Informing Hurricane Flooding and Sea-Level Rise Vulnerability

Submitted by tadamson on

Fusing remote sensing products from different satellite sensors allows the development of enhanced maps of the current distribution of coastal wetland plants and more accurate models of coastal elevations. This critical information about vulnerability to sea-level rise and hurricane flooding is being used by DOI partners and State and local agencies to improve management in a changing climate. In 2021, USGS scientists developed a 10-meter resolution vegetation classification model using wet- and dry-season imagery from Sentinel-1 (SAR data) and Sentinel-2 (multispectral data) acquired from 2015–2017 (before Hurricane Irma in September 2017). The model successfully distinguished mangroves from grassland habitats (88%) with a mean overall accuracy of 81%. The results show where mangroves have expanded their inland range into fresh and salt marshes across National Park Service and U.S. Fish and Wildlife Service lands, impacting important habitats.

Coastline elevation is an important characteristic for understanding floods in tidal wetlands. Even small biases in elevation can translate to large differences in inundation time, which is a crucial metric for estimating future flooding vulnerability.  Elevation data are usually derived from airborne lidar, but dense vegetation blocks the lidar signal from reaching the soil surface, causing a positive bias in bare-earth digital elevation models (DEMs). Using a statistical model (LEAN: Lidar Elevation Adjustment using NDVI) and ground calibration datasets, USGS scientists developed a new DEM for southwest Florida that accounts for the bias from vegetation. The LEAN model assumes that plant density is correlated with the Normalized Difference Vegetation Index (NDVI), calculated from National Agriculture Imagery Program (NAIP) imagery. LEAN effectively calibrates lidar-derived DEMs to the input survey-grade ground elevation data. The use of the LEAN model has resulted in a 51% improvement in DEM accuracy.



Digital elevation models (DEMs) are calculated using airborne lidar, but these data can be biased where vegetation is dense. Using the LEAN model, researchers were able to correct for this bias and generate a more accurate DEM for southwest Florida. A separate vegetation classification model developed using remotely sensed imagery allowed the researchers to generate a vegetation map for the same region.

Author Name
Karen Thorne
Author Email