Hyper-Spatial Multispectral Analysis of Tornado Damage in the High Plains

Multispectral analyses in satellite imagery have proven useful in detecting vegetation damage that would be otherwise undetected. The ability to detect damage in rural areas and to vegetation, however, is limited by spatial and temporal resolution, especially at lower values of the Enhanced-Fujita (EF) scale. Unpiloted Aerial System (UAS) multispectral imagery can better capture storm damage and variability of damage to vegetation because of the hyper-spatial (centimeter scale resolution) information collected in red, red-edge, and near-infrared spectral bands. This research evaluates UAS-based multispectral and true-color analysis in detecting tornado damage in the High Plains, U.S.A. We obtained visible and multispectral imagery of the 28 May 2019 Tipton, KS tornado (EF-2 rating with a path length of 24 miles and width of 0.5 miles) using two UAS platforms respectively: DJI Phantom 4 Pro with a true-color camera, and a DJI Inspire 2 equipped with a Micasense RedEdge MX multispectral camera. We focus on three aspects of the track: beginning, end, and hotspots with the heaviest damaged areas based on ground survey data by National Weather Service (NWS) collaborators. Our results show that multispectral analysis (NDVI) can detect a larger portion of the track and can be used to develop damage indicators that are more reflective of tornado intensity. Multispectral analysis could improve damage detection in rural locations, especially in the High Plains, which have well-documented reporting biases due to low population densities, relatively inaccessible areas, and limited damage indicators for stressed vegetation.