Predicting how much food crops will produce each year is becoming more difficult as weather patterns grow less predictable due to climate change. Accurate, timely forecasts are important for farmers, food suppliers, and policymakers. In this study, I developed a new method to predict corn yields during the growing season using satellite data that tracks how efficiently plants are using sunlight to grow. This measurement, known as solar-induced chlorophyll fluorescence, tells us how active the plant’s photosynthesis is, essentially how hard it’s working.

I collected data every two weeks over five years from a major U.S. farming region and used a form of artificial intelligence called a neural network to learn how these plant signals relate to final crop harvests. I found that this method could predict corn yields with high accuracy weeks before harvest, and it performed better than older techniques that only use satellite images of plant color or structure.

My approach is fast, scalable, and doesn't require local weather data or field surveys. It could be used around the world to help farmers and governments prepare for food supply changes caused by droughts, floods, or other climate-related challenges.