Dr. Hannah Kerner, a NASA Harvest Hub partner, specializes in developing machine learning applications using satellite data for crop mapping and forecasting. She was recently featured in episode 25 “Harvest Hub: Food Security from Space” of Via Satellite’s On Orbit podcast to discuss her work with NASA Harvest, her computer science background, and the Students for the Exploration and Development of Space (SEDS) organization. Dr. Kerner begins by explaining that machine learning, in essence, is the use of computer programs to complete a task based on its experience, where experience is in the form of data.
She describes how satellite data can reveal which types of crops are located in which regions, differentiate between types of land use (e.g., forest versus agricultural versus urban), and help forecast crop conditions. This type of data can be used to inform commodity markets, anticipate food shortages, monitor food insecurity, and help farmers to develop more sustainable, resilient, and productive farming practices.
To train her crop monitoring algorithms, Dr. Kerner uses massive amounts of satellite data provided largely by NASA’s Landsat 8 and the European Space Agency’s Sentinel satellites, highlighting the importance of open access data for food security and agricultural research. Additionally, through the public-private relationships built through the NASA Harvest program, data is also commercially available through industry partners such as Planet who provide data with higher spatial resolution and often with increased temporal frequency. Dr. Kerner points out that NASA Harvest projects put this readily-available data to best use with end-users in mind by working directly with our partners from the beginning in order to inform specific problems that are important to each stakeholder. Projects supported by NASA Harvest take a flexible approach to best achieve the goals of our partners, whether that is governments trying to increase their country’s food production, using crop monitoring to stabilize markets, developing tools for farmers to collect ground data, and anything in between. Dr. Kerner summarizes this cooperative and mutually-beneficial approach, stating that “there is a huge community of users who rely on these products for their decision-making” and the NASA Harvest partners that have been in the industry for many decades are well-equipped to provide support for a wide variety of goals.
Dr. Kerner has an extensive background in developing machine learning algorithms for both Earth and space, but while learning about planets beyond our own is very exciting and important research, “Earth is definitely the best planet and we really really need to pay attention to [it] as well!” She reveals that her passion for Earth-based agricultural science is fostered by the immediacy of this type of research. Because of access to timely Earth Observation data, the NASA Harvest team is equipped to respond readily not only to changes in crop conditions but can also account for economic trends such as worldwide trade circumstances or pest outbreaks. Dr. Kerner enjoys seeing the value of her work come to fruition, explaining that her team is responding to real-world problems immediately and that this sort of immediate response is not common in academia. Furthermore, she finds it exciting to be closer to the type of data and problems that she is working on and seeing how that affects real people; she “ really wanted to have a closer tie to the work [she] was doing and have some impact on the world as [she] was seeing it today.”
Listen to the full podcast for more on Dr. Kerner’s perspective on machine learning for agriculture and her industry experience.