- 212 Downloads
Big Data and (Smart) Agriculture
Big data and digital technology are driving the latest transformation of agriculture – to what is becoming increasingly referred to as “smart agriculture” or sometimes “digital agriculture.” This term encompasses farming systems that employ digital sensors and information to support decision-making. Smart agriculture is an umbrella concept that includes precision agriculture (see “AgInformatics” – De Montis et al. 2017) – in many countries (e.g., Australia) precision agriculture commonly refers to cropping practices that use GPS guidance systems to assist with seed, fertilizer, and chemical applications. It therefore tends to be associated specifically with cropping farming systems and deals primarily with infield variability. Smart agriculture, however, refers to all farming systems and deals with decision-making informed by location, contextual data, and situational awareness. The sensors employed in smart agriculture can range from simple feedback systems, such as a thermostat that acts to regulate a machines temperature, to complex machine learning algorithms that inform pest and disease management strategies. The term big data, in an agricultural context, is related but distinct – it refers to computerized analytical systems that utilize large databases of information to identify statistical relationships that then inform decision support tools. This often includes big data from nonagricultural sources, such as weather or climate data or market data.
An example of how these concepts interact in practice: a large dataset may be established that contains the yield results of many varietal trials across a broad geographical area and over a long period of time (including detailed information pertaining to the location of each trial, such as soil type, climatic data, fertilizer, and chemical application rates, among others). This data could be analyzed to specifically determine the best variety for a particular geographic location and thus form the basis for a decision support system. These two steps constitute the data and the analytic components of “big data” in an agricultural context. The data could then inform other activities, such as the application (location and rate) of chemicals, fertilizers, and seed through digital-capable and GPS-guided farm machinery (precision agriculture).
Applications of Big Data in Smart Agriculture
Big data, and in particular big data analytics, are often described as disruptive technologies that are having a profound effect on economies. The amount of data being collected is increasing exponentially, and the cost of computing and digital sensors is decreasing exponentially. As such, the range of consumer goods (including farm machinery and equipment) that incorporates Internet or network connectivity as a standard feature is growing. The result is a rapidly expanding “Internet of things” (IoT) and large volumes of new data. For example, John Deere tractors are fitted with sensors that collect and transmit soil and crop data, which farmers can subscribe to access via proprietary software portals (Bronson and Knezevic 2016). The challenge in agriculture is reaching a point where available data and databases qualify as “big.” Yield measurements from a single paddock within one growing season are of little value because such limited data does not inform actionable decision taking. But, when the same data is collected across many paddocks and many seasons, it can be analyzed for trends that inform on-farm decision-making and thus becomes much more valuable. This is true across the full agricultural value chain.
Cropping systems: Variable rate application technology (precision agriculture), unmanned aerial vehicles or drones for crop assessment, remote sensing via satellite imagery.
Extensive livestock: Walkover weighing scales and auto-drafting equipment, livestock tracking systems, remote and proximal sensor systems for pasture management, virtual fencing.
Dairy: As for extensive livestock, plus individual animal ID systems and animal activity meters that both underpin integrated dairy and herd management systems.
Horticulture: Input monitoring and management systems (irrigation and fertigation), robotic harvesting systems, automated postharvest systems (grading, packing, chilling).
Overall, while the technology is still relatively new, agriculture is already seeing substantial productivity gains from its use. Further transformative impacts will be felt when real-time information business process decisions and off-farm issues (e.g., postharvest track and trace of products), such as planning, problem-solving, risk management, and marketing are underpinned by big data.
Challenges and Implications
In an agricultural context, there are several challenges.
First, convincing farmers that the data (and its collection) are not merely a novelty but something that will drive significant productivity improvement in the future may be difficult. In many cases the hardware and infrastructure required to collect and use agricultural data are expensive (and prohibitively so in developing countries) or unavailable in rural areas (especially fast and reliable Internet access), and the benefits may not be realized for many years. Similarly, technical literacy could be a barrier in some cases. These issues are, however, common to many on-farm practice change exercises that drive improvements in efficiency or productivity and can generally be overcome.
Second, farmers may perceive that there are privacy and security issues associated with making data about their farm available to unknown third parties (Wolfert et al. 2017). Large proprietary systems from the private sector are available to capture and store significant amounts of data that is then made available to farmers via subscription. But, linking big data systems to commercial benefit raises the possibility of biased recommendations. Similarly, farmers may be reluctant to provide detailed farm data to public or open-source decision support systems because they often do not trust government agencies. These systems also lack ongoing development and support for end users.
Finally, a sometimes-over-looked issue is that of data quality. In the race for quantity, it is easy to forget quality and the fact that not all digital sensors are created equal. Data is generated at varying resolutions, with varying levels of error and uncertainty, from machinery in various states of repair. The capacity of analytical techniques to keep pace with the amount of data, to filter out poor quality data, and to generate information that is suitable at a range of resolutions are all key issues for big analytics. For analyses to underpin accurate agricultural forecasting or predictive services that improve productivity, advancements in intelligent processing and analytics are required.
Ultimately, it is doubtful that farmer knowledge can ever be fully replaced by big data and analytic services. The full utility of big data for agriculture will be realized when the human components of food and fiber production chains are better integrated with the digital components to ensure that the outputs are relevant for planning (forecasting and predicting), communication, and management of (agri)business processes.
- Australian Farm Institute. (2016). The implications of digital farming and big data for Australian agriculture. Surry Hills: NSW Australian Farm Institute. ISBN 978-1-921808-38-8.Google Scholar
- De Montis, A., Modica, G., & Arcidiacono, C. (2017). AgInformatics. Encylopedia of Big Data. https://doi.org/10.1007/978-3-319-32001-4_218-1.