Assessing Changes in Grassland Species Distribution at the Landscape Scale Using Hyperspectral Remote Sensing.

Area of Science:

• Earth and Environmental Sciences
• Remote Sensing
• Ecology

Background:

• Hyperspectral remote sensing advances land cover characterization in complex ecosystems.
• Quantifying fractional abundance of land cover classes is crucial for ecological monitoring.

Purpose of the Study:

• To apply a linear spectral unmixing (LSU) algorithm to NEON hyperspectral imagery for quantifying land cover changes.
• To assess the effectiveness of integrating LSU with UAV data for detecting shifts in semi-arid grassland composition.

Main Methods:

• Linear spectral unmixing (LSU) algorithm applied to NEON hyperspectral imagery from 2018 and 2022.
• UAV imagery used for high-resolution reference data acquisition and model validation.
• Fractional abundance of herbaceous vegetation, mixed forbs, and bare soil quantified.

Main Results:

• Herbaceous cover decreased from 16.02 ha to 11.56 ha; bare soil expanded from 3.37 ha to 6.39 ha.
• Mixed forb cover remained relatively stable (12.38 ha to 13.82 ha).
• Slope-based trend maps revealed localized vegetation loss and regrowth patterns despite non-significant statistical changes in mean abundance.

Conclusions:

• Linear spectral unmixing integrated with UAV data effectively detects subtle, ecologically meaningful shifts in semi-arid grassland composition.
• The study highlights the utility of hyperspectral remote sensing for monitoring rangeland dynamics.
• Findings demonstrate localized vegetation dynamics within the experimental rangeland.