Wetlands help mitigate flooding and provide important habitats. They are transitional between terrestrial and aquatic systems and have one of the three following characteristics: 1) the land supports hydrophytes, at least periodically, 2) the substrate is predominately undrained hydric soil, and 3) the substrate is saturated or covered with water sometime during the growing season. From colonial times to the mid-1980s, the lower 48 states of the U. S. lost over 100 million of the more than 200 million acres of wetlands (>404,700 km of the >809,400 km). Primarily, these losses resulted from wetland drainage and conversion to agriculture. Wetlands are still being lost to conversion, however, loss of coastal wetlands is aggravated by salt water intrusion and water logging.
Generally, wetlands can be grouped into swamps and marshes. Swamps include either inland or coastal bottomland hardwood and cypress/tupelo freshwater forests and coastal saline mangrove forests. Cypress/tupelo forests are permanently flooded, while bottom land hardwood forests are intermittently flooded. Of the original estimated 25 million acres of bottomland hardwood forests, only about 5 million acres now exist (20,235 km2 of the original 101,175 km2).
Marshes are intermittently flooded coastal and inland grasslands that can be fresh or saline. Coastal marshes continuously undergo compaction. To maintain an elevation equilibrium, compaction must be offset by sediment or detritus input. Canal building and channelization can disrupt the flow of sediment into the marsh. For example, channelization of the Mississippi River in the mid-1950's severely reduced the flow of fresh water and sediments to the coastal Louisiana marshes. By the mid-1970s, this disruption, and others, resulted in losing nearly half the coastal Louisiana marshes.
The following examples illustrate three important aspects related to remote sensing of coastal wetlands threatened by global climate change influences; marsh loss detection, flood detection and monitoring, and management practices affecting the utility of remote sensing in detecting marsh loss.
Detecting wetland change: Hurricane Andrew (August 26, 1992) Impacts to the Coastal Louisiana Marshes
Marsh loss due to the hurricane impact was examined by using November 1990, March 1991, October 1992, and January 1993 Landsat Thematic Mapper (TM) images. TM images were separated into pre- and post-hurricane image sets and progressive clustering used to classify each image set into marsh and water. The pre and post classified images were differenced to produce a marsh loss map. To evaluate the accuracy of the classified and change maps generated from TM images, pre- and post-hurricane color infrared photographic images (about 1 m spatial resolution, October 1991 and 1992) were similarly classified. Results of the comparison indicated that in both the pre- and post-hurricane classifications, the TM marsh class was over estimated. The over estimation was attributed primarily to two causes. First, small gaps (<3m) containing water or floating vegetation in the otherwise continuous marsh canopy were misclassified as marsh. Second, more continuous areas containing mixtures of marsh, floating vegetation, and mud within the 30m spatial resolution of the TM sensor were predominantly classified as marsh.
The over estimation of marsh extent in the pre- and post-hurricane TM classifications resulted in about a 50% over estimation of marsh loss. The TM change analysis missed small areas of marsh change, critical to detecting the initiation of marsh loss. Further, larger areas of marsh loss were mostly located correctly in the TM analysis, but were spatially too extensive, causing an over estimation of marsh loss. Finally, higher spectral resolution of the TM sensor was not as important as the high spatial resolution of the photography in detecting marsh loss.
Sea Level Rise: Associating Marsh Flooding With Marsh Type and Biomass
Sea level rise poses one of the greatest threats to the long-term health and stability of coastal wetlands. This susceptibility is primarily an effect of the low shore-normal topographic gradient of coastal marshes (around 12 to 30 cm/km). To predict the possible consequences of sea level rise on a coastal marsh, the relationship between flooding extent, frequency, depth, and duration and marsh type and biomass is being examined in the Big Bend area of coastal Florida. To support this effort, at five marsh field sites, near-continuous recording of flood depth were collected. Concurrently, six consecutive ERS-1 SAR satellite images were collected of the marsh. Comparison of recorded flood depths at the times of the ERS-1 satellite collections, showed SAR returns from flooded marsh were attenuated in comparison to SAR returns from non-flooded marsh (Figure 15.1). Use of this finding allowed flood extent contours to be generated at the times of three ERS-1 SAR collections.
Next, an attempt will be made to convert these flood extent contours to topographic contours and from these generate a micro topographic surface of the marsh. Currently, only 150-cm topographic contours are available for the area. Recorded flood data will then be used to produce flood frequency, depth, and duration surfaces covering the marsh area; storm surge data will also be included where appropriate. If successful, the flood surfaces will be associated with marsh characteristics produced by using sets of optical data collected during the same time period. Ultimately, the connection between marsh and flooding characteristics will be used to simulate the consequences of a rising sea level on marsh type and biomass.
Burn Management Practices and Detecting Marsh Change with Remote Sensing
Large areas of marshes in the Gulf Coast are burned each year. To examine the effects of these burn practices in detecting marsh change, TM images and canopy reflectance spectra (400 to 900 nm with about 2.5 nm spectral bandwidths) of three sites were collected over an eight month period of a coastal Louisiana fresh marsh (about 12km by 12km). A December 1990 classified TM image was used to mask-out water and scrub shrub from all TM classifications.
By March 1991, large expanses of the study area had been burned. These burns were separated as old (January) and new (early March) burns. January reflectance spectra of the three in-situ field sites were spectrally non-distinct; however, reflectances of one site burned early in January were lower in magnitude. March canopy reflectance spectra of the three sites showed the beginnings of a red edge and nir plateau. The spectra associated with the previously burned site, however, was lower in amplitude with a more distinct red edge and nir plateau. Immediately after the March TM collection, another large area of marsh was burned. The extent of this burn was mapped by using the July 1991 TM image, however, most areas previously burned (January to March) were not detected. Inspection of the July canopy spectra from the three sites also indicated little spectral differences between the three sites. All canopy reflectance spectra had nearly the same magnitude and a well developed red edge and nir plateau.
Combining the March and July burn classifications allowed the generation of a burn history map for the area. The burn history map depicted marsh burned in January, in early March, and just after the March TM collection (Figure 15.2). In summary, TM imagery can be used to detect changes in a marsh subjected to burning. Even though advantageous to burn detection and recovery monitoring, these differences could lessen the utility of remote sensing techniques to detect marsh loss. Thus, remote sensing techniques should be developed that can discern changes in marsh canopy reflectance related to short-term management practices (e.g., burning) from longer term changes related to marsh loss.