Detection of invasive plants using NAIP imagery and airborne LiDAR in coastal Alabama and Mississippi, USA

Abstract

Invasive plants have imposed severe threats to native ecosystems worldwide. Triadica sebifera (Tallow tree) and Ligustrum sinense (Chinese privet) are among the most prolific invasive species in the southern United States (US) that needs urgent assessment to protect coastal ecosystems. The lack of spatially explicit assessments of these invasives, coupled with the increasing availability of high-resolution remotely sensed data, represents an opportunity to produce a distribution map for subsequent monitoring. The overall goal of this study was to develop spatially comprehensive maps of Tallow tree and Chinese privet in ecologically sensitive coastal regions, where both invasives have become well established. The study was conducted in three coastal sites within Alabama and Mississippi: (1) Mobile Tensaw River Delta, (2) Bon Secour National Wildlife Refuge, and (3) Mississippi Sandhill Crane National Wildlife Refuge. We implemented three image classification methods, representing unsupervised, supervised, and machine learning techniques, respectively: (1) ISODATA, (2) Maximum Likelihood (ML), and (3) Random Forest (RF). For each classification, a 1 m National Agriculture Imagery Program (NAIP) orthoimage was first examined, then integrated with vegetation structure and topography parameter derived from airborne light detection and ranging (LiDAR). The maximum Overall Accuracy (OA) of 87.5% was obtained using RF model with NAIP stacked image integrated with LiDAR derived variables. Overall, findings highlight the potential for accurately characterizing both Tallow tree and Chinese privet using readily available remote sensing data. Mapped products from this study represent a spatially comprehensive baseline inventory of crucial invasive species and will serve to inform the development of a framework for broader-scale mapping and monitoring efforts.