Data-Driven Fungicide Decisions

Farmers must wear many hats to keep their operations running, but one of their most important tasks is managing variable input levels. This task is further complicated by the dynamic nature of pests and diseases. In southeastern U.S. wheat production, for example, crops are frequently threatened by Stagonospora leaf blotch (SNB), a fungal disease that can result in yield losses of up to 30% under favorable conditions. To combat SNB, farmers rely on resistant cultivars, crop rotation and fungicide applications throughout the season.

Wheat is a relatively low-value crop, generating only a few hundred dollars per acre. With rising input costs and fluctuating commodity prices, the economic return from fungicide use can be highly variable or even negligible. For example, with wheat priced at $6 per bushel and fungicide plus application costs around $30 per acre, growers need a 5.0 bu.\yield increase just to break even. However, because SNB epidemics are sporadic, varying from year to year and across fields, not all areas will achieve the yield increase needed to justify fungicide use. Some fields will benefit, but the challenge lies in identifying which ones.

As a PhD candidate and FFAR Fellow at NC State University, my research centers on developing computer models to predict SNB epidemics across landscapes. These models help farmers target high-risk areas, reducing fungicide use in low-risk zones and improving the chances of achieving a positive return on investment.

How Do Computer Models for Plant Disease Prediction Work?

Computer models can process large volumes of data, execute complex calculations quickly and reveal key insights that might otherwise go unnoticed. In the case of SNB, weather conditions play a key role in disease development. Factors such as temperature, humidity, rainfall and wind influence every stage of the pathogen’s life cycle. For instance, higher temperatures can speed up the pathogen’s growth, while high humidity and extended leaf wetness promote spore germination. Rain also spreads fungal spores between plants, impacting all stages of disease progression.

While the role of weather in SNB development is well-recognized, turning these effects into meaningful variables for disease prediction is complex. Drawing from data on 100 SNB epidemic cases collected over three years and 65,000 hourly weather data points, I developed algorithms to pinpoint key periods of weather-disease interaction. This involved advanced programming, data analytics, and feature engineering—a process that turns raw data into relevant variables for modeling. For example, we found that the number of dawn hours with temperatures between 16°C and 19°C, combined with rainfall over 0.2 mm during a week before and after wheat anthesis, increases the risk of SNB outbreaks. In total, over 100 weather variables were generated, providing a detailed understanding of how environmental factors influence SNB dynamics.

By integrating these weather variables with agronomic data, I developed probabilistic models to predict SNB severity in North Carolina wheat fields. These models utilize advanced statistical techniques, particularly Bayesian methods, which excel at managing uncertainty and providing probabilistic insights into disease occurrence. The Bayesian approach helps us gauge how confident we can be about whether fungicide applications will work or if they might not be worth the cost, making decision-making clearer.

New Frontiers in Agriculture

As agriculture faces the growing challenges of climate change and the increasing demand for higher-quality food, the role of predictive models—like those I am developing during my Ph.D.—cannot be overstated. Creating models that capture the biological processes behind disease and crop development is both an art and a science. With the recent advances in computational power, we have new opportunities to better understand these processes.

Through the FFAR Fellows Program, I have refined my research skills, collaborated with leading plant pathology scientists and enhanced my ability to communicate complex data to benefit farmers directly. I am especially thankful to UPL for sponsoring my fellowship and to my mentors and Ph.D. advisor, Dr. Peter Ojiambo, for their invaluable guidance along the way.

Looking ahead, I am eager to continue exploring the environmental interactions influencing crop and disease dynamics and to further validate my models. My agricultural background drives my passion for transforming scientific discoveries into practical solutions that help farmers remain profitable while reducing the environmental footprint of fungicides in agriculture.