Predicting Where Flooding Will Occur In The West

The new three-pronged approach combines anew numerical computer model with two additional methods,satellite-image analysis and field observations. Each method serves asa check on the other two. The research team focused on the floodplainsfor a 10-year flood, a 100-year flood and a maximum flood for two sitesin Arizona.

In addition to providing better hazard information tothe public, revising the floodplains maps could have major economicimpact in the rapidly growing Southwest. Often, homeowners in areasdeemed to be in a floodplain must buy flood insurance in addition toregular homeowner's insurance.

"I think the important thing isthat we have three methods that give darn near the same result, andit's a way smaller floodplain than the FAN model FEMA has generallyused," said research team leader Jon D. Pelletier, an associateprofessor of geosciences at The University of Arizona in Tucson. "Thesethree independent methods converge on the same answer. … That wasreally surprising to us."

The combined method applies to thefoothills of western mountain ranges such as the Santa Catalinas andthe Tortolitas outside Tucson. Many western cities, including Phoenix,Las Vegas and Denver, have similar foothills.

“You have a lot ofsmall channels draining the mountains, and the floodplains are verycomplex,” said Pelletier. “Back in the day when data that mapped thesesmall channels weren’t available, FEMA called it all floodplain. That'sfine in the Lower Mississippi Valley, but it’s not appropriate outhere. These three methods give us a new level of detail. We’re lookingat a scale of five to 10 meters and can see that the area of thelandscape that’s prone to flooding is often really small. I predict ourresult would apply to many other foothills regions in Arizona.”

Pelletier,Larry Mayer, a UA adjunct professor of geosciences, and Philip A.Pearthree, a research geologist with the Arizona Geological Survey inTucson, and their colleagues have published their findings in thecurrent issue of GSA Bulletin, a publication of the Geological Societyof America. A complete list of authors and the reference for the papercan be found at the end of this release. The research was supported bythe National Science Foundation, the Pima County Flood ControlDistrict, the Arizona Geological Survey and the Flood Control Districtof Maricopa County.

In the region of the West characterized bytowering mountains interspersed with broad, flat valley floors, theland sloping away from the mouth of a mountain canyon, known as thealluvial fan, is criss-crossed by a myriad of small, shallow channelsthat drain into the valley below. Such topography is found in Arizona,Utah, western Colorado, western New Mexico and parts of eastern andsouthern California.

Even in a major flood, some upland areasbetween channels remain dry. However, traditional techniques forassessing the potential extent of floods on alluvial fans oftendesignate those upland areas as flood-prone, thereby overestimating theamount of land in danger of flooding.

Pelletier was embarrassedwhen he had to tell students that such assessments werestate-of-the-art. So he and his colleagues decided to devise a betterway.

To create a computer model to predict flood intensity,Pelletier used very detailed maps of alluvial fans, data from streamgauges and a mathematical analysis that predicted how the water flowedthrough the numerous small channels on a given alluvial fan during agiven storm.

For maps, Pelletier used digital elevation models(DEMs), which are computer-generated maps made from low-altitude aerialphotographs that can show changes in elevation of only 4 inches (10cm). Such maps are available for Wild Burro Canyon in the TortolitaMountains outside of Tucson, Ariz. and Tiger Wash in the HarquahalaMountains west of Phoenix. Because of the complex topography of thosetwo sites, land that hasn’t flooded in thousands of years can beseparated from the active floodplain by an elevation difference of onlyone to two feet. Both areas also have stream gauges or fieldobservations that record the volume of water during past floods.

Pelletierplugged the data for two historic floods -- the September 1997 flood inTiger Wash from tropical storm Nora and the July 1988 flood in WildBurro Canyon from monsoon rains -- into his mathematical model. Thecomputer then created maps that predicted where the waters from thosefloods went.

Pearthree and Mayer had already done detailedstudies of how floods affected those sites. Mayer had developed amethod that uses LANDSAT satellite images to document changes thatoccurred on the landscape during floods. Pearthree had used traditionalfield-based method shortly after the flooding to see where inundationhad occurred.

Pelletier then compared the computer-generated mapsto two retrospective methods of assessing floods on alluvial fans. Thecomputer model slightly underestimated the observed extent of thefloods, matching the other methods for about 85 percent of the time.He's now improving the model to increase its accuracy.

"Numericalmodels like mine are much faster than the other two methods," Pelletiersaid. "But it's important to have independent verification and theother two techniques provide that."

Geologic field-mapping, whilevery accurate, requires that highly trained personnel spend manyman-hours mapping the soils and other landscape features. Tiger Washand Wild Burro Canyon have had such intensive study, but few otherareas have. In contrast, the satellite-based method can be used forextensive regions of the West but can only provide information aboutwhat has occurred within the last 20 years. So if a particular canyonhad only moderate floods during that time, the satellite-based methodcannot indicate what areas will be inundated during a 100-year flood.

Pelletier'scomputer method uses data that are becoming more readily available formore areas. He can then specify the size of flood -- 10-year, 100-yearor maximum level -- for a specific canyon, and the computer predictswhere and how high the floodwaters will go. His model provides aprobability of whether a particular piece of land will be flooded by aspecific size of flood.

He wants to further test thethree-pronged approach in other areas, including Clark County, Nev.,one of the fastest-growing counties in the country, where Las Vegas islocated.

The scientists' coauthors include P. Kyle House of theNevada Bureau of Mines and Geology at the University of Nevada, Reno;Karen A. Dempsey of Portland, Ore.; Jeanne E. Klawon of the U.S. Bureauof Reclamation in Denver; and Kirk R. Vincent of the U.S. GeologicalSurvey in Boulder, Colo. Their paper, "An integrated approach to floodhazard assessment on alluvial fans using numerical modeling, fieldmapping, and remote sensing," is in the September/October issue of theGSA Bulletin.