Day 4

Detailed paper information

Back to list

Paper title Including hydrology in satellite-derived flood mapping
  1. Arjen Haag Deltares Speaker
  2. Gennadii Donchyts Deltares
  3. Kel Markert
  4. Amanda Markert
  5. Ate Poortinga Spatial Informatics Group
  6. Farrukh Chishtie Spatial Informatics Group
  7. Timothy Mayer University of Alabama in Huntsville, NASA SERVIR Science Coordination Office
  8. Chinaporn Meechaiya Asian Disaster Preparedness Center
  9. Nyein Soe Thwal Asian Disaster Preparedness Center
  10. Peeranan Towashiraporn Asian Disaster Preparedness Center
Form of presentation Poster
  • A7. Hydrology and Water Cycle
    • A7.01 Inland Water Storage and Runoff: Modeling, In Situ Data and Remote Sensing
Abstract text Satellite remote sensing is an effective approach to monitor floods over large areas. Ground-based gauges remain a vital instrument for monitoring water levels or streamflow, but they cannot capture the spatial extent of a water body or flood. Numerical models can be an excellent source of such information, but are not readily available in all regions and can be costly to set up. Satellites already orbit and monitor nearly all regions of the globe and can thus provide relevant information where other sources are lacking. However, while earth observation has many advantages, there are also data gaps and challenges, which can be different for each specific sensor.

Flood mapping studies and applications often use imagery from optical, e.g. MODIS, Landsat, Sentinel-2, and/or synthetic aperture radar (SAR) sensors, e.g. ALOS, Sentinel-1. SAR’s cloud penetrating capability is especially important for flood mapping, as clouds are often present over (inland) floods, because these are triggered by rainfall originating from clouds. ESA’s Sentinel-1 constellation has for the first time in history made it possible to provide reliable flood mapping services on a large (even global) scale. The synergistic use of optical imagery can help overcome some of SAR’s known issues regarding flood mapping (such as signals resembling that of water over sandy soils and/or agricultural fallows), as well as help provide more timely flood maps, essential for disaster response and relief efforts. Still, current satellite-derived flood maps are not perfect and under- and overestimations of flood waters are to be expected. This is especially true for areas under thick vegetation canopies, as both optical and (most) SAR sensors cannot penetrate these, and urban areas, where signals can be distorted and data from the freely available satellites mentioned here don’t possess the spatial resolution required to accurately map water between or within urban features.

The HYDrologic Remote sensing Analysis for Floods (HYDRAFloods) tool is already using multiple sensors for the improved capabilities mentioned above, with current research focusing on data fusion of optical and SAR imagery as well as the inclusion of hydrologic information. Hydrology plays an important role in the general water cycle, influences floods and can also be used to constrain or improve satellite-derived flood maps. Low soil moisture values in sandy soils and areas of agricultural fallows can be used to prevent false positives derived from SAR imagery. Hydrologically-relevant topography information can be used in a similar fashion, but also to identify potentially flooded areas that are otherwise obscured from satellite imagery, such as under forest canopies. For this, we link the flood maps to hydrologically connected surface water flow paths.

HYDRAFloods is under active development in the SERVIR-Mekong program, covering a large part of Southeast Asia, by ADPC, SIG, SEI and Deltares, supported by NASA and USAID. It is used operationally by the United Nations World Food Programme (WFP) in Cambodia, being made available in their Platform for Real-time Impact and Situation Monitoring (PRISM), and was field tested during the severe floods that hit the country in October 2020. HYDRAFloods embraces open science and combines relevant algorithms from literature with our own custom developments, which are published in open access journals. It runs on the Google Earth Engine platform to facilitate easy data access and running at scale across the entire South East Asia region. The code itself is hosted on an online repository with open source license, including up-to-date documentation.

HYDRAFloods has been described in general at other conferences, so we will only give a brief overview and instead focus on recent research on including hydrologically relevant information in the processing chain to obtain more accurate flood maps. We hope this can lead to a fruitful discussion on the underlying techniques and assumptions, as well as contribute to a broader discussion on combining data from various sources (e.g. in-situ, models, EO) and its best practices.