops. Pest or disease infestations can rapidly affect, if not wipe out, large quantities of raw material. When we contrast this environment to the carefully controlled (temperature, humidity, etc.) clean-room environment of the semiconductor fabrication industry, we immediately understand that, because of external forces, the levels of precision in crop or herd yield are far lower than in other industries.
Large Geospatial Dispersion
Various points in an agricultural supply chain are widely dispersed, and the overall agricultural system can be segmented into interconnected sub-processes (Figure 1). Because agriculture is land-intensive, it requires significant acreage for the in-field parts of the process chain, which are usually not collocated with upstream or downstream processes.
Another challenge related to geospatial dispersion is that the processing of some raw materials such as livestock and perishable crops is time critical. Thus, long distances between processing points in the supply chain introduce risks into the overall viability of the entire process.
Security and Safety Requirements
Security and safety are paramount for agricultural systems on two separate time scales. On a short time scale of days or weeks, the safety of food is critical because many products are eventually ingested by humans. Protecting human health requires tight process management of the agricultural supply chain on a global scale.
On a longer time scale of years or decades, sustainability of the natural environment is critical to long-term societal health. Agricultural enhancements, such as pesticides or fertilizers, must be used in ways that increase productivity without adversely affecting the overall quality of life. Similarly, resources, such as land and water, must be used in ways that can be sustained indefinitely.
All of these challenges must be met within the constraints inherent in the agricultural process. For example, the amount of arable land is relatively fixed globally, particularly in more developed countries. Time is also a key constraint, particularly for time-sensitive systems, such as livestock and perishable produce. There is a finite window of time during which these agricultural products are viable during processing. Finally, the first four challenges must be met within the constraints of economic viability (i.e., cost).
Opportunities Enabled by Information Technology
The challenges described above also provide significant opportunities for using IT. Indeed, if agriculture can be seen as a supply chain, then many of the existing paradigms for managing supply chains are already available.
Besides upgrading physical infrastructure, the efficiency of supply chains can be improved by the acquisition and exploitation of information resident in the chain itself. Access to these data makes monitoring of existing processes, rapid responses to changes in the supply chain, and optimal allocation of resources more feasible. These approaches can be summed up as information gathering, information processing, and decision making, the same principles that apply to management of supply chains in other industries.
The ability to gather agricultural information through advanced sensing systems has been improving steadily over the past two decades, and the unit cost per bit of information has decreased in line with similar cost decreases in other industries. For agriculture, there are three major types of sensing systems: physical, biological, and human.
Physical sensors, the most prevalent of the three types of sensing systems, are used throughout the chain shown in Figure 1. DNA sequencers for crops gather genetic information about crop varieties and provide input for bio-informatics algorithms that can identify promising strains (e.g., herbicide-resistant soy or wheat) (Gale and Devos, 1998). Moisture sensors correlate soil-to-capacitance changes for the in-field monitoring of moisture. In addition, nitrates and other chemical compositions can be sensed by in-field devices that are commercially available (Ehsani et al., 1999).
Remote sensing is used to collect broad geospatial information about in-field systems (Nowatzi et al., 2004). For example, Figure 2 shows different electromagnetic reflections for two health states of sugar beet crops as the result of complex absorption and reflection of solar energy. A thorough understanding of this spectrum for specific crops can be used to calibrate a sensor that can provide health or growth data on particular agricultural systems. A similar technique can be used to monitor livestock by sensing, for example, methane emissions.
Information gathering in agricultural processing facilities is done with many of the same sensing modalities used for monitoring conditions (e.g., temperature, mass flow, chemical composition, etc.) in chemical processing facilities.
Global information systems (GIS) can pinpoint geospatial locations of agricultural units in a supply chain using radio frequency identification and global positioning systems (Attaran, 2007). These systems can track and monitor crops and livestock as they progress “from farm to fork.”
Biological sensing, sometimes called indirect sensing, is physical sensing augmented by biological cues. The state of produce can be inferred based on the monitoring of a biological agent, such as an insect or bacterial population. This can be done in-field or during post-field transport and processing. An illustrative example would be a model for the relationship between crop yields and the presence of pollinators. This model would both monitor pollinator counts and provide a basis for inferring future crop yields (Garibaldi et al., 2011).
Humans in the Loop
Not all data about agricultural systems come from engineered instrumentation and sensors, such as those mentioned above. Humans in the loop are an important data source that should not be overlooked. Indeed, networking advances in this age of IT can significantly improve our understanding of the workings of highly distributed, poorly automated socio-physical systems that may nevertheless significantly impact the planet’s food supply. This is especially pertinent in developing countries, where cell phone penetration is extensive, especially compared to other technologies such as computers or personal vehicles. Although “crowdsourcing” has been used effectively in other domains,1 its application in the agricultural domain is still largely unexplored.
Opportunities for information gathering in agriculture are primarily in networking individual pieces of information from many sources throughout the supply chain. The individual pieces of sensed data contain local information at time and spatial resolutions that allow for increasing levels of “separability” or granularity. Taken together, these data provide a global view that is more useful than any individual unit of information. The combination of simultaneously separating and aggregating information is the real power behind information gathering (Sonka et al., 1999).
Processing of information entails manipulating data to change it from a less useful form to a more useful form. This computational activity turns the binary 1s and 0s of the data world into information that can be used by humans. In this section we describe modeling and data mining, two major aspects of information processing related to agriculture.
The purpose of modeling is to represent the physical world in computational simulations, which can provide a basis for making predictions. Currently, many different types of models are being developed, each one tailored to a specific use and community. For example, models of genetics are associated with bio-informatics. Models of nutrients and water flow (Robinson et al., 2000) throughout a field may be cued to in-field growth models and used to predict yields. Models can also be used to predict the behavior of the overall supply chain (Stutterheim et al., 2009). Even though the accuracy of these models is often debated, they are useful for scenario-based planning and decision making.
Data mining is a technique for extracting knowledge and information from unstructured data sets. Data-mining tools have been used in limited ways for agriculture, but there are still enormous untapped opportunities. Much like modeling, data mining has so far been focused largely on individual uses and communities.
Today, data-tagging protocols and search patterns are being used to tease out hidden relationships to predict trends in agricultural processes (Mucherino et al., 2009). A pure data-mining approach would make possible inferences based solely on data, thereby eliminating the need for developing infrastructure associated with modeling.
Opportunities for the Future
Although data mining based on specific knowledge of a particular agricultural system usually provides more insight than completely unstructured searches, modeling tools, used in conjunction with data mining, provide a framework for searching for hidden correlations or relationships. The coupling of models and data frameworks also has the potential to provide insights into the entire supply chain as a system.
By using a combination of these technologies, we should be able not only to predict behavior, but also to gather enough information to determine whether the predictions are accurate. Moreover, macro-scale models and data frameworks can be incorporated into local, high-resolution models and data frameworks to clarify overall supply chain behavior.
Bringing together disparate types of information processing will be challenging but could provide many benefits. For example, we would be able to “zoom in” to see how micro-scale activities, such as individual systems in the supply chain, might be affected by macro-scale processes acting as boundary conditions to those systems. Simultaneously, this modeling and data-mining capability would allow us to “zoom out” and see how macro-scale processes influence, or even dominate, interconnected networks of micro-scale sub-systems.
Moving effectively from data to decisions can often mean the difference between success and failure for an enterprise. Many individual organizations have integrated information systems to assist in making decisions on a space and time scale relevant to their particular interests. For example, manufacturers of agricultural equipment have adopted systems that allow GIS sensor data to flow to machines and control the site-specific applications of seeds, fertilizers, or pesticides. Similarly, seed companies have the ability to monitor crop growth trials in several different fields and select useful genetic varieties for market in a time frame suitable for future planting seasons.
However, agriculture has several domains and several stakeholders within these domains (Figure 1), and our current ability to gather information across the supply chain outpaces our ability to synthesize and make time-critical decisions that affect the system as a whole.
Levels of Decision Making
Decision making in agricultural systems occurs on many levels. On the micro-scale, individual economic and technical decisions can be made down to the level of an individual animal or plant. For example, a decision can be made about how much of a fertilizer to apply to a single plant based on the cost of the fertilizer and the potential yield of the plant. Currently, micro-scale decisions at one point in the supply chain are made independently of micro-scale decisions at other points in the supply chain.
On the so-called macro-scale, global decisions that affect multiple aspects of the overall supply chain must be made about agricultural policy. Decisions at this level include how much surplus of a staple crop to store in reserve or whether to allow genetically modified foods into a given country. Currently global decision making and local decision making are only loosely coupled, particularly in developing countries where there is relatively little use of IT in agriculture.
The Need for Coordinated Decision Making
Opportunities for decision making in agriculture are enhanced and enabled by the ubiquity of information gathering and the power of modern information processing. If, for example, data analytics associated with data mining tools can be combined with contextual or situational information from models or sensor data, the combined data could provide input for powerful computing algorithms that could provide support for the right decision choice at the right time. Moreover, such decisions could be made scalable so decisions for one geospatial location or one agricultural product would not conflict with other decisions.
This level of coordinated decision making on a regional basis is currently lacking in modern agricultural practice, despite the significant benefits it would provide in yield and efficiency. Lessons can be learned from other fields, such as the U.S. Department of Defense (DOD), which often has to make large-scale, time-critical, distributed decisions that affect a value chain or supply chain.
An illustrative example (Figure 3) shows how DOD fuses information from multiple intelligence, surveillance, and reconnaissance (ISR) platforms to come up with near real-time predictions and decisions to commit resources (Lemnios, 2010). The goal of this program, named Thunderstorm, is to demonstrate the identification of fast-moving threats (drug runners) from large data sets and make decisions about where to place interception points. This requires massive amounts of information from maritime and satellite ISR platforms, intensive computing based on oceanographic models and data-mining techniques, and decision-making tools for detecting threshold levels for taking action.
An analogy to agriculture would be the sudden emergence of crop pests or livestock disease that must be detected (information gathering), damage levels and transmission paths that must be predicted (information processing), and global or local interdiction strategies that must be created and implemented (decision making).
Although current capabilities in agriculture are not on the level of the DOD example given above, the future of agriculture and IT lies in this direction. Specifically, the integration of information can help address many of the challenges identified in the opening section of this article: inherent heterogeneity; unanticipated disturbances; geospatial dispersion; security and safety requirements; and constraints.
Exploiting Opportunities and Meeting Challenges
Capitalizing on the kinds of opportunities described above will require using technologies in a coordinated way. Information gathered must be fed to processing systems housed in data centers of respective agribusinesses. The data can then be mined and used in decision making, either automated or by humans. Relevant decisions could include the type and quantity of information to be gathered in the first place.
By taking advantage of the combined capabilities of information gathering and information processing techniques, decisions can be made that address the challenge of inherent heterogeneity by accommodating it, and even exploiting it when this would be advantageous. For example, GIS and site-specific application can exploit heterogeneities in soil in a field by spatially varying the application of seed or fertilizer to minimize the total amount used.
Unanticipated disturbances, such as changes in weather or outbreaks of disease, are inevitable occurrences in the agricultural setting. As genetic modifications to crops and livestock produce increasingly monocultural, biological systems, susceptibility to such disturbances increases. One can react to these disturbances, or one can predict and anticipate them. In either case, information can inform and accelerate the response.
On a shorter time scale, such as months, an unexpected flood might cause local food shortages creating subsequent food security issues. In this situation, resources can be re-routed based on supply chain information.
On a longer time scale, computational modeling can be used to predict climatic changes and their effects, and sensing can be used to validate and calibrate modeling predictions and extrapolations. This information capability can then be used in scenario-based planning to guide decisions about the types of genetic enhancements of crops or livestock that would be most suitable for coping with predicted environmental changes.
Currently, the challenge of geospatial dispersion is addressed by using IT to “tag” individual crops or animals in the supply chain and geo-locate them in space and time. As the price of tags and readers comes down, the ubiquity of sensed information will certainly increase.
Security and safety challenges can be met by improving awareness of all aspects of the supply chain. Continuous monitoring of crops or livestock can alert us to the presence of potentially threatening biological elements, enabling a prompt response. If a malicious activity occurs, data-mining techniques coupled with appropriate information gathering along the supply chain can be used to pinpoint the location of the security breach and provide information about how it propagates. This information can then be used to make a decision about the appropriate response.
As IT is increasingly used to exploit opportunities and address challenges, agricultural output will increase, despite the challenging, inherent constraints. Higher output per unit time, unit cost, or square meter of land will follow, along with faster and better time-critical decisions.
Many of the technologies and techniques discussed above are already being pursued by individual organizations and companies. What we need now is closer collaboration and integration among stakeholders to accelerate the introduction of IT into agricultural processes.
Industries such as telecommunications have benefited greatly from collaboration leading to standardization in the IT space. We believe that agriculture would benefit similarly if industry-government partnerships were to define agriculture-specific goals and standards for information thereby introducing uniformity and unleashing innovation similar to innovation in the telecom and computing industries. This could be particularly important in developing countries where modern IT infrastructure (e.g., the cellular communication network) tends to leapfrog established frameworks.
Agriculture is a critical human activity that will become increasingly important as the world population grows. In fact, given the necessity of feeding a growing planet, we have no choice but to increase our output with available resources, and IT, which offers many opportunities for addressing challenges to increasing global agricultural output, must play a central part in meeting that goal. Although IT has influenced many industries and greatly improved the management of supply chains, to date it has only influenced agriculture in a relatively localized way. It will take buy-in by all stakeholders and coordinated, communal efforts on a global level to integrate agriculture and IT to meet the needs of a hungry world.
Attaran, M. 2007. RFID: an enabler of supply chain operations. Supply Chain Management: An International Journal 12(4): 249–257.
Ehsani, M.R., S.K. Upadhyaya, D. Slaughter, S. Shafii, and M. Pelletier. 1999. A NIR technique for rapid determination of soil mineral nitrogen. Precision Agriculture 1(2): 219–236.
Gale, M., and K. Devos. 1998. Comparative genetics in the grasses. Proceedings of the National Academy of Sciences 95(5): 1971–1974.
Garibaldi, L., M. Aizen, A. Klein, S. Cunningham, and L. Harder. 2011. Global growth and stability of agricultural yield decrease with pollinator dependence. Proceedings of the National Academy of Sciences 108(14): 5909–5914.
Lemnios, Z.J. 2010. Transforming U.S. Defense R&D to Meet 21st Century Challenges. Plenary presentation to SPIE Defense Security and Sensing Symposium, Orlando, Fla., April 2010.
Mucherino, A., P. Papajorgji, and P. Pardalos. 2009. Data Mining in Agriculture. Springer Optimization and its Applications, Vol. 34. New York: Springer Verlag.
NAE (National Academy of Engineering). 2000. Greatest Engineering Achievements of the 20th Century. Available online at http://www.greatachievements.org/.
Nowatzi, J., R. Andres, and K. Kyllo. 2004. Agricultural Remote Sensing Basics. NDSU Extension Report AE 1262. Available online at http://www.ag.ndsu.edu/pubs/ageng/gis/ae1262.pdf.
Robinson, B., H. Viswanathan, and A. Valocchi. 2000. Efficient numerical techniques for modeling multicomponent ground-water transport based upon simultaneous solution of strongly coupled subsets of chemical components. Advances in Water Resources 23(4): 307–324.
Shannon, C. 1948. A mathematical theory of communication. Bell System Technical Journal 27(July and October): 379–423 and 623–656.
Sonka, S., R. Schroeder, S. Hoving, and D. Lins. 1999. Production agriculture as a knowledge creating system. International Food and Agribusiness Management Review: 165–178.
Stutterheim, R., C.N. Bezuidenhout, and P.W.L. Lyne. 2009. A framework to simulate the sugarcane supply chain, from harvest to raw sugar. International Sugar Journal 111(1324): 250–255.
1 See http://www.google.org/flutrends/.