RUSH: Rapid remote sensing updates of landcover for storm and hurricane forecasts

Research has shown that vegetation cover in coastal settings significantly controls flooding, erosion, and barrier island breaching during extreme storms through spatially variable wave energy dissipation. Coastal modelers require simple and fast ways to obtain up-to-date high resolution coastal vegetation cover for modeling coastal impacts.

To address this critical need, a tool for production of rapid-repeat high-resolution coastal vegetation maps has been developed. A Jupyter Notebook Application and a Graphical User Interface use Planet Labs Super Dove8-band, 3-meter multispectral imagery and a machine learning classification model to deliver high-resolution maps of coastal vegetation showing near real-time conditions. The remote sensing modeling approach employs a reliable random forest machine learning technique for image classification that is transferable across a large geographic region. The model has 93% overall accuracy and is trained on project areas within three states (North Carolina, Louisiana, and Florida). The application consists of two modules: the satellite imagery download module automatically downloads Planet imagery and the remote sensing module generates coastal vegetation raster data products using the pre-trained machine learning algorithm.

For further information and access to the software, visit this webpage.

Cheang, C.W., Byrd, K.B., Enwright, N.M., Buscombe, D.D.,and Gesch, D.B., 2024, A Tool for Rapid-Repeat High-Resolution CoastalVegetation Maps to Improve Forecasting of Hurricane Impacts and CoastalResilience (Version 1.0.0): U.S. Geological Survey software release, https://doi.org/10.5066/P13FJCK5.