Close Window

Disentangling the Spectra of Flowers to Map Landscape-Scale Blooming Dynamics

Yoseline Angel,  NASA Goddard Space Flight Center - UMD,  yoseline.b.angellopez@nasa.gov (Presenter)
Dhruva Kathuria,  NASA Goddard Space Flight Center,  dhruva.kathuria@morgan.edu
Katherine Dana Chadwick,  Jet Propulsion Laboratory, California Institute of Technology,  katherine.d.chadwick@jpl.nasa.gov
Philip G Brodrick,  Jet Propulsion Laboratory, California Institute of Technology,  philip.brodrick@jpl.nasa.gov
Alexey N Shiklomanov,  NASA Goddard Space Flight Center,  alexey.shiklomanov@nasa.gov

Remotely modeling landscape flowering dynamics (e.g., blooming) could reveal hints on ecological processes, such as pollination patterns and floral adaptations to environmental changes. Image spectroscopy can measure the optical properties of flower pigments across different spatio-temporal scales and their contribution to canopy spectra. Our research investigates methods that account for the relative contributions of flowers, leaves, and soil to typical hyperspectral scenes, leading to characterizing within-pixel spectral variability, mapping flowering areas, and revealing specific phenophases across species. Weekly series imagery from the airborne imaging spectrometer AVIRIS-NG and field spectra were collected and processed as part of the NASA SBG High-Frequency Time Series (SHIFT) campaign carried out in The Nature Conservancy's Dangermond Preserve, CA, during 2022's spring (March-May). The processed data was input to investigate flowering species' spectro-temporal variation and spatial distribution using the Spectral Mixture Residual, Bayesian clustering techniques, and a proposed hyperspectral flowering index. Spectral unmixing allowed the computation of residual signal that comprises subtle spectral features linked to biophysical processes. Principal Components of the mixture residual spectra reduced the dimensionality of the data to identify flowering clusters and their uncertainty based on an unsupervised Gaussian Mixture Model. Mapping flowering events throughout the season allowed us to identify average spectral gradient variations linked to flowering pigments and water content changes. Time series of the Modified Enhanced Blooming Index and the Red-Edge Normalized Difference Vegetation Index revealed flowering and greenness cycles across species. Our approach for mapping flowering events opens opportunities for future hyperspectral satellite missions(e.g., SBG) to monitor floral biodiversity at broader scales.

Poster: Poster_Angel_3-5_226_35.pdf 

Associated Project(s): 

Poster Location ID: 3-5

Presentation Type: Poster

Session: Poster Session 3

Session Date: Thu (May 11) 3:00-5:00 PM

CCE Program: TE

Close Window