Member Login Menu

Global Satellite Coverage Improves Flood Mapping

By Catherine A. Cardno, Ph.D.

The increase in satellite coverage surrounding the globe has made it possible to use existing data sets to create flood maps quickly and inexpensively.

featured image
Among the images from the USGS’s Landsat program that the professors are working with is an image of Byrd Glacier, a 15 mi wide, 100 mi long ice stream in Antarctica that cuts through a valley in the Transatlantic Mountains. Courtesy of USGS/Landsat

 January 5, 2016—Creating historical flood maps of regions is typically a costly, time-consuming process. But, as is true of so much, new technology has streamlined the process, improving the speed and lowering the cost. Now, researchers at Deltares—the Delft, the Netherlands-based applied water and subsurface research institute—have created a new method for mapping historical river flooding. The ever-increasing number of satellites encircling the globe have captured a rich treasure trove of free images that can now be analyzed to create regional flood maps—with details as small as 30 m is size—that date back to 1999.

The river flood mapping capability is the work of Gennadii Donchyts, a researcher and senior consultant, and Jaap Schellekens, Ph.D., an expert in hydrological modeling, watershed management, and flood forecasting, both with Deltares. As part of his Ph.D. research in the field of civil engineering at the Technical University of Delft, Donchyts works with the free remote sensing images available from NASA as well as from two projects administered by the United States Geological Survey (USGS): Landsatand the Shuttle Radar Topography Mission. Donchyts created an algorithm that married the data from these data sets to create site-specific, historical flood maps of rivers and their surroundings. 

The data sets are available via Google Earth Engine, a system that has eliminated the need to download and locally process data. Instead, the data is stored and the calculations are completed using Google's considerable computing power. 

Donchyts' work built upon his 15 years of prior experience with surface and subsurface modeling. "In most cases, special skills are required to set up these kinds of models, but also, a lot of input data need to be gathered and converted into proper formats," Donchyts explained in written answers to questions posed by Civil Engineering online. "My main goal is to significantly speed up this process of model setup, [and] it appears that satellite imagery can be perfectly used to achieve this," he said. 

The Landsat data set contains millions of images at 15 to 30 m resolution that date back more than 40 years from which surface water dynamics can be extracted, according to Donchyts. "The main challenge was to process that huge amount of data (about 1 petabyte) using existing software tools," he noted. Using the Google Earth Engine to process the calculations from his new algorithm made it possible, however. "The platform allows us to perform planetary-scale geospatial analysis without the need to download data locally," Donchyts explained. "Everything is processed in the 'cloud,' allowing researchers to focus on their algorithms." 

Schellekens, who also responded in writing to questions posed by Civil Engineering online, explains it this way: "We bring the calculations to the data instead of downloading the data locally. The probabilistic approach we use here was simply not possible before the introduction of Google Earth Engine, as the amount of data to process is just too large for most institutes," he said. Creating the historical river flood maps in this entirely new manner takes just one-tenth of the time that it previously took to create regional flood maps, according to Schellekens. 

The resulting flood maps are "not based on single images but are a probabilistic approach based on a time series of images," Schellekens explained. "As such, the flood maps we create are maximum flood maps over the time period selected."

Donchyts tested and refined his algorithm by integrating it into two existing projects that Schellekens is leading. This includes the SERVIR-Mekong project launched by NASA and the U.S. Agency for International Development (USAID) to improve regional environmental monitoring in the five countries of the lower Mekong region in Southeast Asia, for which Schellekens is the project lead for Deltares. Tests were also carried out on data from the Murray-Darling basin in Australia as part of the European Union-funded global water resource observation program, earthH20bserve, for which Schellekens is the project coordinator.

The SERVIR-Mekong project is developing a better understanding of seasonal flood pulse dynamics, and "providing technical inputs to hydropower dam siting, design, and operation in Cambodia and Laos with the objective to minimize the ecological impacts of these dams on fisheries, sediment transport, and other aspects of the river system's complex hydrology and ecology," Schellekens explained. The work in the Murray-Darling basin in Australia is detecting and establishing permanent water as a prerequisite for later floodwater detection work, according to Schellekens.

"Once a person has access to Google Earth Engine, the scripts, and the supporting data sets, the process of creating maps is a simple matter of zooming to the area of interest using the Google interface and setting the period for analysis," Schellekens said. "After we have done the preparations—creating a number of supporting data sets such as a Height Above Nearest Drain map—the actual mapping can be done nearly instantly." 

While the data can vary through the years because newer satellites tend to have better sensors and older data sets can have gaps, the process is still far easier—and faster—than was previously possible and the maps provide a greater level of detail. As remote sensing is increasingly used for earth observation services, Donchyts said, that level of detail will improve dramatically. "The number of free and commercial satellites is continuously increasing and very soon—thanks to startups like Planet Labs[based in San Francisco] and SkyBox Imaging[in Mountain View, California]—it won't be a problem to get a 3 m resolution image every day for any place on the Earth," he said.

Additional testing and refinement of Donchyts' algorithm will take place as additional regions are mapped. "At present, we are working on a better accuracy of our method of water detection," Donchyts noted. "One of the possible extensions is to develop an automated data-fusion algorithm, which will combine imagery from multiple satellite products," he explained. "Additionally, we're trying to develop tools which can be used for more detailed studies, such as observation of the surface water changes in reservoirs, [and] quantification of long-term morphodynamic changes in rivers and coastal areas." In the long term, the pair plans to analyze surface water changes by integrating the available remote sensing images with locally gathered observation data. 

The flood mapping products have many uses. In addition to site-specific construction, environmental, or preservation projects, the maps can also be used to create communications programs to educate stakeholders and decision makers about how river dynamics change over time, Schellekens noted. The visual results enabled by drawing from the data sets' images can be striking. Such is already the case with the time lapses from around the worldcreated by Google.



Read Civil Engineering magazine on your smart device: download our apps.

app store play store