| Big sagebrush (Artemisia tridentata) is a keystone species across the western U.S and has adapted
to a wide range of environments resulting in complex communities. Big sagebrush is a polyploidy
species, which refers to the number of genome duplicates present in a given individual. Genome
duplication may increase the adaptive potential for an individual to survive stressful
environmental conditions and occupy novel ecological niches. Big sagebrush are either diploid or
tetraploid, with polyploid individuals often being indistinguishable from their diploid progenitors.
The complexity of the relationships between big sagebrush polyploidy and local environmental
conditions complicates our ability to understand how a changing environment will influence the
distribution of sagebrush subspecies and restoration efforts. We used uncrewed aerial systems
(UAS) equipped with a hyperspectral sensor to model and predict big sagebrush ploidy levels
based on spectral response along an elevational gradient in Castle Rocks State Park. Data
collection of near infrared spectroscopy (NIRS) spectroscopy for ground truthing, and
hyperspectral UAS imagery occurred in June and September from 2021 to 2023 to capture peak
leaf greenness and seed production. Principal component spectral analysis yielded wavelengths
350-700 nm and 750-1200 nm were key features in big sagebrush spectra. We used convolutional
neural networks (CNN) and support vector machines (SVM) to model NIRS and UAS
hyperspectral spectral response to predict big sagebrush ploidy level. Our most successful models
NIRS and hyperspectral yielded classification accuracies of 82% and 90%.
Keywords: AI Modeling, Big sagebrush, Neural Networks, Remote Sensing, Uncrewed Aerial
Systems |