UMaine-led Study: Imaging Spectroscopy can Predict Water Stress in Wild Blueberry Fields

Imaging spectroscopy can help predict water stress in wild blueberry barrens, according to a University of Maine-led study.

The technology involves measuring the light reflected off of objects depicted in images captured by drones, satellites and other remote sensing technology to classify and gather pertinent information about the objects. According to researchers, it can precisely measure light across dozens, if not hundreds, of bands of colors. The reflectance spectra can depict nutrient levels, chlorophyll content and other indicators of health for various crops, according to researchers.

Scientists from UMaine, the Schoodic Institute and Wyman's, one of the world's largest purveyors of wild blueberries and the number one brand of frozen fruit in the country, found in their research that when incorporated into models, imaging spectroscopy can help predict whether wild blueberry fields will lack sufficient water for growing. Not only can the technology help inform growers as they evaluate irrigation routines and manage their water resources in a way that avoids damaging the crop, researchers say.

The team collected imaging spectroscopy data by deploying a drone equipped with a spectrometer for capturing visible and near-infrared light to photograph wild blueberry fields owned by Wyman's in Debois, Maine. Researchers then processed the images to measure reflected light spectra from the plants for indications of chlorophyll levels and other properties that would help estimate their water potential, which, they say, is the primary force driving water flow and an indicator of water stress. At the same time, the group collected small branches with leaves from wild blueberry plants in the plots to assess their water potential and validate the spectra-based estimation. Pictures and samples were collected in the spring and summer of 2019 when the plants experienced peak bloom, green fruit and color break.

The data from both drone images and ground samples were incorporated into models, which they developed using machine learning and statistical analysis, to estimate water potential, and thereby predict water stress, of the plants in the barrens. Models from the ground sample data were used to help guide the development of and validate the model created with data from the images. The results of both sets of models were comparable, demonstrating that imaging spectroscopy can accurately predict water stress in wild blueberry barrens at different times of the growing season. With the efficacy of the technology confirmed, researchers say scientists can capitalize on the benefits of it, such as conducting repeated measurements on small objects like blueberry leaves with ease.

Graduate student Catherine Chan led the study, joined by UMaine faculty Daniel Hayes and Yongjiang Zhang, Schoodic Institute forest ecologist Peter Nelson and Wyman's agronomist Bruce Hall. The journal Remote Sensing published a report of their findings.

"We couple spectral data and areas of known water potential in wild blueberry fields through machine learning, creating a model to further predict areas that may be water stressed," Chan says.

Understanding how to sustainably manage water resources to mitigate risk associated with current and increasing drought frequency is crucial to wild blueberry growers, researchers say.

"This research provides key learnings to ensure the continued viability of wild blueberry crops for generations to come," Hall says.