Agriculture is a major driver of New York State’s economy. Given this, Cornell’s Nutrient Management Spear Program (NMSP) seeks to develop management tools that help farmers realize more profit through sustainable approaches. Under the leadership of Sunoj Shajahan, formerly a Cornell research associate now faculty at the University of Illinois, and Subha Srinivasagan, Atkinson postdoctoral fellow, NMSP uses precision agriculture technologies to help meet this goal.
Precision agriculture harnesses technology and data science to precisely predict crop productivity and discovers optimal areas where fertilizer is needed — ultimately increasing crop yield. To estimate and predict corn grain and silage yields, Shajahan and Srinivasagan use time-series drone and satellite imagery. “Drones and satellites are remote sensing technologies that have paved the way for precision agriculture,” said Shajahan. “The images captured from these tools give us a birds-eye view of farm fields and data across the growing season, from planting until harvest and from year to year.”
While yield monitors are increasingly used to record yield, not all farms can afford the equipment or effectively manage the data generated. Using images from drones and satellites can be an alternative way of precisely mapping yield and may be used to estimate yields from past years as well. "Satellites enhance precision agriculture by providing real-time visual insights of crop health, weather, and soil conditions from above, and it is freely available,” Srinivasagan said.
“Drones and satellites are remote sensing technologies that have paved the way for precision agriculture.”
Shajahan joined NMSP in 2019. His research tested use of drones at different scales, from small plot to much large commercial corn fields. His research, published in Remote Sensing in 2021 and most recently in Biosystems Engineering in 2023, used multispectral images from drones to predict corn yield and compare the reliability of different machine learning models. The research showed that drone images can predict yield with considerable accuracy. Capturing drone images later in the season, when plants are in the reproductive and more mature growing stages, allows for more accurate estimates than flying drones early in the season when plants are still small.
Given the variability and complexity of the data collected from multiple sources (yield monitors and multiple imagery sources), analyses can be challenging. “Both drone and satellite data are messy in their raw format,” Shajahan explained. “For example, clouds and shadows can get in the way or the time-series imageries might not align well. And, since we are bringing together all of these data, it can be complicated to standardize the various data layers.”
