Photograph of flooded post office  

High Water:

Building A Global Flood Atlas

  • April 6, 2005
  • by Rebecca Lindsey
  • design by Robert Simmon
  • adapted from ‘Flood Hunters,’ by Laura Naranjo, in Supporting Earth Observing Science, 2004


The small plane droned loudly as it made its way from western Minnesota toward St. Louis in June 1993. Through his window, Dartmouth College geologist Bob Brakenridge glimpsed the white dome of a grain silo that seemed to float like a bleached lily pad on a muddy, inland sea. Everywhere he looked water sloshed up the sides of houses, swallowed fields full of early-season crops, transformed roadways into canals. It was tempting to just stare out the window and gawk at the greatest flood in the recorded history of the Upper Mississippi.

  Page 2
  Aerial photograph of a flooded power plant along the Mississippi River, July 1993.

But he hadn’t bought eight hours of flying time just for sightseeing. He was a scientist on a mission. Brakenridge, a geologist who specializes in studying the shape and structure of the topography produced by rivers, wanted to see how well satellite imagery of the floods matched up with the conditions on the ground. The idea had been floating around in the back of his mind since 1991, when he had spent a year as a visiting senior scientist at NASA Headquarters in Washington, D.C.

All day he had been looking for boundaries—locations where the line between flooded areas and dry land was obvious. He directed the plane’s young pilot to an area that looked promising. A powerline whizzed by the window as they dropped down from 900 feet for a closer look. He found their current location on a grainy radar image, and nodded with satisfaction at the pattern that was emerging: against the lighter colors of dry land, flood waters stood out in black.


The great floods of 1993 inundated 80,000 square kilometers (20,000,000 acres) of land along the Missouri and Mississippi Rivers. This aerial photograph of a flooded power plant along the Mississippi only hints at the extent of the disaster. Because of the vast area covered by water, scientists turned to satellite remote sensing to map the floods. Remote-sensing techniques developed to study the 1993 Mississippi floods are now used to map floods worldwide. (Photograph courtesy FEMA photo library)

  Radar image of flooded areas near the confluence of the Des Moines and Mississippi Rivers

“When that flood began, so many scientists just dropped whatever they were doing and focused on the Mississippi,” Brakenridge recalls. “I contacted NASA and asked for help in getting imagery from ERS-1 [European Remote-sensing Satellite-1], and then I quickly applied for a small grant from the Association of American Geographers. A week later, I was in a plane flying over the Upper Midwest.”

His study of the Great Mississippi Flood of 1993 with remote-sensing data was the launching off point for what has become Brakenridge’s passion for the last decade: creating a satellite-based, global atlas of flood hazards and putting the information into the hands of disaster relief and planning organizations. Brakenridge is motivated by the hope that current information will help people cope with today’s flood disasters, but also by the hope that as his flood archive covers more of the globe over a longer time, patterns will emerge that will help people prepare for floods more confidently.


Geologist Bob Brakenridge used radar data to study areas covered by floods. Waterways and flooded areas are black in this radar image, which shows the region near the confluence of the Des Moines and Mississippi Rivers on July 16, 1993. (Image courtesy Dartmouth Flood Observatory)

  Map of major flooding, March 21-28, 2005

Predicting precisely when a flood is going to occur or how severe it will be is mostly a matter of assessing the current weather and environmental conditions in an area, and that isn’t the kind of information you can get from an archive. “What a global archive of flood information can do,” says Brakenridge, “is to help disaster management planners and re-insurance companies to constrain flood hazards, giving a clear indication of what areas are susceptible to floods, how severe those are, and how often severe flooding returns to the same area.” [See Sidebar: Teleconnections]


Satellite sensors allow Bob Brakenridge to map floods on a worldwide basis. Red areas correspond to regions that experienced significant flooding in the last week of March 2005. (Image adapted from the Dartmouth Flood Observatory)


Teleconnections: Long-distance Relationships


Predicting the flooding of a specific river several months in advance might never be possible, but a global, long-term archive of flood events could reveal possible teleconnections between flooding events in different locations. Teleconnections are climate anomalies that are related but often widely spaced in distance and/or time.

  page 1
  Map of El Nino's teleconnections

The relationship between the two climate patterns is not necessarily cause and effect. Often both out-of-the-ordinary climate phenomena are caused by some third factor, such as when El Niño events increase the chance of above-average precipitation in the U.S. Southwest from January through March and increase the chance of drought in Indonesia from June through August.

Brakenridge said that a few teleconnections between flooding events seem to be emerging from the archive, but that they need more time and data to establish a solid correlation. “We also see somewhat predictable locations of where floods occur during El Niño years,” he added. If the search for teleconnections is successful, the relationships could one day be used to make regional predictions of seasonal flooding several months in advance.


The El Niño/Southern Oscillation is the most prominent of Earth’s teleconnections. The eastern Pacific (left of image center) and the western Pacific (right of center) are connected over thousands of miles by an inverse relationship in which the central and western Pacific off South America get much more rain than normal during El Niño events, while the eastern Pacific between Southeast Asia and Australia often suffers drought. The reverse pattern occurs during La Niña years. (Image courtesy JPL TRMM Team)


Hurdles and Milestones


Brakenridge’s success with the study of the Great Mississippi Flood in 1993 paved the way in 1995 for a NASA grant to study “extreme hydrological events” with remote-sensing data. Brakenridge says that before this activity there were only a few ad hoc, commercial flood-hazard applications. The Dartmouth Flood Observatory, as he named the project, was the first providing a global perspective.

In the early stages, says Brakenridge, “The project was a preservation activity—just like when you go into towns and you see where people have put signs of high water marks on stream gauges and buildings to preserve the flood history—but we were using satellite data, whatever we could get our hands on.”

  page 1Page 3
  Satellite images showing the Mississippi floods of 1993 compared to normal conditions in1991

Knowing what to do with the satellite data once he had it was another matter. “When the current generation of Earth-observing satellites was being planned, people pretty much assumed that runoff from rivers and streams was just too small to be observed from satellite,” explains Brakenridge. “Everyone just assumed that was one part of Earth’s cycles that we would have to understand through models.”

In the absence of specific sensors and technologies intended for flood mapping, Brakenridge and his team—usually no more than a graduate assistant or two—have been pioneers, evaluating available satellite resources and tinkering with the kinds of imagery that they could make from each. During the life of the project, they have raided the immense storehouse of information at several of NASA’s specialized data centers: Earth Sciences at Goddard Space Flight Center, Land Processes in South Dakota, and Snow and Ice in Boulder, Colorado. They have used satellite observations from a host of NASA and National Oceanic and Atmospheric Administration (NOAA) satellites, including the original efforts with the European Remote Sensing satellite, Landsat, QuikSCAT, Terra, Aqua, and the NOAA-NASA Pathfinder.

“When I think about the work I’m able to do, it is really amazing. All scientists get to use special instruments, lab or field instruments, for example. But NASA has these instruments in space that cost hundreds of millions of dollars, and we get to use them! Sometimes we can have one of them turned on for us,” he says, as though even after all this time, he can’t believe his good luck.

“NASA has really been leading the way [in Earth observation], especially with their open data policy. It’s really exciting, taking these systems that haven’t really been designed for us and having the access to the data and the opportunity to figure out how they can help us.”


Scientists at the Dartmouth Flood Observatory use data from many satellites to map historical floods. These images show the confluence of the Mississippi, Illinois, and Missouri Rivers just upstream of St. Louis. The upper image shows the rivers in normal summer conditions (August 1991), while the lower image shows the rivers at the end of the great floods of 1993 (August 1993). (Image by Jesse Allen, based on data provided by the Landsat Science Team)


Flooded With Information


The real explosion in satellite-based flood information came in 2000, when the Moderate Resolution Imaging Spectroradiometer (MODIS) on NASA’s Terra satellite began collecting 250-meter-spatial-resolution imagery of almost the entire globe every day. A second MODIS instrument was launched on NASA’s Aqua satellite in 2002. Although MODIS wasn’t specifically designed to observe floods, Brakenridge says it’s turned out to be the perfect tool.

“Before the era of satellite remote sensing, and particularly before MODIS, there was no way to map flooding over large regions,” he says. The limitation is due to the inescapable trade off between the level of detail a satellite can provide and the size of the area it can observe at one time. As the level of detail goes up, the area covered goes down. For large floods, says Brakenridge, “A satellite with high spatial resolution isn’t very useful if it means you need 30 scenes to get a complete image of the flooded area.”

  page 2Page 4
  Wide-scale satellite image of floods in China, with area covered by high-resolution sensor for comparison

MODIS provides the right balance of coverage and detail, showing large floods as they are happening all over the world. Brakenridge uses the imagery provided by the MODIS Rapid Response Team, a group at NASA’s Goddard Space Flight Center that processes and distributes global MODIS imagery and customized products in near-real time. He uses natural-color imagery that looks like a digital photo and a kind of false-color imagery that adds color to non-visible wavelengths of energy detected by MODIS, such as infrared. Standing water absorbs infrared wavelengths of energy, so in the imagery it appears almost black.

Sometimes even those kinds of image enhancements aren’t enough. After all, where there are floods, there is rain, and where there is rain, there are clouds. Son Nghiem of NASA’s Jet Propulsion Laboratory at California Institute of Technology helps Brakenridge accommodate those unwelcome visitors in many flood-related images. He processes radar data collected by NASA’s QuikSCAT satellite, which sends out microwave pulses that can penetrate clouds. The timing and strength of the return signal reveal the surface beneath. Brakenridge combines those observations with the images from MODIS and other sensors to map floods even when the clouds linger for days or weeks.

Even when they can see through the clouds to see how big an area is underwater during a flood, Brakenridge points out that they are still missing a key piece of information: depth. Recently Brakenridge has been investigating how to use topographic information collected by radars on the Space Shuttles to get an idea of how deep standing water might be. Brakenridge has been combining the high-quality topographic maps produced by the Shuttle Radar Topography Mission (SRTM) with the flood imagery from MODIS and other sensors. The SRTM maps tell him the altitude of various features on the surface, and he can gauge how high the water is by what features have been covered up.


The wide area seen by sensors like MODIS allows flood monitoring over the entire globe in near-real time, while high-resolution satellites such as Landsat can zoom in to a specific area. This partial MODIS image shows flooding along China’s Huai River on July 23, 2003. The white outline shows the smaller area covered by a single Landsat scene. (Image by Jesse Allen, based on data provided by the MODIS Rapid Response Team)

  Topographic map of the Tondano River, with flood extent  

Brakenridge and his colleagues determine the depth of flood waters by matching maps of flooded areas (blue) with topographic data (gray shades, lighter is higher elevation) that reveals the elevation at the water’s surface. The graph below the image shows the topography (dark blue line) and depth of flooding (light blue fill) along the cross section of terrain marked with a white line in the top image. Scientists derive flood depth by matching the water extent visible in recent satellite images with the topography along the edges of the flooded area. (Image courtesy Dartmouth Flood Observatory)


Knowing What’s Normal


Because of MODIS’ near-daily observations of every spot on Earth, Brakenridge has the opportunity to construct baseline maps of watersheds when they are not flooded. “It’s important to have a standard reference so that when we map a flood, we don’t mistakenly include areas that are normally under water, such as lakes, reservoirs, or wetlands,” he explains. [See Sidebar: Monitoring Wetlands.]

Brakenridge is also interested in documenting another normal reference: the normal flood. One thing you notice when you look through the Dartmouth Flood Observatory’s archives is that there are places on Earth—for example, northern India, China, and Bangladesh—where large-scale flooding is surprisingly regular. As more people crowd into flood-prone areas the normal cycles of flooding can begin to be seen as dangerous or unusual. “We don’t want to make a big fuss about major flooding along a particular river when, in fact, it floods every year,” he says.

  page 3Page 5
Graph of annual flooding in Bangladesh

Sharing What They Know

Monitoring the floods and creating the global atlas of flood hazards is only half of the Dartmouth Flood Observatory’s mission. They also want to get their information into the hands of disaster management planners and relief agency workers as quickly as they can when a disastrous flood occurs.

“Within large parts of Africa, Asia, and South America, there are no flood warning systems,” says Brakenridge. Getting the word out to aid organizations that this online resource is available to them, however, wasn’t an easy process. Brakenridge has now teamed up with the AlertNet Website to help get flood maps to the international aid community. Established by the Reuters Foundation, AlertNet provides news and information to international humanitarian aid and disaster relief organizations.


Monsoon rains bring floods to low-lying Bangladesh every summer. Although the amount of land covered is large and the damage is often severe, annual flooding is the normal behavior of the Ganges River Delta. This graph shows how the amount of standing water across the region peaks and ebbs from season to season. The surface area covered by standing water was measured by the QuikSCAT satellite from mid-1999 until the end of 2004. (Graph courtesy Dartmouth Flood Observatory)

  High-resolution satellite image of the Betsiboka River in flood during March 2004

Even though satellite data are freely available from NASA, humanitarian relief agencies often don’t have the expertise or resources to directly download and process data. AlertNet presents satellite data as imagery that is easy to understand and more meaningful to emergency managers. “It has taken a long time for the international aid community to recognize the value of satellite imagery, but this is beginning to change because of services such as AlertNet,” says Oliver Greening, a data visualizer with ESYS Consulting who has been funded by the European Space Agency to produce satellite images for AlertNet.

Brakenridge hopes that making flood maps publicly available online in near-real time will overcome technological, financial, and infrastructure barriers in developing countries that prevent aid agencies there from taking advantage of satellite data.


Satellites observe the world globally, enabling relief workers to monitor floods in remote areas, such as Madagascar’s Betsiboka River. This image shows flood water (turquoise) spilling out from the river’s channel on March 10, 2004. Vegetated areas are red and bare ground appears tan. (Image by Jesse Allen based on data provided the NASA/GSFC/MITI/ERSDAC/JAROS ASTER Science Team)


Monitoring Wetlands


Monitoring wetlands was a natural expansion of the Dartmouth Flood Observatory effort to document the background state of rivers. Wetlands are part of a river’s natural flood control system, taking in water and then slowly releasing it. Some wetlands are permanent, but others are subject to regular seasonal flooding. The distinctive types of soil and vegetation in seasonal wetlands affect not only local or regional biogeochemical cycles, but global ones as well.

  page 4
  Satellite image of the Tonle Sap Wetlands during the wet season   Satellite image of the Tonle Sap Wetlands during the dry season

Wetlands store large quantities of carbon in both vegetation and in their mucky soil, which means that as they change—getting bigger or smaller, or warming up and drying out—their influence on the global carbon cycle changes. Also, wetlands are the largest source of methane emissions worldwide, and methane is a powerful greenhouse gas. How much methane wetlands emit can vary depending on how flooded they are. Keeping tabs on where and when seasonal wetland flooding is underway will help scientists create more realistic models of the role of naturally occurring greenhouse gases in the climate system.


The Tonle Sap Wetlands in Cambodia experience dramatic seasonal change. During the wet season (left) water carried from the interior of Asia overflows Tonle Sap Lake. Several months later, the water recedes, fertilizing the soil. (Images courtesy MODIS Rapid Response System)


Returning To His First Love


It’s not hard to imagine how rewarding this work must be to Brakenridge. In the summer of 2004, he was contacted by the U.S. State Department, who had seen some of his images on the Internet of the massive flooding in late July and early August that had covered nearly two-thirds of Bangladesh with water.

“They wanted a high-quality annotated version of the map, which we quickly made and printed out. We delivered the map, which quickly made it into the hands of the Ambassador and into a diplomatic mail pouch. Within two days, it was in the hands of aid workers in Bangladesh,” said Brakenridge.

  page 4
  Map of flooding in Bangladesh during 2003

“Now I am a scientist,” he continued, “and as a scientist I know that I am judged successful or not by my peers in the traditional scientific manner, through publishing research in journals and speaking at conferences. But personally, to know that the maps we made have helped some person or group in a time of crisis is just as rewarding to me as the more traditional measures of professional success.”

“It’s kind of funny, actually, my doing this work. I grew up in the 60s, and the Apollo program was underway. I was very taken by space. In fact, my first desire was to be an astronomer, but instead I became a geologist. I went from wanting to study things ‘way out there’ to having my nose pressed into the dirt. Now I am using space-based technology to observe these ground-based events. It’s like coming full circle.”


Maps provided by the Dartmouth Flood Observatory to aid organizations like the U.S. State Department and AlertNet enable non-technical people to utilize sophisticated satellite data. This map shows flooded areas detected by MODIS in 2003 (red shades) combined with previous years’ flooding (blue shades). Light blue areas are permanent water bodies. (Map copyright Dartmouth Flood Observatory)


Currently the Dartmouth Flood Observatory receives financial support predominantly from NASA, but they also have some small grants from the United Nations World Meteorological Organization and also from NATO’s Science for Peace Programme.

“Today I have two young scientists from Romania across the hall working with my research assistant. They had never been to the U.S. before. They will be here 10 days, courtesy the NATO grant, working out how to integrate flood mapping techniques we developed here into their prediction and flood disaster response efforts.”

“I do worry very much about continuity of this project, dependent as it is on grant support. On the other hand, given the degree of interest that has been shown from a long list of nations, agencies, NGOs [non-governmental organizations], and academic institutions, I do feel confident that this effort will survive. I just never realized it would become my full-time job making sure that happens,” Brakenridge said. It’s not a job he expects to quit anytime soon. Despite the worries, he says, “I absolutely love this work.”