Identifying Black Ash Wetlands Regionally using Visible and Synthetic Aperture Radar Imagery

Black ash (Fraxinus nigra Marsh) is a dominant canopy species in forested wetlands from Minnesota to Maine and south to Ohio. The future of these forested wetlands is in doubt due to the continued expansion of Emerald ash borer (EAB, Agrilus planipennis Fairmaire), an invasive insect first discovered in 2002. The ecological, geochemical, and management implications are the subjects of ongoing research. Expanding site-scale study results to regional impacts and identifying existing and sensitive black ash stands for pre-infestation management require detailed knowledge of the species' distribution. Existing maps of black ash are too small-scale to be used for on-the-ground management or for modeling the regional impacts of EAB on black ash wetlands. Accurate mapping at a finer scale requires plot-level forest inventory data from across the range of black ash and computationally intensive image processing and classification over a broad geographic extent. The scale of data collection and image processing required for this type of species mapping is often a barrier for ecosystem process and forest management studies, making evaluation of regional impacts difficult. To overcome these barriers we use existing spatially-referenced forest inventory data from a variety of sources, and perform our analysis using the Google Earth Engine geospatial analysis platform. Contributed datasets were standardized and combined into a master data set to be used for training and testing the classifier. Recorded sensor values and derived measures from visible, near-infrared (LANDSAT 7), and synthetic aperture (PALSAR) imagery were used from multiple periods within each growing season. Pixel-wise mapping was performed using models developed with partial least squares regression (PLS). The PLS models were used to identify pixels with forest canopy that is dominated by black ash. Following the pixel-wise classification cluster-analysis will be used to identify existing black ash stands. The immediate goal of our research is to generate a publicly-available map of black ash at a scale fine enough for use in management decisions and calculations of regional impacts of EAB. Our approach of utilizing crowd-sourced data and cloud-based processing may also serve as a template to improve species-level forest mapping.