Abstract
Invasive plant species remain one of the largest threats to ecosystems, and a major limitation to predicting and managing invasives is the lack of updated and spatially relevant species distribution data. Remote sensing has emerged as an option for providing wall-to-wall mapping of species distributions, of which the most successful approaches employed hyperspectral imagery primarily limited to airborne platforms. We explore the newly launched hyperspectral Earth surface Mineral dust source InvesTigation (EMIT) mission on the International Space Station (ISS) to detect invasives compared to the multispectral satellites Sentinel-2, Landsat 8, PlanetScope SuperDove, and imagery from the National Agricultural Imagery Program (NAIP). We focused on detecting the presence of Brazilian peppertree ( Schinus terebinthifolia ) in Everglades National Park in February 2024 using a pixel-based support vector machine classifier, comparing spectral and spatial resolutions. With a 60 m spatial resolution and 223 spectral bands, EMIT imagery enhanced invader detection, achieving an average overall accuracy of 80.2% and captured the known distribution of Brazilian peppertree the best. Sentinel-2 resampled to 60 m spatial resolution had the next highest accuracy with 78.8%, but failed to capture the known distribution. The classifications show that additional spectral bands, rather than spatial size, yield improved results when detecting this shrub/small tree invader, especially at locations with lower invader density, which are often a target for Early Detection and Rapid Response (EDRR) management efforts. There is still spectral confusion, most likely due to the limited phenological differences between the invader and native communities at the time of year when the imagery was collected, which could be potentially overcome using imagery from the fall or time series. The most recent work detecting invasives using spaceborne hyperspectral imagery was concluded with the decommissioning of Hyperion in 2017, so the launch of EMIT revitalizes opportunities for plant invader detection from space. With the irregular orbit of the ISS, EMIT imagery may not capture key phenological events that would increase the spectral separation between invaders and native communities. However, optimally timed collection will be possible from future global hyperspectral missions due to consistent image revisit times, which will significantly improve detection capabilities by capturing the dynamic nature of invasives. This work showcases that the new wave of hyperspectral spaceborne sensors can improve invasive plant detection even at coarse spatial resolutions, and significantly improve detection for early detection and rapid response management efforts.