
	1999 Summer Scholars Program

Previous: Alyssa D. Gregory

Next: Glenda-Gayle Haskin

Radar Modeling of Post-1993 Missouri River Floodplain Vegetation Changes

By Laurel Griggs
Mentor: Dr. Raymond E. Arvidson
Department of Earth and Planetary Sciences, Washington University, St. Louis, MO

For thousands of years, man has attempted to manipulate his environment to improve his quality of life. For example, engineers in the early twentieth century manipulated the Missouri River and its floodplains by diverting its braided path into a single channel suitable for navigation and by building levees and wing dams to maintain the channel. In doing so, they hoped to lessen the threat of the seasonal floods that occasionally plagued the farmers along the floodplain. In 1993, however, efforts at containment were in vain as the river rose to a height surpassing other peak water levels recorded in this century, breaching the levees and flooding acres of valuable agricultural land. Scour zones and sand deposits were left after floodwaters receded. The U.S. Fish and Wildlife Service took advantage of the extensive damage by purchasing agricultural land from some farmers. Armed with recent scientific findings on the importance of wetlands and riparian natural habitats, the USFWS reconnected the floodplain to the river on newly- acquired "green beads" of floodplain lands to form the Big Muddy Fish and Wildlife Refuge. The Fish and Wildlife Service decided to let Jameson Island, Lisbon Bottoms, and Arrow Rock Bottoms, an area of the floodplain comprising one such "green bead," return to its natural state without intervening to repair the levees and roads that the floodwaters had damaged. This area of the floodplain has become the subject of study of Dr. Raymond Arvidson and those who work in his laboratory in an effort to better understand floodplain dynamics and biogeomorphology so that officials can make more informed decisions regarding river management in the future.

This particular study focused on using radar to model changes in the vegetation of the study area. AIRSAR radar images of the study areas from the years 1995, 1996, and 1998 were first acquired. For each year, three different images generated by the emission and return of radar waves of bands C (l = 3.8 cm), L (l = 24.0 cm), and P (l = 68.0 cm) were available but not co-registered. Each image in turn consisted of bands of three different polarizations: HH (horizontal transmit and horizontal return), VV (vertical transmit and vertical return), and HV (horizontal transmit and vertical return). To obtain a color composite of each polarization for each year, the three different images (C, L, and P band) for each year were geo-referenced to a LANDSAT TM (thematic mapper) scene to convert the radar images into common Universal Transverse Mercator parameters. This conversion made it possible to overlay the HV polarization for the 1995 C band, for example, with the HV polarization for the 1995 L and P bands to create a color composite in which the intensity of the return signal for the P, L, and C bands determined the intensity of the red, green, and blue components, respectively, of the color composites. Mathematical models were then generated to determine how the intensity of a radar signal returned from stands of vegetation varied with wavelength, density of the vegetation stand, and polarization. These models showed that vegetation could be most accurately modeled using the color composites from the HV polarization, which exhibited the greatest contrast in the intensity of the return signal between vegetation stands of various densities.

The color composites could then be interpreted based on the intensity of color in different areas of the image. Because of the relatively short wavelength, the C band best responds to roughness at all scales, while the L band (medium wavelength) best responds to roughness at a medium scale and the P band (long wavelength) best responds to roughness at a large scale. Therefore, areas of the image appearing white are likely to contain vegetation that exhibits roughness on a large scale (such as trees), while areas of the image appearing blue-green are likely to contain vegetation that exhibits roughness on an intermediate scale (saplings), and blue areas are covered by sands and sparse grasses. Also, areas of the image appearing brighter contain vegetation at a greater density than darker areas of the image. Interpretation of the images according to these guidelines yielded the conclusion that in general, the vegetation of the study area has greatly increased since 1995, growing both in density and height. Riparian forests are expanding into areas that had been cleared prior to the flood of 1993, and vegetation is beginning to colonize and populate areas of thick sand deposits. Exceptions to the trend of generally increasing vegetation include areas of new channel formation as a result of unrepaired levee breaks and seasonal floods in 1995 and 1998. For example, the chute that had formed in the Lisbon Bottoms area by 1998 washed away several stands of trees. Similarly, the progressive erosion of the northeastern corner of Jameson Island obliterated some of the vegetation that had colonized the area after the 1993 flood in addition to trees that had been present in the area prior to the flood. On the other hand, in other areas (such as the southern portion of Lisbon Bottoms bordering the levee break), vegetation colonized the previously saturated areas and increased in density from 1995 to 1998. As vegetation succession proceeded over time, grasses and shrubs yielded to willows in wet areas and cottonwoods in drier areas.

Previous: Alyssa D. Gregory

Scholars Index

Next: Glenda-Gayle Haskin

This page was last updated on Fri, Jun 16, 2000 at 2:15:55 PM by Tom Elgin with Userland Frontier.