Multi-temporal RADARSA T-1 and ERS Backscattering Signatur es of Coastal Wetlands in Southeaster n Louisiana

Using multi-temporal European Remote-sensing Satellites (ERS-1/-2) and Canadian Radar Satellite (RADARSAT-1) synthetic aperture radar (SAR) data over the Louisiana coastal zone, we characterize seasonal variations of radar backscattering according to vegetation type. Our main findings are as follows. First, ERS-1/-2 and RADARSAT-1 require careful radiometric calibration to perform multi-temporal backscattering analysis for wetland mapping. We use SAR backscattering signals from cities for the relative calibration. Second, using seasonally averaged backscattering coefficients from ERS-1/-2 and RADARSAT-1, we can differentiate most forests (bottomland and swamp forests) and marshes (freshwater, intermediate, brackish, and saline marshes) in coastal wetlands. The student t-test results support the usefulness of season-averaged backscatter data for classification. Third, combining SAR backscattering coefficients and an opticalsensor-based normalized difference vegetation index can provide further insight into vegetation type and enhance the separation between forests and marshes. Our study demonstrates that SAR can provide necessary information to characterize coastal wetlands and monitor their changes. Introduction Coastal wetlands constitute important ecosystems in terms of flood control, water and nutrient storage, habitat for fish and wildlife reproduction and nursery activities, and overall support of the food chain (Karszenbaum et al., 2000). The integrity of such wetlands has significant ecologic and economic implications. Louisiana has one of the largest expanses of coastal wetlands in the conterminous United States, and its coastal wetlands contain an extraordinary diversity of habitats. The unique habitats of upland areas and the Gulf of Mexico, complex hydrological connections, and migratory routes of birds, fish, and other species place the coastal wetlands of Louisiana among the nation’s most productive and important natural assets (USACOE, 2004). However, the balance of Louisiana’s coastal systems has been upset by a combination of natural processes and human activities. Massive coastal erosion probably started around 1890, and more than one million acres or about Multi-temporal RADARSA T-1 and ERS Backscattering Signatur es of Coastal Wetlands in Southeaster n Louisiana Oh-ig Kwoun and Zhong Lu 20 percent of the coastal lowlands (mostly wetlands) have eroded in the past 100 years (LCWCRTF/WCRA, 1998). For example, the loss rate for Louisiana’s coastal wetlands was as high as 25,200 and 15,300 acres per year in 1970s and 1990s, respectively (Barras et al., 2003). Massive environmental changes have significant impacts on the coastal ecosystem of Louisiana, including effects from frequent natural disasters such as the Hurricane Katrina in 2005. Therefore, an effective method of mapping and monitoring coastal wetlands is essential to understanding the status and influence of environmental changes and human activities on wetlands. Optical satellite images such as those from Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM ) have been used to map coastal wetlands (Homer et al., 2004; Loveland and Shaw, 1996). For example, National Land-cover Data 1992 was created from Landsat TM data; the producer’s accuracy for wetlands over the southeastern region of the United States was assessed at 46 77 percent (Stehman et al., 2003). A unique characteristic of synthetic aperture radar (SAR) in monitoring wetlands over cloudprone subtropical regions is the all-weather and day-andnight imaging capability. In addition, SAR can provide key descriptors for a wetland environment, such as ground vegetation structure, inundation, topographic features, and moisture content (Sadre et al., 1995). Waite and MacDonald (1971) first reported that flooded forests in “leaf off” conditions in Arkansas showed up as anomalously bright areas on K-band radar images. Since then, several studies have demonstrated that satellite SAR can map and monitor forested and non-forested wetlands occupying a range of coastal and inland settings (Ramsey III, 1998 and 1999; Ramsey III et al., 2006). Many of those studies relied on the fact that when standing water is present beneath the vegetation canopies, the radar backscattering signal changes with water level changes, depending on vegetation type and structure. Therefore, they used SAR backscattering signals to monitor floods and dry conditions, temporal variations in the hydrological conditions of wetlands, including classification of wetland vegetation at various study sites (Baghdadi et al., 2001; Bourgeau-Chavez et al., 2001; Bourgeau-Chavez et al., 2005; Costa, 2004; Costa et al., 2002; Grings et al., 2006; Hess and Melack, 1994; Hess PHOTOGRAMMETRIC ENGINEER ING & REMOTE SENS ING May 2009 607 Oh-ig Kwoun is with Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, and formerly with SAIC at the U.S. Geological Survey, EROS Data Center, Sioux Falls, SD 57198 ([email protected]). Zhong Lu is with the USGS EROS Center and Cascades Volcano Observatory, Vancouver, WA 98683 ([email protected]). Photogrammetric Engineering & Remote Sensing Vol. 75. No. 5, May 2009, pp. 607–617. 0099-1112/09/7505–0607/$3.00/0 © 2009 American Society for Photogrammetry and Remote Sensing 607-617_07-097.qxd 4/16/09 9:14 PM Page 607