Examining the digital transformation in agriculture

The Business Challenge:

Responding to the global challenges, agriculture must improve on all aspects: Smarter resource use, increasing yields, increased operational efficiency, and sustainable land usage. Big data is expected to have a large impact on "smart farming" and involves the whole supply chain, from biotechnology and plant development to individual farmers and the companies that support them. On the ground, smart sensors and devices produce big amounts of data that provide unprecedented decision-making capabilities. Working through partnerships to prototype, organizations can take advantage of the growing technology to augment their ability and gain an advantage.

As in many industries, agriculture is also experiencing a shift towards Artificial Intelligence and predictive analytics methods through the adoption of machine and deep learning algorithms. These business and process improvements build upon automated systems to utilize new information sources and larger datasets to augment human decision making, with the goal to generate improved growth of crops or animal husbandry practices. The agriculture industry is uniquely positioned to utilize such solutions because large-scale farming efforts have the capacity to generate massive quantities of data that can be captured and analyzed to use for such decision-making. These data can be used for a variety of tasks from predicting fecundity, to tracking animal health and behavior, and automating yield management by the incorporation of novel data sources, such as satellite images, historical weather data, and sensor information. Further, these data can be naturally predictive, with repetitive patterns and seasonal events to classify. To kickoff off our 2018 article series, SFL Scientific looks at solutions for Smart Farming & AgTech and how to improve outcomes for farmers, crop insurers, and companies.

Resource Efficiency

Agriculture is widely acknowledged as the industry that may benefit most from the adoption of IoT technologies. The scale of modern agricultural techniques means that any small amount of a resource (water, fertilizer, feed) saved on a given day, can result in massive savings for the year. Maximizing resource efficiency by tracking precise resource allocation while correlating crop outcomes is now possible using sensor networks to track resources distributed. Precise monitoring coupled with image recognition, as discussed below, allows for hybrid and automated optimization of crop fields. Machine learning applied to sensor data can be used to monitor crop success and detect anomalies allowing for accurate, fast, and precise interventions. Soil humidity sensors, for instance, can be used to create a controlled smart irrigation system to automatically monitor water levels, releasing the quantity of water required for crop success. Soil nutrient levels can also be monitored, allowing for controlled release of fertilizer using the minimum amount necessary to maintain the desired crop yield. Maximizing resource efficiency is an important solution to the modern farmer in order to achieve the highest ROI, especially when factoring the investment of time and equipment.

Crop Assessments

Assessing and monitoring crop success is a major part of a farmer’s job, which was traditionally accomplished with a “boots-on-the-ground” approach in order to monitor crops and check for signs of pests and diseases, requiring a significant time investment. These labor-intensive techniques can be updated using satellite and drone imaging, as well as locally mounted camera systems, which provide high-resolution data for decision making. Satellite and drone image data is a powerful resource for the smart farmer’s data pipeline, allowing for accurate prediction and measurement capabilities on their land.

Using drones, aerial cameras, or satellites to image fields in multiple spectrums, provides high resolution, multimodal data, which can be used to assess crops without having to send out surveyors to manually assess crop health. For example, machine vision can be used to analyze crops for signs of detrimental pests and point out spots for pesticide application to control pest populations. Deep learning frameworks can be deployed on such image datasets to identify common plant types with high accuracy. Image data can also be used to identify crop diseases, allowing for treatment in early stages to minimize crop loses. Further, crop insurers can use imagery to improve underwriting performance, quickly surveying damaged areas, and keeping a historical record of problem areas (flood zones, hurricane, etc.). Use of image data for field analysis and crop assessment is a cheap and precise alternative to traditional methods of crop assessment and provides a higher resolution map and historical record.

Combining these techniques with farming equipment, such as irrigation systems, we can create systems to monitor and prevent problems before they arise, such as plant stress. Using thermal and NDVI imaging, we can monitor and predict hot zones and malfunctioning irrigation equipment.

Use of drone systems coupled to smart detection software and time-lapse studies, allows individuals to supervise and monitor large fields, improving yields and detecting problems before they threaten plant health.

Robotics, Machinery, and Maintenance

Robotics and the data associated with their use are positioned to improve crop ROI by reducing labor costs and providing precise data for decision making. Investment in machinery and maintenance accounts for a large portion of a farmer’s budget for a given year, outpacing spending on fertilizer or chemicals. Automation reduces the expense associated with operating machinery by eliminating the need to have an operator on each vehicle and by maximizing the efficiency of the machine use. Keeping equipment in working condition requires a constant labor to upkeep and can be unpredictable, potentially impacting other operations. In order to reduce maintenance cost, IoT and sensor technology can be used to predict maintenance of farm machinery by combining operating data with historical machine data. This would allow owners to perform maintenance on machinery before more costly repairs or replacements are necessary at a time that is controlled by the owner.  

Livestock Monitoring

Raising livestock is a labor-intensive segment of the agricultural marketplace requiring a staff to track animals and maintain livestock. IoT and imaging technologies can also be used to monitor animal behavior, movement, and health outcomes remotely and maintain a single data pipeline. Automated animal monitoring has the potential to save owners in manual animal maintenance costs by detecting anomalous behavior in animals, allowing for earlier intervention and improved preventative care. IoT sensors can be used to monitor animals health, from movement sensors to physical characteristics, and report anomalies for human verification. Sensors can also in combination with machine learning to modify the diets of specific animals to increase lactation and maximize production of top producing animals.

These are just a subset of use cases for the modern agriculture market moving forward. With the adoption of new technologies, the agricultural sector can look to improved ROI on their crops and herds by modeling the minimum input necessary, at the right time, and being proactive about detrimental events. An analytics first approach in combination with new data streams will ensure that every farmer and organization can achieve the most for what they put in. The best in class organizations will innovate and integrate data science and technology into every business process.



About SFL Scientific

SFL Scientific helps firms accelerate accurate decisions that drive revenue. We’re a data science consulting firm offering custom development and solutions in data engineering, machine learning, predictive analytics, and Artificial Intelligence. SFL uses specific domain knowledge to solve complex and novel business problems, specializing in helping organizations develop a full1y integrated approach to leverage data-driven systems. By employing a holistic process to our services, we combine the latest technical advances, real-world expertise, innovation, and an understanding of core business goals to generate tangible and immediate value.


More from Michael Segala, Ph.D.:

Case Study: The Business Value ofUsing Data Science in Pharmaceuticals, A Machine Learning Pipeline for Automatic Text Extraction

Michael Segala, Ph.D. on LinkedIn