Every year, over 25% of the world’s annual output of agricultural goods are ruined or damaged by pests and diseases. Pests typically eat the agricultural goods, rendering them unfit for use. Meanwhile, fungal diseases tend to produce various mycotoxins, many of which are powerful carcinogens, contaminating the agricultural products so that they are unsafe for either human or animal consumption. Both agricultural pests and diseases cause hundreds of billions of dollars (USD) in economic damage per year, and are a major impediment to feeding a growing human population moving forward.
For many of these pests and diseases, the conditions that are particularly conducive to an outbreak remain unknown. In the available academic literature, researchers often focus on finding the minimum conditions that allow the pest or disease to survive in the wild (i.e. whether enough heat units or GDDs have accumulated that season for the eggs/spores to have hatched), but offer little insight as to why one field can be infested while a nearby field shows no signs of a problem. Differences in farming practices (i.e. cultivar planted, field sprayed, number of tillage operations, etc.) across fields contribute to some of the variability, but local differences in weather conditions and the state of the surrounding vegetation are also an important, yet often overlooked, piece of the puzzle.
Researchers at aWhere Inc. have developed a number of statistical models to help farmers, researchers, and business people predict when and where they are likely to have issues with various pests or diseases. These models were created using information from both the research community and the production field history of farms from across the world. These datasets were integrated using machine learning algorithms to provide insights that only a big data approach to agriculture can accomplish. Comparisons done by aWhere researchers have found that these insights are often more accurate than the predictions made using other available tools.
Figure 1: Summary of Field Results for Gray Leaf Spot
As an example, gray leaf spot is one of the most economically costly diseases of corn in the United States. Early blighting of the corn can lead to yield losses of more than 50%. A statistical model developed by aWhere, and tested across the entirety of the corn belt, was able to successfully identify over 99.4% of the fields that were infected by the gray leaf spot disease while still maintaining a false positive rate of just 17%. The figure to the right shows how this model was able to identify the diseased fields (blue bars) from the healthy fields (gray bars) the vast majority of the time. The data used for this analysis was from the actual observation reports of a major crop scouting company in the corn belt. This index is able to assess weather patterns over a growing season to derive a single value that predicts whether a field is likely to be diseased or not (value of x axis in figure above). Researchers at aWhere have had similar success with other economically important fungal diseases, including Northern Corn Blight, Common Corn Rust, Soybean Rust, Fusarium, Aspergillus, and others, in a number of crops, such as corn, soybean, wheat, and cotton, amongst others.
aWhere combines precision agriculture with big data analytics to provide actionable insight to farmers. These insights allow farmers to make faster, more efficient, and more profitable interventions and actions. These insights can help farmers to know how and where a problem is most likely to occur so that it can be dealt with proactively, saving them both time and money in the long run. aWhere continues to develop new models, algorithms, and insights to support the ever evolving world today’s farmers find themselves in.
Edited by Leila Al-Hamoodah