Skip to content Where Legends Are Made
Cooperative Institute for Research to Operations in Hydrology

Remote Sensing FIM Workshop

Remote Sensing Flood Inundation Mapping and Enhancement with High-Resolution DEM

Day 4 Session 1 (9:10 AM MDT)

Presenters:

Dan Tian
The University of Alabama

Sagy Cohen
The University of Alabama

This workshop includes two components, a Google Earth Engine web app that can automatically delineate flood inundation areas from satellite images using a pretrained multilayer perceptron (MLP) model and an algorithm to enhance the flood maps directly derived from remote sensing images.

Quickly mapping the flood is critical for rescue during a flooding event and assessment of property loss and damage due to the hazards. Google Earth Engine (GEE) provides a cloud-based geospatial processing platform for timely environmental monitoring and analysis. However, most current flood mapping applications on GEE rely on simple thresholding techniques or spectral index-based approaches. The GEE web app introduced in this workshop use a multilayer perceptron (MLP) model trained locally based on a global dataset. It can quickly and automatically delineate inundation areas based on Sentinel-1 Synthetic Aperture Radar (SAR) and Height Above the Nearest Drainage (HAND). The only user inputs are the interested date and location. This application can be used by both scientists and non-scientists, including those without knowledge and skills of remote sensing image processing.

Flood maps directly derived from remotely sensed images are often imperfect with various errors, owing to the landscape complexity and image separability limitation. Environmental factors, such as clouds, terrain, and tree shadows on the remote sensing images, often result in information gaps. Flooded tree canopy and buildings tend to be erroneously classified as non-flooded areas, causing serious omission errors on image-derived flood maps. In the second part of this workshop, we introduce a hydrologically guided region growing algorithm to enhance remote sensing flood maps by incorporating high-resolution Digital Elevation Models (DEMs). This hydrologically guided region growing algorithm can detect flooded forests and buildings, and fill out data gaps caused by clouds, and hence greatly improve the accuracy and reliability of the initial flood map derived from remote sensing imagery. This algorithm also produces a flood depth map in addition to the flood extent.

Learning Outcomes:

  • From the GEE web application component:
    • Understand the theoretical foundation of SAR flood mapping.
    • Gain experience on searching, filtering, and processing satellite and GIS dataset on GEE using the Code Editor.
    • Gain experience on developing an Earth Engine web app.
  • From the flood mapping enhancing component:
    • Understand the workflow of the hydrologically guided region growing algorithm.
    • Gain experience on working with user installed packages in ArcGIS Notebooks.
    • Run the python code to improve the FIM for an example case.
    • Understand the meaning of each step and parameter in the example and be able to apply it to other flooding cases.

Prerequisites:

Knowledge:

  • Having a fundamental knowledge of the Sentinel-1 SAR platform at the level of the ESA User Guides is preferred but not required.
  • Being familiar with ordinary Python syntax is preferred.
  • Fundamental GIS knowledge and skills is preferred.

Software:

Accounts: