Airborne Hyperspectral Data Application in Health Stress Detection of Blueberry Fields and Ash Trees

Sponsor: Maine Space Grant Consortium

Daniel Hayes and Catherine Chan - University of Maine
Peter Nelson - University of Maine at Fort Kent

Hyperspectral datacube displayed as different band indices overlaid on a NAIP image

Advanced detection of health stress in agricultural fields and forests can prompt management responses to mitigate detrimental conditions such as nutrient deficiencies, disease, and mortality. New applications in hyperspectral data and imaging spectroscopy on agriculture and forests have shown the potential for early stress detection. We build on previous work by assessing two different systems; drought stress in wild blueberry vegetation through multi-temporal observations, and health classifications in ash trees of varying emerald ash borer (EAB) infestation levels. Our aim is to use hyperspectral imaging processes to measure health stress while associating the characteristics to drought stress in blueberries and health vigor classifications in ash trees. This will allow provisions in land cover assessments and mitigation options for Maine land management. Our objectives include:

1. Collect airborne hyperspectral data over blueberry fields and infested ash sites in New Hampshire
2. Collect ground data including ground spectral reflectance, water potential in blueberries, and health ratings in ash trees encompassing the classification spectrum
3. Generate predictor maps to determine locations of drought or infestation, susceptible areas, and potential solutions

Airborne data was collected using a Headwall Micro A-Series imaging spectrometer (400-1000 nm, 326 bands, 1024 pixel scanline, 3-10 cm GSD) affixed to an unmanned aerial vehicle (UAV). Remotely sensed data will be validated through ground measurements using a Spectral Evolution spectroradiometer (350-2500 nm). Airborne scans were collected over two 40-acre wild blueberry fields (irrigated and non-irrigated) in Downeast Maine three times throughout the 2019 growing season. Spectroradiometer scans and water potential measurements were taken of 20 samples in each field. Airborne and ground scans were collected over infested ash sites in southern New Hampshire in July 2019. Ground data sampling including health classification ratings was conducted in November of the same year.

Using methods in machine learning and statistical analysis, we aim to show patterns of early-onset drought in blueberries, showcasing crop characteristics (e.g. elevation, road-side, pesticide treatment) susceptible to drought stress. In the ash sites, reflectance measurements will be related to different intensities of EAB infestation to understand tree stress signals, particularly in early stages to prevent irreversible damage. Ideally, our results will provide accurate classifications of drought and health vigor in their respective field categories. A further objective is assessing the data’s potential in providing pre-visual drought and emerald ash borer symptoms. Having the capability in early detection provides benefits in crop management in blueberries and monitoring of those areas in Maine where EAB infestation is primeval. If these aspects show promise, the potential of imaging spectroscopy over large spatial extents could provide great strides in practical monitoring and mitigation.