Using Airborne LIDAR as an Approach for Detection of Understory Invasive Shrubs in Southern Illinois

Author

Elizabeth James, Southern Illinois University CarbondaleFollow

Date of Award

12-1-2025

Degree Name

Master of Science

Department

Forestry

First Advisor

Pease, Brent

Abstract

Mapping and managing the spread of non-native invasive (NNI) shrub species across large areas can be highly time-consuming and resource intensive, often requiring extensive data collection periods and mitigation efforts to slow the spread. One potential solution for mapping NNI is LiDAR (Light Detection and Ranging), an active remote sensing method that uses laser pulses of light to measure environmental features such as topography and hydrology. While the LiDAR is typically targeting surface-level information such as elevational gradients, the laser interference with vegetation results in a three-dimensional characterization of forest structure that can be developed to assist land managers in the mapping of vegetation, potentially including NNI shrub species. We used LiDAR point cloud data collected by the state of Illinois to derive forest metrics based on a segmented height stratum to detect understory invasive shrub species, autumn olive (Elaeagnus umbellata) and bush honeysuckle (Lonicera maackii), in an upland hardwood forest in southern Illinois. To provide an assessment of LiDAR for detecting NNI shrubs, forest inventory data was collected at 91 circular plots (0.2 ha) during May to August 2024, which included identifying the native and NNI shrub species present and their associated average and maximum height, diameter at breast height (DBH), proportion area occupied, forest stand structure characteristics, and overstory canopy openness. Overstory forest conditions consisted of mixed hardwood species, dominated by Quercus spp. and Carya spp., while the midstory and understory were dominated by Ulmus spp. and Asimina triloba. To test the accuracy of LiDAR to predict occupancy of NNI species in forested stands, I first assessed five NNI occurrence thresholds within the focal height stratum to determine sensitivity of LiDAR to NNI presence (1%, 5%, 10%, 25%, and 50% occupancy). The 10% threshold—indicating that 10% of the LiDAR returns occurred in the focal stratum—was determined as best predictor for NNI occupancy based on its precision (0.87), accuracy (0.73), F1 harmonic mean (0.77), and Mathew’s Correlation Coefficient (MCC) (0.45). In addition to using LiDAR to predict where NNI species may be found, I also tested the relationship of two other metrics (canopy openness and land-use history) to NNI species coverage. The area occupied by NNI species was regressed against canopy openness within a plot using Poisson generalized linear regression, with results suggesting a canopy that is more open is likely to have a larger abundance of NNI present. Next, I assessed whether land-use history, as assessed by aerial imagery captured in 1937-1947, and documented those areas previously “Forested” had no significant, current indication of NNI presence (F3,87 = 1.19, p-value = 0.47), whereas areas previously “Non-Forested” did statistically indicate the presence of NNI (F3,87 = 1.19, p-value < 0.01). Combined, these results suggest that LiDAR may be a promising method for detection and mapping of NNI species, particularly when it is combined with other forest characteristic measurements and land use history.