Abstract:
Twenty-five georeferenced sets of images and maps estimating marsh vegetation cover and vulnerability were derived from Landsat Thematic Mapper (TM) Path 022 and Rows 039 and 040 digital number data sets collected between 1984 and 2011 for a time-series analysis to evaluate the effect of the Deepwater Horizon oil spill (April 20 - September 19, 2010) on the Terrebonne, Barataria, and Breton Sound basin emergent marshes. Fourteen datasets were classified as clear-sky (UDI R4.x264.000:0049), ten as cloudy-sky (UDI R4.x264.000:0050) and one dataset containing the September 2, 2009 clear-sky dataset that as a three to four pixel-wide line of missing data located over the southern Terrebonne and Barataria marshes (UDI R4.x264.000:0051). The projection is the Universal Transverse Mercator (UTM Zone 15 North). The file format is georeferenced ENVI data (.dat) and header (.hdr) files. Surface reflectance values were used to calculate the Normalized Difference Vegetation Index (NDVI), the Normalized Difference Water Index (NDWI), and the Normalized Difference Vegetation Index (NDSI) for each Normalized Difference Composition (NDXI) clear-sky data sets. Linear spectral unmixing of the three stacked normalized difference indices data sets (NDXI) with image-derived spectral endmembers of marsh vegetation, water, and marsh substrate/soil derived from the composite NDX data sets were used to estimate of the percentage of marsh vegetation, water, and marsh substrate for each marsh pixel in the Terrebonne, Barataria, and Breton Sound basins. The classified maps consist of three classes: (1) non-marsh or unclassified pixels have a value of 0 and are black; (2) pixels with 40% or less vegetation have a value of 1 and are black or gray; and (3) pixels with more than 40% marsh vegetation or are classified as intact marsh. Intact marsh pixels have a value of 2 and are white. Additional masks were used to estimate the percentage of intact marsh area in four twenty-km wide zones oriented roughly parallel to the coast and twelve three-km wide segments located immediately adjacent to the gulf waters in Terrebonne Bay, Barataria Bay and Breton Sound. The four zones and twelve coastline segments were used to evaluate the effects of Deepwater Horizon oil spill and sea level rise on the Terrebonne, Barataria, and Breton Sound basin emergent marshes.
Suggested Citation:
Riter, Joyce Christine Alexis. 2017. Landsat Thematic Mapper-Derived Estimates of Marsh Cover and Vulnerability in Terrebonne, Barataria, and Breton Sound Basin Marshes, Southeastern Louisiana: Clear-sky Data Set with Missing Pixels. Distributed by: GRIIDC, Harte Research Institute, Texas A&M University–Corpus Christi. doi:10.7266/N7Z31X2K
Methods:
The criteria used for the selection of the Landsat TM data sets (Path 22 and Row 39; Path 22 and Row 40) collected between 1984 and 2011 were: 1) dates coinciding with peak or near-peak vegetation growth based on image appearance on the United States Geological Survey GLOVIS website (http://www.glovis.usgs.gov); 2) a tidal stage near Mean Lower Low Water; 3) high atmospheric clarity (little haze or cloud cover); 4) regular intervals between observations; and 5) the inclusion of scenes bracketing major hurricanes. Rarely were all criteria met, which is typical of most remote sensing studies; twenty-five total datasets chosen for the final study were judged to be acceptable. The major limiting factor of data set selection was finding cloud- and haze-free images during the summer and early fall time period, when vegetation approached its peak growth period. Ten data sets with haze and/or clouds were used if no other imagery for the year or period of interest was available. Most cloudy or hazy images had less than five percent cloud or haze cover. The water level/tidal stage was determined to be the least important criterion used in image selection. The effect of water level on changes in image quality for analysis of marsh vegetation is diminished since the resolution of Landsat TM data (30 square meters) cannot typically detect the moderate changes in spatial extent of water coverage between high and low tide where tidal ranges are low (32.3 cm at Grand Isle, Louisiana station #8761724). (http://tidesandcurrents.noaa.gov/stationhome.html?id=8761724). Radiance was calculated for the co-registered Landsat TM data sets using standard remote sensing techniques and equations (Jenson, 2002; Chander et al., 2009) and ENVI-IDL software. The Exelis FLAASH module was used to perform the atmospheric correction and to calculate surface reflectance (Berk et al., 2002). The atmospherically-corrected Path 014 and Rows 32 and 33 were then mosaicked together. A small triangular gap exists over the southwestern Breton Sound basin due to a mistake in the mosaicking. The absence of these data, which account for less than 0.59% of the total marsh area, should not significantly affect the results of the analysis. Landsat TM surface reflectance data for bands 3, 4, and 5 were masked to reset all surface reflectance values equal to or less than 0.0000 to 0.00001. All surface reflectance values greater than 1.00000 were masked and reset to 1.00000. The spectral indices for vegetation, water, and soil were calculated by normalizing the difference between Landsat TM bands 3, 4, and 5 surface reflectance values to produce a spectral space (NDXI) that approximates the optimal spatial model for spectral unmixing. The three normalized indices are defined as follows: NDVI = (band 4NIR – band 3Red)/(band 4NIR + band 3Red); NDWI = (band 3Red – band 5SWIR)/(band 3Red + band 5SWIR); and NDSI = (band 5SWIR – band 4NIR)/(band 5SWIR + band 4NIR) (Rogers and Kearney, 2004). Wavelengths of Landsat TM band 3 (the visible red band) = 0.63 -0.69 micrometers. Wavelengths of Landsat TM band 4 (the Near Infrared (NIR) band) = 0.76 -0.90 micrometers, and wavelengths of Landsat TM band 5 (the Shortwave Infrared 1 (SWIR 1) band) = 1.55 – 1.75 micrometers. Spectral signatures of the three end members, vegetation, water, and soil were extracted from each Landsat TM image-derived NDXI data set (composite NDVI, NDWI, and NDSI data set). Each NDX data set was unmixed using the spectral signatures of the three end members to produce three new data sets: percent vegetation, percent water, and percent soil. The classification of the percent vegetation data set produced a new data set consisting of areas of marsh dominated by the spectral signature of vegetation, estimated as more than 40% vegetation based on previous work on Delaware coastal marshes (Kearney and Riter, 2011). Any land cover type other than emergent marshland was masked and eliminated from analysis. The non-marsh land covers include New Orleans, farmland, major roads and infrastructure. A water mask was created for the waters of the Terrebonne Bay, Barataria Bay, Breton Sound, and the Gulf of Mexico and the large lakes to eliminate sun glint off waves causing incorrect identification of lake, bay and gulf waters as soil. The non-marsh mask has a value of 0 and the pixels are black. Two masks were created for each cloudy and hazy data set. A ‘cloud-off’ mask was made to identify every cloud or cloud-shadowed pixels over marshes with the ROI tool. The cloud-off mask was applied to the s_yyyy-mmdd_veg data set, with all missing pixels reset to a value of 0.00000. A second ‘cloud-on mask’ was applied to the s_yyyy_mmdd_veg data set that had a comparable marsh/water land cover spatial pattern adjacent to the cloud- and cloud-shadow covered pixels. Application of the cloud-on-mask to the clear-sky data set resets all pixels other than those under the line-on-mask to 0.00000. The two masked data sets were stacked and added to produce one composite data set. The estimate of the s_yyyy_mmdd_com image is subject to greater error than a clear-sky data, but the total error is likely to be less than if this procedure was not followed. Zone boundaries were created in ESRI ArcGIS by buffering around a line drawn along the Gulf of Mexico shoreline on the Landsat images. Due to the irregular shape of the coastline, the width of zone 1 is irregular. Two to four zones (depending on the length of the basin coastal marshes) were defined for the southeastern Mississippi delta plain to evaluate the effect of distance from the Gulf on marsh vegetation and land area. Zones 1 to 4 are approximately 20 km-wide. The Terrebonne, Barataria and Breton Sound basins consist of three, four, and two 20-km wide zones respectively. The zone number increases as distance from the gulf increases. One three km-wide-zone immediately adjacent to the Gulf of Mexico is within the boundaries of zone 1. The three km-wide zone is divided into twelve segments defined by the orientation of the coastline and adjacent islands. Berk, A., Adler-Golden, S.M., Ratkowski, A.J., Felde, G.W. et al (2002) Exploiting MODTRAN radiation transport for atmospheric correction: The FLAASH algorithm, in Proceedings of the Fifth International Conference on Information Fusion, pp. 798- 803, vol. 2, Institute for Electrical and Electronic Engineering, New York. Chander, G., Markham, B.L., Helder, D.L. (2009) Summary of current radiometric calibration coefficients for Landsat MSS, TM, ETM+, and EO-1 ALI sensors, Remote Sensing of Environment 113(5): 893-903. DOI: 10.1016/j.rse.2009.01.007 Jensen, J.R. 2007 Remote Sensing of the Environment: An Earth Resource Perspective, 2nd Edition, Prentice-Hall, Inc. Upper Saddle River, NJ, 592 pp. Kearney, M.S. and Riter, J.C.A. (2011) Inter-annual variability in Delaware Bay brackish marsh vegetation, USA, Wetlands Ecology and Management 19(4): 373-388. DOI: 10.1007/s11273-011-9222-6
Instruments:
Processing by the Kearney research group was performed primarily with Exelis/Harris Geospatial ENVI-IDL software. The zones and segments were created with ArcGIS software (http://www.esri.com/arcgis/about-arcgis). The Landsat Thematic Mapper data used in this study were produced by the Landsat Ecosystem Disturbance Adaptive Processing System (LEDAPS) by NASA (Masek et al., 2006). LEDAPS processes Landsat digital numbers to surface reflectance, using atmospheric correction routines developed for the Terra MODIS instrument (Vermote et al., 1997). Masek, J.G., Vermote, E.F., El Saleous, N., Wolfe, R., Hall, F.G., Huemmrich, F., Gao, F., Kutler, J., and Lim, T.K. (2006) A Landsat surface reflectance data set for North America, 1990-2000, Geoscience and Remote Sensing Letters 3: 68-72. Vermote, E.F., El Saleous, N., Justice, C.A., Kaufman, Y.J., Privette, J.L., Remer, L., Roger, J.C., Tanre, D. (1997) Atmospheric correction of visible to middle-infrared EOS- MODIS data over land surfaces: Background, operational algorithm, and validation, Journal of Geophysical Research 102: 17131-17141.
View Metadata