Early detection of invasive Phragmites australis at the tidal marsh-forest ecotone with airborne LiDAR

• LiDAR-derived spatial metrics can detect understory invasive Phramites australis in forest; • These metrics include Mean distance, Point density, Scatter, Omnivariance, and Eigentropy; • Phramites australis detection algorithm we developed outperformed current map products; • Phramites australis map in North America can be generated with open source airborne LiDAR. Wetlands across North America are invaded by an introduced lineage of the common reed Phragmites australis , and sea level rise has exacerbated the spread of this species. P. australis at tidal marsh-forest ecotones has rapidly been expanding into deteriorating forest, colonizing understory environments ahead of native marsh species. Early detection of P. australis at the ecotone will be critical to the management of this invasive species in coming decades. In this study, we develop and validate a new method for early detection of P. australis , using open access airborne LiDAR data that can uniquely penetrate the tree canopy and detect P. australis within the forest understory. The method was designed for areas of sparse to moderate tree cover, such as the forest edge where trees are dying and P. australis is expanding, where understory species mapping was previously impossible with most spectral data. To differentiate P. australis from shrubs and other understory herbaceous plants, we tested the effectiveness of several LiDAR-derived spatial metrics, including Mean distance, Point density, Scatter, Omnivariance, and Eigentropy, as inputs to a Support Vector Machine (SVM) classifier, followed by a smoothing algorithm to avoid occasional obstacles or disturbances. We compare among metrics and single- vs. multiple- metric-based classifications. The resulting best early detection method of P. australis achieved a classification accuracy of 91.48% at the development site, and between 56.16% and 80.65% accuracy at other test sites. This algorithm provides a cost-effective and high accuracy method of detecting understory P. australis using public airborne LiDAR data. Larger-scale application of this method will provide coastal resource managers and policy-makers with distribution maps of P. australis through time in open environments and the forest understory. More generally, this approach provides a framework for mapping understory species and plant functional groups using LiDAR-derived metrics.