a baseline study of histo

2 downloads 0 Views 20MB Size Report
that the road (red arrow) is not present and that the boat launch parking (blue arrow) was minimal, there appears to be only ...... Alberti, M., R. Weeks and S. Coe 2004. Urban land .... Environmental Microbiology 72(5):3357-3366. Costa C.S. ...
THE UNIVERSITY OF SOUTH ALABAMA COLLEGE OF ARTS AND SCIENCES NATURAL DISASTERS AND LONG-TERM RECOVERY: A BASELINE STUDY OF HISTORICAL CHANGE AND HABITAT STRUCTURE OF JUNCUS ROEMERIANUS MARSH AT GRAND BAY, MISSISSIPPI BY Tami Maureen Wells A Dissertation Submitted to the Graduate Faculty of the University of South Alabama in partial fulfillment of the requirements for the degree of Doctorate of Philosophy in Department of Marine Science Approved:

May 2010

Co-Chair of Dissertation Committee: Just Cebrian, Ph.D. Co-Chair of Dissertation Committee: Anne A. Boettcher, Ph.D. Member of Committee: Robert L. Shipp, Ph.D. Member of Committee: Shelia A. Brown, Ph.D. Member of Committee: Jack L. Gallagher, Ph.D. Chair of Department: Robert L. Shipp, Ph.D. Director of Graduate Studies: S. L. Varghese, Ph.D. Dean of the Graduate School: B. Keith Harrison, Ph.D.    

 

Date:

NATURAL DISASTERS AND LONG-TERM RECOVERY: A BASELINE STUDY OF HISTORICAL CHANGE AND HABITAT STRUCTURE OF JUNCUS ROEMERIANUS MARSH AT GRAND BAY, MISSISSIPPI

A Dissertation Submitted to the Graduate Faculty of the University of South Alabama in partial fulfillment of the requirements for the degree of Doctorate of Philosophy in Department of Marine Science

by Tami Maureen Wells B.S., Louisiana State University, 1997 M.S., the University of Southern Mississippi, 2005

Copyright c 2010_Tami Maureen Wells All rights reserved.

 

I would like to dedicate this publication to all whom have assisted in data collection in what were often grueling climatic conditions and to the many non-profit organizations, Federal and State agencies, and environmentally concerned citizens that supported our efforts. Thank you, for your time and interest in this research project.

ii  

ACKNOWLEDGEMENT

I would like to thank my committee members for their guidance and expertise in the preparation of this dissertation. I would especially like to thank the many students from the Undergraduate Research Programs at the University of South Alabama that assisted in the data collection and analyses of this research. I want to thank Dr. Bob Shipp and the Alabama Center for Estuarine Studies for providing funding for this research. Additionally, I would like to thank Radiance Technologies, Army Corps of Engineers, Weeks Bay and Grand Bay National Estuarine Research Reserves (NERRs), United States Department of Agriculture - Natural Resource Conservation Service, Durwin Carter, U.S. Fish and Wildlife, Wolf Bay Water Watch, Baldwin County Master Gardeners, Alabama Cooperative Extension System, Coden Community, LSU Department of Horticulture, Dr. Susan Rees, John Olive, Randy Roach, Lee Yokel, Dr. Cynthia Moncreiff, Dr. Ed Perkins, Dr. Richard Lance, Dr. Doug Haywick, Dr. Hugh MacIntyre, Dr. Judy Stout, Dr. Patrick Biber, Dr. John Dindo, Dr. George Croizer, Charlyn Partridge, Dan Martin, Emily Boone and Laura Linn for their support. Finally, I would like to thank my dear friends Debra and Danny Hood for their encouragement.

iii  

TABLE OF CONTENTS

Page LIST OF TABLES ..............................................................................................vii LIST OF FIGURES.............................................................................................ix LIST OF SYMBOLS............................................................................................ix LIST OF ABBREVIATIONS ...........................................................................xvii LIST OF NOMENCLATURE ..........................................................................xxii LIST OF EQUATIONS ...................................................................................xxiii ABSTRACT...................................................................................................... xxiv CHAPTER I – BACKGROUND ..........................................................................1 Introduction.....................................................................................................1 Salt Marshes ..............................................................................................2 Research Site..............................................................................................8 Remote Assessment of Marsh Damage and Loss ...................................10 Hurricane Impacts Along the Northern Gulf Coast ..............................14 Research Summary ......................................................................................16 Scope and Organization of Dissertation .....................................................16 CHAPTER II – GRAYSCALE IMAGE TECHNIQUES FOR DETERMINING LAND COVER CHANGE IN A GRAND BAY, MISSISSIPPI SALT MARSH ............................................................................24 Introduction..................................................................................................24

iv  

Problem...................................................................................................28 Question..................................................................................................29 Methods ........................................................................................................29 Study Area..............................................................................................29 Description of Imagery and Acquisition Platforms ..............................30 Image Normalization and Georeferencing............................................31 Change Detection ...................................................................................34 Vegetation Transects........................................................................34 Write Function Memory Insertion...................................................37 Difference Mapping ..........................................................................37 Post Classification Difference Image...............................................39 Supervised Mahalanobis Distance Classification...........................42 Results ..........................................................................................................43 Image Normalization and Georeferencing.............................................43 Change Detection ....................................................................................44 Vegetation Transects.......................................................................44 Write Function Memory Insertion ..................................................45 Difference Mapping .........................................................................46 Post Classification Difference Image..............................................47 Supervised Mahalanobis Distance Classification .........................48 Conclusion .....................................................................................................50 CHAPTER III –NATURAL STORM EFFECTS AND LANDSCAPE CHANGE DETERMINED FROM DIFFERENCE MAPPING AND SUPERVISED MAHALANOBIS CLASSIFICATIONS .................................102 Introduction.................................................................................................102 Study Area..............................................................................................105 Methods .......................................................................................................107 Image Greyscale Normalization............................................................107 Supervised Mahalanobis Distance Classification ................................108 Difference Mapping ...............................................................................109

v  

Results ........................................................................................................109 Image Greyscale Normalization............................................................109 Supervised Mahalanobis Distance Classification ................................110 Difference Mapping ...............................................................................111 Conclusion ...................................................................................................112 CHAPTER IV– CONCLUSION.......................................................................136 REFERENCES.................................................................................................143 APPENDICES Appendix I: Tables A-1 through A-3 ..........................................................178 Appendix II: Figures A-1 and A-2 ..............................................................182 Appendix III: Table A-4 ..............................................................................184 Appendix IV: Figure A-3.............................................................................186 Appendix V: Table A-5 ................................................................................187 Appendix VI: Table A-6...............................................................................188 Appendix VII: Table A-7 .............................................................................189 BIOGRAPHICAL SKETCH.............................................................................190

vi  

LIST OF TABLES

Table I-1. II-1. II-2. II-3. II-4.

II-5. II-6. II-7. II-8. II-9.

Page A comprehensive recording of Gulf of Mexico (GOM) hurricanes that significantly affected coastal Alabama and Mississippi were recorded from 1950 - 2005........................ 18 Statistical results from comparison of image normalization stretch procedures for the natural site subsections measuring 197 rows x 138 columns = 27,186 pixels................... 61 Statistical results from comparison of image normalization stretch procedures for the impacted site subsections measuring 527 rows x 400 columns = 210,800 pixels................. 62 A table showing statistical results from Gaussian normalized greyscale change detection. ...................................... 63 Percent frequency data (%) compiled from line intercept sampling at each transect at the Kreole, Mississippi study site. The frequency represents the percent presence and absence of each vegetative and non-vegetative land cover within the 50- intercept crossings (samples) of the six random transects (25m). The percent land cover frequency is calculated as the number of samples containing land cover / total number of intervals x 100........................................ 64 Vegetative relative frequency data (%) compiled from line intercept sampling at each transect at the Kreole, Mississippi study site. .................................................................. 65 Vegetative relative cover (%) compiled from line intercept sampling at each transect at the Kreole, Mississippi study site. .............................................................................................. 66 A table of the vegetation biodiversity indices for six random transects at the study area in Kreole, Mississippi. .................... 67 Transect elevation and average stem count and height of Juncus roemerianus are recorded from in situ investigations. .............................................................................. 68 A table showing statistical results from Mahalanobis Classification change detection. .................................................. 69

vii  

IIII-1.

Percent change computed from Mahalanobis distance supervised classification of pre and post-hurricane imagery. ..................................................................................... 118

Appendix Table A-1. A-2. A-3. A-4. A-5.

A-6. A-7.

Line Transect geolocation for Kreole, Mississippi study site. .................................................................................. 178 Vegetation transects biodiversity statististics for Kreole, Mississippi (Grand Bay National Research Reserve Study Site). ............................................................................... 179 Juncus roemerianus average stem count (m2) and height (cm) compiled based on line transect (25m) sampling (random quadrat) at Kreole, Mississippi. ................. 181 Percent nitrogen and carbon ratios recorded from ten Juncus roemerianus tissue samples within the study area at Kreole, Mississippi. . ..................................................... 184 Surface Water Elevations collected on February 11, 2008 during a Neap Tide (approaching 1st quarter) with high at 13:53 and low at 23:51 and a range of 0.30 meter (1.0 ft) (Mississippi Department of Marine Resources, Biloxi, MS). ................................................................................ 187 A table of pore water seasonal results for the Kreole, Mississippi study site. .............................................................. 188 Percent land-cover change comparisons between Grand Bay National Estuarine Research Reserve (Wells, T. 2010 Dissertation, University of South Alabama) and Louisiana Coastal Assessment Research by Barris et al. 2006. ................................................ 189

viii

LIST OF FIGURES

Figure I-1.

I-2.

I-3. I-4.

I-5.

II-1.

II-2. II-3. II-4.

Page The flower of Juncus roemerianus Scheele (photo a), 1500X magnification (Zeiss) of J. roemerianus seed encased in pod (photo b), and magnification of seeds adjacent to a U.S. minted dime (photo c), photo: TamiWells, University of South Alabama. ................................. 19 A TerraMetrics Landsat image of the northern Gulf of Mexico shows the location of the Grand Bay National Estuarine Research Reserve (GNDNERR) site at (30.37N, -88.40W), source: Google Earth 2006. ......................................... 20 A map of the Grand Bay National Estuarine Research Reserve located in the SE Quadrat in Jackson County, MS. ................................................................................................ 21 A 2009 United States Department of Agriculture (USDA) Farm Service Agency (FSA) image of the natural (solid line) and impacted (dashed line) research sites at Grand Bay National Estuarine Research Reserve (GNDNERR) located at (30.37N, -88.40W), source: Google Earth 2009. ......... 22 An eastern map of the United States of America dentifying the major landfalling hurricanes between the years of 1900-2005, source: (September 2007) Map © Michael A. Grammatico -2004 http://www.geocities.com/hurricanene/gulfcoast.htm................. 23 A comparison of Landsat Thematic Mapper 5 imagery of West Pearl River and Highway 190 (near the White’s Kitchen landing) located at the Louisiana and Mississippi boarder before and after Hurricane Katrina. ............................. 70 A graph shows the projected coastal land loss 1956-2050. ........ 71 A subsection of 2006, U.S. Geological Survey RGB (Red, Green, Blue), True Color Image (TCI) of a boat launch (red arrow), at Heron Bayou in Kreole, Mississippi. .................. 72 The Non-Processed Greyscale Imagery from 1940 (Left), 1985 (Center) and 2008 (Right) documents the change over a 68 year, time-frame. .......................................................... 73

ix  

II-5. II-6. II-7.

II-8.

II-9.

II-10. II-11.

II-12.

II-13.

II-14.

II-15. II-16.

A block design illustrating the change detection procedure utilizing greyscale geo-referenced imagery. .............. 74 ENVI produced Change Detection Difference Map of 1940greyscale aerial photo image (Left) and 2002 DOQQ greyscale image (Right)................................................................ 75 This graph depicts the Gaussian filter as the best fit normalization stretch techniques in relation to mean pixel relationships between multiple years (1940, 1985, and 2006) of imagery. .......................................................................... 76 Figure shows 2D scatter plots of greyscale pixel intensities for 2006 (y-axis) and 1940 (x-axis) subsection of imagery with man-made impacts (left); and 2006 (yaxis) and 1940 (x-axis) subsection of imagery without man-made impacts (right)............................................................ 77 Figure shows 2D scatter plots comparing the greyscale pixel intensities of 2006 (y-axis) and 1940 (x-axis) Gaussian normalized subsection with impacts (left) and a 2006 (y-axis) and 1940 (x-axis) of greyscale Gaussian normalized subsection without impacts (right). ......................... 78 Image normalization characteristics of the natural subsection of greyscale imagery 1940 (Left Column), 1985 (Center Column) and 2006 (Right Column).. .............................. 79 A series of image normalization projections of impacted (man-made construction) subsection of the greyscale imagery 1940 (Left Column), 1985 (Center Column) and 2006 (Right Column). ................................................................... 80 Change detection difference maps of Gaussian normalized greyscale (natural subsection) using a simple value difference for image comparisons 1940-1985 (Left), 19852006 (Center) and 1940-2006 (Right). ......................................... 81 Change detection difference maps of Gaussian normalized greyscale (impacted subsection) using a simple value difference for image comparisons 1940-1985 (Left), 19852006 (Center) and 1940-2006 (Right). ......................................... 82 True color image showing all transects sampled at Grand Bay NERR, Kreole, MS, transect geo-location data in NAD83, source: USDA Farm Service Area (FSA) Imagery (August 7, 2007), 1m resolution, Google Earth........................... 83 A cross section of Juncus roemerianus dominated marsh at Grand Bay NERR determined from baseline data collections...................................................................................... 84 Multi-Date Change Detection (MCD) using Write Function Memory Insertion of 1940, 1985 and 2006

  x  

II-17. II-18.

II-19. II-20. II-21.

II-22. II-23. II-24.

II-25.

II-26.

II-27.

greyscale imagery to produce a RGB composite image – , R1940 G1985, B2006 (Right). ...................................................... 85 ENVI produced Greyscale Difference Maps are created between two image comparisons 1940-1985 (Left), and 1940 – 2006 (Right). ..................................................................... 86 ENVI produced Change Detection Difference Map created from 1940 greyscale aerial photo image and 2006 DOQQ greyscale imagery using 11 classes, simple difference, 0-1 normalization and user defined threshold adjustments algorithm....................................................................................... 87 Image on the (Left) is a 1985 RGB Image showing glint (red circle) on the surface of the water due to solar azimuth angle at time of data collections. .................................. 88 Gaussian greyscale change detection statistics for 4 classes within the natural image subsection (Top) and 5 classes within the impacted image subsection (Below).89 A synthetic color transformation from greyscale is automated in ENVI v. 4.2. The results are generated from Gaussian normalized greyscale natural subsection imagery and are shown as follows 1940 image (left), 1985 (center), and 2006 (right). ............................................................ 90 A synthetic color transformation from greyscale is automated in ENVI v. 4.2. ........................................................... 91 Supervised Mahalanobis distance classification percent accuracy (top) and Kappa coefficient (Below) for both impacted and natural sites. ......................................................... 92 Three Supervised Mahalanobis Distance Classifications from the transformed synthetic color images utilized the 2006 image and vegetation transect data as the baseline dataset for a regions of interest point collection......................... 93 Three Supervised Mahalanobis Distance Classifications from the transformed synthetic color images utilized the 2006 image and vegetation transect data as the baseline dataset for a regions of interest point collection......................... 94 Change detection difference maps (natural subsection) created from simple value difference of Mahalanobis Supervised Classification images for 1940-1985 (left), 1985-2006 (center) and 1940-2006 (right). .................................. 95 Change detection difference maps (impacted subsection) created from simple value difference of Mahalanobis Supervised Classification images for 1940-1985 (left), 1985-2006 (center) and 1940-2006 (right) ................................... 96

  xi  

II-28. II-29. II-30.

II-31.

II-32.

III-1.

III-2. III-3.

III-4. III-5. III-6.

III-7.

Supervised classification Change detection statistics for 4 classes within the natural image subsection (Top) and 5 classes within the impacted image subsection (Below). ............. 97 Elevation data were averaged from NED Contiguous U.S. 1/9 E. Arc Second Elevation Data (USGS NRCS, Seamless Data Accessed: 08Sept2009). ....................................................... 99 National Oceanic and Atmospheric Administration (NOAA), National Climatic Data Center (NCDC), Mississippi StatewidePalmer Hydrological Drought Index (PHDI) for years 1900 – 2007. ..................................................... 99 National Oceanic and Atmospheric Administration (NOAA), National Climatic Data Center (NCDC), map depicting the Palmer Hydrological Drought Index (PHDI) for long-term conditions (above) and Standardized Precipitation Index (SPI) for six-month conditions (below). .... 100 National Oceanic and Atmospheric Administration (NOAA), National Climatic Data Center (NCDC), map depicting the Palmer Z Index (PZI) for short-term conditions in January 2004 (above) and March 2006 (below). ........................................................................................ 101 GOES-12 Satellite Image of Hurricane Elena on 02 September 1985, source, http://www1.ncdc.noaa.gov/pub/data/images , Accessed: 03 August 2009. ............................................................................... 119 Hurricane Elena tracking and rainfall data, Source, http://www.hpc.ncep.noaa.gov/tropicalrain/elena1985.htm Accessed 03 August 2009. .......................................................... 120 GOES-12 Satellite Image of Hurricane Katrina on 28 August 2005, source, http://www1.ncdc.noaa.gov/pub/data/images , Accessed: 03 August 2009. ............................................................................... 121 Hurricane Katrina tracking and rainfall data, Source, http://www.hpc.ncep.noaa.gov/tropical rain/elena1985.html Accessed 03 August 2009. ....................... 122 A chart showing the population level for Mississippi Coastal Counties in 1980 versus 2008. Data source, US Census, Washington DC. ........................................................... 123 Hurricane Elena greyscale imagery, pre-hurricane, (a) March 25, 1985 and post-hurricane, (b) October 6, 1985. Hurricane Elena Gaussian normalized imagery (c) March 25, 1985 and Gaussian normalized (d) October 6,1985. ........... 124 Hurricane Katrina greyscale imagery, pre-hurricane, (a) 2004and post-hurricane, (b) 2006. ............................................. 125

 xii  

III-8.

III-9.

III-10.

III-11.

III-12.

III-13.

III-14.

III-15. III-16.

III-17.

A two dimensional (2D) scatter plot of greyscale imagery (left) March 25, 1985 - x-axis, and October 6, 1985 - y-axis and 2D scatter plot of Gaussian normalized greyscale imagery (right) March 25, 1985 - x-axis, and October 6, 1985 - y-axis. ............................................................................... 126 A two dimensional (2D) scatter plot of greyscale imagery (left) January 2004 - x-axis, and March 2006 - y-axis and 2D scatter plot of Gaussian normalized greyscale imagery (right) January 2004 - x-axis, and March 2006 - y-axis. .......... 127 Synthetic (RGB) pseudo-color imagery created from Gaussian enhanced greyscale imagery collected on (a) March 25, 1985, (b) October 6, 1985, (c) January 2004, and (d) March 2006. ................................................................... 128 Change detection utilized color imagery dated (a) March 25, 1985 and (b) October 6, 1985 representing pre and post-impacts from Hurricane Elena that made landfall in Biloxi, Mississippi on September 2, 1985.................................. 129 Change detection utilized color imagery dated (a) January 2004 and (b) March 2006 representing pre and postimpacts from Hurricane Katrina that made landfall in Pearlington, Mississippi on August 28, 2005............................ 130 Mahalanobis distance supervised classification based on class size from the regions of interest (Juncus roemerianus marsh, pine hummock, open-water, and salt panne).......................................................................................... 131 Mahalanobis distance supervised classification based on class size from the regions of interest (Juncus roemerianus marsh, pine hummock, open-water, and salt panne).......................................................................................... 132 Two-dimensional scatter plots display the spectral variability in the Gaussian normalized greyscale imagery...... 133 Change detection utilized color imagery dated (a) March 25, 1985 and (b) October 6, 1985 representing pre and post-impacts from Hurricane Elena that made landfall in Biloxi, Mississippi on September 2, 1985.................................. 134 Change detection utilized color imagery dated (a) January 2004 and (b) March 2006 representing pre and postimpacts from Hurricane Katrina that made landfall in Pearlington, Mississippi on August 28, 2005............................ 135

 xiii  

Appendix Figure A-1. A-2. A-3.

Multispectral Image showing the location of surface water wells and carbon and nitrogen tissue sampling at Kreole, Mississippi. ................................................................................. 182 Annual percent mean of carbon and nitrogen ratios for ten, Juncus roemerianus tissue, collection sites at the Kreole, Mississippi study area................................................... 183 Ten plant tissue samples were collectedd at each study area during four seasonal timeframes. ..................................... 186

 xiv  

LIST OF SYMBOLS

a

– alpha

.evf

– ENVI vector file

m

– meter

m2

– meter square

km

– kilometer

km/h

– kilometer per hour

km2

– square kilometer

mb

– millibars

kn

– knots

mph

– miles per hour

min

– minimum

sec –

– second

oC

– Celsius

C:N

– Carbon to Nitrogen Ratio

Hg

– Mercury

C

– Carbon

N

– Nitrogen

 xv  

p(.05)

– p-value of 5 % or 95% confidence

2D

– two dimensions

hPa

– hectoPascals (newtons per square meter)

log10(x)

– The base-10 function, natural logarithm used by biologists

log2(x)

– The base 2 function, natural logarithm used by computer scientists



– Sigma, summation, the addition of a set of numbers

30o 22’ 42.18 “

– 30 degrees 22 minutes 42.13 seconds

CaCO3

– Calcium Carbonate

SQRT

– square root

STDEV

– standard deviation

ARCSIN

– ArcSine

mm

– Millimeters

mg

– Milligrams

mi2

– square miles

xvi

LIST OF ABBREVIATIONS

ANOVA



Analysis of Variance

B&W



Black and White

BMP



Bitmap

BV



Brightness Value

CDMO



Centralized Data Management Office

CHNS-O



Carbon Hydrogen Nitrogen Sulfur – Oxygen

CM



Change Map

CZMA



Coastal Zone Management Act

(DD)



Decimal Degrees

DIN



Dissolved Inorganic Nitrogen

DGPS



Differential Global Positioning System

(DMS)



Degrees Minutes Seconds

DN



Digital Number

DOQQ



Digital Orthophoto Quarter Quadrangle

DPI



Dots Per Inch

ECS



Elemental Combustion System

EDM



Electronic Distance Meter

  xvii  

ENVI



Environment for Visualizing Images

EPA



Environmental Protection Agency

EROS



Earth Resources Observation and Science

FSA



Farm Service Area

FWS



Fish and Wildlife Service

GBEP

-

Galveston Bay Estuary Program

GC



Gas Chromotography

GCP



Ground Control Points

GIS



Geographic Information System

GPS



Global Positioning System

GNDNERR



Grand Bay National Estuarine Research Reserve

GOM



Gulf of Mexico

HSV



Hue, Saturation and Value

IMS



Internet Management Service

IACD



Image Algebra Change Detection

JPEG



Joint Photographic Experts Group

LAT



Latitude

LCA



Louisiana Coastal Area

LCT



Land Cover Type

LIDAR



Light Detection and Ranging

LON



Longitude

  xviii  

LUC



Land Use Change or Land Use Cover

LWC



Louisiana Wetland Center

MCD



Multi-Date Change Detection

MSI



Multi-Spectral Imagery

NAD83



North American Datum 1983

NAPP



National Aerial Photography Program

NCDC



National Climate Data Center

NDVI



Normalized Difference Vegetation Index

NED



National Elevation Data

NERR



National Estuarine Research Reserve

NESDIS



National Environmental Satellite Data and Information Service

NHAP



National High Altitude Photography

NLM



National Landscape Model

NMFS



National Marine Fisheries Service

NOAA



National Oceanic and Atmospheric Administrative

NOS



National Ocean Service

NRCS



National Resource Conservation Service

NRT



Near Real-Time

NSSDA



National Standards for Spatial Data Accuracy

NWC



National Weather Center

NWI



National Wetland Inventory

 xix  

NWRC



National Wetlands Research Center

NWS



National Weather Service

OBS



Office of Biological Sciences

O.C.



Outside Circumference

OCRM



Office of Ocean and Coastal Resource Management

PHDI



Palmer Hydrolic Drought Index

PSC



Photo Scanner Copier

PV



Pixel Value

PVC



Polyvinyl Chloride

PZI



Palmer Z Index

REDOX



Reduction and Oxidation

RGB



Red, Green, Blue (Spectral Bands)

ROI



Regions of Interest

ROW



Right-of-Way

SE



Southeast

SPI



Standardized Precipitation Index

STEM



Science Technology Engineering and Mathematics

SWBI



Shannon-Weiner Biodiversity Index

SWMP



System Wide Monitoring Program

TCE



Texas Cooperative Extension

TCWP



Texas Coastal Watershed Program

TDS



Tripod Data Systems

 xx  

TIFF



Tagged Image File Format

TM



Thematic Mapper

TS



Tropical Storm

USD



US Dollar

USDA



United States Department of Agriculture

USFWS



United States Fish and Wildlife Service

USGS



United States Geological Survey

UTM



Universal Transverse Mercator

VCD



Visual Change Detection

WAAS



Wide Area Augmentation System

 xxi  

LIST OF NOMENCLATURE

B. maritima

– Batis maritime L.

– turtleweed

B. frutescens

– Barrichia frutescens (L.) DC.

– bushy seaside tansy

D. spicata

– Distichlis spicata (L.) Greene

– saltgrass

I. (D.) notablis

– Isotoma (Desoria) notablis (Schf.) – springtail

J. roemerianus – Juncus roemerianus Scheele

– black needle rush

P. palustris

– Pinus palustris Mill.

– longleaf pine

S. alterniflora

– Spartina alterniflora Loisel.

– saltmarsh cordgrass

  xxii  

LIST OF EQUATIONS

Equation

Page

II- 1

Computation of pixel ground resolution

32

II- 2

A standard calculation for changing an image from RGB composite image to greyscale

32

II- 3

Species frequency

35

II- 4

Relative species frequency

35

II- 5

Relative species density

35

II- 6

Percent species cover

35

II- 7

Relative percent cover

35

II- 8

Shannon-Wiener, H(s)

36

II- 9

Pielou’s evenness index, j

36

II- 10

Simpson’s dominance index, c

36

II- 11

A calculation for determining spectral change Modified from Virag and Colwell, (1987)

38

II- 12

True color composite function expression

39

II- 13

Kappa Coefficient

41

III- 1

Simple difference change mapping

109

III- 2

Percent change map

109

  xxiii  

ABSTRACT

Wells, Tami Maureen, Ph.D., University of South Alabama, May 2010. Natural Disasters and Long-Term Recovery: A Baseline Study of Historical Change and Habitat Structure of Juncus roemerianus Marsh at Grand Bay, Mississippi. Co-Chairs of Committee: Just Cebrian, Ph.D., and Anne A. Boettcher, Ph.D. Ecosystem structure and function in a Juncus roemerianus dominated marsh in coastal Mississippi was investigated using remote sensing techniques. Rate of change in land cover types were computed from historical aerial photography (1940 to 2006) of the study area located in Kreole, MS. Baseline biodiversity and vegetation coverage were determined from in situ line intercept vegetation analyses. Post-classification difference maps for intervals between image time periods and baseline parameters were compared using 2006 vegetation transects and a 1 m resolution Digital Ortho Quarter Quadrangle (DOQQ) image as ground-reference. The ancillary photographs and digital imagery were rectified, digitized and interpreted from both analog and digital file formats using a combination of geospatial techniques. A Gaussian normalization technique provided a minimum standard deviation in pixel brightness among 1940, 1985 and 2006 images as compared to other image normalization techniques. Additional investigations were made from images with dates significant to Hurricane Elena (1985) and

xxiv  

Hurricane Katrina (2005) storm events in coastal Mississippi. Data collected from Mahalanobis distance supervised image classification and simple value difference maps demonstrated that change in marsh structure did occur between pre and post-storms at the study area in Mississippi. A classification error matrix showed accuracies ranging from 46.80% in the 1940 classification to 91.96% in the 2006 ground reference. The supervised post classification percent change of natural marsh for the Hurricane Elena (1985) imagery was + 5.25% Juncus roemerianus marsh, +2.47% pine hummock, +0.91% open-water, -7.14% salt panne, and +0.34% unclassified. The Hurricane Katrina (January 2004-March 2006) supervised post classification percent change was +13.62% Juncus roemerianus marsh, 13.3230% pine hummock, -2.90% open-water, +2.61% salt panne, and 0.0% unclassified. The percent change difference map for Hurricane Elena (natural) was 0%, 10.81%, 88.85%, and 0.34% respectfully with a mean of 2.8952% and standard deviation of 1.0683% across all classes. The Hurricane Katrina percent change difference was 0%, 11.68%, 88.32%, and 0% respectively with a mean of 2.8832% and standard deviation of 0.3212 %. The overall pixel change in pre and post storm images from Hurricane Elena were 10.34% of the pixels (2811 m2) and 11.68 % (3175 m2) for Hurricane Katrina.

xxv  

CHAPTER I – BACKGROUND

Introduction Coastal marshes are among the most important and highly productive ecosystems on earth (Keefe 1972; Cowardin et al. 1979; Peterson and Peterson 1979; Odum et al. 1984; Mitsch and Gosselink 1993; Adam1990; Finkl 2004). Only within the past forty years have coastal marshes received national attention and appreciation for their contribution to shoreline ecology (Knutson et al 1982) and capacity to protect inland areas by absorbing pollutants, trapping sediments and buffering waves (Valiela et al. 1978; Knutson et al. 1981; Boesch et al. 1994; O’Callaghan 1996; Baldwin and Mendelssohn 1998; Ainsfeld et al. 1999; Möller et al. 2002). Marsh functions that have received the most focus are fisheries habitat (Bortone 1976; Portier et al. 2000), storm protection (Zedler 2000), and mineral extraction such as crude oil, natural gas, sulfur and salt (Chabreck 1988; Beck et al. 2003). Previously treated as nuisances to be drained and filled, marshes are now protected by federal (Public Law 92500 Section 404) and state regulations (United States Army Corps of Engineers 1983). The loss of coastal marshes is well documented and is the result of both natural and human-induced impacts (Ensminger and Nichols 1957; Chabreck

  1  

and Palmisano 1973; Mendelssohn et al. 1983; De Leo and Levin 1997; Roman et al. 1997; Davis et al. 2004; Dahdouh-Guebas and Koedam 2006; Stone 2006). Hurricane damage (Dingler and Reiss 1995; Baldwin et al. 2001; Piou et al. 2006; Park et al. 2007; Edminston et al. 2008), land use change (Kim and Weaver 1994), and development (Antrop 2001) within coastal boundaries increase the need to develop linkages between ecological theory, conservation and restoration of critical salt marsh habitat (Lewontin 1969; Whisenant 1999; Valentine et al. (In Press)). Land that historically was functional tidal marshes is the primary aim for marsh re-establishment and restoration projects (Clements 1916; Odum 1969; Broome et al. 1988; Picket and McDonnell 1989; Bakker et al. 2002).

Salt Marshes Salt marshes are defined as the transitional areas between land and water where salinity ranges occur from near oceanic, 35, to brackish, 5 (Odum et al. 1984; Finkl 2004). Salt marshes in the northern Gulf of Mexico usually experience daily flooding tides that lead to fluctuations in salinity and temperature (Stumpf and Haines 1998). The frequency and extent of flooding (Wiegert et al. 1981; Seliskar and Gallagher 1983; Stout 1984) varies seasonally on the lunar schedule and based on storm events (Walker 2001; Peng and Pietrafesa 2003). In a salt marsh, the physical arrangement (spatial pattern) of flora and fauna (LaSalle et al. 1991; Levine 1998) is

  2  

dependent on an organism’s tolerance (Cowardin et al. 1979; Chabreck 1988; Fahrig 1997; Costa et al. 2003; Lindig-Cisneros et al. 2003) to periodic flooding (Zeff 1999) and hydrological gradients (Bolton 1991; Hughes et al. 1998, Kuhn et al. 1999) associated with changes in salinity, nutrients, oxygen availability and temperatures (Craft et al. 1988; Cai and Wang 1998). Salt marsh environmental gradients are complex and extremely variable (Chapman 1973; Stout et al. 1982; Bertness 1992; Pennings and Bertness 2001; Wallington et al. 2005; Benincá et al. 2008), where a slight change can alter the dominant vegetation (Chapman 1976; Brinson and Christian 1999; Mallin and Corbett 2006) and corresponding animal habitats (Adam 1990; Reed and Rozas 1995; Wilcox and Meeker 1995; Chambers et al. 1999; Angradi et al. 2001). The typical vegetative cover of a northern Gulf of Mexico salt marsh consists of thick mats of grasses, sedges and rushes. In most coastal regions, the prevalent vegetation associated with marshes commonly inundated or saturated by surface or ground water is Spartina alterniflora Loisel., smooth cordgrass (Rozema et al. 1985). S. alterniflora is a C4 plant that exhibits morphological adaptations (Seliskar et al. 2002) such as deep roots and rhizomes (Gallagher 1974), salt glands, thick foliage and strong culms with specialized vascular (aerenchyma) tissue that enables the plant to survive high temperature and frequent exposure to high salinity water (Sutcliffe 1962) found in the low marsh (Lindberg and Harriss 1973). However, along

  3  

the tidally influenced coastal marshes of Mississippi and Alabama, Juncus roemerianus Scheele, commonly called black needle rush, is the predominant vegetation (Eleuterius and Caldwell 1985; Tiner 1993; LaSalle 1996). Unlike the neighboring Spartina alterniflora dominated marshes of Louisiana, J. roemerianus marshes have a lower tidal inundation rate (Koretsky and Miller 2008) due to higher elevation of the marsh surface. J. roemerianus serves as an important stabilizer of shoreline structure and provides habitat for a variety of commercially and ecologically important fin and shellfish (Barnette and Crewz 1997). J. roemerianus is an easily recognized C3 rush that grows from horizontal rhizomes (Eleuterius 1976) and supports upright leaves. The leaves emerge from the marsh in solid stands and have a sharp point on the terminal end. The inflorescence extends from a simple bract supported by a single leaf (Duncan and Duncan 1987). J. roemerianus has a typical zonation pattern where in high salinity soil (Bouma 1963; Drury and Nisbet 1973; Eleuterius and Caldwell 1985; Dudgeon and Petraitis 2001) it is frequently interrupted by salt pannes and succulent vegetation adapted to high salinity (Woodell 1985; Woerner and Hackney 1997; Pennings et al. 2005). J. roemerianus has a high tolerance to calcium carbonate (CaCO3) and anaerobic conditions (USDA Plants, www.plants.usda.gov). In the northern Gulf of Mexico, J. roemerianus produces seed (Figure I–1) during the spring (April – May) (USDA Plants, www.plants.usda.gov). Commercial propagation for restoration includes seed germination (Somers and Grant

  4  

1981; Shumway and Bertness 1992; Huiskes et al. 1995; Noe 2002; Luna and Zedler 2007), tissue culture (Wang et al. 2005), and vegetative division. In the northern Gulf of Mexico, researchers are observing critical transitions in salt marsh ecosystems that may be largely due to external drivers (Dudgeon and Petraitis 2001; Suding 2001; Carpenter 2003) such as repetitive cycles of mild to severe hurricane (Zedler et al. 1986; Donnelly et al. 2001; Blake et al. 2005; Mallin and Corbett 2006) events compounded by changes in land use patterns (Rogers et al. 2009). Natural storms can often generate a regime shift (Wiegert et al. 1983; Paerl et al. 2001; Walker 2001; Mayer and Rietkerk 2004) in what is considered an already fragile and/or recovering marsh system (Scheffer and Carpenter 2003; Scheffer 2008). Inference to critical transitions in marsh ecosystems can be modeled (Scheffer and Beets 1994; Whisenant 1999; Carpenter 2003; Scheffer and Carpenter 2003; Carpenter and Brock 2006; van Nes and Scheffer 2007) by establishing vegetation thresholds based on biodiversity (Hacker and Bertness 1999; Beisner et al. 2003; Elmqvist et al. 2003; Wallington et al. 2005; Bestelmeyer 2006; Ives and Carpenter 2007; Walker et al. 2007), nutrient flux (Thoedose and Roths 1999; Pennings et al. 2005), herbivory (Cebrain and Duarte 1994; Salgado and Pennings 2005), and sediment (Gallagher 1980) profiles. In situ research coupled with both recent and archived remotely sensed imagery (Ramsey and Laine 1997, Kleinen et al. 2003) can enhance the understanding of complex ecosystem function (Collie et al. 2004) and structure (Holling and

  5  

Allen 2002; Jackson and Bartolome 2002). An example of in situ evaluations coupled with archived imagery is demonstrated in the adaptive management (Holling 1978; Thorn 1997) and ecosystem modeling (Leemans 1999) approach for conservation of longleaf pine (Pinus palustris) forest at Elgin Air Force base in Florida (Hardesty et al. 2000; Groffman et al. 2006). An ecosystem model was designed to capture the dynamics of landscape change over-time utilizing ancillary imagery and peer-reviewed field data collected by forest ecologists and managers (Peterson 2002). The outcome included integration of a computer model and collaborative baseline datasets which led to new management policies for longleaf pine forests based on predictability and ecological threshold-based (May 1977; Brown et al. 1999) strategies (Groffman et al. 2006) in both the ecological and social-ecological system (Walker and Meyers 2004; Groffman et al. 2006). Ecological thresholds are important in understanding the connectivity and spatial patterns (Burd 1992; Tyler and Zieman 1999; Ludwig et al. 2000) associated with ecological processes (Pennings and Bertness 2001; Biles et al. 2002; Groffman et al. 2006). The percolation theory (Stauffer and Anharony 1992) and neutral landscape model (NLM) (With and King 1997; Groffman et al. 2006) have provided insight into patch size (Krummel et al. 1987; Lindenmeyer and Fischer 2006), vegetation change (Roman et al. 1984; Roman et al. 2002), and habitat structure and function (Niering and Warren 1980; Roman et al. 1995; Hacker and Bertness 1999; Pennings and Bertness

  6  

2001) of coastal landscapes. In coastal remote sensing, understanding vegetative cover, successional trends (Del Moral 2000) and patch size can be the first steps to modeling and predicting (Bortone 2006; Edminston et al. 2008) the trends in conservation and restoration of coastal ecosystems. Marsh Restoration Restoration efforts attempt to amend degraded marshes returning them to naturally sustainable ecosystems by removing human-made hydrological controls (Bedford 1999), introducing dredge material (Lewis and Lewis 1977; Ford et al. 1999; Short et al. 2000), and trapping sediment transported from adjacent water bodies that will function as soil substrate (Cebrian and Duarte 1995) for the restored marsh (Gallaher 1974; Stovall 1982; Steiger and Gurnell 2003). The new sediment and containment structures must have suitable characteristics for the planned restoration project (Gallagher 1980). Improperly designed containment structures could increase surface elevations to levels that are unsuitable for the desired restoration vegetation (Orson et al. 1992). In addition, since the vegetative structure of intertidal marshes changes with elevation and environmental gradients (Mendelssohn and Marcellus 1976; Seliskar 1980), competition and diversity of species are both important factors in salt marsh restoration (Choi and Wali 1995; Mullineaux et al. 2003; Seabloom et al. 2003; Williams 2003; Lytle and Poff 2004; Lulow 2006). Thus, for restored marshes to reach levels of productivity (Gallagher et al. 1980) and biological diversity similar to those of native

  7  

marshes, native environmental parameters must be closely replicated (Young et al. 2005; Hobbs and Suding 2008). As a consequence, an understanding of environmental conditions in both natural marshes (Fench and Moore 2003; Corbosiero et al. 2005) and restoration sites will play a vital role in the restoration design (Gallagher 1975; Roman and Dailber 1989; Louisiana Executive Summary 1997; Mullineaux et al. 2003; Lytle and Poff 2004; Suding et al. 2004; Scröder et al. 2005; Wallington et al. 2005, King and Hobbs 2006). Research Site The Grand Bay National Estuarine Research Reserve Systems (GNDNERR) is a partially enclosed system where salt and fresh water mixes to form brackish water. The National Estuarine Research Reserve (NERR) program evolved from the Coastal Zone Management Act (CZMA) in 1972, as a partnership between the National Oceanic and Atmospheric Administration (NOAA) and coastal states. The NERRS are a network of 27 protected estuarine habitat areas established for long-term research and protect more than 5,394 square kilometers (km2) of essential coastal habitat (National Estuarine Research Reserve System-Strategic Plan 2005-2010). The GNDNERR study area is located in the Grand Bay southeast (SE) Quadrat in Jackson County, Mississippi (Figure I–2). The reserve encompasses a coastal plain that exhibits gradual sloping in elevation extending from north to south. It is comprised of approximately 72 km2,

  8  

which includes both tidal and non-tidal wetland habitats (Figure I–3). The specific site for this research is located in Kreole, MS. The elevation at the center Global Positioning System (GPS) reference point is approximately 1 meter (m), the Universal Transverse Mercator (UTM) projection zone is 16, the grid coordinates used to subsection the study area (Figure I–4) are; Xmin 364929.7600, Xmax 365125.7600, Ymin -3365107.4000, and Ymax 3365303.4000 (NAD 83 Decimal Degree DD). The site elevation was obtained from the National Elevation Data (NED) Contiguous U. S. 1/3E Arc Second Elevation Data, (USGS Seamless Data Portal – file: USGS 16R CU 65106 3365201). The study area was selected because of the extensive monoculture stands of J. roemerianus and salt pannes both of which are adjacent to a pine hummock. Tidally influenced bayous surround the study area on all three sides. The study area is classified as a buffered area by the GNDNERR management team and is considered to provide habitat protection for estuarine dependent species. The study area is part of the Grand Bay National Wildlife Refuge and is managed by the U.S. Fish and Wildlife Service (USFWS). The GNDNERR provides researchers with a landscape mosaic that is ideal for exploring scientific hypotheses and quantitative simulations, or models, of landscape processes such as soil erosion, hydrological processes, and vegetation succession (Ervin and Hasbrouck 2001). Remote sensing

  9  

imagery and Geographic Information Systems (GIS) initiatives are significant to the advancement of landscape models especially when defining influences such as coastal development and natural storm events (Adger et al. 2005). Remote Assessment of Marsh Damage and Loss By mapping and monitoring the distribution of coastal vegetation, the effects of natural and anthropogenic impacts and interactions (McDonnell and Pickett 1993; Swart et al. 2001; Weinstein 2007) on a salt marsh ecosystem can be better addressed (Pfadenhauer 2001; Litaker and Tester 2003) and the potential for restoration evaluated (Levine et al. 2003; Faulk et al. 2008). Remote sensing of digital imagery improves the accuracy in detection of plant populations (NOAA 2007) and former habitat (Everitt et al. 1995; Allen-Diaz et al. 1998), and can assist in targeting areas for conservation and restoration (Tuxen et al. 2008). Remote sensing instruments provide a spatially referenced representation of vegetation diversity by measuring the flux of radiation from the earth’s surface (Curran 1994; Shanmugam et al. 2003). Remote sensing and GIS are designed to acquire, process, and manage spatial information (Michalak 1993; Caloz and Collet 1997; Williams et al. 2003). The recent advancement in these computer-based technologies has stimulated researchers to develop new methods (Wheeler 1993; Lehmann and Lachavanne 1997) for detection of slight changes to the environment (Dobson 1993; Hinton 1996; Ramsey et al.

 10  

1997; Elvidge et al. 1998; Hudak and Wessman 1998). Airborne, hyperspectral remote sensing imagery is a useful tool in collecting near to real-time environmental monitoring data in coastal marshes (Ramsey 1998; Lunetta et al. 2002; Williams et al. 2003). The capacity to obtain image data in many spectral bands is beneficial to species classification (Goetz et al. 1985; Cocks et al. 1998) and automated mapping of vegetation. The interfacing of remote sensing and GIS data have been challenging (Westmoreland and Stow 1992; Lehmann and Lachavanne 1997), yet the potential to automate the extraction of geographical information from remote sensing data (Lehmann et al. 1994) have become more practical with the development of new pixel (raster) based, GIS modules (Michalak 1993; Wilkinson 1996; Hudak and Wessman 1997; Tischendorf 2001). Algorithms designed to adequately map the distribution and abundance of aquatic plant species are needed so that associations between environmental biodiversity and biological and physical factors can be made (Lunetta et al. 2002; Williams et al. 2003). Geospatial technology offers descriptive mapping applications that contribute to integrated approaches to ecological modeling (Delaney and Webb 1995; Lehmann et al. 1997; Hanna and Kulpepper 1998). Traditionally, ground based measurements were used to survey biodiversity of aquatic vegetation (Chapman 1976) but today remote sensing combined with GIS mapping has the potential to offer a more efficient and time saving method for achieving a large scale picture (Goetz 1992; Green and Morton

 11  

1994; Goetz 1995; Ramsey et al. 1997; Green et al. 1998; Hirano et al. 2003; Valta-Hulkkonen 2003). Yet it is important to note that limitations to remote observations exist where image interpretation algorithms may differ greatly in regards to spatial resolution, spectral wavelength and temporal timeframes. Aerial imagery from decades ago is an excellent source for compilation of historic geological data and changes in shorelines, (Nicke et al. 1997) however the spectral information is often lost by years of decay and during digitization of the imagery (Singh 1989; Green et al. 1998). In addition, Normalized Difference Vegetation Index (NDVI) algorithms cannot be performed on ancillary aerial imagery and some present day electro – optical (EO) multispectral imagery (MSI) due to the absence of spectral bands in the visible and red wavelengths (Fung and Siu 2000; Jensen 2000; Stefanov and Netzband 2005). Change information is important for local and regional environmental monitoring and decision-making (Estes 1992; Dobson et al. 1995). The United States has lost more than 53 percent of wetland habitat to agriculture, residential development and commercial land use in the 200 years spanning 1780’s-1980s (Dahl 1990). To determine what successional changes are taking place in salt marsh vegetation (Elvidge 1990; Hobbs 1993), scientists are utilizing seasonal and multiple year digital imaging for analyzing change detection and patch densities of landscapes (Jensen et al. 1987; Jensen et al. 1995; Fahrig 1997; Ramsey and Laine 1997).

 12  

Ancillary in situ datasets are equally critical to investigations concerning change detection in coastal marshes (Christensen et al. 1987; Häme et al. 1998; Mas 1999) allowing for ground-referencing of images with regional topographical maps considered one of the most important datasets in coastal science research (Chavez and MacKinnon 1994; Ramsey et al. 1998; Ramsey et al. 2001). Topographical maps provide information regarding elevations and geomorphology of coastal as well as predicted plant species distributions for an area such as GNDNERR. Research that correlates imagery and a priori environmental information into the investigation will normally provide a better understanding of ecosystem function and structure over time (Gustafson and Parker 1992). In an effort to support field and remote sensing investigations, the National Estuarine Research Reserve (NERR) System has created a Centralized Data Management Office (CDMO) located at North Inlet-Winyah Bay NERR in South Carolina. The CDMO supports the NERRs System-wide Monitoring Program (SWMP) at 27 reserves in the U. S. and Puerto Rico. This near realtime (NRT) monitoring system allows scientists to collect monthly nutrient, water quality and meteorological information (NOAA 2004) that will greatly enhance interpretation of real-time or near-real-time imagery (National Research Council 2004; Small 2004; Sault et al. 2005; NOAA 2007a; NOAA 2007b; Scott et al. 2009) and support research efforts regarding coastal

 13  

restoration and NOAA Height Modernization efforts (Zilkoski and Hothem 1989; Zilkoski et al. 1997; Shinkle and Dokka 2004). The GNDNERR staff continues to advance technologies that will better predict the variables associated with natural trends, storm events and coastal development. Baseline datasets coupled with near-real-time monitoring sites and ecosystem models will certainly provide important information regarding impacts associated with specific storm events when compared to those associated with coastal development and global climate change dynamics. The GNDNERR study site has experienced several natural storm events in the past decade. Although, these storms were estimated to have extensive winds and storm surge, the GNDNERR ecosystem appears to have buffered the impacts in that it remains a diverse and fully functional ecosystem. Further study into the resiliency of these marshes will enable better conservation and restoration planning for similar marshes in the northern Gulf of Mexico. Hurricane Impacts Along the Northern Gulf Coast Coastal wetlands have sustained damage due to increased development along our coastal shores as well as the impact of storm events (Engle 1948; White and Pickett 1985; Tedesco 1995). In the past 100 years, the coast of Mississippi and Alabama have experienced five major hurricane events (Table I–1, Figure I–5), Elsner and Kara 1997). For the purpose of this research, we have identified “major storms” as those reported as category 3-5

 14  

according to the Saffir-Simpson Hurricane Scale when making landfall along the Gulf coast (Saffir et al. 1972; Ho et al. 1975, Ho and Tracey 1975; United States Weather Bureau 1986). Geological and verbal records suggest that in 1906 and 1916 (Mitchell 1928; United States Weather Bureau 1986; Kleinberg 2004), unnamed storms, impacted the coast of Mississippi with extreme force. Limited data are available as to how much damage was incurred from the impact of these two storms. In 1969, Hurricane Camille devastated coastal Mississippi (Cristwell 1970; Hsu et al. 1970; Saffir et al. 1972) and Hurricane Elena damaged the Mississippi coastal marshes in 1985 (Sullivan 1986; Conner et al. 1989; Sparks 1991), with high winds and high storm surge. Hurricane Katrina made landfall on the Gulf coast of Mississippi at Pearlington, near the Louisiana-Mississippi boarder. Katrina severely impacted the Mississippi coast by bringing with it a storm surge estimated at 26-28’ nearest landfall and 10-14’ at the Mississippi-Alabama boarder (Holmes et al. 2006; Blackwell et al. 2007). The largest ecosystem impact of Hurricane Katrina was from the excessive storm debris (i.e. household, man-made infrastructure, and organic) deposited into the sensitive coastal marshes (Mitchell 2006; Poirrier et al. 2008; Rodgers et al. 2009) of Mississippi. The GNDNERR as noted earlier has remained functional and the structure of the marsh system appears to exhibit natural succession despite the Reserve’s proximity to landfall of recent hurricanes in both Alabama and

 

15  

Mississippi (Stewart 2004; Chen et al. 2008). This research further investigates the natural trends and potential storm dynamics associated with a Juncus roemerianus marsh ecosystem located within the tidal boundary of GNDNERR in Mississippi.

Research Summary This research uses a geospatial approach to examine the distribution (Hupp et al. 1995; Friedrichs and Perry 2001) of physical and biological factors associated with the structure of J. roemerianus marsh (Teal 1962; Pomeroy et al. 1972; Gabriel and de la Cruz 1974). Remote sensing technology was correlated to vegetation diversity and patch density. The in situ information enhanced quantitative and descriptive approaches to mapping the boundary of marine and terrestrial biomes (Zhan et al. 2002; van de Koppel et al. 2005). Finally, change of land cover was investigated from images at timeframes that included hurricane impacts that brought forth extreme floods and winds to the region.

Scope and Organization of Dissertation The primary goals of this study were to: 1) establish a baseline dataset of ancillary imagery and elevation dynamics of J. roemerianus vegetation in a salt marsh ecosystems, 2) investigate the function and structure dynamics of J. roemerianus marshes using applied geospatial technologies and standard

 

16  

vegetation biodiversity indices, and 3) detect and quantify the change in stands of J. roemerianus associated with hurricanes using remote sensing of multispectral imagery. The following hypotheses were tested: Ho1: Taken together  grayscale  digital  imagery  from  an  airborne  platform  scaled   to  ~1  meter  resolution  and  in  situ  data  provide  valuable  information   regarding  alterations  and  historical  trends  of  coastal  marshes  at  Grand  Bay   National  Estuarine  Research  Reserve  (GRDNERR)  in  Kreole,  Mississippi. Ho2: Change  detection  map  produced  from  greyscale  supervised  Mahalanobis   distance  classifications  will  successfully  detect  impacts  to  the  landscape  post-­‐ hurricane  events.  

 

17  

Table I-1. A comprehensive recording of Gulf of Mexico (GOM) hurricanes that significantly affected coastal Alabama and Mississippi were recorded from 1950 - 2005. Saffir-Simpson Hurricane Intensity Scale: 1 to 5 rating utilizes the standard wind measure, as storm surge values are dependent on the geological slope associated with the continental shelf in the landfall region. Scale number represents: TS – Tropical Storm; 1– central pressure >980 millibars (mb), winds 64-82 knots (kn), 2 – central pressure 965-979 mb, winds 83-95 kn; 3 – central pressure 945-964 mb, winds 96-113 kn; 4 – central pressure 920-944 mb, winds 114-134 kn; and 5-central pressure 134 kn.

Date 31-Aug-50 15-Sep-60 17-Aug-69 23-Sep-75 12-Sep-79 02-Sep-85 03-Aug-95 04-Oct-95 20-Jul-97 28-Sep-98 16-Sep-04 10-Jul-05 30-Aug-05

Name Baker Ethel Camille Eloise Fredric Elena Erin Opal Danny George Ivan Dennis Katrina

Landfall Gulf Shores, AL Biloxi, MS Waveland, MS Fort Walton Beach/Panama City, FL Mobile (Dauphin Island), AL Biloxi, MS Santa Rosa Island, FL Navarre Beach, FL Fort Morgan, AL Ocean Springs, MS Gulf Shores, AL Santa Rosa Island, FL Pearlington, MS

 18  

Lowest Pressure (mb) 991 981 909 955 946 959 976 942 984 961 910 930 928

Wind (kn) 74 51 139 90 114 91 70 109 70 91 100 100 90

SaffirSimpson Scale 2 TS 5 3 4 2 1 3 1 3 3 3 3

 

 

 

 19  

Figure I-1. The flower of Juncus roemerianus Scheele (photo a), 1500X magnification (Zeiss) of J. roemerianus seed encased in pod (photo b), and magnification of seeds adjacent to a U.S. minted dime (photo c), photo: Tami Wells, University of South Alabama.

C

B

A

20 Figure I-2. A TerraMetrics Landsat image of the northern Gulf of Mexico shows the location of the Grand Bay National Estuarine Research Reserve (GNDNERR) site at (30.37N, -88.40W), source: Google Earth 2006.

   

Figure I-3. A map of the Grand Bay National Estuarine Research Reserve located in the SE Quadrat in Jackson County, MS. The lands within the Grand Bay NERR are classified as either core areas or buffer areas. The core area (outlined by yellow on the above map) consists of approximately 12, 800 acres of estuarine tidal marsh, tidal creeks or bayous, shallow, open-water habitats, oyster reefs, seagrass beds, maritime forests, salt panne, sandy beaches and shell middens. The buffer area (outlined by blue above) consists of approximately 5,600 acres of tidal marsh, scrub shrub, pine flatwood and wet pine savanna habitats, source: Grand Bay NERR.

 21  

22   Figure I-4. A 2009 United States Department of Agriculture (USDA) Farm Service Agency (FSA) image of the natural (solid line) and impacted (dashed line) research sites at Grand Bay National Estuarine Research Reserve (GNDNERR) located at (30.37N, -88.40W), source: Google Earth 2009.

   

Figure I-5. An eastern map of the United States of America identifying the major landfalling hurricanes between the years of 1900-2005, source: (September 2007) Map © Michael A. Grammatico -2004 http://www.geocities.com/hurricanene/gulfcoast.htm

 23  

CHAPTER II – GRAYSCALE IMAGE TECHNIQUES FOR DETERMINING LAND COVER CHANGE IN A GRAND BAY, MISSISSIPPI SALT MARSH

Introduction Coastal marshes are one of the most significant components in supporting equilibrium among the land, ocean and atmosphere (Palmer et al. 2004). Therefore, conservation and restoration of coastal marsh habitat are a high priority locally, nationally and globally (Bortone 2005). Impacts to salt marsh vegetation can occur from a number of events both natural (Donnelly et al. 2001; French and Moore 2003) and anthropogenic (Bedford 1999; Alberti et al. 2004). The destruction and resiliency of salt marsh vegetation is difficult to measure especially when the impact experienced is short lived (Bortone 2006; Edminston et al. 2008). Simply meaning that the storm impact is not ongoing as would be the impacts from land use change or impaired waterways. Today, non-native species competition (Choi and Wali 1995; Mullineaux et al. 2003; Seabloom et al. 2003; Lytle and Poff 2004; Lulow 2006), marsh fragmentation, and adverse climate change (Duarte et al. 1992) are accelerating (Hobbs 1993; Hobbs and Suding 2008), causing regime shifts, (Cebrian 1999; Córdova-Kreylos et al. 2006) oscillatory behavior and cross-

 24  

scale interactions (Adams and Bortone 2005). Indications of these shifts include changes in ecosystem dynamics (Benincá et al. 2008) such as, increased nutrient loading (Cebrian et al. 1998), fire frequency (Mack et al. 2001; Williamson et al. 2006; Walker and Reddell 2007) and grazing intensity (Zedler and Callaway 1999; Bartha et al. 2003; Hobbs and Suding 2008) all of which are considered natural disturbances. Pollution, development, recreational use, dredging and construction are only a few of the anthropogenic impacts that alter coastal marsh systems (Davis et al. 2004; Adger et al. 2005). Anthropogenic impacts are increasing due to increased population and development of coastal lands (Davis et al. 2004). Ground reference data are extremely important to studies of land cover change. The accuracy of change detection from remote sensing interpretation can be over or under-estimated (Foody 2009) depending on the error in ground reference data, lack of data, time between ground reference and image collection, or correlation of ground reference data to the image spectral or spatial quality. Interesting correlations between vegetation biodiversity and the landscape mosaic (Phinn et al. 1999) can be made when sampling in coastal waters (Eastwood 1997; Underwood et al. 2000). Biodiversity refers to the variety of life forms within an ecosystem (Straker and Platt 2001). Specifically, vegetation biodiversity refers to the genetic species and ecosystem diversity of plants found within a defined area. Understanding

 25  

the biodiversity of a salt marsh is important when describing ecosystem function and structure, biological resources, and social benefits (i.e. commerce, esthetical, and recreational) of surrounding townships or communities. Research suggests that high biodiversity is correlated directly to an ecosystem’s resiliency to damage and destruction from natural storm events, and essential habitat recovery (Steneck et al. 2002; Allison 2004; Beaumont et al. 2008). Landscape mosaic refers to the patchiness of vegetation, sediment, manmade impacts such as utility right of ways, roads, parking areas, residential homes or camps, and industrial footprints such as offshore oil and gas reserves, pipelines and commercial fishing docks. There are many man-made features that can change a wavy, meandering natural flow to an obvious linear and geometric pattern in the landscape. Landscape change and modification are rarely random and often target the most productive areas of the ecosystem (Ervin and Hasbrouck 2001; Lindenmayer and Fischer 2006). Determining the amount of change and visualizing alterations in coastal marshes is an expensive and time-consuming task (Jensen 1996). Remote sensing via aerial photography provides a way to both quantify and visualize these changes (Gallagher et al. 1980; Burd 1992; Schmidt and Skidmore 2003) without disturbing the marsh or undertaking lengthy on site observations and measurements. As previously mentioned, the topography of coastal landscapes are constantly changing due to effects of natural storm

 26  

events, ecosystem dynamics and anthropogenic influences associated with coastal development (Boumans et al. 2002). Landscape scale change detection describes the alteration of habitat structure and function over a specific timeframe (Silvestri and Marani 2004). This ecological concept is a valuable tool for environmental managers, researchers and planners (Estes 1992; Dobsen et al. 1995). Most often, change in vegetation is a response to disturbance and successional trends for a given location (Schmidt and Skidmore 2003). In ecosystems that are relatively untouched by management, a shifting landscape mosaic (Aubreville 1938) remains relatively constant where a balance is maintained between disturbance and succession on a large landscape scale. In this scenario, the human impacts do not exhibit shifting mosaic tendency, yet can be a catalyst in disrupting population structure, resource availability and other ecosystem dynamics (Leitão et al. 2006). To better understand the dynamics of coastal marsh ecosystems, scientists should examine successional trend (Parker 1997), effects of both natural and human induced disturbances, as well as other drivers of landscape change (Tuxen et al. 2008). Today, researchers are utilizing temporal and multiple years of digital imaging as a tool for analyses of baseline features and identification of substantial alterations of the ecosystems over time (White and Pickett 1985; Christensen et al. 1987; Phinn et al. 1996; Smith et al. 1998; Simenstad et al. 2006).

 27  

Several image interpretation techniques (Civco et al. 2002; Coppin et al. 2004) have been utilized to detect change in landscape matrices (Singh 1989; Baker et al. 2007; Tuxen et al. 2008). In 2005, post Hurricane Katrina, the United States Geological Service (USGS) began baseline studies to identify the direct impacts associated with marsh shearing, ripping and scouring by utilizing Landsat Thematic Mapper (TM) images to detect landscape change over-time (Figure II–1).

Problem In 2002, a project entitled “100 years of land change in Louisiana” was initiated as a collaborative effort between the USGS National Wetland Research Center (NWRC) in Lafayette and Baton Rouge, and the Louisiana Coastal Area (LCA) Land Change Study Group. This project utilized moderate resolution satellite imagery to quantify trends of land and water changes over time. The study group determined that difference and classification mapping from digital imagery provide a good estimate of land change over time. In the LCA study, land loss in Louisiana between the years 1956-78 was quantified to be 102.04 km2 (39.4 mi2)/yr; years 1978-90 as 90.39 km2 (34.9 mi2)/yr, and 1991-2000 as 113.96 km2 (44.0 mi2)/yr. (Figure II–2). The quantification of land loss in Louisiana was an image-derived assessment and extremely valuable information due to the inability to traverse and survey the extensive marshes in South Louisiana. The LCA

 28  

group also noted that difficulty of interpretation existed when distinguishing between ephemeral and permanent trends after storm events (Barras 2006). Other disadvantages acknowledged in the study were the difficulties associated with identifying trends during extreme events including water level differences, seasonal patterns such as senescence of vegetation, conventional weather events, and limited availability of datasets and metadata. Image interpretations in change detection analyses are often purely observational and difficult to quantify due to limited control in data collection and experimental design (Fisher and Pathirana 1990). Question In this research, I tested the hypothesis that together grayscale digital imagery from an airborne platform scaled to ~1 meter resolution and in situ data provide valuable information regarding alterations and historical trends of coastal marshes at Grand Bay National Estuarine Research Reserve (GRDNERR) in Kreole, Mississippi.

Methods

Study Area The study area for this research is located within the Grand Bay National Estuarine Research Reserve Systems (GNDNERR) that is a partially enclosed system where salt and fresh water mixes to form brackish water.

 29  

The GNDNERR encompasses a coastal plain that exhibits gradual sloping in elevation extending from north to south. The study site encompasses a tidally influenced salt marsh where homogeneous stands of Juncus roemerianus, black needle rush dominate the low marsh with a gradual mixing of J. roemerianus and Distichlis spicata found in the mid marsh and adjacent to an extensive salt panne. The upper marsh is a pine hummock with a scrub shrub understory. The apparent zonation of the tall and short vegetation types of J. roemerianus mediate the landscape transition from salt marsh to pine hummock. The study site is considered to be a low impact environment and represents the full range of vegetation types (Margules and Pressey 2000, Lindenmayer and Burgman 2005, Lindenmayer and Fischer 2006). Description of Imagery and Acquisition Platforms Multi-year imagery of the Kreole, Mississippi research area was acquired from local conservation offices, federally sponsored Internet Management Services (IMS), and other government sponsored geospatial data warehouses. Hardcopies of ancillary aerial photography were acquired from United States Department of Agriculture (USDA) Natural Resource Conservation Service (NRCS) in Mobile, Alabama (1940 & 1980 imagery) and the United States Geological Survey (USGS) Earth Resources Observation and Science (EROS) in Sioux Falls, SD (1985, 1992, 1996, and 1997 imagery). The Digital Ortho Quarter Quadrangle (DOQQ) data (2002 & 2008) were obtained from USGS via the Louisiana Wetland Center (LWC) Internet Management System

 30  

(IMS) portal. The USDA Farm Service Agency (2009) imagery (1.0 m) were accessed from Google Earth (earth.google.com). The earliest imagery acquired included data from 1980 – 1987 National High Altitude Photography (NHAP) and 1:40,000 scale imagery from 1987 – 1997 National Aerial Photography Program (NAPP). The NHAP and NAPP imagery were collected on film from a fixed winged aircraft. This data as originally collected was not in a digital format. Image Normalization and Georeferencing The vertical black and white (B&W) hardcopy 1940, USDA/NRCS and 1980, NHAP were scanned and digitized (Novak 1992) at 500 dots per inch (dpi) resolution using a Hewlett Packard Photo/Scanner/Copier (PSC) 1350 linear array desktop scanner. The 1980 NAHP imagery were scanned from 1:24000 hardcopies at 500 dpi. The converted hardcopy to digital imagery were saved in both bitmap (.bmp) and Joint Photographic Experts Group (.jpeg) and later converted to a geoTIFF file format containing the georeferenced information (.jpeg2000 is another file option). Most of the NAHP and NAPP imagery (1982-1997) were digitized at EROS using a microdensitometer (Light 1993) and delivered digitally in Tagged Image File Format (TIFF). The 2006 imagery were downloaded from the USGS Seamless Data IMS portal (USGS 2010) at a 1m digital resolution. The 2008 image was downloaded from Google Earth IMS portal (Google 2010) at a 1m digital resolution.

 31  

The computation for determining pixel ground resolution (Jensen 1996) from scanned hardcopy imagery is as follows: Equation II-1

PM = (S /DPI) /39.37



where, PM is the pixel size (meters), S is the image scale (i.e. 1:20,000, 1:58,000), DPI is the dots per inch scan rate, and 39.37 is the conversion representing inch = meters x 39.37. A greyscale subset of 2006, digital orthophoto quarter quadrangle (DOQQ) imagery (Figure II–3) was obtained at 1m resolution by saving red, blue and green (RGB) visual band channel versions of each image as a TIFF file and converted from the TIFF files to greyscale using Environment for Visualizing Imagery (ENVI) ™ v. 4.2 (ITT, Inc, Boulder, CO) software. A greyscale image has a single value per pixel that represents the reflectance intensity where white represents the brightest intensity and black the lowest intensity. Greyscale images are also known as monochromatic images. The greyscale imagery used in this research is scaled to 8 bits per sampled pixel lending to a linear range of intensities or brightness values (BV) from 0 (the lowest) to 255 (the highest). A standard calculation for changing an image from a RGB composite image to greyscale (Johnson 2006) is as follows:

Greyscale( LinearBV ) = 29.9%( Rband ) + 58.7%(Gband ) +11.4%(Bband )



 32  

Equation II-2

where; Linear BV = Linear Brightness Value, and Rband, Gband, and Bband represent Red, Green and Blue (RGB) spectral bands. All of the un-registered images dating from 1940 to 1997 were georeferenced with ENVI v.4.2 software where the 2006 DOQQ imagery served a baseline for image-to-image (Richards 1999) coordinate (latitude and longitude) matching. Ground control points (GCP) of like objects were selected in each image. The un-registered image (warp image) was matched to the 2006 DOQQ (base image) by aligning the GCP between the nonreferenced (warp image) and the 2006 DOQQ geo-referenced image. The warp method utilized a second-degree polynomial (GCPs) > (degree + 1)2 across all pixels using an area-weighted mean and nearest neighbor distance metric algorithm that is automated in the ENVI v. 4.2 software package (Jensen 2000, Leitão et al. 2006). Many of the GCPs were selected from “soft” match points (trees, middens or other geological anomalies) because of the nature of the marsh and lack of manmade objects found within the earliest dated images. Therefore, our effort was focused on selecting as many “hard” match points (bridges, buildings, towers, and other manmade infrastructure) as possible from each image. A minimum of 98 match points were used for geo-referencing and enforcing my analyst rule that a minimum of two points must fall within each square of a 7x7 grid overlaying the image. The images analyzed in this research were geo-referenced to the North American Datum 1983 (NAD 83) coordinate system and projected as a

 33  

Universal Transverse Mercator (UTM) -Zone 16 (secant) map or image. After accurately geo-referencing the images (1940 – 2008), three spatially referenced subsections (1940, 1985 and 2006) with identical geo-referencing corner points were observed side by side for manual on screen change detection. This technique is beneficial in small-scale observations and between long periods of time. Changes in land-cover type (LCT) were observed among the 1940, 1985 and 2008 greyscale images prior to processing with automated algorithms (Figure II–4). Many changes could be visualized prior to computer processing and served as analyst reference of change and aided in accuracy assessments in the write function memory insertion, difference mapping and supervised post-classification ENVI automated algorithms described in more detail below. Change Detection Vegetation Transects. Six transects measuring 25 meters in length were randomly placed within the study area (Table A-1). To prevent biased interaction with naturally occurring mosaic in the vegetation, standard transect directionality used in most transect sampling schemes was avoided. A plant species inventory was collected at 0.5 meter intervals using a standard, continuous line intercept sampling technique (Pielou and Routledge 1976; Burnham et al. 1980). The density of J. roemerianus was extrapolated from each transect utilizing stem count data collected from three quadrats measuring 0.25 m2 that were randomly tossed along each

 34  

transect (Hill 1973). Vegetation height of three bracts was randomly sampled within three quadrats along each of the six transects. The only vegetation measured was J. roemerianus and each measurement was made from the soil surface to the tip of the bract. The vegetation transects completed on June 3, 2006, were georeferenced and digitally represented in a Geographic Information System (GIS) vector graphic overlaid on August 7, 2007, Farm Service Agency (FSA) true color imagery and also correlated to digital imagery, both ancillary and newly acquired, of the study areas. Species frequency, density and percent cover were calculated from the in situ vegetation data by applying the following equations to each transect:

Frequency = __number of intervals containing species__ total number of intercept crossing intervals

Equation II-3

Relative = Frequency

frequency of species total frequency of all species

Equation II-4

Equation II-5

Relative Density

=

total # of individual species total # of individuals of all species

Percent Cover

=

total intercept length of a species total length of transect

X 100

Equation II-6

Relative Cover

=

total intercept length of a species X 100 total intercept length of all species

Equation II-7

 35  

A species density eveness index (Pielou 1966) and biodiversity (Lloyd et al. 1968; Shannon and Wiener 1963) and dominance (Simpson 1949) calculations were utilized to determine how similar or dissimilar the vegetation was among transects. The equations utilized to determine biodiversity and density were as follows: Shannon-Wiener, H(s):

Equation II-8

  S $ ' H(s) = C /N %(N log10 N) " # n i log10 n i ( & ) i"1

!

where C = 3.321928 (constant used in converting log10 to log2), N = the total number of individuals, ni = the number of individuals in the ‘ith’ species, and S = the total number of species.   Pielou's evenness index, j:

Equation II-9

  j = H(s) /H(max)  

!

where H(s) = the Shannon-Wiener function, H(max.) = the theoretical maximum value for H(s) if all species in a transect are equally abundant. Simpson's dominance index, c:

Equation II-10

s

c = # (n i / N) 2 i"1

where  ni  =  number  of  individuals  in  the  ‘ith’  species,  N  =  total  number  of  individuals,   !

and  S  =  total  number  of  species.

 36  

Write Function Memory Insertion. ENVI v.4.2 was used to perform an on screen change detection by layering greyscale images into a color image. Each image was opened and saved as an ENVI image (.img) and header (.hdr) file and viewed in the Available Band List of the software function. In the Available Bands List, the RGB Color toggle button was selected and each image was assigned to an individual R-G-B channel in a sequence of the earliest dated (time t) image as band 1 (red), the mid date image (time t+1) as band 2 (green) and the latest dated image (time t+2) as band 3 (blue) to create a RGB image. The RGB image then represented a false color display where blue, green and red image bands are associated with different dates. The false color image display was saved for later use in change detection analyses as .jpeg and ENVI .img data file formats. In this research, multidate imagery change detection investigated a write function memory insertion of 1940, 1985 and 2006 greyscale imagery to produce a RGB image. This technique allowed us to view and discriminate among three years of remotely sensed data with one displayed image. For the write function memory insertion change algorithm, the 8-bit image normalization is automated in ENVI v 4.2 and is accomplished by subtracting the minimum intensity number (ranging from 0-255) in the greyscale image from the value of each image pixel and then dividing by the range of pixel values within the 8-bit image (ENVI, 2003). Difference Mapping. In this technique, the greyscale images (1:24,000 scale)

 37  

were scaled to the same spatial resolution (pixel = 1.1063 x 1.0300 meters or 1.1395 m2) with the imagery covering only the natural site (natural occurring materials) measuring 197 rows x 138 (27,186 pixels) and the natural site plus anthropogenic impacts (man-made components) measured as 527 rows x 400 columns (210,800 pixels). All of the imagery for difference mapping and classification was normalized so that the data value of each pixel ranged between 0 and 255. ENVI v. 4.2 automated Gaussian, equalization, square root and 2% linear normalization contrast stretch techniques were applied to 1940, 1985, and 2006 imagery and basic statistical information (range, mean and standard deviation) were calculated for each image to determine the best contrast stretch among the three dates. After selecting a filter normalization technique for the imagery, a difference maps was created comparing the images from 1940-1985, 19852006, and 1940-2006. The calculation for determining spectral change vectors was modified from Virag and Colwell, (1987), where a change detection map is equal to the sum of the absolute value of the differences between corresponding pixels in two time dated greyscale images (Figures II– 5 and II–6). The modified calculation is as follows: Equation II-11

    where PVijk is the change map pixel value, [BVijk]t and [BVijk]t+1 are the

brightness values for time (t) and time (t+1) in the greyscale imagery, and c is

 38  

the mean value of the greyscale range (127), i is the line number, j is the column number, and k is equal to one band (greyscale 0-255) (Singh 1989; Jensen 1996). Post Classification Difference Image. In order to investigate the change differences associated with Mahalanobis Distance classification mapping, all imagery must be transformed to a color RGB image. Because some of the earlier imagery (1940-1985) utilized in this research was only available in greyscale (0-255), it is mathematically impossible for us to convert back to a RGB image because of the three variables involved in creating a color image, R (red band), G (green band) and B (blue band). The true color composite function expression is as follows: Equation II-12     where, BV = brightness value of pixel in greyscale (0-255), A is the coefficient 29.9%, B is the coefficient 58.7%, and C is the coefficient 11.4%. ENVI v. 4.2 has an automated synthetic color algorithm that transforms a greyscale image into a color image by applying high pass and low pass filters to the greyscale image that separates the frequency data into low and high values. The low frequency data are assigned to the hue (H) and the high frequency data are assigned to the value (V) and a final fixed saturation (S) level is determined (ENVI 2003). These hue, saturation, and value (HSV) data are transformed to represent the red, green and blue (RGB) bands,

 39  

producing a synthetic image. (Daily 1983). Synthetic color images were created for both greyscale subsections (natural and impacted) and for the years 1940, 1985 and 2006. The high kernel was set at 15 and the low pass kernel was set at 10 and the saturation was set at 0.5. Difference images were created from the color-transformed imagery (1940, 1985 and 2006, natural and impacted subsets) using Mahalanobis distance post image classification in ENVI v 4.2. The Mahalanobis distance supervised classification is a direction sensitive distance classifier that utilizes the statistical differences between the endmembers from specific regions of interest and the associated pixel collection. The Mahalanobis distance classification is very similar to the maximum likelihood classification except that in the Mahalanobis classification the classes containing the regions of interest (ROIs) are assumed to have equal covariance. (ENVI 2003). Five ROI classes were selected from the large subsection (including the man-made construction) of the 2006 image. This image was utilized as the ground-truth image (for each subsection group) based from vegetation transects data and other field observations that were collected in the same year. The ROIs for the impacted subsection included: Juncus roemerianus marsh (191 pixels), pine hummock (352 pixels), openwater (414 pixels), salt panne (500 pixels), and human impacts (703 pixels). The ROIs of all classes were geo-located and found within the subsection groups. Only four (less the human impacts) were collected from the 2006

 40  

natural subsection. The ROIs for the natural subsection included: Juncus roemerianus marsh (228 pixels), pine hummock (211 pixels), open-water (261 pixels), and salt panne (320 pixels). The supervised classification selected for four classes (Juncus roemerianus marsh, pine hummock, open-water, and salt panne) in the natural subsection imagery. Next, the process was repeated on the man-made impacted subsections with the previously mentioned classes and an additional class of ROIs that included the man-made components associated with footprint of camps and other man-made infrastructures (human impacts). The areas that satisfy the Mahalanobis Distance criteria for a class are carried over as a classified area into the classified image. An  ENVI  v.  4.2  error  matrix  was  performed  to  determine  the  percent  accuracy  and   Kappa  coefficient  of  classification  for  each  ROI  in  each  image.    Standard  deviation   and  percent  accuracy  from  a  pixel-­‐to-­‐pixel  comparison  will  estimate  the  degree  of   error  in  the  image.    A  calculation  used  for  the  Kappa  coefficient  in  remote  sensing   classification  is  found  in  the  literature  of  Bishop  et  al.  (1975)  and  Stehman  (1996).     The  calculation  as  implemented  in  this  research  is  shown  below:  

Equation II-13

where pij = the population error matrix parameters Nij/N, pi+ = Ni/N, and p+I = Mi/N. Because Kappa is a parameter of the population, it is unchanged by the chioce of sampling design (Stehman 1996).

 41  

A Kappa coefficient of 1.0 indicates that the classification is accurate, and a lower Kappa coefficient near 0 indicates that the classification is no better than the random agreement between the regions of interest and the classified pixels in the image (Kulkarni, A. 2004 – Thesis Louisiana State University). Finally, a difference map was created from the Mahalanobis Distance supervised classifications using a 50% threshold (Richards 1999). Percent change was determined and comparisons (Tou and Gonzalez 1974) were recorded for the years 1940-1985, 1985-2006, and 1940 and 2006. Supervised Mahalanobis Distance Classification. To determine a baseline vegetation map of the study area in 2006, and determine vegetation trend from 1940-2006, the histogram of pixels in the greyscale imagery (1940, 1985 and 2006) were clustered into groups based on ground-reference data in 2006 and the Gaussian pixel intensities for specific classes. A supervised Mahalanobis distance classification of the four basic matrices: Juncus roemerianus marsh, pine hummock, open-water, and salt panne all found within the imagery were created for the 2006 Gaussian normalized greyscale image of the salt marsh study area in Kreole, Mississippi. To investigate the percent cover associated with Mahalanobis distance classification mapping, the image was transformed from greyscale (0-255) to a synthetic color RGB image. A supervised classification of the study area was created from the color-transformed using Mahalanobis Distance classification in ENVI v 4.2. Four Regions of Interest (ROIs) were collected from the 2006 natural

 42  

subsection image. The ROIs for the natural subsection included: Juncus roemerianus marsh (228 pixels), pine hummock (211 pixels), open-water (261 pixels), and salt panne (320 pixels). The supervised classification selected for the same four classes (Juncus roemerianus marsh, pine hummock, openwater, and salt panne) in the imagery. The percent cover of the four ROIs were recorded and served as a baseline for vegetation distribution at the study site. The supervised Mahalanobis distance classification assumes normal distribution of the data and determines the probability that a given pixel should be consigned into a specific class. Each pixel is grouped into the class with the highest probability of likeness. If the highest probability is smaller than a specific class (region of interest), the pixel will be unclassified. Additionally, a comparison of supervised image classifications were investigated at a smaller timeframe (January 2004 - March 2006) to determine if class changes were associated with drought conditions (NOAA 2009). Climate conditions for January 2004 and March 2006 were determined from archived meteorological data on the National Climate Data Center (NCDC), Internet Management System (IMS).

Results

Image Normalization and Georeferencing The Gaussian contrast stretch technique provided the best normalization

 43  

(Figures II–7, II–8, and II–9) among the 1940, 1985 and 2006 imagery used in difference mapping (Tables II–1 and II–2). The Gaussian stretch was centered at a mean digital number (DN) of 127 with the data values 3 standard deviations set to 0 and 255 in the image. The Gaussian contrast normalization was sufficient for images in years 1940, 1985 and 2006 for both the natural subsection and impacted subsection encompassing the manmade components.(Figures II–10 and II–11). The change detection maps created from the Gaussian stretched images are shown in Figures II–12 and II–13. Table II–3 provides percentage change of pixels found within each class for the date comparisons of 1940-1985, 19852006, and 1940-2006. Change Detection Vegetation Transects. The vegetation transects were mapped to true color geo-referenced imagery so that the associated vegetation type could be viewed as the transect lines crossed varying patches of vegetation as well as barren sediment. The transect data were compiled (Table A-2) and vegetation frequency (Tables II–4 and II–5) and percent cover (Table II–6) were calculated for each transect. Transects measured in the low marsh indicated the lowest vegetation biodiversity with the highest biodiversity was recorded in the upper marsh (Figure II-14). As one would expect, the percent cover was lowest near the salt panne (Hackney and Adams 1992).

 44  

The adjusted Simpson Index for the vegetation transect data at the study area were 0 for transect 1, 0.2791 for transect 2, 0.4465 for transect 3, 0.4808 for transect 4, 0.5069 for transect 5, and 0.3722 for transect 6. The adjusted Shannon-Wiener Index for the vegetation transect data at the study area were 0 for transect 1, 64.09% for transect 2, 58.10% for transect 3, 95.44% for transect 4, 99.54% for transect 5, and 79.50% for transect 6. The Eveness Index for the vegetation transect data at the study area were 0 for transect 1, 0.3080 for transect 2, 1.1167 for transect 3, 0.4586 for transect 4, 0.4782 for transect 5, and 0.3820 for transect 6 (Figure II–15, Table II–7). The mean stem count (per m2 ) and mean height (Table II–8) of J. roemerianus for each random line transect were transect 1, 75 ± 24 stem count, 53.00 ± 2.00 cm height; transect 2, 59 ± 57 stem count, 52.33 ± 2.89 cm height; transect 3, 73 ± 33 stem count, 48.67 ± 3.21 cm height; transect 4, 60 ± 52 stem count, 53.33 ± 5.51 cm height; transect 5, 57 ± 41 stem count, 48.67 ± 2.52 cm height; and transect 6, 132.00 ± 34 stem count, 52.00 ± 7.55 cm height (Table A-3). Write Function Memory Insertion. The results of the write function memory insertion change detection (Figure II–16) expanded on the observed changes in the geo-referenced images by visualizing the significant change in the roads and boat launch parking area occuring between 1940 – 2006 in blue. Change in the boat launch slips and J. roemerianus salt marsh vegetation in green and an extensive change (1940 – 2006) in salt panne patch size in red.

 45  

Difference Mapping. ENVI produced change detection difference mapping of greyscale images for years (1940 and 2002) resulted from subtracting the initial and final state images where a positive change between two images is denoted in bright red and identified areas where the final state image is brighter than the initial state image. A negative change between two images is denoted in bright blue and identifies areas where the final state image was darker than the initial state image (Figure II–17). The difference map intensities that denote change can be regulated by adjusting the threshold at which change will be indicated (Figure II–18). Setting a high threshold could detect too much change in the image such as minor surface water glint and setting a low threshold could suppress significant changes in the imagery. The Change difference mapping technique was performed on all image (1940-2002) combinations with the thresholds set at 50%. At the Kreole, Mississippi study site, blue normally indicates increased density in vegetation, shoreline erosion or open water due to subsidence, flooding and scalding of the surface sediment. Red normally indicates human impacts (parking areas, roads, hiking trails, and utility lines), natural loss of vegetation, or deposition of wrack and sandy sediment due to normally occurring storm events and flooding. Some anomalies in the imagery could be associated with differences in aperture speed of the camera system causing over or under exposure of the area of interest; in percent cloud cover over the area causing shading or reduction in irradiance, in solar azimuth

 46  

angle causing glint (Figure II–19) on the surface of vegetation and water, and in pitch, roll and yawl of the aircraft during collection. A difference map was also created from the Gaussian normalized greyscale imagery using a 50% threshold, 4 classes for the natural and 5 classes for the impacted subsections and a percent change mapping threshold. Percent change was determined and comparisons were recorded for the years 1940-1985, 1985-2006, and 1940 and 2006. The percent change was limited to two classes (or regions) at the 50% threshold. The change occurred in the brightest (>.50) and darkest pixels (< 0) with very little change in the middle (≥ 0) range (Figure II–20). Post Classification Difference Image. Synthetic color images were created for both greyscale subsections (natural and impacted) and for the years 1940, 1985 and 2006 (Figures II–21 and II–22). Three supervised Mahalanobis distance classifications created from the transformed synthetic color images, utilized the 2006 vegetation transect as the baseline dataset for selecting the regions of interest point collection. An automated ENVI v. 4.2 error matrix was conducted on each classified image to test the accuracy of the class selection (Figure II–23). For the 1940 (natural subsection) supervised classification the accuracy was 874/1020 pixels (85.6863%) overall with a Kappa coefficient of 0.8077; 1985 (natural subsection) was 754/1020 pixels (73.9216%) overall with a Kappa coefficient of 0.6487; and 2006 (natural subsection) is 938/1020 pixels (91.9608%) overall with a Kappa coefficient of

 47  

0.8919 (Figure II–24). For the 1940 (impacted subsection) supervised classification the accuracy was 1011/2160 pixels (46.8056%) overall with a Kappa coefficient of 0.3333; 1985 (impacted subsection) was 1316/2160 pixels (60.9167%) overall with a Kappa coefficient of 0.4863; and 2006 (natural subsection) is 1467/2160 pixels (67.9167%) overall with a Kappa coefficient of 0.5955 (Figure II–25). The percent change for post-classification Mahalanobis distance supervised classification (Figures II–26 and II–27) change detection (simple value) is provided in Table III–9. The classes represent four ranges (natural subsection) and five ranges (impacted subsection) of pixel values found within the supervised Mahalanobis images based on a 50% threshold of percent change (Figure II–28). The classes that reported 0 change were grouped into the nearest class during the percent change difference detection. The four class ranges for the natural subsection were class 1 (> 0.50), class 2 (> 0 ≤ 0.50), class 3 (= 0 = 0), and class 4 (< 0). The five classes of the impacted subsection are class I, (> 0.50), class 2 (> 0 ≤ 0.50), class 3 (= 0 = 0), class 4 (< 0 ≥ -0.50), and class 5 (< -0.50). These classes represent a 50% threshold for change detection. Supervised Mahalanobis Distance Classification. An automated ENVI v. 4.2 error matrix was conducted on the 2006 baseline classified image to test the accuracy of the supervised classification. The accuracy was 910/1020 pixels (89.2157%) overall with a Kappa coefficient of 0.8551. The supervised Mahalanobis distance classification results for class percent cover were

 48  

estimated as 45.16% (12,277 m2) Juncus roemerianus marsh, 16.61% (4,515 m2) pine hummock, 5.97% (1623 m2) open-water, and 32.27% (8773 m2) salt panne. The January 2004 and March 2006 visual imagery analyses were correlated to the short-term Palmer Z-Index (drought index) for the same timeframe revealed that in January 2004, the coast of Southern Mississippi were experiencing mid range conditions (-1.24 to +0.99) and in March 2006 the conditions indicated extreme drought (-2.75 and below). The percent change of open-water (wet condition) and salt panne (dry condition) computed from Mahalanobis supervised classification using the same four regions of interest (Juncus roemerianus marsh, pine hummock, open-water and salt panne) indicated that the January 2004 imagery consisted of 8.8685% openwater and 5.9663% salt panne with a Kappa coefficient of 0.8460, and the March 2006 image consisted of 5.9663% open-water and 32.2666% salt panne with a Kappa coefficient of 0.8551. These results indicated a change from 2004 to 2006 in open-water at -2.9022% (-788.9921 m2) and salt panne at +2.6080% (+709.0109 m2). It is important to note that Hurricane Katrina made landfall approximately 120 km from the study site on August 28, 2005. The Standardized Precipitation Index (SPI) for August 2005 is very moist (+2.50 to +3.49) yet quickly returned to severe drought (-2.00 to -2.74) conditions in September 2006.

 49  

Conclusion The purpose of the study was to investigate grayscale digital change detection algorithms (Johnson and Kasischke 1998) capable of allowing visualization and evaluation of alterations in coastal marshes (Michalek 1993). Specifically, I was interested in documenting the ecological change in Juncus roemerianus Steele, the common salt marsh vegetation in coastal Mississippi using multiple years of remote sensing and in situ biodiversity data. A geomorphic framework for salt marsh characterization and habitat assessment was created in 2006 using a habitat mapping procedure adapted from Thompson (2002). In situ line transects revealed biodiversity and aerial imagery provided baseline data regarding landscape patterns in a J. roemerianus dominated salt marsh systems. Other baseline data were utilized to enhance descriptions of the structure and function (Bortone 2005) and provide ground reference for remote sensing of the study area in coastal Mississippi. This research is focused on techniques using a low-cost method to scan ancillary aerial imagery as opposed to scanning the imagery with a microdensitometer (Light 1993) which is a more expensive to implement. These techniques are developed for quick-look analyses capable of providing relative change over-time. Due to the limitations in spectral and spatial resolution of the imagery, the interpretation of marsh change is hindered and should be viewed as a generalization not a detailed habitat map except when

 50  

field reference data are available and regions of interest (ROIs) can be selected. A good example of the limitations of detecting change without good ground reference data are the (2003) Case Study on Changes in Freshwater Wetland Habitat (EPA 2006) at Galveston Bay. In a collaborative effort between Galveston Bay Estuary Program (GBEP), Texas Coastal Watershed Program (TCWP), and Texas Cooperative Extension (TCE), a status and trends of the wetlands surrounding Galveston Bay were conducted with an objective to map and quantify changes in regional habitats, land loss, and land use changes (Jacob and Lopez 2005). In 1992, a NWI was conducted by GBEP. This data were utilized as a baseline for comparing aerial photos taken in 2000 and 2002. According to the ancillary NWI data of 1992, the Galveston Bay watershed consisted of 194,556 acres of palustrine, lacustrine, and riverine wetlands. The 2000 and 2002 change analyses data revealed that 285,432 acres remained indicating a loss of 3.1% in freshwater, non-tidal wetlands as based on the 1992 NWI. Experts believe that these reported loss estimates are low due to the potential of wetlands areas being missed in the 1992 NWI conducted by GBEP (EPA 2006). The partners in this project agreed that methods of wetland identification have improved greatly since 1992 and that some of the change is associated with areas that were missed in the 1992 survey.

 51  

The Kappa coefficients in my research did show that relative to the 2006 ground reference data, the 1940 and 1985 imagery declined in accuracy of classifications. This would certainly result in a degree of error that should be recognized in the final percent change determination over-time. Unfortunately, there is no accurate way of estimating the degree of error without ground-reference data from 1940 and 1985. Additional error would result from the gains and offsets of the sensor or camera system where some bright or dark pixels are saturated in the 0-255 range. This was evident in my research where I attempted to delineate the high reflectivity of the unvegetated salt panne from a nearby boat launch parking area. This clearly shows the limitations in multispectral imagery even when applying normalization strategies. In a comparative study of remote sensing change detection methods by Shaoqing and Lu 2008, an image subtraction (initial state – final state imagery) method after classification reduced the probability of errors. The value of change of natural images did not always detect change of objects due to factors such as atmospheric conditions, solar irradiance, sensor calibration and ground water conditions. Additionally, the image subtraction method was not suitable for change detection in urban areas where threshold levels are quite difficult to determine resulting in loss of information. These results do correlate with my results where accuracy of detection was less when

 52  

human impacts were included in the image subset and compared to natural image subsets. This investigation was similar to the previously mentioned Louisiana Coastal Area (LCA) pre and post-Hurricane Katrina study, in that I was interested in trends between land and water, and vegetated versus nonvegetated land-cover change over time. However, the resolution of the aerial imagery utilized in my change detection research was scaled to ~1 meter and spanned the years from 1940 to 2006. The imagery from the LandSat Thematic Mapper 5 (TM-5) consisted of 6 reflective bands at 30 meters resolution and 1 thermal infrared band collected at 120 meters resolution and scaled down to 30 meters during the rectification process. Additionally, the Landsat TM-5 satellite image collection began in 1984 (NASA 2010) and therefore is unable to provide information concerning ecological change over the past 25 years. The  LCA  study  provided  relative  change  pre  and  post-­‐Hurricane   Katrina  (2004-­‐2006)  associated  with  a  large  region  of  the  Louisiana  coast. My interest included ecological change from natural succession and storm impacts in a tidal marsh located approximately 2.75 km due north of the northern most shoreline at Middle Bay and approximately 7.33 km due north from the extent of the southeasterly shoreline of Pt. Aux Chenes Bay and the Mississippi Sound. The 1985 imagery were resampled from 1:58,000 scale imagery to ~1 m for pixel-to-pixel change detection. Certainly, the higher resolution imagery (1:20,000 scale) provided for better accuracy (determined

 53  

from error matrix, accuracy assessments and Kappa coefficient) and improved spectral comparisons for detecting change over the 66 years timeframe. Improved sensor technologies since the 1940’s have increased remote sensing accuracy and contribute to a better an higher level of understanding in marsh fragmentation and change during non-storm years and where limited man-made impacts such as roads, piers, camps and boat launches were present in the imagery. The results of vegetation biodiversity for the six transects were as expected where homogenous stands of J. roemerianus occurred in the low marsh adjacent to the tidal creeks (Figure II-29). The stem height and count was increased in the low marsh (73 ± 33, 75 ± 24, 132 ± 34) where nutrient availability from the tidal creeks would be increased as well lending to higher productivity. The salt tolerant plants were found on the fringe and within the salt panne. Low frequency of J. roemerianus was found within the salt panne, however some patches did exist on a small scale. The J. roemerianus found within or near the salt panne exhibited low stem height (48 ± 3.21, 48 ± 2.52) cm and count (73 ± 33, 57 ± 40). The upper marsh transect data did exhibit higher biodiversity with moderate stem height and fairly high stem counts. I did notice that the sediment around these plants remained moist from what I assume to be surface flow coming from the pine hummock. Additionally, the vegetation in the low marsh remained continually moist from tidal flow. I selected the majority of the J. roemerianus regions of

 54  

interest for the supervised classification from the low marsh and homogeneous stands near the salt panne. This was to prevent bias selection of only plant material with higher moisture content and turgor pressure which could increase reflectivity and adsorption of solar irradiance and reduce transmissivity of solar irradiance through the above ground tissue. Therefore, the ground-reference pixel brightness values of J. roemerianus in the remote sensing research consisted of plant material undergoing various plant physiological conditions such as high and low turgor, redox potentials, salinity and other plant water status and biochemical processes. The  difference  map  technique  for  change  detection  (Virag  and  Colwell  1987)   works  well  with  greyscale  or  post-­‐classification  images.    Taken together  grayscale   digital  imagery  from  an  airborne  platform  scaled  to  ~1  meter  resolution  and  in  situ   data  provides  valuable  information  regarding  alterations  and  historical  trends  of   coastal  marshes. An advantage to the difference map technique is that the computational algorithm is a relatively simple set-up comparing negative and positive values. The result of the computed comparison creates an easily interpreted image map where positive and negative values reflect the changes while a value of zero indicates no change. Threshold range is often driven by the research interest and the absorption, emmissivity and reflectance properties of the materials being investigated. Setting the threshold is a user based observation requiring the image analyst or software designer to set the range

 55  

based on a priori knowledge of marsh dynamics and/or in situ irradiance and reflectivity measurements. Therefore, analyses of change detection is enhanced with field experience and collaborative efforts between computer scientists, engineers and biologists. There are some disadvantages to these methods such as issues with sensor capabilities, solar azimuth angle and atmospheric conditions that are not accounted for in the difference map protocol (Klemas et al. 1993). The most prominent disadvantage to using greyscale imagery is that detailed spectral information capable of identifying specific abiotic and biotic change is unavailable and must be determined manually via in situ sampling which is much more labor intensive. These projects that require extensive fieldwork are most often linked to an augmented impact to ecosystems and contribute to an increase in the project’s cost/benefit ratio as the landscape scale increases. However, in some cases, in situ data have provide better accuracy in Geographical Information System (GIS) mapping especially when coupled to high-level Global Positioning System (GPS) receivers (Welch et al. 1992) and field data loggers (Dobsen et al. 1995). Another disadvantage of image mapping is that the accuracy (Wilkie and Finn 1996) and quality is specific to the scale (Welch et al. 1992) and resolution of the imagery (Bailey et al. 1978; Barras 2006) and the availability of this data to researchers, managers, planners and the general public. The post classification difference image works well to determine changes of large and connected patches however, the

 56  

degree of error is likely to be higher in fragmented marshes (Fisher and Pathirana 1990; Wang 1990) and in landscapes where anthropogenic infrastructure exist (Barras 2006). Classification utilizing mean pixel brightness values could be skewed due to Raleigh scattering of solar irradiance from elevated and man-made objects. Therefore, understanding the spectral histogram dynamics of compared imagery is a first step in determining the relative value of lengthy computational endeavors in change detection comparisons (Bailey et al. 1978; Cowardin et al. 1979). The accuracy in our supervised Mahalanobis distance classifications from the transformed synthetic color images indicates that image classifications are skewed by pixel saturation of man-made impacts such as shell parking lots that were classified in the same range as the salt pannes. This research has indicated that Gaussian normalized greyscale (~1 meter pixel resolution) imagery from varying platforms and sensor systems can provide information on marsh change overtime. I determined that the best technique for image subsets that contain man-made structures such as piers, camps, and shell parking areas is a greyscale difference map. However, if the image subset consists of natural materials such as vegetation, salt pannes, open-water and forests, a Mahalanobis supervised classification using ROIs from a recent ground-referenced image would likely provide the best change detection accuracy (Barras 2006).

 57  

The supervised Mahalanobis classification did not classify the soil moisture previously mentioned as “water”, because the year that the imagery was collected (2006), we were in a severe drought (NOAA 2009). National Oceanic and Atmospheric Administration (NOAA) National Climate Data Center (NCDC) reported that during March 2006, Mississippi coastal lands were under extreme drought conditions of -3.00 to -3.99 (Figure II-30) according to the long-term (Karl 1986), Palmer Hydrolic Drought Index (PHDI), and March – August 2006, the northern Gulf of Mexico (S.E. Louisiana, S. Mississippi, S. Alabama, and the N.W. panhandle of Florida) were experiencing severely dry conditions of -1.50 to -1.99 according to the Standardized Precipitation Index (SPI) a standard measure used for shortterm comparisons (Figures II-31 and II-32). The perimeter of the “wet” (open-water) classification located in the salt panne of the January 2004 image measured approximately 97.28 meters (397.41 m2) or 1.4% change. The supervised classification used as a baseline did appear accurate in detecting wet and dry conditions when compared to visual field observations from 2004 - 2006 and the imagery collected January 2004- March 2006. The imagery did have some spectral artifacts associated to abnormal reflectance (glint) and solar angles (shadows) especially within the pine hummock tree canopy. Some of the shaded areas from the trees were classified as openwater (blue color class).

 58  

The greyscale and classification image (1 meter spatial scale) used in this research for landscape change interpretation revealed that the study area is following a Gleasonian succession trend (Gleason 1917, van der Valk 1981) where dry marsh is converting to a regenerated marsh with normal rainfall (Middleton 1999). The Clementsian succession model would not apply to this study site due to the magnitude of disturbance from natural storms, fires, and herbivory. Additional ecosystem variables that could potentially distort the expected landscape trends are changes in sea level and ground water flow (Hadley et al. 1987), elevation, invasive species competition, human impacts and many other site specific conditions (Figures A-1, A-2, and A-3; Tables A4, A-5, A-6, and A-7). Further baseline in situ studies and long-term aerial observations at higher resolution could overtime indicate that the marsh is undergoing a transition and trending from a regenerating marsh to a degenerating marsh (van der Valk and Davis 1978; Middleton 1999). Additional, correlations to long-term ecological studies may better correlate and identify if degeneration is associated with devastating natural storms, anthropogenic stressors, plant pathogens, herbivory, tidal inundation period and duration of flooding due to land use change and/or hydrological alterations in the watershed. Future research in salt marsh change detection using ancillary imagery should consider accuracy compared to 1 meter, 10 meter and 30 meter resolutions. Additional ground reference data incorporating sediment and

 59  

elevation dynamics would enhance interpretation of the changes detected. Historical greyscale imagery provides researchers with relative trends in marsh succession and identifies impacts from man-made features that may change marsh function and structure over time; however, multispectral (RGB) data will provide better classification accuracy resulting in better interpretation of change detection maps and imagery.

 60  

Table II-1. Statistical results from comparison of image normalization stretch procedures for the natural site subsections measuring 197 rows x 138 columns = 27,186 pixels.  

61

Year 1940 1985 2006 1940 1985 2006 1940 1985 2006 1940 1985 2006 1940 1985 2006

Normalization Linear 0-255 8 bit Linear 0-255 8 bit Linear 0-255 8 bit Gaussian Gaussian Gaussian Equalization Equalization Equalization Square Root Square Root Square Root 2% Linear 2% Linear 2% Linear

Min 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

   

Max 255 255 255 200 207 187 244 248 234 251 252 248 255 255 255

Mean 132.7199 69.2665 112.3642 126.2791 126.2086 124.0413 126.2606 126.1989 125.6891 170.9329 204.0745 146.9297 131.3028 69.4432 112.3642

STDEV 72.0183 75.5843 85.2459 41.5804 43.1532 44.5194 73.4840 74.4223 73.1772 56.7017 43.6815 66.4903 72.9313 75.8807 85.2459

Table  II-­‐2.    Statistical  results  from  comparison  of  image  normalization  stretch  procedures  for  the  impacted  site  subsections   measuring  527  rows  x  400  columns  =  210,800  pixels.    

62

Year 1940 1985 2006 1940 1985 2006 1940 1985 2006 1940 1985 2006 1940 1985 2006

Normalization Linear 0-255 8 bit Linear 0-255 8 bit Linear 0-255 8 bit Gaussian Gaussian Gaussian Equalization Equalization Equalization Square Root Square Root Square Root 2% Linear 2% Linear 2% Linear

Min 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

   

Max 255 255 255 196 188 184 241 235 231 248 222 240 255 255 255

Mean 135.8624 61.1877 117.7893 131.6570 122.0078 129.7975 138.6808 118.7942 137.6108 100.5403 175.0784 157.5588 135.8624 61.1877 117.7893

STDEV 69.3895 68.5342 72.4920 41.9338 39.8724 42.4324 71.3545 69.9054 69.8670 56.8679 54.0925 57.8245 69.3895 68.5342 72.4920

Table II-3. A table showing statistical results from Gaussian normalized greyscale change detection.   Subsection - Natural Year (Initial State) - (Final State) 1940 - 1985 1985 - 2006 1940 - 2006

Percent Pixels Class 1 49.1172 51.0115 51.1011

Class 2 0 0 0

Subsection-Impacted

63

Year (Initial State) - (Final State) 1940 - 1985 1985 - 2006 1940 - 2006

Class 3 1.3279 1.3026 1.2056

Class 4 49.5549 47.6859 47.6933

Percent Pixels Class 1 22.7818 31.0676 20.5339

Class 2 0 0 0

Class 3 28.5375 26.9836 25.9242

Class 4 0 0 0

Class 5 48.6807 41.9488 53.5419

Table II-4. Percent frequency data (%) compiled from line intercept sampling at each transect at the Kreole, Mississippi study site. The frequency represents the percent presence and absence of each vegetative and nonvegetative land cover within the 50- intercept crossings (samples) of the six random transects (25m). The percent land cover frequency is calculated as the number of samples containing land cover / total number of intervals x 100. NonVegetative Land-cover

Mud

Vegetative Land-cover Juncus roemerianus Spartina alterniflora Distichlis spicata Batis maritima Borrichia frutescens

T-1 100

T-2 72

Random Transects T-3 T-4 72 50 2 30 20

14 6 14

20

64

 

 

T-5 54

T-6 76

46

24

Mean Total Land-cover (%) 70.67 16.00 30.00 14.00 6.00 22.00

Standard Deviation 17.87 19.80 14.00 0 0 11.31

Table II-5. Vegetative relative frequency data (%) compiled from line intercept sampling at each transect at the Kreole, Mississippi study site. The frequency represents the percent presence and absence of each vegetative and non-vegetative land cover within the 50-intercept crossings (samples) of the six random transects (25m). The Percent relative vegetation frequency is calculated as the frequency of samples / total frequency of species x 100.   Vegetation Species Juncus roemerianus Spartina alterniflora Distichlis spicata Batis maritima Borrichia frutescens

T-1 100

T-2 83

RandomTransects T-3 T-4 72 62 2 38 20

16 6

65  

T-5 54 46

T-6 76 24

Mean Relative Frequency (%) 74.5000 28.6667 22.0000 16.0000 6.0000

STDEV Relative Frequency (%) 16.1710 23.4379 2.8284 0.0000 0.0000

Table II-6. Vegetative relative cover (%) compiled from line intercept sampling at each transect at the Kreole, Mississippi study site. The frequency represents the percent presence and absence of each vegetative land cover within the 50-intercept crossings (samples) of the six random transects (25m). The percent relative vegetation cover is calculated as the total intercept length of a species / total frequency of species x 100.  

66

Vegetation Species Juncus roemerianus Spartina alterniflora Distichlis spicata Batis maritima Borrichia frutescens Total % Vegetative Cover Total % Non-Vegetative Cover

T-1 1.00

1.00

Random Transects T-2 T-3 T-4 .72 .72 .50 .02 .30 .20 .14 .06 .86 1.00 .80 .14 .20

 

T-5 .54

T-6 .76

.46

.24

1.00

1.00

% Cover Mean .71 .05 .15 .02 .01 .94 .06

% Cover STDEV 0.1787 0.1980 0.1400 0.0000 0.0000 0.0898 4.2426

Table II-7. A table of the vegetation biodiversity indices for six random transects at the study area in Kreole, Mississippi. The Simpson Index is subtracted from 1. The adjusted Simpson Index (1 times 100%) is equal to infinite diversity and 0 = no diversity. The adjusted Shannon-Wiener Index represents the percent of the maximum diversity possible where the maximum is the ln (total number of species):1.099. Eveness is a measure that represents the similarity of the abundances of species that are found in the landscape. The Eveness Index values range between 0-1 , where less community variation between species increases the value of the Eveness Index (similar proportions =1.0). The Eveness index can be above 1.0 because of the ln(total number of species):1.099 ratio in the Shannon-Wiener Index.

67

Biodiversity Indices Simpson Index Simpson Index (adjusted) Shannon-Wiener Index Shannon-Wiener Index (adjusted) Eveness Index

T-1 1.0000 0.0000 0.0000 0.0000 0.0000

T-2 0.7209 0.2791 0.4443 0.6409 0.308

Random Transects T-3 T-4 0.5535 0.5192 0.4465 0.4808 0.8055 0.6616 0.5810 0.9544 1.1167 0.4586

 

T-5 0.4931 0.5069 0.6899 0.9954 0.4782

T-6 0.6278 0.3722 0.5511 0.7950 0.382

Index Mean 0.6524 0.3476 0.5254 0.6611 0.4062

Index STDEV 0.1892 0.1892 0.2854 0.3632 0.3667

Table  II-­‐8.    Transect  elevation  and  average  stem  count  and  height  of  Juncus  roemerianus  are  recorded  from  in  situ investigations. The percent cover (J. roemerianus only) is calculated from line intercept data and represents the relative density of J. roemerianus at the study site in Kreole, Mississippi. Line intercept is not a quadrat count, but rather a survey of the number of individual plants that cross specific transect sampling locations (50 samples per transect). Random Transect T-1 T-2 T-3 T-4 T-5 T-6

Midpoint Latitude 30.41186 30.41156 30.41108 30.41150 30.41172 30.41183

Midpoint Longitude -88.40450 -88.40450 -88.40450 -88.40551 -88.40501 -88.40523

Midpoint Elevation (m) 0.66 0.35 0.34 0.48 0.31 0.52

2

Mean (m ) Stem Count 75 59 73 60 57 132

68    

STDEV Stem Count 24 57 33 52 41 34

Mean (cm) Stem Height 53.00 52.33 48.67 53.33 48.67 52.00

STDEV Height 2.00 2.89 3.21 5.51 2.52 7.55

J. roemerianus % Cover 100 72 72 50 54 76

Table II-9.

A table showing statistical results from Mahalanobis Classification change detection.

Subsection - Natural Year (Initial State) - (Final State) 1940 - 1985 1985 - 2006 1940 - 2006

Percent of Pixels Class 1 38.4242 15.7584 26.4594

Class 2 0 0 0

Subsection-Impacted Year (Initial State) - (Final State) 1940 - 1985 1985 - 2006 1940 - 2006

Class 3 44.6517 47.1447 49.8694

Class 4 16.9242 37.0969 23.6712

Percent of Pixels Class 1 0 0 0

Class 2 0 0 0

69  

Class 3 6.9739 21.0588 6.6075

Class 4 92.9554 78.9016 93.3529

Class 5 0.0707 0.0396 0.0396

70   Figure II-1 A comparison of Landsat Thematic Mapper 5 imagery of West Pearl River and Highway 190 (near the White’s Kitchen landing) located at the Louisiana and Mississippi boarder before and after Hurricane Katrina. The marsh ripping, shearing, and scouring is evident in the post Hurricane Katrina image. Source, USGS 2005.

   

Figure II-2. A graph shows the projected coastal land loss 1956-2050. The curve is derived from quadratic s-line: y=a+bx+cx2, where y is the land in square miles and x is he duration in years. The LCA Benefit represents an estimate of land accretion due to enhanced conservation management strategies; source, USGS National Wetlands Research Center, Lafayette and Baton Rouge, Louisiana and the Louisiana Coastal Area (LCA) Land Change Study Group.

 71  

Figure II-3. A subsection of 2006, U.S. Geological Survey RGB (Red, Green, Blue), True Color Image (TCI) of a boat launch (red arrow), at Heron Bayou in Kreole, Mississippi. This study area was selected as a research site for change detection mapping because of the natural succession of vegetation and large expanse of homogeneous vegetation exhibited in the area, image source: U.S. Geological Survey 2006.

 72  

73 Figure II-4. The Non-Processed Greyscale Imagery from 1940 (Left), 1985 (Center) and 2008 (Right) documents the change over a 68 year, time-frame. The changes in roads (red arrow), camps (yellow arrow) and boat launch parking (blue arrow) indicate that land use change (LUC) has increased relative to time. The 1940 image shows that the road (red arrow) is not present and that the boat launch parking (blue arrow) was minimal, there appears to be only one camp (yellow arrow). The 1985 image shows that the road (red arrow) is present, the boat launch parking (blue arrow) has expanded, and more structures at the water’s edge (yellow arrow) are present. The 2008 image shows that the road and utility line (red arrow) are beginning to reduce the patch size of the vegetation adjacent the linear features of the road and utility lines, the boat launch parking (blue arrow) has expanded and infrastructure such as piers are now altering the hydrology at the shoreline of Heron Bayou, and two camps now exist and patch size of adjacent vegetation has shrunk due to human impact (yellow arrow).

 

 

  Figure II-5. A block design illustrating the change detection procedure utilizing greyscale geo-referenced imagery. The change detection calculation is explained in detail within the text of this document. Source: Tami Wells, University of South Alabama.

 74  

75 Figure II-6. ENVI produced Change Detection Difference Map of 1940 greyscale aerial photo image (Left) and 2002 DOQQ greyscale image (Right). The difference map (Center) is the result of subtracting the initial and final state images where blue indicates a change to lower brightness value (dark) and red indicates a change in higher brightness value (light). At the Grand Bay site, blue normally indicated an increase in vegetation density, shoreline erosion, or open-water due to subsidence, flooding and scalding. Red normally indicates human impacts (parking areas, roads, utility lines) or natural loss of vegetation. This map is created using a simple value difference.

   

Normalization Stretch Techniques Natural Subsection

Impacted Subsection

Mean Pixel (0-255)

250 200 150 100 50

2% Linear

2% Linear

2% Linear

Square Root

Square Root

Square Root

Equalization

Equalization

Equalization

Gaussian

Gaussian

Gaussian

Linear 0-255 8 bit

Linear 0-255 8 bit

Linear 0-255 8 bit

76

0

1940 1985 2006 1940 1985 2006 1940 1985 2006 1940 1985 2006 1940 1985 2006 Image Date and Normalization Technique

Figure II-7. This graph depicts the Gaussian filter as the best fit normalization stretch techniques in relation to mean pixel relationships between multiple years (1940, 1985, and 2006) of imagery. The solid line represents the smaller (27,186 pixels) subsets containing natural emissivity of irradiant energy and the slashed line representing the larger (210,800 pixels) subsets containing anthropogenic (man-made construction) impacts within the imagery.

Normalization Stretch Techniques Natural Subsection

Impacted Subsection

Mean Pixel (0-255)

250 200 150 100 50 2% Linear

2% Linear

2% Linear

Square Root

Square Root

Square Root

Equalization

Equalization

Equalization

Gaussian

Gaussian

Gaussian

Linear 0-255 8 bit

Linear 0-255 8 bit

Linear 0-255 8 bit

0

1940 1985 2006 1940 1985 2006 1940 1985 2006 1940 1985 2006 1940 1985 2006 Image Date and Normalization Technique

Figure II-7. This graph depicts the Gaussian filter as the best fit normalization stretch techniques in relation to mean pixel relationships between multiple years (1940, 1985, and 2006) of imagery. The solid line represents the smaller (27,186 pixels) subsets containing natural emissivity of irradiant energy and the slashed line representing the larger (210,800 pixels) subsets containing anthropogenic (man-made construction) impacts within the imagery.

 76  

77 Figure II-8. Figure shows 2D scatter plots of greyscale pixel intensities for 2006 (y-axis) and 1940 (x-axis) subsection of imagery with man-made impacts (left); and 2006 (y-axis) and 1940 (x-axis) subsection of imagery without man-made impacts (right).

 

78 Figure II-9. Figure shows 2D scatter plots comparing the greyscale pixel intensities of 2006 (y-axis) and 1940 (xaxis) Gaussian normalized subsection with impacts (left) and a 2006 (y-axis) and 1940 (x-axis) of greyscale Gaussian normalized subsection without impacts (right).

 

 

 

 

 

 

 

 

 

  Figure II-10. Image normalization characteristics of the natural subsection of greyscale imagery 1940 (Left Column), 1985 (Center Column) and 2006 (Right Column). The filtering techniques are as follows: standard linear 0255 greyscale (Top Row), linear 2% (Second Row), Gaussian (Third Row), equalization (Fourth Row) and square root (Bottom Row).

 79  

 

 

 

 

 

 

 

 

  Figure II-11. A series of image normalization projections of impacted (manmade construction) subsection of the greyscale imagery 1940 (Left Column), 1985 (Center Column) and 2006 (Right Column). The filtering techniques are as follows: standard linear 0-255 greyscale (Top Row), linear 2% (Second Row), Gaussian (Third Row), equalization (Fourth Row) and square root (Bottom Row).

 80  

81  

 

Figure II-12. Change detection difference maps of Gaussian normalized greyscale (natural subsection) using a simple value difference for image comparisons 1940-1985 (Left), 1985-2006 (Center) and 1940-2006 (Right).

   

 

82  

 

Figure II-13. Change detection difference maps of Gaussian normalized greyscale (impacted subsection) using a simple value difference for image comparisons 1940-1985 (Left), 1985-2006 (Center) and 1940-2006 (Right).

 

 

  Figure II-14. True color image showing all transects sampled at Grand Bay NERR, Kreole, MS, transect geo-location data in NAD83, source: USDA Farm Service Area (FSA) Imagery (August 7, 2007), 1m resolution, Google Earth.

 83  

84   Figure II-15. A cross section of Juncus roemerianus dominated marsh at Grand Bay NERR determined from baseline data collections. Adjusted Shannon-Weiner Biodiversity Index (SWBI) indicates the homogeneity of vegetation with transect number 1, representing the northern extent; transect number 2, representing the midsection; and transect number 3, representing the southern extent of the study area. Percent land-cover for J. roemerianus were calculated from relative frequency data. Average stem count was calculated from in situ counts from random quadrat (m2) sampling within each transect. Geospatial coordinates (lat, lon) represent the midpoints of each transect within a cross-section gradient from North to South.

 

85 Figure II-16. Multi-Date Change Detection (MCD) using Write Function Memory Insertion of 1940, 1985 and , 2006 greyscale imagery to produce a RGB composite image – R1940G1985,B2006 (Right). This technique allows the analyst to view several years of remotely sensed data in one image (Jensen et al. 1993b). The RGB image indicates a change between 1940-2006 in salt panne patch size (red circles); change in boat launch slips and water quality in the1985 image (green circles); and change in roads, boat launches, parking areas, and decrease in vegetation patch size within the salt panne in the 2006 image (blue circles).

   

86 Figure II-17. ENVI produced Greyscale Difference Maps are created between two image comparisons 1940-1985 (Left), and 1940 – 2006 (Right). Red indicates a brighter pixel intensity (red circles) of high reflectivity materials such as roads and piers, and blue indicates a darker intensity change (blue circles) of high absorptive materials such as trees, water and other vegetative cover. Setting thresholds and limits within the software change detection algorithm controls the degrees of change depicted on the change map. The change difference was generated utilizing percent change.

   

87  

    Figure II-18. ENVI produced Change Detection Difference Map created from 1940 greyscale aerial photo image and 2006 DOQQ greyscale imagery using 11 classes, simple difference, 0-1 normalization and user defined threshold adjustments algorithm. The difference map (Left) is the result of threshold ranging from 0.20 to -0.20, and the difference map (Right) is the result of threshold ranging from 0.80 to -0.80. At the Grand Bay site, blue normally indicates a low pixel brightness value indicated an increase in vegetation density, shoreline erosion, or open-water due to subsidence, flooding and scalding. Red normally indicates a high pixel brightness value associated with human impacts (parking areas, roads, utility lines) or natural loss of vegetation.

 

88  

 

Figure II-19. Image on the (Left) is a 1985 RGB Image showing glint (red circle) on the surface of the water due to solar azimuth angle at time of data collections. The image on the right is an unsupervised K-means classification using 5 classes, 3.0 standard deviation and 50 iterations. The water surface glint (black circle) in the image is detected the same as the salt panne, roads and parking areas in the imagery. This classification indicates that anomalies in the image can pose a higher degree of error in classification mapping and change detection. The best time for data collections at Kreole, MS is from 1000-1400 GMT when the solar angle cast fewer shadows and glint.

   

60

Percent (%)

50 40 1940 - 1985

30

1985 - 2006 1940 - 2006

20 10 0 Class 1

Class 2

Class 3

Class 4

 

60

Percent (%)

50 40 1940 - 1985

30

1985 - 2006 1940 - 2006

20 10 0 Class 1

Class 2

Class 3

Class 4

Class 5

  Figure II-20. Gaussian greyscale change detection statistics for 4 classes within the natural image subsection (Top) and 5 classes within the impacted image subsection (Below). The results are from the simple value change detection using Gaussian normalized greyscale imagery.

 89  

90  

 

Figure II-21. A synthetic color transformation from greyscale is automated in ENVI v. 4.2. The results are generated from Gaussian normalized greyscale natural subsection imagery and are shown as follows 1940 image (left), 1985 (center), and 2006 (right).

   

 

91  

 

Figure II-22. A synthetic color transformation from greyscale is automated in ENVI v. 4.2. The results are generated from Gaussian normalized greyscale impacted (with man-made features) subsection imagery and are shown as follows 1940 (left), 1985 (center), and 2006 (right).

   

 

100 90

Percent (%)

80 70 60 50 40

Impacted

30

Natural

20 10 0 1940

1985

2006

Imagery (Year)

1 Kappa Coefficient (0-1)

0.9 0.8 0.7 0.6 0.5 0.4

Impacted

0.3

Natural

0.2 0.1 0 1940

1985

2006

Imagery (Year)

  Figure II-23. Supervised Mahalanobis distance classification percent accuracy (top) and Kappa coefficient (Below) for both impacted and natural sites.

 92  

93

 

 

 

Figure II-24. Three Supervised Mahalanobis Distance Classifications from the transformed synthetic color images utilized the 2006 image and vegetation transect data as the baseline dataset for a regions of interest point collection. 1020 pixels were collected to represent the classification ground truth dataset for the subsection (natural site) at Grand Bay NERR. The classes for ground reference are Juncus roemerianus marsh (228 pixels), pine hummock (211 pixels), open-water (261 pixels) and salt panne (320 pixels) for a total of 1020 pixels. The classification image dates are 1940 (left), 1985 (center) and 2006 (right). An automated ENVI v. 4.2 error matrix was conducted on each classified image to test the accuracy of the class selection. For the 1940 classification the accuracy was 874/1020 pixels (85.686%) overall with a Kappa coefficient of 0.8077; 1985 is 754/1020 pixels (73.9216%) overall with a Kappa coefficient of 0.6487; and 2006 is 938/1020 pixels (91.9608%) overall with a Kappa coefficient of 0.8919.

 

94

 

 

Figure II-25. Three Supervised Mahalanobis Distance Classifications from the transformed synthetic color images utilized the 2006 image and vegetation transect data as the baseline dataset for a regions of interest point collection. A total of 2160 pixels were collected to represent the classification ground truth dataset for the subsection (impacted site) at Grand Bay NERR. The classes for ground reference are Juncus roemerianus marsh (191 pixels), pine hummock (352 pixels), open-water (414 pixels), salt panne (500 pixels), and human impacts (703 pixels) for a total of 2160 pixels. The classification image dates are 1940 (left), 1985 (center) and 2006 (right). An automated ENVI v. 4.2 error matrix was conducted on each classified image to test the accuracy of the class selection. For the 1940 classification the accuracy was 1011/2160 pixels (46.8056%) overall with a Kappa coefficient of 0.3333; 1985 is 1316/2160 pixels (60.9259%) overall with a Kappa coefficient of 0.4863; and 2006 is 1467/2160 pixels (67.9167%) overall with a Kappa coefficient of 0.5955.

 

 

95  

 

Figure II-26. Change detection difference maps (natural subsection) created from simple value difference of Mahalanobis Supervised Classification images for 1940-1985 (left), 1985-2006 (center) and 1940-2006 (right).

 

 

96  

 

Figure II-27. Change detection difference maps (impacted subsection) created from simple value difference of Mahalanobis Supervised Classification images for 1940-1985 (left), 1985-2006 (center) and 1940-2006 (right)

   

 

60

Percent (%)

50 40 1940 - 1985

30

1985 - 2006 1940 - 2006

20 10 0 Class 1

Class 2

Class 3

Class 4

100 90 80 Percent (%)

70 60

1940 - 1985

50

1985 - 2006

40

1940 - 2006

30 20 10 0 Class 1 Class 2 Class 3 Class 4

Class 5

 

Figure II-28. Supervised classification Change detection statistics for 4 classes within the natural image subsection (Top) and 5 classes within the impacted image subsection (Below). The results are from the simple value change detection using Gaussian normalized greyscale imagery.

 97  

Figure II-29. Elevation data were averaged from NED Contiguous U.S. 1/9 E. Arc Second Elevation Data (USGS NRCS, Seamless Data Accessed: 08Sept2009). Juncus roemerianus tissue Carbon:Nitrogen ratios are reported in percent and correlated to productivity and nutrient availability in the plants (Appendices II-III).

 98  

 

  Figure II-30. National Oceanic and Atmospheric Administration (NOAA), National Climatic Data Center (NCDC), Mississippi Statewide Palmer Hydrological Drought Index (PHDI) for years 1900 – 2007. The graph indicates that extreme drought conditions also occurred at the time of the 1940 image acquisition. Source, www.ncdc.noaa.gov/SOTC/ accessed 27Sept2009.

 99  

 

  Figure II-31. National Oceanic and Atmospheric Administration (NOAA), National Climatic Data Center (NCDC), map depicting the Palmer Hydrological Drought Index (PHDI) for long-term conditions (above) and Standardized Precipitation Index (SPI) for six-month conditions (below). The maps show severely dry conditions of the northern Gulf of Mexico (GOM) from March-August of 2006. The 2006 imagery utilized in this research was collected in March during drought conditions. Source, http://www.ncdc.noaa.gov/SOTC/index; accessed 27Sept2009.

  100  

 

Figure II-32. National Oceanic and Atmospheric Administration (NOAA), National Climatic Data Center (NCDC), map depicting the Palmer Z Index (PZI) for short-term conditions in January 2004 (above) and March 2006 (below). The maps show mid-range conditions at the Kreole, MS study site in January 2004 and extreme drought conditions in March 2006. Source, www.ncdc.noaa.gov/SOTC/index accessed 27Sept2009.

  101  

CHAPTER III –NATURAL STORM EFFECTS AND LANDSCAPE CHANGE DETERMINED FROM DIFFERENCE MAPPING AND SUPERVISED MAHALANOBIS CLASSIFICATIONS

Introduction Marshes are areas that are saturated with surface water or groundwater and support vegetation adapted for life in exceedingly moist conditions. These aquatic lands are considered among the most important and highly productive (Odum 1996) ecosystems on Earth (Mitsch and Gosselink 1993). Coastal salt marshes assimilate pollutants; reduce erosion by trapping sediments; buffer inland properties from storms and serve as an important breeding, nursery and feeding ground for fish and other wildlife, all important to the economy (Lynne et al. 1981; Bell 1989). Coastal marshes are vulnerable to disturbance by natural storms such as hurricanes and tropical storms (Michener et al. 1997; Lugo 2000; Hayes and Sader 2001). Some of the outcomes of post-storm events include: changes in successional direction, erosion, selective pressure on organisms, convergence of community structure and function, and increased turnover rates in biomass and nutrients (Ramsey et al. 1997; Lugo 2000). The ecological impacts of these disturbances are often predicted based on remote sensing of ancillary imagery (Cablk et al. 1994; Ramsey et al. 2001). The spatial

  102  

(Goodchild and Mark 1987; Read and Lam 2002; Goodchild 2003) and spectral scale of the imagery utilized in determining impacts (or change) must be considered in the remote sensing inquiry for pre and post-storm events. Research by Read and Lam (2002) summarizes the need for finer scale remote sensing observations that utilize techniques capable of automated change detection and repeatability of methods in coastal ecosystems. An important consideration is the spectral and spatial resolution of the sensor system. Multispectral imagery (imagery with few spectral bands of data) is less likely to provide detailed classifications in coastal ecosystems resulting in a generalized assumption of ecosystem dynamics. However, hyperspectral imagery (imagery with many spectral bands of data) with both higher spatial and spectral resolutions (Cocks et al. 1998) will provide better classification and accuracy in change analyses and repeatability in coastal wetlands (Ramsey and Laine 1997). In Blake et al. (2005), the incidence of major hurricanes with direct hits on the Gulf of Mexico states between 1851-2006 were reported as: Texas (19), Louisiana (20), Mississippi (9), Alabama (6) and Florida’s-west coast (26). The National Weather Service (NWS), National Hurricane Center (NHC), National Oceanic and Atmospheric Administration (NOAA), Technical Memorandum NWS TPC-5 (2007), noted that the Gulf of Mexico (GOM)

  103  

states are likely underrepresented due to the lack of coastal population before 1900. To assess change in marsh habitat structure due to hurricane impacts, I investigated change in a salt marsh located in Kreole, Mississippi, at a spatial resolution of approximately 1 meter where the primary study site (natural) measured approximately 26,398 m2. Most remote sensing research is hindered by the lack of vegetation inventory and ground reference data. Therefore, I first conducted a baseline inventory of vegetation biodiversity and other in situ measurements within the study area in July of 2006 (postHurricane Katrina). At a larger scale, this vegetation survey would be similar to a National Wetland Inventory (NWI) as conducted by the United States Fish and Wildlife Service (USFWS). Because of the degree of accuracy associated with scale and image resolution, I decided to only look at a small subset of the images, measure the accuracy of classification based on 2006 vegetation transects, and compare the degree of change associated with two category 3 storms making landfall in Mississippi. This research unlike larger scale projects that utilize 3, 5, 10 and 30 meter resolution imagery, should provide some insight into the accuracy and dynamics of change in small subsets of coastal ecosystems (Jensen 1996; Sawaya et al. 2003). In this research, I tested the hypothesis that a change detection map produced from greyscale supervised Mahalanobis distance classifications will

  104  

successfully detect impacts to the landscape post-hurricane events. My selection of change mapping and supervised Mahalanobis distance classification comparison was based on previous investigations of greyscale imagery and percent accuracy of various methods including those at a larger scale (Tami Wells dissertation chapter II). For the purpose of this study, I concentrated on a subset of the imagery that did not include human impacts such as roads, camps, piers or utility right of ways (ROW) with a spatial resolution of approximately 1 m. The objectives for this study were: 1)

Identify the changes associated with four habitat classes (Juncus roemerianus marsh, pine hummock, open-water and salt panne) in pre and post-hurricane grayscale imagery.

2)

Compare the results of landscape change using Mahalanobis distance supervised classification and difference mapping for pre and post-hurricane events in a natural environment.

3)

Compare the percent accuracy of Mahalanobis Distance classifications associated with time lapse between ground-referenced data and image collection.

Study Area The study area was a typical tidal salt marsh ecosystem located at Grand Bay National Estuary Research Reserve (GNDNERR) in Kreole, Mississippi.

  105  

The study area measured approximately 26,398 m2. The georeferenced corner points are Xmin 364929.7600, Xmax 365125.7600, Ymin 3365107.4000 and Ymax 3365303.4000 (NAD 83 Decimal Degree DD). The Grand Bay NERR provides a diverse landscape mosaic that is ideal for exploring remote sensing techniques. The study area exhibits extensive Juncus roemerianus marsh, salt pannes, and a pine hummock. Tidally influenced bayous surround the study area on all three sides. This study area is located approximately 7.33 km from the Mississippi Sound and has experienced several recent storm events. My research imagery of the study area lent to the study of pre and post-impacts of category 3 hurricanes. Hurricane Elena which made landfall in Biloxi, Mississippi, approximately 50 km from the study area and Hurricane Katrina which made landfall in Pearlington, Mississippi, approximately 120 km from the study area. Hurricane Elena made landfall at Biloxi, MS on September 2, 1985 as a category 3 storm (Figures III–1, III–2). The statistics recorded by NOAA on Hurricane Elena state that the maximum winds were 205 km/h (125 mph), lowest pressure 953 mbar (hPa 28.14 in Hg) with an overall estimate of damage at 2.5 billion (2009 USD). Another category 3 storm that made landfall in Pearlington, MS on August 28, 2005 is Hurricane Katrina (Figures III–3, III–4). NOAA statistics report that Hurricane Katrina had maximum winds of 280 km/h (175 mph), lowest pressure 902 mbar (hPa 26.64 in Hg) with an overall estimate of damage at 90.9 billion (2009 USD). The damage

  106  

estimates should be correlated to population growth (Sage and Galloway 2002; Jacob and Lopez 2005) in coastal Mississippi (Figure III–5) as well as storm intensity.

Methods

Image Greyscale Normalization Greyscale images were created from true color composite imagery using ENVI v.4.2 software. The imagery dated March 25, 1985 and October 6, 1985 represent pre and post-impacts from Hurricane Elena that made landfall in Biloxi, Mississippi on September 2, 1985. Color composite imagery from January, 2004 and March, 2006 (specific dates are unknown) represent the impacts pre and post-Hurricane Katrina that made landfall in Pearlington, Mississippi on August 28, 2005. The final images were scaled from 1:52,000 (Elena) and 1:24000 (Katrina) imagery to 1 m2 resolution imagery and encompassed the natural materials (vegetation, sediment, and water) within the study area. The greyscale imagery was normalized using a Gaussian normalized technique that is automated in ENVI v.4.2. Two dimensional scatter plots were created to show the normalization of the data (greyscale versus Gaussian normalized greyscale).

  107  

Supervised Mahalanobis Distance Classification Supervised Mahalanobis distance classifications of the four matrices: Juncus roemerianus marsh, pine hummock, open-water and salt panne were created for each of the Gaussian normalized greyscale images. To investigate the change differences associated with Mahalanobis Distance classification mapping, all imagery was transformed from greyscale (0-255) to synthetic color RGB images. Difference images were created from the colortransformed imagery (March 25, 1985, October 6, 1985, January 2004 and March 2006, all consisting of natural subsets) using Mahalanobis distance post image supervised classification in ENVI v 4.2. The ROIs for the natural subsection included: Juncus roemerianus marsh (228 pixels), pine hummock (211 pixels), open-water (261 pixels), and salt panne (320 pixels). Both simple value and percent change detection maps were created for comparisons of pre and post hurricane events. Hurricane Elena imagery March 25, 1985, (prehurricane) and October 6, 1985 (post-hurricane); and Hurricane Katrina, January 2004, (pre-hurricane) and March 2006, (post-hurricane) classifications were compared to determine the change for each hurricane event. An ENVI v. 4.2 error matrix was performed to determine the percent accuracy and Kappa coefficient of classification for each ROI in each image. Standard deviation and percent accuracy from pixel-to-pixel image comparisons were used to estimate the degree of error in the image.

  108  

Difference Mapping A difference map was created from the Mahalanobis distance supervised classifications using a 50% threshold (Richards 1999). Landscape change was determined and change maps were created (Tou and Gonzalez 1974) for the classified image comparisons spanning March 25, 1985 – October 6, 1985, and January 2004 – March 2006 using percent change and simple value detection. The simple difference change mapping utilized the logic:

SVChangeMap = [ InitialStateImage] − [ FinalStateImage]



The percent change map utilized the logic:

%ChangeMap =

  €

Equation III-1

[ IntialStateImage] − [ FinalStateImage] [InitialStateImage]

Equation III-2

Results

Image Greyscale Normalization The Gaussian stretch was centered at a mean digital number (DN) of 127 with the data values 3 standard deviations set to 0 and 255 in the image. The Gaussian contrast normalization was performed for image comparison between pre and post hurricane imagery for each storm. The greyscale normalization is shown in figures III–6 and III–7. Two dimensional scatter-

  109  

plots provided a visual comparison for normalization and raw greyscale data (Figures III–8 and III–9). Supervised Mahalanobis Distance Classification Synthetic color images were created for greyscale subsections for the dates of March 25, 1985, October 6, 1985, January 2004 and March 2006 (Figure III–10). Supervised Mahalanobis distance classifications from the transformed synthetic color images utilized the 2006 vegetation transect as the baseline dataset for a regions of interest point collection (Figures III-11 and III–12). An automated ENVI v. 4.2 error matrix was conducted on each classified image to test the accuracy of the class selection. For the March 25, 1985 supervised classification, the accuracy was 880/1020 pixels (86.2745%) overall with a Kappa coefficient of 0.8155; the October 6, 1985 subsection was 828/1020 pixels (81.1765%) overall with a Kappa coefficient of 0.7328; the 2004 subsection was 903/1020 pixels (88.5294%) overall with a Kappa coefficient of 0.8460, and 2006 subsection is 910/1020 pixels (89.2157%) overall with a Kappa coefficient of 0.8551. The percent change for postclassification Mahalanobis distance supervised classification change detection (Figures III–13 and III–14). The post classification percent change of natural marsh (excluding man-made artifacts) for the Hurricane Elena (1985) imagery was + 5.25% Juncus roemerianus marsh, +2.47% pine hummock, +0.91% open-water, -7.14% salt panne, and +0.34% unclassified pixels. The Hurricane Katrina (January 2004-March 2006) supervised post

  110  

classification percent change was +13.62% Juncus roemerianus marsh, 13.3230% pine hummock, -2.90% open, 2.61% salt panne, and 0% unclassified (100% of the pixels were classified). The overall pixel change in pre and post-storm images from Hurricane Elena were 10.34% of the pixels (2811 m2) and 11.68% (3175 m2) for Hurricane Katrina indicating that changes do occur and can be detected at a small scale when comparing imagery of the same original resolution provided geo-referencing is accurate and imagery is Gaussian normalized prior to the change analyses. Difference Mapping The difference map of the natural Mahalanobis distance classifications utilized a 50% threshold of 4 classes (Table III – 1) with coordinating ranges as class 1 (> 0.50), class 2 (> 0 and ≤ 0.50), class 3 (=0 ), and class 4 (< 0). A 50% threshold was utilized because the images were normalized using a Gaussian technique. The peak of the Gaussian curve would represent the 50% threshold in greyscale imagery with resolution based on (0-255). By setting the threshold at 50% we are more likely to get most of the vegetation into a class at the mid-level brightness range. The open-water would likely fall into one class at the lower percent range and non-vegetated areas in the higher percent range. Depending on the resolution of the imagery the threshold may be set higher or lower. The determination of where to set the threshold range values are also dependent on the image analysts level of skill and the level field data available to correlate to the result.

  111  

Conclusion The change detection was similar for both of the hurricane events where a range of 49-59% of the pixels experienced no change, 14-21% of the pixels experienced change to a brighter intensity and 25-29% of the pixels experienced change to a darker intensity. These results would indicate within 81 -86% accuracy approximately 49-59% of the study area is capable of buffering the storm impacts with little change to the landscape, 14-21% of the area experienced scald, sediment deposition, or loss of vegetation, and 25-29% of the area experienced erosion (Morton et al. 2005), increased productivity, or perhaps deposition of wrack (Barras 2006; Barras 2007). The error matrix results indicate that a supervised Mahalanobis distance classification change detection methods will have error associated with the final change detection map. This error can be introduced by tidal range, photoperiod, rainfall, drought or other environmental differences that occur at the time of the image collection. The two-dimensional scatter plots of the Gaussian normalized greyscale imagery identifies the distribution patterns of 1985 imagery that was likely collected with a lower spectral resolution sensor and 2004 - 2006 imagery which is certainly an improved sensor system (Figure III–15). The resolution of the sensor would be relevant only if comparing the accuracy of change detection between each storm event. The Hurricane Elena change analysis was obvious in the pine hummock (upland) area where it appears that storm surge damaged understory and

  112  

possibly trees from high winds. Hurricane Elena’s storm track (Figure III-4) passed approximately 50 km south of Mobile, Alabama and made landfall approximately 50 km southeast of the study site. The site was located in the northeast quadrant of the hurricane. Change also appeared in the Spartina alterniflora fringe and shoreline. S. alterniflora was not included in the classification regions of interest due to the poor pixel purity in the georeference data. S. alterniflora fringe was classified as pine hummock in the March 25, 1985 image and was unclassified in the October 6, 1985 image. Other change is likely from scald (vegetation loss), wrack and sediment deposition (Figure III–16). The Hurricane Katrina change (Figure III–17) was affected by severe drought in 2006. A good indication is observed in the salt panne that normally remained moist (post-Hurricane Katrina). In the 2004 image, the lowest area of the salt panne is classified as blue (open-water) and the water is not present in the 2006 image due to the drought. Other change observed is the impact to the tree canopy and understory of the pine hummock. Additionally, an impact created from our research was detected in the 2006 image. The impact was created from our boat and off-loading of field equipment. The shoreline indicates a break in the fringing vegetation and a small puncture in the sediment surface creating an impacted and extremely wet area. I am confident in explaining the relative change from the January 2005-

  113  

March 2006 images. However, I am uncertain of the collection date for these years and cannot determine the tide, rainfall, and seasonal trends that may affect the percent change and extent of natural versus storm induced change. Accurate quantification of hurricane change would require additional field data and higher resolution imagery. In Barras (2006), the Louisiana Coastal Assessment (LCA) trend area (Barras 2003) served as the study site for determining baseline data of total land and water change pre and post-Hurricanes Katrina (August 29, 2005) and Rita (September 24, 2005). This research utilized LandSat-TM midinfrared (30 meter) imagery and ancillary LCA classification data (Twilley and Barras 2003; Morton et al. 2005; Bernier et al. 2006) as baseline land cover and hydrological information. The total LCA study area measured approximately 33,458 km2 (12,918 mi2 ) and the land – water changes were evaluated between Fall of 2004 and mid-October 2005 The total percent change reported for LCA were 2.3193% (776 km2 , 483 mi2) in land (vegetated to non-vegetated) and 1.6780% (562 km2 , 217 mi2) in fragmentation to open-water. The research estimated approximately 212 km2 (82 mi2) of new open-water area increase caused by direct effects from Hurricane Katrina, 303 km2 (117 mi2) from Hurricane Rita, and 46.6 km2 (18 mi2) resulting from both storms. Additionally, the salt marsh land areas decreased by +0.2167% (72.5 km2 , 28 mi2) from the direct impacts of both storms. The permanent loss of land in this research was thought to have

  114  

occurred due to the direct removal of wetland sediment and vegetation. Transitory water area changes are caused by “remnant flooding, removal of aquatic vegetation, scouring of marsh vegetation, and water-level variation attributed to normal tidal and meteorological variation between satellite images. Permanent losses cannot be estimated until several growing seasons have passed and the transitory impacts of hurricanes are minimized” (Barras 2006). In comparison, the percent land-cover change of the (Fall 2004-October 2005) LCA LandSat thematic imagery revealed 2.3193% change from vegetated to non-vegetated landscape. The (January 2005- March 2006) greyscale 1:24000 scale imagery of the GNDNERR study area revealed 2.3137% change from vegetated to non-vegetated landscape. The percent change in land to water of the LCA LandSat thematic imagery revealed 1.6780%. Whereas, the percent change in land to water of the greyscale imagery was 2.9022% (Table III-2). The LCA study utilized land-use and land-cover inventories for image classification lending to a higher degree of accuracy in the change detected. The GNDNERR study area was delineated using in situ vegetation transect data, Mahalanobis distance classifications and included ancillary meteorological data. The NOAA meteorological data for Grand Bay and the 1 meter resolution classification identified approximately 1.5% error in land-cover of un-vegetated to open-water in a salt panne (Wells, T. 2010 - Dissertation-Chapter II University of South

  115  

Alabama). Therefore, a correction should be made to the previously mentioned percent change for the GNDNERR study site. The meteorologically corrected percent change results would be 3.8137% change from vegetated to non-vegetated landscape, and 1.4022% from land to openwater (Table A-10). The methods utilized in my research provided similar results as the LCA study during a similar timeframe of 2004-2006. The methods developed in this research will provide baseline data at a scale smaller than 30 meters by generating a classification of the landscape mosaic utilizing 1 meter resolution imagery at greyscale resolution (0-255 range). In Johansen et al. (2006), change detection rule sets utilizing Definiens Developer ® v.7.0 (Definiens Inc, Parsippany, New Jersey) were established for mapping and monitoring riparian zone land cover classes within two high resolution QuickBird® (Digital Globe, Longmont, Colorado) images. The results of four, object-oriented (determined regions of interest) and pixelbased (i.e. nearest neighbor, maximum likelihood) change detection algorithms were compared. The change detection classifications included five land-cover classes: (1) riparian vegetation; (2) stream bed; (3) woodland; (4) rangeland, and (5) barren ground. The change maps were derived from both object-oriented and pixel-based image regressions. The final change map results from the object-oriented and pixel-based classifications were similar except that a small degree of landscape change occurred in the pixel-based

  116  

approach. The slight change was assumed to be from differences in sensor viewing geometry and a slight degree of image mis-registration. Compared to image regression and tassel cap transformation, both of the postclassification comparisons and image difference maps produced a higher degree of accuracy in determining change in the landscape over-time. The object-oriented, post-classification results implied that phenological change were not misclassified as change from one land-cover class to another. Research presented by Im et al. (2008) found that object-oriented change classification achieved higher accuracy when detecting vegetative change in urban landscapes. Desclee et al. (2006) and Middleton et al. (2008) investigated forest change detection methods and found that object-oriented change maps provided more accuracy in forests land-cover than that of pixelbased change maps. Future research should include patch analyses to determine the structure and function of salt marshes over time. Patch analyses would provide increased information about diversity, fuel models, habitats, and elevations capable of driving natural successional trends and ecosystem success and resiliency.

  117  

Table III-1. Percent change computed from Mahalanobis distance supervised classification of pre and posthurricane imagery. The classes are computed from a 50% threshold for percent change and simple value difference mapping.   25 MAR 1985 – 06 OCT 1985 Hurricane Elena Percent Change Simple Value Change

Class 1 00.0000 21.7980

Class 2 10.8144 00.0000

Class 3 88.8472 49.1613

Class 4 00.3384 29.0407

MEAN 2.8952 2.8544

STDEV 0.3171 1.0683

JAN 2004 – MAR 2006 Hurricane Katrina Percent Change Simple Value Change

Class 1 00.0000 15.0390

Class 2 11.6827 00.0000

Class 3 88.3173 59.8265

Class 4 00.0000 25.0000

MEAN 2.8832 2.9506

STDEV 0.3212 0.9222

118  

Figure III-1. GOES-12 Satellite Image of Hurricane Elena on 02 September 1985, source, http://www1.ncdc.noaa.gov/pub/data/images , Accessed: 03 August 2009.

  119  

  Figure III-2. Hurricane Elena tracking and rainfall data, Source, http://www.hpc.ncep.noaa.gov/tropical /rain/elena1985.html Accessed 03 August 2009.

  120  

  Figure III-3. GOES-12 Satellite Image of Hurricane Katrina on 28 August 2005, source, http://www1.ncdc.noaa.gov/pub/data/images , Accessed: 03 August 2009.

  121  

  Figure III-4. Hurricane Katrina tracking and rainfall data, Source, http://www.hpc.ncep.noaa.gov/tropical /rain/elena1985.html Accessed 03 August 2009.

  122  

200000 180000 160000 Population

140000 120000 100000

1980

80000

2008

60000 40000 20000 0 Hancock

Harrison

Jackson

Figure III-5. A chart showing the population level for Mississippi Coastal Counties in 1980 versus 2008. Data source, US Census, Washington DC.  

  123  

 

a  

b  

 

124     c   d       Figure III-6. Hurricane Elena greyscale imagery, pre-hurricane, (a) March 25, 1985 and post-hurricane, (b) October 6, 1985. Hurricane Elena Gaussian normalized imagery (c) March 25, 1985 and Gaussian normalized (d) October 6,1985.

 

 

 

a  

b  

 

125     c   d       Figure III-7. Hurricane Katrina greyscale imagery, pre-hurricane, (a) 2004 and post-hurricane, (b) 2006. Hurricane Katrina Gaussian normalized imagery (c) January 2004 and Gaussian normalized (d) March 2006.

   

126         Figure III-8. A two dimensional (2D) scatter plot of greyscale imagery (left) March 25, 1985 - x-axis, and October 6, 1985 - y-axis and 2D scatter plot of Gaussian normalized greyscale imagery (right) March 25, 1985 - x-axis, and October 6, 1985 - y-axis.

   

127  

 

Figure III-9. A two dimensional (2D) scatter plot of greyscale imagery (left) January 2004 - x-axis, and March 2006 - y-axis and 2D scatter plot of Gaussian normalized greyscale imagery (right) January 2004 - x-axis, and March 2006 - y-axis.

 

 

a  

b  

 

128     c   d       Figure III-10. Synthetic (RGB) pseudo-color imagery created from Gaussian enhanced greyscale imagery collected on (a) March 25, 1985, (b) October 6, 1985, (c) January 2004, and (d) March 2006. Synthetic color image transformations utilized ENVI v.4.3 software developed by ITT (Boulder, CO).

   

a  

c  

e  

 

 

 

b  

d  

f  

 

 

 

  Figure III-11. Change detection utilized color imagery dated (a) March 25, 1985 and (b) October 6, 1985 representing pre and post-impacts from Hurricane Elena that made landfall in Biloxi, Mississippi on September 2, 1985. Supervised Mahalanobis distance classification of the four matrices: (red) Juncus roemerianus marsh, (yellow) salt panne, (blue) open-water, and (green) pine hummock in the (d) March 1985 and (e) October 1985 imagery. A (e) percent change difference map and (f) simple value difference map were created from the two supervised classifications.

  129  

a  

c  

e  

 

 

 

b  

d  

f  

 

 

 

Figure III-12. Change detection utilized color imagery dated (a) January 2004 and (b) March 2006 representing pre and post-impacts from Hurricane Katrina that made landfall in Pearlington, Mississippi on August 28, 2005. Supervised Mahalabolis distance classification of the four matrices: (red) Juncus roemerianus marsh, (yellow) salt panne, (blue) open-water, and (green) pine hummock in the (c) January 2004 and (d) March 2006 imagery. A (e) percent change difference map and (f) simple value difference map were created from the two supervised classifications.

  130  

15.00 10.00

Percent

5.00 0.00 -5.00 131

-10.00 -15.00 Elena

Katrina

Unclassified 0.34 0.00

J. roemerianus Marsh 5.25

13.62

Hummock

Water

Salt Pan

-13.32

-2.90

2.61

2.47

-0.91

-7.14

Figure III-13. Mahalanobis distance supervised classification based on class size from the regions of interest (Juncus roemerianus marsh, pine hummock, open-water, and salt panne). The graph indicates the percent change.

5000.00

Square Meters

4000.00 3000.00 2000.00 1000.00

0.00

132

-1000.00 -2000.00 -3000.00 -4000.00 Elena

Katrina

Unclassified 92.00 0.00

J.roemerianus Marsh

1426.01 3702.00

Hummock 672.01

-3621.99

Water

-248.02 -788.99

Salt pan

-1942.00 709.01

Figure III-14. Mahalanobis distance supervised classification based on class size from the regions of interest (Juncus roemerianus marsh, pine hummock, open-water, and salt panne). The graph indicates the change in meter2.

  Figure III-15. Two-dimensional scatter plots display the spectral variability in the Gaussian normalized greyscale imagery. The Hurricane Elena scatter plot (top) and the Hurricane Katrina scatter plot (below) shows variability in the four approximated class (circles). The simple value change at a 50% threshold setting for four classes (1-4) resulted in 22%, 0%, 48%, and 29% for each class in the Hurricane Elena data, and 15%, 0%, 60% and 25% for each class in the Hurricane Katrina data. Class 2 and 3 were grouped together based on the threshold value of 50%.

  133  

a  

c  

e  

 

 

 

b  

d  

f  

 

 

 

Figure III-16. Change detection utilized color imagery dated (a) March 25, 1985 and (b) October 6, 1985 representing pre and post-impacts from Hurricane Elena that made landfall in Biloxi, Mississippi on September 2, 1985. Supervised Mahalanobis distance classification of the four matrices: (red) Juncus roemerianus marsh, (green) pine hummock, (blue) open-water, and (yellow) salt panne in the (d) March 1985 and (e) October 1985 imagery. A (e) percent change difference map and (f) simple value difference map was created from the two supervised classifications. A large change in pine hummock (or upland) in the (red circle) October 1985 image, small change in (green circle) Spartina alterniflora fringe and (black circle) likely scald, wrack and sediment deposition.

  134  

 

a

b

c

d

Figure III-17. Change detection utilized color imagery dated (a) January 2004 and (b) March 2006 representing pre and post-impacts from Hurricane Katrina that made landfall in Pearlington, Mississippi on August 28, 2005. Supervised Mahalanobis distance classification of the four matrices: (red) Juncus roemerianus marsh, (green) pine hummock, (blue) open-water and (yellow) salt panne in the (d) January 2004 and (e) March 2006 imagery. A (e) percent change difference map and (f) simple value difference map was created from the two supervised classifications. Change in (red circle) pine hummock (or upland) in the imagery, change in (green circle) Spartina alterniflora fringe and impact from marsh access and change in (black circle) sediment moisture associated with drought.

  135  

CHAPTER IV– CONCLUSION

When computing a difference map from greyscale imagery, the analyst must consider the differences associated with the collection instrument (sensor), the collection time (solar angle), the collection date (earth’s tilt, temporal), the atmospheric conditions (especially when using satellite data), the final image calibration and resolution, and the co-registration (georeference) accuracy. Additionally, the analyst must determine the class thresholds for the difference mapping. The default classification thresholds are evenly spaced ranging from -1 to +1 for simple value difference mapping, and -100 and + 100 for percent difference mapping. In ENVI v.4.2 software user’s Guide (2003), the class definition and difference procedure is as follows: “For n classes, where n is odd, the first (n/2) class represents positive changes, starting with the largest positive change and ending with the smallest. The middle class, (n/2)+1, represents no change. The last n/2 classes represent negative changes, starting with the smallest negative change and ending with the largest. For even number of classes the definitions remain the same except that one reduces the

  136  

number of negative classes. In short, the default class definitions range from positive to negative, with the magnitude of the change increasing with distance from the middle of no change class. A simple value difference is computed by subtracting the initial state image from the final state image. The percent difference is the simple difference divided by the initial state value.” Selection of the class number and the threshold range is very important to the computer-generated outcome. And, temporal field observations, baseline and quantitative environmental data, and georeferenced data will greatly enhance the accuracy of remote sensing and change detection research. However, long-term environmental research sites such as Grand Bay National Estuarine Research Reserve and increased funding for sensor development and software design is equally as important. The result of this study accepted the null hypothesis, Ho1, taken together grayscale digital imagery from an airborne platform scaled to ~1 meter resolution and in situ data provided valuable information regarding alterations and historical trends of coastal marshes at Grand Bay National Estuarine Research Reserve (GRDNERR) in Kreole, Mississippi. A Multi-Date Change Detection (MCD) using Write Function Memory Insertion of 1940, 1985 and 2006 greyscale imagery to produce a RGB composite image allowed me to view several years of remotely sensed data in one image. These results are based completely from my visual observation of

  137  

change using ENVI v.4.0 software. The change is determined by discrimination between timelines using visual cues associated with each color filter (red, green and blue) and the association that color has with image date. The RGB image created from years 1940 (red), 1985 (green) and 2006 (blue) indicated a change between 1940-2006 in salt panne patch size and in man-made features such as boat launch slips. In the1985 image change occurred again in roads, boat launches, parking areas reducing standing stock of natural vegetation in these areas. A decrease in vegetation patch size within the salt panne was apparent in the subjective estimates and comparison of 1885 and 2006 imagery. An ENVI produced percent change difference map from 1940 (initial state) and 2006 (final state) Gaussian normalized greyscale images revealed that 51.10% of the pixels changed at > 0.50 threshold range (from dark to light reflectance), 0 pixels changed at > 0 ≤ 0.50, 1.21% at = 0 and = 0 (no change), and 47.69% at < 0.50 (from light to dark reflectance). This data indicates that extreme change in vegetation has occurred in the 66 years time frame. The main differences were from the error in the classification where human impacts revealed a higher reflectance value in the overall image. These bright areas such as, shell roads and parking areas, metal roofing and other reflective materials created an over-estimate in salt panne. The imagery did not distinguish the non-vegetated areas from human induced impacts. Perhaps, pattern analyses coupled with reflective analyses would

  138  

help to delineate and mask the anthropogenic impacts in the imagery. In imagery at 1m scale, masking the impacts would be extremely important to the total measurements. However, at a 30 meter scale, the pixel averages may provide better classification results because the pixel averages would have been less revealing of the human induced impacts at this site which is predominately natural. The in situ vegetation line transects provided valuable ground reference and supervised classification accuracy. Biodiversity of the study area was determined from standard procedures involving vegetation assessments and is reported as follows. The Simpson Index is subtracted from 1 therefore the result is not counterintuitive. 1 (100%) is the adjusted Simpson Index equal to infinite diversity and 0 = no diversity. The adjusted Shannon-Wiener Index represents the percent of the maximum diversity possible where the maximum is the ln(total number of species):1.099. Eveness is a measure that represents the similarity of the abundances of species that are found in the landscape. The Eveness Index values range between the 0-1, where less variation between communities of species increase the value of the Eveness Index (similar proportions =1.0). The adjusted Simpson Index for the vegetation transect data at the study area were 0 for transect 1, 0.2791 for transect 2, 0.4465 for transect 3, 0.4808 for transect 4, 0.5069 for transect 5, and 0.3722 for transect 6. The adjusted Simpson Index results demonstrate a large difference in transect biodiversity where transect 1 has no diversity and

  139  

transect 5 has 50% more diversity than transect 1. This would be a normal distribution of vegetation at this location. The low marsh would exhibit low diversity in a natural succession seaward. The high marsh has evolved over time from a barren landscape to a more diverse landscape with pioneering species. Additionally,  the  results  of  the  study  accepted  the  null  hypothesis,  Ho2,  change   detection  map  produced  from  greyscale  supervised  Mahalanobis  distance   classifications  will  successfully  detect  impacts  to  the  landscape  post-­‐hurricane   events.    A  Mahalanobis  distance  classification  was  utilized  to  determine  the  spatial   extent  of  J.  roemerianus  at  the  study  site  pre  and  post-­‐hurricane  events.    The   accuracy  of  the  supervised  Mahalanobis  distance  classification  utilized  in   determining  the  spatial  extent  of  J.  roemerianus  pre  and  post-­‐storm  events  were   recorded.    For  the  March  25,  1985  supervised  classification,  the  accuracy  was   880/1020  pixels  (86.2745%)  overall  with  a  Kappa  coefficient  of  0.8155;  the  October   6,  1985  subsection  was  828/1020  pixels  (81.1765%)  overall  with  a  Kappa   coefficient  of  0.7328;  the  2004  subsection  was  903/1020  pixels  (88.5294%)  overall   with  a  Kappa  coefficient  of  0.8460,  and  2006  (ground-­‐reference)  subsection  was   910/1020  pixels  (89.2157%)  overall  with  a  Kappa  coefficient  of  0.8551.    The  post   classification  percent  change  of  natural  marsh  (excluding  man-­‐made  artifacts)  for   the  Hurricane  Elena  (1985)  imagery  was  +  5.25%  Juncus  roemerianus  marsh,   +2.47%  pine  hummock,  0.91%  open-­‐water,  -­‐7.14%  salt  panne,  and  0.34%   unclassified.    The  Hurricane  Katrina  (2004-­‐2006)  classification  change  was  

  140  

+13.62%,  -­‐13.3230%,  -­‐2.90%,  2.61%  respectively,  with  100%  of  the  pixels   classified.    The  simple  value  difference  mapping  revealed  that  the  21.78%  of  the   pixels  changed  at  >  0.50  threshold  range,  0  pixels  changed  at  >  0  ≤  0.50,  49.16%  at  =   0  (no  change),  and  29.04%  at  <  0.50.    These  results  indicate  that  21.78%  of  the   pixels  changed  from  dark  intensity  to  light  intensity  displayed  as  “  bright  red”  on  the   difference  map;  49.16%  of  the  pixels  had  no  change  or  very  little  change  and  were   grouped  from  class  2  (>  0  ≤  0.50)  into  class  3  (=  0,  no  change)  and  will  be  displayed   as  “grey,  light  red,  or  light  blue”  in  the  difference  map;  29.04%  of  the  pixels  changed   from  light  intensity  to  dark  intensity  displayed  as  “dark  blue”  in  the  difference  map.   A distinct change in the extent of J. roemerianus marsh was detected in the pre and post hurricane comparison data. An increase of approximately 5.25% (1426 m2) in the J. roemerianus marsh occurred from March 25, 1985 (pre Hurricane Elena) to October 6, 1985 (post Hurricane Elena); and an increase of approximately 13.62% (3702 m2) J. roemerianus marsh occurred from 2004 (pre Hurricane Katrina) to 2006 (post Hurricane Katrina). This data revealed that the extent of J. roemerianus marsh is not static and does change over time. The study site encompasses 27,186 m2 (2.7186 hectares). Finally, this study provided valuable information regarding the baseline assessment of coastal salt marsh in Mississippi. However, careful consideration should be taken when detecting change in coastal marshes with other dominant emergent vegetation and land elevation. The accuracy of classification and thus change detection is directly correlated to the

  141  

improvement of electro-optical sensor capabilities and the availability of ground reference data. The data acquired in this research provides a general baseline of the trend in land-cover and Kreole, Mississippi over time. This trend does follow traditional succession in Juncus roemerianus dominated marshes in the Northern Gulf of Mexico where the J. roemerianus salt marsh and pine hummocks have increased seaward and J. roemerianus is slow to change compared to other species and remains a relic population.

  142  

REFERENCES

REFERENCES

Adam, P. 1990. Saltmarsh Ecology. Cambridge University Press, New York, NY. Adams, S.M. and S.A. Bortone 2005. Future Directions for Estuarine Indicator Research, p. 503-50. In S.A. Bortone (ed.), Estuarine Indicators. CRC Press, Inc., Boca Raton, Florida. Adger, W.N., T.P. Hughes, C. Folke, S.R. Carpenter and J. Rockström 2005. Social-ecological resilience to coastal disasters. Science 309(5737):10361039. Alberti, M., R. Weeks and S. Coe 2004. Urban land cover change analysis for the Central Puget Sound: 1991-1999. Journal of Photogrammetry and Remote Sensing 70:1043-1052. Allen-Diaz, B. and J.W. Bartolome 1998. Sagebrush-grass vegetation dynamics. Comparing classical and state-transition models. Ecological Applications 8:795-804. Allison, G. 2004. The influence of species diversity and stress intensity on community resistance and resilience. Ecological Monographs, 74(1):117134. Angradi, T.R., S.M. Hagan, and K.W. Able 2001. Vegetation type and the intertidal macroinvertebrate fauna of a brackish marsh: Phragmites vs. Spartina. Wetlands 21:75-92. Aubreville, A. 1938. La foret coloniale: les forets de l’Africque occidentale francaise. Annis Acadmy Science 9:1-245. Anisfeld, S.C., M.J. Tobin and G. Benoit 1999. Sedimentation rates in flowrestricted and restored salt marshes in Long Island Sound. Estuaries 22:231-244.

  143  

Antrop, M. 2001. The language of landscape ecologists and planners. Landscape and Urban Planning 55(3):163-173. Bailey, J., D.H.P. Jones, G.E. Parkes, and K.O. Mason 1978. Polarimetry and photometry of AM Herculis Mon. Not. R. Astron. Society 184:73-77. Baker, C., R.L. Lawnce, C. Montagne and D. Patten 2007. Change detection of wetland ecosystems using Landsat imagery and change vector analysis. Wetlands 27(3):610-619. Bakker J.P., P. Esselink, K.S. Dijkema, W.E. van Duin, and D.J. de Jong 2002. Restoration of salt marshes in the Netherlands. Hydrobiologia 478: 29-51. Baldwin, A.H., M.S. Egnotovich and E. Clarke 2001. Hydrologic change and vegetation of tidal freshwater marshes: Field, greenhouse and seed bank experiments. Wetlands 21:519-531. Baldwin, A.H. and I.A. Mendelssohn 1998. Effects of salinity and water level on coastal marshes: An experimental test of disturbance as a catalyst for vegetation change. Aquatic Botany 61:255-268. Barnette M.R, and D.W. Crewz 1997. Common coastal plants in Florida: A guide to planning and maintenance. Florida Sea Grant Program. University Press of Florida. Gainesville. 110pp. Barras, J.A. 2007. Satellite images and aerial photographs of the effects of Hurricanes Katrina and Rita on coastal Louisiana: U.S. Geological Survey Data Series 281, at http://pubs.usgs.gov/ds/2007/281. Barras, J.A. 2006. Land area change in coastal Louisiana after the 2005 hurricanes – a series of three maps: U.S. Geological Survey Open-File Report 206-1274, http://pubs.usgs.gov/of/2006/1274/ accessed 01 Sept 2009. Barras, J.A. 2003. Land area changes in coastal Louisiana after Hurricanes Katrina and Rita: U.S. Geological Survey Science and the Storms: the USGS Response to the Hurricanes of 2005. pubs.usgs.gov/circ/1306/pdf/c1306_ch5_b.pdf accessed 01 Seot 2009. Bartha, S., S.J. Meiners, S.T.A. Pickett and M.L. Candenasso 2003. Plant colonization windowns in a mesic old field succession. Applied Vegetation Science 6:205-212.

  144  

Beaumont, N.J., M.C. Austen, S.C. Mangi and M. Townsend 2008. Economic valuation for the conservation of marine biodiversity. Marine Pollution Bulletin 56(3):386-396. Beck, M.W., K.L. Heck, Jr., K.W. Able, D.L. Childers, D.B. Eggleston, B.M. Gillanders, B.S. Halpern, C.G. Hays, K. Hoshino, T.J. Minello, R.J. Orth, P.F. Sheridan and M.P. Weinstein 2003. The role of nearshore ecosystems as fish and shellfish nurseries, 1112 pp. Issues in Ecology, Volume 11. Ecological Society of America, Washington, D.C. Bedford, B.L. 1999. Cumulative effects on wetland landscapes: links to wetland restoration in the United States and southern Canada. Wetlands 19:775-778. Beisner, B.E., D.T. Haydon and K. Cuddington 2003. Alternative stable states in ecology. Frontiers in Ecology and Environment 1:376-82. Bell, S.S. 1980. Meiofauna-macrofauna interactions in a high salt marsh habitat. Ecological Monographs 50:487-505. Benincá, E., J. Huisman, R. Heerkloss, K.D. Jöhnk, P. Branco, E.H. van Nes, M. Scheffer, and S. P. Ellner 2008. Chaos in a long-term experiment with a plankton community. Nature 451:822-25. Bernier, J.C., R.A. Morton, and J.A. Barras 2006. Constraining rates and trends of historical wetland loss, Mississippi River delta Plain, SouthCentral Louisiana, In Coastal Environment and Water Quality, Y.J. Xu and V.P. Singh (editors), 371-382. Bertness ,M.D. 1992. The ecology of a New England salt marsh. American Scientist 80:260-268. Bestelmeyer, B.T. 2006. Threshold concepts and their use in rangeland management and restoration: The good, the bad, and the insidious. Restoration Ecology 14:325-29. Biles, C.L., D.M. Paterson, R.B. Ford, M. Solan and D.G. Raffaelli 2002. Bioturbation, ecosystem functioning, and community structure. Hydrology and Earth System Service 6(6):999-1005. Bishop, Y.M.M., S.E. Fienberg and P.W. Holland 1975. Discrete Multivariate Analysis Theory and Practice. MIT Press, Cambridge, Massachusetts 557p.

  145  

Blackwell K.G., P. Fitzpatrick, C. Velden, and T. Wimmers 2007. Double (concentric) eyewalls in Hurricane Katrina at landfall: A key to storm’s huge size and devastating impact over a three-state coastal region. 61st Interdepartmental Hurricane Conference, Office of the Federal Coordinator for Meterology, New Orleans, LA, March 2007. Blake, Eric S., E.N. Rappaport, J.D. Jarrell and C.W. Landsea 2005. The Deadliest, Costliest, and Most Intense United States Tropical Cyclones from 1851 to 2004 (And Other Frequently Requested Hurricane Facts). NOAA Technical Memorandum NWS TPC-4, Tropical Prediction Center, Miami, FL. Boesch, D.F., M.N. Josselyn, A.J. Mehta, J.T. Morris, W.K. Nuttle, C.A. Simenstad and D.J.P. Swift 1994. Scientific assessment of coastal wetland loss, restoration and management in Louisiana. Journal of Coastal Research Special Issue 20:103. Bolton H.S. (Ed.) 1991. Coastal Wetlands. New York: American Society of Civil Engineers, 515 pp. Bortone, S.A. 2006. Recommendations on establishing a research strategy in the Gulf of Mexico to assess the effects of hurricanes on coastal ecosystems. Estuaries and Coasts 29(6A):1062-1066. Bortone, S.A. 2005. Estuarine Indicators. CRC Press, Inc., Boca Raton, Florida. Bortone, S.A. 1976. The effects of a hurricane on the fish fauna at Destin, Florida. Florida Scientist 39:245-248. Bouma, A.H. 1963. Sampling and treatment of unconsolidated sediments for study of internal structures. Journal of Sedimentary Research 34(2):349354. Boumans, R.M.J, D.M. Burdick and M. Dionne 2002. Modeling habitat change in salt marshes after tidal restoration. Restoration Ecology 10(3):543-555. Brinson, M.M. and R.R. Christian 1999. Stability of Juncus roemerianus patches in a salt marsh. Wetlands 19(1):65-70. Broome S.W., E.D. Seneca, and W.W. Woodhouse, Jr. 1988. Tidal salt marsh restoration. Aquatic Botany 32: 1-22.

  146  

Brown, J.R., J. Herrick and D. Price 1999. Managing low-output agroecosystems sustainability: the importance of ecological thresholds. Canadian Journal of Forest Restoration 29:1112-9. Burd, F. 1992. Erosion and vegetation change on the saltmarshes of Essex and North Kent between 1973 and 1988. Research and Survey in Nature Conservancy Council, Peterborough, UK, report No. 42. Burnham, K.P., D.R. Anderson and J.L. Laake 1980. Estimation of density from line transect sampling of biological populations. Wildlife Monographs. 72:1-202. Cablk, M.E., B. Kjerfve, W.K. Michener, and J.R. Jensen 1994. Impacts of Hurricane Hugo on a coastal forest: Assessment using Landsat TM data. Geocarto International 2: 15-24. Cai, W.J. and Y. Wang 1998. The chemistry, fluxes, and sources of carbon dioxide in the estuarine waters of the Satilla and Altamaha Rivers, Georgia. Limnology and Oceanography 43(4):667-688. Caloz R. and C. Collet 1997. Geographic information systems (GIS) and remote sensing in aquatic botany: methodological aspects. Aquatic Botany 58: 209-228. Carpenter, S.R. 2003. Regime shifts in lake ecosystems: Pattern and variation. Oldendorf/Luhe, Germany: Ecology Institute. Carpenter, S.R. and W.A. Brock 2006. Rising variance: A leading indicator of ecological transition. Ecology Letters 9:308-15. Cebrian, J. and C. M. Duarte.1994. The dependence of herbivory on growth rate in natural plant communities. Funct. Ecol. 8: 518-525 Cebrian, J. and C. M. Duarte. 1995. Plant growth-rate dependence of detrital carbon storage in ecosystems. Science 268: 1606-1608. Cebrian, J., M. Williams, J. MacClelland and I. Valiela. 1998. The dependence of heterotrophy and carbon accumulation on autotrophic nutrient concentration in ecosystems. Ecology Letters 1:165-170. Cebrian J. 1999. Patterns in the fate of production in plant communities. Am. Nat. 154:449-468.

  147  

Chabreck, R.H. and A.W. Palmisano 1973. The effects of hurricane Camille on the marshes of the Mississippi River delta. Ecology 54:1118-23. Chabreck, R.H. 1988. Coastal Marshes: Ecology and Wildlife Management. University of Minnesota Press, Minneapolis, MN. Chambers, R.M., L.A. Meyerson and K. Saltonstall 1999. Expansion of Phragmites australis into tidal wetlands of North America. Aquatic Botany 64:261-273. Chapman, C.R. 1973. The impact of estuaries and marshes on modifying tributary runoff. Pp. 235-58, In Chabreck, R.H. (ed.). Proceedings of 2nd Coastal Marsh and Estuary Management Symposium, Louisiana State University, Baton Rouge, Louisiana. Chapman, S.B. 1976. Methods in Plant Ecology Oxford Blackwell Scientific Publications, New York, p. 589. Chavez, P.S., Jr., and D.J. MacKinnon 1994. Automatic detection of vegetation changes in the Southwestern United States using remotely sensed images. Journal of Photogrammetric Engineering and Remote Sensing 60(5):571-583. Chen, Q., L. Wang and R. Tawes 2008. Hydrodynamic response of Northeastern Gulf of Mexico Hurricanes. Estuaries and Coasts 31(6):1098-1116. Choi, Y.D. and M.K. Wali 1995. The role of Panicum virgatum (switch grass) in the revegtation of iron-mine tailings in northern New York. Restoration Ecology 3:123-32. Christensen, E.J., J.R. Jensen, E.W. Ramsey III and H.E. Mackey 1987. Aircraft MSS data registration and vegetation classification of wetland change detection. International Journal of Remote Sensing 9(1):23-28. Civco D.L., J.D. Hurd, E.H. Wilson, M. Song and Z. Zhang 2002. A comparison of land use and land cover change detection methods. American Congress on Surveying and Mapping. American Society for Photogrammetry and Remote Sensing. Annual conference Proceedings. Clements, F.E. 1916. Plant Succession: An Analysis of the Development of Vegetation. Carnegie Institute of Washington, Publication 242. 512pp.

  148  

Cocks, T., J. Jensen, A. Stewart, I. Wilson and T. Shields 1998. HyMap airborne hyperspectral sensor: the system, calibration, and performance. In M. Schaepman. D. Schlpfer, and K.I. Ltten (Eds.). First EARSel Workshop on Imaging Spectroscopy (pp. 37-43). Paris EARSel. Collie, J.S., K. Richardson and J.H. Steele 2004. Regime shifts: Can ecological theory illuminate the mechanisms? Progress in Oceanography 60:281-302. Conner, W.H., J.W. Day, R.H. Bauman and J.M. Randall 1989. Influences of hurricanes on coastal ecosytems along the Northern Gulf of Mexico. Wetlands Ecology and Management Vol. 1. Coppin P., I. Jonckheere, K. Nackaerts, B. Muys and E. Lamloin 2004. Digital change detection methods in ecosystem monitoring: a review. International Journal of Remote Sensing 25:1565-96. Corbosiero, K.L., J. Molinari and M.L. Black 2005. The structure and evolution of Hurricane Elena (1985). Part I: Symmetric Intensification. American Meteorological Society Monthly Weather Review 133:2905-2921. Córdova-Kreylos, A.L., Y. Cao, P.G. Green, H.M. Hwang, K.M. Kuivila, M.G. LaMontagne, L.C. Van De Werfhorst, P.A. Holden and K.M. Scow 2006. Diversity, composition, and geographical distribution of microbial communities in California salt marsh sediments. Applied and Environmental Microbiology 72(5):3357-3366. Costa C.S., J.C. Marangoni and A.M. Azevedo 2003. Plant zonation in irregularly flooded salt marshes; relative importance of stress tolerance and biological interactions. Journal of Ecology 91:951-965. Cowardin, L.M., V. Carter, F.C. Golet and E.T. LaRoe 1979. Classification of wetlands and deepwater habitats of the United States FWS/OBS-79/31, U.S. Fish and Wildlife Service, Washington, D.C. Craft, C.B., S.W. Broome and E.D.Seneca 1988. Nitrogen, phosphorus and organic carbon pools in natural and transplanted marsh soils. Estuaries 11:272-280. Criswell, M.E. and U.S. Army Engineer Waterways Experiment Station. 1970. Survey of Gulf Coast structural damage resulting from Hurricane Camille, August 1969. Technical Report N, 70-10. Vicksburg, Mississippi: U.S. Army Engineer Waterways Experiment Station.

  149  

Curran, P.J. 1994. Imaging spectrometry. Progress in Physical Geography 18(2):27-266. Dahdouh-Guebas, F. and N. Koedam 2006. Coastal vegetation and the Asian tsunami. Science 311:37. Dahl T.W. 1990. Wetland losses in the United States 1780’s to 1980s. Washington D.C.: US Department of Interior, U.S. Fish and Wildlife Service, 21 p. Daily, M. 1983. Hue-saturation-intensity split-spectrum processing of Seasat radar imagery Photogrammetric Engineering and Remote Sensing 49(3): 349-355. Davis, J. L., P. Elósegui, J. X. Mitrovica, and M. E. Tamisiea 2004. Climatedriven deformation of the solid Earth from GRACE and GPS, Geophys. Res. Lett., 31, Delaney, T.P. and J.W. Webb 1995. Methods for using a GIS in a quantitative analytical study of estuarine marshes. In GIS/LIS ’95 Annual Conference and Exposition Proceedings, pp.277-286. De Leo, G.A. and S. Levin 1997. The multifaceted aspects of ecosystem integrity. Conservation Ecology 1(1):3 Available for the Internet. URL://www.consecol.org/vol1/iss1/art3/ Del Moral, R. 2000. Succession and species turnover on Mount St. Helens, Washington. Acta Phytogeographica Suecica 85:53-62. Desclee, B., P. Bogaert and P. Defourny 2006. Forest change detection by statistical object-based method. Remote Sensing of Environment 102:1-11. Dingler J.R. and T.E. Reiss 1995. Beach erosion on Trinity Island, Louisiana caused by Hurricane Andrew. Journal of Coastal Research. Special Issue 21: 254-264. Dobsen J.E., R.L. Ferguson, D.W. Field, L.L. Wood, K.D. Haddad, H. Iredale, J.R. Jensen, V.V. Klemas, R.J. Orth and J.P. Thomas 1995. NOAA Coastal Change Analysis Project (C-CAP): Guidance for Regional Implementation, National Oceanic and Atmospheric Administration, NMFS 123, 92p. Dobsen, J.E. 1993. Commentary: A conceptual framework for integrating

  150  

remote sensing, GIS and geography. Photogrammetric Engineering and Remote Sensing 59(10):1491-1496. Donnelly, J.P. , S.S. Bryant, J. Butler, J. Dowling, L. Fan, N. Nausmann, P. Newby, B. Shuman, J. Stern, K. Westover, and T. Webb III 2001. 700 yr sedimentary record of intense hurricane landfalls in southern New England. GSA Bulletin 113:714-727. Drury, W.H. and I.C.T. Nisbet 1973. Succession. Journal of the Arnold Arboretum 54:331-68. Duarte, C.M., J. Cebrian and N. Marbá 1992. Global change in the marine environment: the life-history of long-term monitoring programmes in Europe. Nature 356: 190. Dudgeon, S. and P.S. Petraitis 2001. Scale-dependent recruitment and divergence of intertidal communities. Ecology 82:991-1006. Duncan W.H., and M.B. Duncan 1987. The Smithsonian guide to seaside plants of the Gulf and Atlantic coasts from Louisiana to Massachusetts, exclusive of lower peninsular Florida. Washington, D.C., Smithsonian Institution Press. Eastwood, J.A., M.G. Yates, A.G. Thomson and R.M. Fuller 1997. The reliability of vegetation indices for monitoring saltmarsh vegetation cover. International Journal of Remote Sensing 18:3901-3907. Edminston, H.L., S.A. Fahrny, M.S. Lamb, L.K. Levi, J.M. Wanat, J.S. Avant, K. Wren and N.C. 2008. Tropical storm and hurricane impacts on a Gulf Coast estuary: Apalachicola Bay, Florida. Journal of Coastal Research, SI(55), 38–49. West Palm Beach (Florida), ISSN 0749-0208. Eleuterius, L.N. 1976. Vegetative morphology and anatomy of the salt marsh rush, Juncus roemerianus. Gulf Research Reports 5:1-10. Eleuterius, L.N. and J.D. Caldwell 1985. Soil characteristics of four Juncus roemerianus populations in Mississippi. Gulf Research Reports 8(1):9-13. Elmqvist, T., C. Folke, M. Nystrom, G. Peterson, J. Bengtsson, B. Walker and J. Norberg 2003. Response diversity, ecosystem change, and resilience. Frontiers in Ecology and the Environment 1:488-94. Elsner J.B. and A.B. Kara 1997. Hurricane return periods along the Gulf

  151  

Coast and Florida: NOAA Technical Memorandum NWS SR, 192. Fort Worth, Texas: Scientific Services Division, Southern Region, U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Weather Service. Elvidge, C.D., Miura, T., Jansen, W.T., Groeneveld, D.P. and Ray, J. 1998. Monitoring trends in wetland vegetation using a Landsat MSS time series. In R. S. Lunetta, & C. D. Elvidge (Eds.) Remote sensing change detection: environmental monitoring methods and applications (pp.191-210). Chelsea, MI: Ann Arbor Press (318-illustrations, co-published in Europe by Taylor and Francis, UK). Elvidge, C.D. 1990. Visible and near-infrared characteristics of dry plant materials. International Journal of Remote Sensing 11(10):1775-1795. Engle J.B. 1948. Investigations of the oyster reefs of Mississippi, Louisiana, and Alabama following the hurricane of September 19, 1947. Special Scientific Report: (U.S. Fish and Wildlife Service), no. 59. Washington D.C.: U.S. Department of Interior. Ensminger A.B. and L.G. Nichols 1957. Hurricane damage to Rockefeller Refuge. Proceedings of Annual Conference of Southeast Association of Game and Fisheries Commission 11:52-56. Environment for Visualizing Images 2003. User’s Guide v.4.0 . Research Systems, Inc. Boulder, CO. pp 1084. Environmental Protection Agency 2006. National Estuary Coastal Condition Report EPA-842/B-06/001 Office of Water/ Office of Research and Development, Washington DC pp 445. Ervin S. and H. Hasbrouck 2001. Landscape Modeling: Digital Techniques for Landscape Visualization. McGraw Hill, New York pp 290. Estes, J.E. 1992. Technology and policy issues impact global monitoring, GIS World 5(10):52-55. Estes, R.J. 1992. At the Crossroads: Dilemmas in Development Toward the Year 2000 and Beyond. Praeger, New York. Everitt J.H., D.E. Escobar and M.R. Davis 1995. Using remote sensing for detecting and mapping noxious plants. Weed Abstracts 4(12):639-49.

  152  

Fahrig, L. 1997. Relative effects of habitat loss and fragmentation on population extinction. Journal of Wildlife Management 61(3):603-610. Faulk, D.A., M.A. Palmer and J.B. Zedler (Eds.) 2008. Foundations of Restoration Ecology, Island Press Washington D.C. pp. 364. Finkl, C.W. 2004. Coastal classification: Systematic approaches to consider in the development of a comprehensive system. Journal of Coastal Research 20(1):166-213. Fisher, P.F., and S. Pathirana, 1990. The evaluation of fuzzy member- ship of land cover classes in the suburban zone. Remote Sensing of Environment 34:121-132. Foody, G.M. 2009. The impact of imperfect ground reference data on the accuracy of land cover change estimation. International Journal of Remote Sensing 30(12):3275-3281. Ford, M.A., D.R. Cahoon and J.C. Lynch 1999. Restoring marsh elevation in a rapidly subsiding salt marsh by thin-layer deposition of dredged material. Ecological Engineering 12(3-4):189-205. French, G.T. and K. Moore 2003. Interactive effects of light and salinity stress on the growth, reproduction, and photosynthetic capabilities of Vallisneria americana (wild celery). Estuaries 26(5):1255-1268. Friedrichs, C.T. and J.E. Perry 2001. Tidal saltmarsh morphodynamics: a synthesis. Journal of Coastal Research SI27:7-37. Fung, T. and W. Siu 2000. Environmental quality and its changes, an analysis using NDVI. International Journal of Remote Sensing 21(5):10111024. Gabriel, B.C. and A. de la Cruz 1974. Species composition, standing stock, and net primary production of a salt marsh community in Mississippi. Chesapeake Science 15:72-77. Gallagher, J.L. 1980. Salt marsh soil development. Rehabilitation and creation of selected coastal habitats: proceedings of a workshop, FWS/OBS-80/27:28-34 Gallagher, J.L. 1975. Effect of an ammonium nitrate pulse on the growth and elemental composition of natural stands of Spartina alterniflora and

  153  

Juncus roemerianus. American Journal of Botany 2:644-648. Gallagher, J. L. 1974. Sampling macro-organic matter profiles in salt marsh plant root zones. Proc. Soil Sci. America 38:155-155. Gleason, H.A. 1917. The Structure and Development of the Plant Association. Bulletin of the Torrey Botanical Club 43:463-481. Goetz, A.F.H. 1992. Imaging Spectrometry for Earth Remote Sensing, in Imaging Spectroscopy: Fundamentals and Prospective Applications, F. Toselli and J. Bodechtel, Eds., pp. 1-19, Brussels and Luxembourg. Goetz, A.F.H. 1995. Imaging spectrometry for remote sensing: vision to reality in 15 years. In M. R. Descour. J. M. Money. D. L. Perry and I. Illing (Eds.) Imaging Spectrometry, vol. 2480 (pp. 2-13). Bellingham, WA: The International Society for Optical Engineering (SPIE). Goodchild, M.F. 2003. Geospatial data in emergencies. In S.L. Cutter, D.B. Richardson, and T.J. Wilbanks, editors, The Geographical Dimensions of Terrorism. New York: Routledge, pp. 99–104. Goodchild, M.F. and D.M. Mark 1987. The fractal nature of geographic phenomena. Annals, Association of American Geographers 77: 265–278. Google 2010. Google Enterprise: Earth and Maps. Google Earth URL: http://earth.google.com/ Green, R.O., M.I. Eastwood, C.M. Sarture, T.T. Chrien, M. Aronsson, H.J. Chippendale, J.A. Faust, B.F. Pavri, C.J. Chovit, M. Solis, M.R. Otah and O. Williams 1998. Imaging spectroscopy and the airborne visible/infrared imaging spectrometer (AVIRIS). Remote Sensing of Environment 65:227-248. Green, D.R. and D.C. Morton 1994. Acquiring Environmental Remotely Sensed Data from Model Aircraft for Input to Geographic Information Systems. In Proceedings AGI'94 Conference. November 15th-17th 1994. pp. 15.3.1-15.3.28. Groffman, P.M., J.S. Baron, T. Blett, A.J. Gold, I. Goodman, L.H. Gunderson, B.M Levinson, M.A. Palmer, H.W. Paerl, G.D. Peterson, N.L. Poff, D.W. Rejeski, J.F. Reynolds, M.G. Turner, K.C. Weathers and John Wiens 2006. Ecological thresholds: the key to successful environmental management or an important concept with no practical application.

  154  

Ecosystems 9:1-13. Gustafson, E.J. and G.R. Parker 1992. Relationships between landcover proportion and indices of landscape spatial change. Landscape Ecology 7:101-110. Hacker, S.D. and M.D. Bertness 1999. Experimental evidence for factors maintaining plant species diversity in a New England salt marsh. Ecology 80(6):2064-2073. Hackney, C.T., and S.M. Adams 1992. Aquatic Communities of the Southeastern United States: Past, Present, and Future. In Biodiversity of the Southeastern United States: Aquatic Communities. C. T. Hackney, S. M. Adams, and W. H. Martin (eds.). John Wiley and Sons, Inc., New York, NY, p. 747-760. Hadley, R.F., M.R. Karlinger, A.W. Burns, and T.R. Eschner 1987. Water development and associated hydrologic changes in the Platte River, Nebraska, U.S.A. Regulated Rivers: Research and Management 1:331– 341. Häme, T., I. Heiler and J. Miguel-Ayanz 1998. An unsupervised change detecting and recognition system for forestry. International Journal of Remote Sensing 19(i6):1079-1099. Hanna, K.C. and B. Kulpepper 1998. GIS in Site Design: New Tools for Design Professionals. New York: John Wiley and Sons. Hardesty, J., J. Adams, D. Gordon, and L. Provencher 2000. Applying science to management: Restoring ecological integrity at Eglin Air Force Base. Conservation Biology in Practice 1: 26-31. Hayes, D. and S. Sader 2001. Comparison of change-detection techniques for monitoring tropical forest clearing and vegetation regrowth in a time series Photogrammetric Engineering & Remote Sensing 67(9):1067-1075. Hill, M.O. 1973. Diversity and evenness: a unifying notation and its consequences. Ecology 54:427-431. Hinton, J.C. 1996. GIS and remote sensing integration for environmental applications. International Journal of Global Information Systems 10(7):877-890.

  155  

Hirano, A., M. Madden and R. Welch 2003. Hyperspectral image data for mapping wetland vegetation. Wetlands 23(2):436-448. Ho, F.P., R.W. Schwerdt and H.V. Goodyear 1975. Some climatological characteristics of hurricanes and tropical storms, Gulf and east coasts of the United States. NOAA Technical Report NWS, 15. Washington D.C.: U.S. Department of Commerce: NOAA: NWS. p. 76. Ho, F.P. and R.J. Tracey 1975. Storm tide frequency analysis for the Gulf coast of Florida from Cape San Blas to St. Petersburg Beach. NOAA Technology Memorandum NWS HYDRO20, Silver Springs, Maryland: Office of Hydrology, National Weather Service. Hobbs, R.J. 1993. Effects of landscape fragmentation on ecosystem processes in the Western Australian Wheatbelt. Biological Conservation 64:193-201. Hobbs, R.J. and K.N. Suding (Eds.) 2008. New Models for Ecosystem Dynamics Island Press: The Science and Practice of Ecological Restoration Series. Washington D.C. 366p. Holling, C.S. and C.R. Allen 2002. Adaptive inference for distinguishing credible from incredible patterns in nature. Ecosystems 5:319-28. Holling C.S. Eds. 1978. Adaptive Environmental Assessment and Management. London: Wiley. Holmes, J., K.G. Blackwell, R.A. Wade and S.K. Kimball 2006. Collapsing precipitation cores in open-eyewall hurricanes at landfall: Are these cores actually downbursts associated with extreme surface wind gusts? 27th Conference of Hurricanes and Tropical Meteorology, American Meteorology Society, Monterray, CA April 2006. Paper 7B.7. http://ams.confex.com/ams.27 Hurricanes/techprogram/paper_108947.htm Hsu, Shih-Ang et al. 1970. Coastal Studies Bulletin. Baton Rouge, Louisiana: Coastal Studies Institute. Louisiana State University. Hudak, A.T. and C.A. Wessman 1998. Textural analysis of aerial photography to characterize large scale land cover change http://gis.esri.com/library/userconf/proc97/proc97/to650/pap643/p643.htm Hughes, C.E., P. Binning and G.R. Willgoose 1998. Characterization of the hydrology of an estuarine wetland. Journal of Hydrology 211(1-4):34-9.

  156  

Huiskes A.H.L., B.P. Koutstall, P.M. J. Herman, W.G. Beeftink, M.M. Markusse and W. DeMunck 1995. Seed dispersal of halophytes in tidal salt marshes. Journal of Ecology 83:559-567. Hume, N.P., M.S. Fleming and A.J. Horne 2002. Denitrification potential and carbon quality of four aquatic plants in wetland microcosms. Soil Science Society of America Journal 66:1706-1712. Hupp C.R., W.R. Osterkamp and A.D. Howard, (EDds.) 1995, Biogeomorphology, Terrestrial and Freshwater Systems the Engineer. Pp. 35, Elsevier Science Publication Co., 1995. Ives, A.R. and S.R. Carpenter 2007. Stability and diversity of ecosystems. Science 317:58-62. Jackson, R.D. and J.W. Bartolome 2002. A state-transition approach to understanding nonequilibrium plant community dynamics in Californian grasslands. Plant Ecology 162:49-65. Jacob, J. and R. Lopez 2005. Freshwater, Non-Tidal Wetland Loss – Lower Galveston Bay Watershed, 1992 -2002. Unpublished report. Galveston Bay Estuary Program, Webster, TX. Jensen, J.R. 2000. Remote sensing of the environment: An Earth Resource Perspective. New Jersey: Prentice Hall, 544 p. Jensen, J.R. 1996. Introduction to Digital Image Processing: A Remote Sensing Perspective. Prentice Hall, Upper Saddle River, NJ, USA pp. 318. Jenson, J.R., K. Rutchey, M.S. Koch, and S. Narumalani 1995. Inland wetland change detection in the everglades water conservation area 2A using a time series of normalized remotely sensed data. Photogrammetric Engineering and Remote Sensing 61(2):199-209. Jensen, J.R., E.W. Ramsey, H.E. Mackey, E. Christensen and R. Sharitz 1987. Inland wetland change detection using aircraft MSS data. Photogrammetric Engineering and Remote Sensing 53(5):521-529. Johansen, K., and S. Phinn 2006. Linking riparian vegetation spatial structure in Australian tropical savannas to ecosystem health indicators: Semivariogram analysis of high spatial resolution satellite imagery. Canadian Journal of Remote Sensing 32:228−243.

  157  

Johnson, S. 2006. Stephen Johnson on Digital Photography. O’Reilly Media, Sebastopol, CA 305 pp. Johnson R.D. and E.S. Kasischke 1998. Change vector analysis: a technique for the multispectral monitoring of land cover and condition. International Journal of Remote Sensing 19(3):411-426. Karl, T.J. 1986. The sensitivity of the Palmer Drought Severity Index and Palmer’s Z-Index to their calibration coefficients including potential evapotranspiration. Journal of Applied Meteorology 25:77-86. Keefe, C.W. 1972. Marsh production: A summary of the literature. Contributions in Marine Science 16:163-181. Kim, K.C. and R.D. Weaver (eds). 1994. Biodiversity and Landscapes: A Paradox of Humanity. NY: Cambridge University Press. King, E.G. and R.J. Hobbs 2006. Indentifying linkages among conceptual models of ecosystem degradation and restoration towards an integrative framework. Restoration Ecology 14:369-378. Kleinberg, E. 2004. Black cloud: The Greatest Hurricane of 1928. Carroll and Graft. New York. Kleinen, T., H. Held and G. Petschel-Held 2003. The potential role of spectral properties in detecting thresholds in the earth system: Application to the thermohaline circulation. Ocean Dynamics 53:53-63. Klemas V., J.E. Dobson, R.L. Ferguson and K.D. Haddad 1993. A coastal land cover classification system for the NOAA CostWatch Change Analysis Project. Journal of Coastal Research 9(3):862-872. Knutson, P.L., J.C. Ford, M.R. Inskeep and J. Oyler 1981. National survey of planted salt marshes (vegetative stabilization and wave stress). Wetlands 1:129-157. Knutson, P.L., R.A. Brouchu, W.N. Seelig and M. Inskeep 1982. Wave damping in Spartina alterniflora marshes. Wetlands 2:87-104. Koretsky, C.M. and D. Miller 2008. Seasonal influence of the needle rush Juncus roemerianus on saltmarsh pore water geochemistry. Estuaries and Coasts: J. CERF 31:70-84.

  158  

Krummel, J.R., R.H. Gardner, G. Sugihara, R.V. O’Neill and PR. Coleman 1987. Landscape patterns in a disturbed environment. Oikos 48:321-4. Kuhn, N.L, I.A. Mendelssohn and D.J. Reed 1999. Altered hydrology effects on Louisiana salt marsh function. Wetlands 19(3):617-626. Kulkarni, A. 2004. Evaluation of the impacts of hurricane hugo on the land cover of Francis Marion National Forest, South Carolina using remote sensing Louisiana State University Thesis, Department of Geography and Anthropology 115 pp. LaSalle, M.W., M.C. Landin and J.G. Sims 1991. Evaluation of the flora and fauna of a Spartina alterniflora marsh established on dredged material in Winyah Bay, South Carolina. Wetlands 11:191-208. LaSalle, M.W. 1996. Assessing the functional level of a constructed intertidal marsh in Mississippi. Technical Report, WRP-RE-15. U.S. Army Engineer Waterways Experiment Station, Vicksburg, MS. Leemans, R. 1999. Modeling for species and habitats: new opportunities for problem solving. Science for Total Environment 240:51-73. Lehmann, A., J.M. Jaquet and J.B. Lachavanne 1994. Contribution of GIS to submerged macrophyte biomass estimation and community structure modeling, Lake Geneva, Switzerland. Aquatic Botany 47:99-117. Lehmann, A., J.M Jaquet and J.B. Lachavanne 1997. A GIS approach of aquatic plant spatial heterogeneity in relation to sediment and depth gradients, Lake Geneva, Switzerland. Aquatic Botany 58:347-361. Lehmann, A. and J.B. Lachavanne 1997. Geographic information systems and remote sensing in aquatic botany. Aquatic Botany 58:195-207. Leitão A.B., J. Miller, J. Ahern, and K. McGarigal 2006. Measuring Landscapes: A Planner’s Handbook. Island Press, Washington 245 pp. Levin, J.M., J.S. Brewer and M.D. Bertness 1998. Nutrients, competition and plant zonation in a New England salt marsh. Journal of Ecology 86(2):285-292. Levine, J.M., M. Vila, C.M. D’Antonio, J.S. Dukes, K. Grigulis and S. Lavorel 2003. Mechanisms underlying the impacts of exotic plant invasions. Proceedings of The Royal Society, Series B 270:775-781.

  159  

Lewontin, R.C. 1969. The bases of conflict in biological explanation. Journal of History of Biology 2:35-45. Lewis, R.R., III and C.S. Lewis 1977. Tidal marsh creation on dredged material in Tampa Bay, Florida. Proceedings of the fourth annual conference on restoration of coastal vegetation in Florida., Tampa, FL, 4667. Light, D.L. 1993. The National aerial photography program as a geographic information system resource. Photogrammetric Engineering and Remote Sensing 48(1):61-65. Lindberg, S.G. and R.C. Harriss 1973. Mechanisms controlling pore water salinities in a salt marsh. Limnology and Oceanography 18:788-791. Lindenmayer, D.B. and J. Fischer 2006. Habitat Fragmentation and Landscape Change: An Ecological and Conservation Synthesis. Island Press, Washington D.C. 328pp. Lindenmayer, D.B. and M.A. Burgman 2005. Practical Conservation Biology. CSIRO Publishing, Melbourne, Australia. Lindig-Cisneros, R.J. Desmond, K.E. Boyer and J.B. Zedler 2003. Wetland restoration thresholds: Can a degradation transition be reversed with increased effort? Ecological Applications 13:193-205. Litaker, R.W. and P.A. Tester 2003. Extreme events and ecological forecasting, P. 85-91. In N. Valette-Silver and D. Scavia (eds.), Ecological Forecasting: New Tools for Coastal and Marine Ecosystme Management. National Oceanic and Atmospheric Administration Technical Memorandum NOS-NCCOS-1. Washington D.C. Lloyd H, J.H. Zar, and J.R. Karr 1968. On the calculation of informationtheoretical measures of diversity. Am Mid Nat. 79(2):257–272. Louisiana Executive Summary 1997. U.S. Congress on effectiveness of Louisiana Coastal Wetland Restoration Projects. Coastal Wetlands Planning, Protection and Restoration Act. Public Law 1001-646, Title III. Ludwig, J.A., J.A. Wiens and D.J. Tongway 2000. A scaling rule for landscape patches and how it applies to conserving soil resources in savannas. Ecosystems 3:84-97.

  160  

Lugo, A.E. 2000. Climate change and tropical forests—uncertainty and fear. Journal of Environmental Quality 29(1): (book review). Lulow, M.E. 2006. Invasion by non-native annual grasses: The importance of species biomass, composition and time among California native grasses of the Central Valley. Restoration Ecology 14:616-26. Luna, H.N.M. and J.B. Zedler 2007. Does seed availability limit plant establishment during salt marsh restoration? Estuaries and Coasts 30(1):12-25. Lunetta, R.S., J. Ediriwickrema, D. M. Johnson, J. G. Lyon and A. McKerrow 2002. Impacts of vegetation dynamics on the identification of land-cover change in a biologically complex community in North Carolina, USA. Remote Sensing of Environment 82:258-270. Lynne, G.D., P. Conroy, and F. J. Prochaska 1981. Economic valuation of marsh areas or marine production processes. Journal of Environmental Economic Management 8(2): 175-186. Lytle, D.A. and N.L. Poff 2004. Adaptation to natural flow regimes. Trends in Ecology and Evolution 19:94-100. Mack, M.C., C.M. D’Antonio and R.E. Ley 2001. Alteration of ecosystem nitrogen dynamics by exotic plants: A case study of C-4 grasses in Hawaii. Ecological Applications 11:1323-35. Mallin, M.A. and C.A. Corbett 2006. How hurricane attributes determine the extent of environmental effects: Multiple hurricanes and different coastal systems. Estuaries and Coasts 29(6)1046-1061. Mas, J. 1999. Monitoring land-cover changes: a comparison of change detection techniques. International Journal of Remote Sensing 20:139-152. May, R.M. 1977. Thresholds and breakpoints in ecosystems with a multiplicity of stable states. Nature 269:471-7. Mayer, A.L. and M. Rietkerk 2004. The dynamic regime concept for ecosystem management and restoration. Bioscience 54:1013-20. McDonnell, M.J. and S.T.A. Pickett (Eds.). 1993. Humans as Components of Ecosystems: The Ecology of Subtle Human Effects and Populated Areas. Springer-Verlag, New York.

  161  

Mendelssohn, I.A. and K.L. Marcellus 1976. Angiosperm production of three Virginia marshes in various salinity and soil nutrient regimes. Chesapeake Science 17:15-23. Mendelssohn, I.A., R.W. Turner and K.L. McKee 1983. Louisiana’s eroding coastal zone: Management alternatives. Journal of the Limnology Society of South Africa 9:63-75. Michalak, W.Z. 1993. GIS in land use change analysis: integration of remotely sensed data into GIS. Applied Geography 13: 28-44. Michener, W. K., E. R. Blood, K. L. Bildstein, M. M. Brinson, and L.R. Gardner 1997. Climate change, hurricanes and tropical storms, and rising sea level in coastal wetlands, Ecological Applications, 7, 3, 770-801. Middleton B. 1999. Wetland Restoration: Flood Pulsing and Disturbance Dynamics John Wiley and Sons, New York 388 pp. Middleton, M., P. Närbi, M-L. Sutinen and R. Sutinen 2008. Object based change detection of historical aerial photographs reveals altitudinal forest expansion, G.J. Hay, Blaschke and D. Marceau (Eds.) GEOBIA 2008Pixels, Objects, Intelligence. GEOgraphic Object Based Image Analysis for the 21st Century. University of Calgary, Calgary Alberta, Canada, August 05-08. ISPRS Vol. NO. XXXVIII-4/C1. Archives ISSN No. : 16821777. 373p. Mitchell J. T. 2006. Hurricane Katrina and Mississippi’s invisible coast: a photographic journal Southeast Geographer 42(6):181-188. Mitchell, C.L. 1928. The West Indian hurricane of September 10-20, 1928. Monthly Weather Review 56:347-350. Mitsch W.J., and J. G. Gosselink 1993. Wetlands (Second Edition). New York, NY: Van Nostrand Reinhold. pp 722. Möller, I., T. Spencer, J.R. French and D.J. Leggett 2002. Wave Transformation over salt marshes: A field and numerical modeling study from north Norfolk, England. Estuarine, Coastal and Shelf Science 49:411426.. Morton, R., J. Bernier, J. Barras and N. Fernia 2005. Rapid subsidence and historical wetland loss in the Mississippi Delta Plain, likely causes and future implications: U.S. Geological Survey Open-File report 2005-1216,

  162  

42 p., http://pubs.usgs.gov/of/2005/1216/ accessed 01 Sept 2009. Mosteller, F. and C. Youtz 1961 Biometricka 48:433-440. Mullineaux, L.S., C.H. Peterson, F. Micheli and S.W. Mills 2003. Successional mechanism varies along a gradient in hydrothermal fluid flux at deep-sea vents. Ecological Monographs 73:523-42. National Aeronautical and Space Administration (NASA) 2010. U.S. Department of Interior. The LandSat Program URL: http://glovis.usgs.gov National Oceanic and Atmospheric Administration 2009. National Climatic Data Center. Palmer Drought Severity Index. URL http://wf.ncdc.noaa.gov/oa/climate/research/prelim/drought/palmer.htm. National Oceanic and Atmospheric Administration 2007a. The National Estuarine Research Reserve’s System-Wide Monitoring Program (SWMP): A Scientific Framework and Plan for Detection of Short-Term Variability and Long-Term Change in Estuaries and Coastal Habitats of the United States. Technical Report. Silver Spring, Maryland: National Oceanic and Atmospheric Administration, 39p. National Oceanic and Atmospheric Administration 2007b. Bringing the Big Picture into Focus: The Future of Remote Sensing at NOAA. [Accessed 30 November 2007.] Available from: http://celebrating200years.noaa.gov/visions/remote_sensing/welcome. html#chal National Oceanic and Atmospheric Administration, Office of Ocean and Coastal Resource Management, National Estuarine Research Reserve System-wide Monitoring Program. 2004. Centralized Data Management Office, Baruch Marine Field Lab, University of South Carolina http://cdmo.baruch.sc.edu. National Research Council 2004. A Geospatial Framework for the Coastal Zone, National Needs for Coastal Mapping and Charting Committee, National Research Council of the National Academies. Washington, DC: National Academies Press p. 168. Nicke, J., H. Schwarzer, A. Neumann and G. Zimmerman 1997. Imaging spaceborne and airborne sensor systems in the beginning of the next century. In H. Fujisada (Ed.), SPIE, Sensors, Systems and Next

  163  

Generation Satellites III. The European Symposium on Aerospace Remote Sensing (IEE), vol. 3221. Ipp. 581-592). London. UK: IFEE. Niering, W.S. and R.S. Warren 1980. Vegetation patterns and processes in New England salt marshes. BioScience 30:301-307. Noe G.B. 2002. Temporal variablility matters: effects of constant vs. varying moisture and salinity of germination. Ecological Monographs 72:427-443. Novak, K. 1992. Rectification of digital imagery Photogrammetric Engineering & Remote Sensing 58(3):339-344. O’Callaghan E. 1996. Land use. The Interaction of Economics, Ecology and Hydrology, Chapman Hall, London, UK. Odum, E.P. 1969. The strategy of ecosystem development. Science 164:262270. Odum, W.E., T.J. Smith, III, J.K. Hoover and C.C. Melvor 1984. The ecology of tidal freshwater marshes of the United States east coast: A community profile. FWS/OBS-83/17, U.S. Fish and Wildlife Service. Odum, H.T. 1996. Environmental Accounting: Energy and Environmental Decision Making. John Wiley, New York 370 pp. Orson, R.A., R.I. Simpson and R.E. Good 1992. The paleoecological development of late Holocene, tidal freshwater marsh of the upper Delaware River estuary. Estuaries 15:130-146. Paerl, H.W., J.D. Bales, L.W. Ausley, C.P. Buzzelli, L.B. Crowder, L.A. Eby, J.M. Fear, M. Go, B.L. Peierls, T.L. Richardson and J.S. Ramus 2001. Ecosystem impacts of three sequential hurricanes (Dennis, Floyd, and Irene) on the United States’ largest lagoonal estuary, Pamlico Sound, NC. Proceedings of the National Academy of Sciences of the United States 98:5655-5660. Palmer, M.R., H.M. Nepf, T.J.R. Pettersson, and J.D. Ackerman 2004. Observations of particle capture on a cylindrical collector: Implications for particle accumulation and removal in aquatic systems. Limnology and Oceanography 49:76-85. Park, K., J.F. Valentine, K.R. Weis, and M.R. Dardeau. 2007. Effects of

  164  

Hurricane Ivan on Mobile Bay, Alabama. Journal of Coastal Research 23:1332-1336. Parker, V.T. 1997. The scale of successional models and restoration objectives. Restoration Ecology 5:301-306. Peng, M., L. Xie and L.J. Peitrafesa 2003. A numerical study of storm surge and inundation in the Croatan-Albemarle-Pamlico Estuary System. Estuarine Coastal and Shelf Science 59:121-137. Pennings, S.C., C.M. Clark, E.E. Cleveland, S.C. Collings, L. Gough, L. Gross, D.G. Milchunas and K.N. Suding 2005. Do individul plant species show predictable responses to nitrogen addition across multiple experiments? Oikos 111:547-555. Pennings, S.C. and M.D. Bertness 2001. Salt Marsh Communities. Marine Community Ecology (ed. M.E. Hay), Sinaur Associates, Sunderland. MA Pp. 289-316. Peterson, G.D. 2002. Forest Dynamics in the Southeastern United States: Managing Multiple Stable States Pages 227-246 in "Resilience of large scale resource systems" Pages 3-20 in Resilience and behaviour of largescale systems. L.H. Gunderson and L. Pritchard Jr., eds. Island Press, Washington, DC. Peterson, C.H. and N.M. Peterson. 1979. The ecology of intertidal flats of North Carolina: a community profile. U.S. Fish and Wildlife Service Biological Services Program FWS/OBS-79/39. 73 pp. Pfadenhauer J. 2001. Some remarks on the socio-cultural background of restoration ecology. Restoration Ecology 9:220-229. Phinn, S.R., D.A. Stow and D.V. Mouwerik 1999. Remotely sensed estimates of vegetation structural characteristics in restored wetlands, Southern California. Photogrammetric Engineering and Remote Sensing 65:485493. Phinn, S.R., D.A. Stow and J.B. Zedler 1996. Monitoring wetland habitat restoration in southern California using airborne multispectral video data. Restoration Ecology 4:412-422. Pickett, S.T.A., and M.J. McDonnell 1989. Changing perspectives in community dynamics: A theory of successional forces. Trends in Ecology

  165  

and Evolution 4:241-245. Pielou E.C. 1966. The measurement of diversity in different types of biological collections. J Theor Biol.13:131–144. Pielou, E.C. and R. Routledge 1976. Salt marsh vegetation: latitudinal gradients in the zonation patterns Oecologia 24:311-321. Piou, C., Feller, I.C., Berger, U., Chi, F. 2006. Zonation patterns of Belizean offshore mangrove forests 41 years after a catastrophic hurricane. Biotropica 38, 365–372. Pomeroy, L.R., L. Shenton, R. Jones and R. Reimold 1972. Nutrient flux in estuaries. Limnology and Oceanography 1:274-291. Poirrier, M.A., Z. Rodriguez del Ray and E. Spalding 2008. Abiotic disturbance of Lake Ponchartrain Benthic Communities by Hurricane Katrina. Estuaries and Coasts 31(6):1221-1228. Portier, K.M., G. Fabi and P.H. Darius 2000. Study Design and Data Analysis Issues, p. 21-50, In /w. Seaman Jr. (ed.), Artificial Reef Evaluation – With Application to Natural Marine Habitats. CRC Press, Inc. Boca Raton, Florida. Ramsey, E.W., III, G.A. Nelson and S.K. Sapkota 2001. Coastal change analysis program implemented in Louisiana. Journal of Coastal Research 17(1):53-71. Ramsey, E.W., III, D.K. Chappell, D.M. Jacobs, S.K. Sapkota and D.G. Baldwin 1998. Resource management of forested wetlands: hurricane impact and recovery mapped by combining landsat TM and NOAA AVHRR data. Photogrammetric Engineering and Remote Sensing 64(7):733-738. Ramsey, E.W., III, D.K. Chappell and D. Baldwin 1997. AVHRR imagery used to identify Hurricane Andrew damage in a forested wetland of Louisiana. Photogrammetric Engineering and Remote Sensing 63(3):293297. Ramsey, E.W., III and S.C. Laine 1997. Comparison of landsat thematic mapper and high resolution photography to identify change in complex coastal wetlands. Journal of Coastal Research 13(2):281-292.

  166  

Read T.S. and N.S. Lam 2002. Spatial methods for characterizing land cover and detecting land-cover changes for the tropics. International Journal of Remote Sensing. v23 i12. 2457-2474. Reed, D. J. and L.P. Rozas 1995. An evaluation of the potential for infilling existing pipeline canals in Louisiana coastal marshes. Wetlands 15:149158. Richards, J. A. 1999. Remote Sensing Digital Image Analysis: An Introduction. Springer-Verlag, New York, p. 240. Rogers, J.C., A.W. Murrah and W.H. Cooke 2009. The impact of Hurricane Katrina on the coastal vegetation of the Weeks Bay Reserve Alabama from NDVI data. Estuaries and Coasts 32:496-507 Roman, C.T., K.B. Raposa, S.C. Adamowiz, M.J. James-Pirri and J.G. Catena 2002. Quantifying vegetation and nekton response to tidal restoration of a New England salt marsh. Restoration Ecology 10:450-460. Roman, C.T., M.J. James-Pirri and J.F. Heltshe 2001. Monitoring salt marsh vegetation. Technical Report, Long-term coastal Ecosystem Monitoring Program, Cape Cod National Seashore, Wellfleet, Massachusetts. Roman, C.T., J.A. Peck, J.R. Allen, J. W. King and P.G. Appleby 1997. Accretion of a New England (U.S.A.) salt marsh in response to inlet migration, storms, and sea-level rise. Estuarine, Coastal and Shelf Science 45:717-727. Roman, C.T., R.W. Garvine and J.W. Portnoy 1995. Hydrologic modeling as a predictive basis for ecological restoration of salt marshes. Environmental Management 19:559-566. Roman, C.T. and F.C. Daiber 1989. Organic carbon flux through a Delaware Bay Saltmarsh: tidal exchange, particle size distribution, and storms. Marine Ecology Progress Series. 54:149-156. Roman, C.T., W.A. Niering and R.S. Warren 1984. Salt marsh cegetation change in response to tidal restriction Environmental Management 19:559-566. Rozema, J.P., P. Bijwaard, G. Prast and R. Broekman 1985. Ecophysical adaptations of coastal halophytes form foredunes and salt marshes. Vegetation 62:499-521.

  167  

Saffir, H.S., J.F. Halsey and R.S. Cummins 1972. Wind and surge damage due to Hurricane Camille. ASCE Journal of Waterways, Harbors and Coastal Engineering. Sage, T. and A. Gallaway 2002. The state of the bay: a characterization of the Galveston Bay Ecosystem. Edited by L.J. Lester and L. Gonzalez. GBEP # T-7. Galveston Bay Estuary Program, Webster, TX. Salgado, C.S. and S.C. Pennings 2005. Latitudinal variation in palatability of salt-marsh plants: Are differences of salt-marsh plants: Are differences constitutive? Ecology 86(6):1571-1579. Sault, M., C. Parrish, S. White, J. Sellars and J. Woolard 2005. A sensor fusion approach to coastal mapping. Proceedings of the 14th Biennial Coastal Zone Conference, New Orleans, Louisiana. Sawaya K.E., L.G. Olmanson, N.J. Heinert, P.L. Brezonik, and M.E. Bauer 2003. Extending satellite remote sensing to local scales: land and water resource monitoring using high-resolution imagery Remote Sensing of Environment 88:144-156. Scheffer, M. 2008. Critical Transitions in Nature and Society. Princeton, NJ: Princeton University Press. Scheffer, M. and S.R. Carpenter 2003. Catastrophic regime shifts in ecosystems: Linking theory to observation. Trends in Ecology and Evolution 18:648-56. Scheffer, M. and J. Beets 1994. Ecological models and the pitfalls of causality. Hydrobiologia 276:115-24. Schmidt, K.S. and A.K. Skidmore 2003. Spectal discrimination of vegetation types in a coastal wetland. Remote Sensing of Environment 85:92-108. Schröder, A., L. Persson and A.M. De Roos 2005. Direct experimental evidence for alternative stable states: A review. Oikos 110:3-19. Scott, G., N. Wijekoon, S. White 2009. Multi-sensor mapping: Integrating data streams for coastal science and management. Hydro International 13(2). Seabloom, E.W., W.S. Harpole, O.J. Reichman, and D. Tilman 2003. Invasion, competitive dominance, and resource use by exotic and native

  168  

California grassland species. Proceedings of the National Academy of Sciences 100:13384-13389. Seliskar, D.M. 1980. Morphological and anatomical responses of selected coastal salt marsh plants to soil moisture Thesis (M.S.) Oregon State University, p. 186. Seliskar, D.M. and J.L. Gallagher 1983. The ecology of tidal marshes of the Pacific northwestern coast: A community profile, 65 ppp. FWS/OBS-82/32, U.S. Fish and Wildlife Service, Division of Biological Services, Washington, D.C. Seliskar, D.M., J.L. Gallagher, D.M. Burdick and L.M. Mutz 2002. The regulation of ecosystem functions by ecotypic variation in the dominant plant: A Spartina alterniflora salt marsh case study. Journal of Ecology 90:1-11. Shanmugam, S., N. Lucas, P. Phipps, A. Richards and M. Barnsley 2003. Assessment of remote sensing techniques for habitat mapping in coastal dune ecosystems. Journal of Coastal Research 19(1):64-75. Shannon, C.E., and W. Weiner 1963. The Mathematical Theory of Communication. University of Illinois Press, Urbana, Illinois, USA. Shaoqing Z., and X. Lu 2008. The comparative study of three methods of remotes sensing image change detection. International Society for Photogrammetry and Remote sensing (ISPRS) Congress Beijing Proceedings of Commission XXXVII(B7). Shinkle, K.D. and R.K. Dokka 2004. Rates of vertical displacement at benchmarks in the lower Mississippi Valley and the Northern Gulf Coast. U.S. Department of Commerce, NOAA Technical Report NOS/NGS 50. Short, F.T., D.M. Burdic, C.A. Short, R.C. Davis and P.A. Morgan 2000. Developing success criteria for restored eelgrass, salt marsh, and mudflat habitats. Ecological Engineering 15:239-252. Shumway S.W. and M.D. Bertness 1992. Salt stress limitation of seedling recruitment in a salt marsh plant community. Oecologia 92: 490-497. Silvestri S. and M. Marani 2004. Salt-Marsh Vegetation and Morphology: Basic Physiology, Modeling and Remote Sensing Observations. Published in Ecogeomorpholgy of Tidal Marshes Eds.: S. Fagherazzi, L. Blum, M.

  169  

Marani, American Geophysical Union, Coastal and Estuarine Monograph Series. Simenstad C., D. Reed and M. Ford 2006. When is restoration not? Incorporating landscape-scale processes to restore self-sustaining ecosystems in coastal wetland restoration. Ecological Engineering 26:2739. Simpson, E H. 1949. Measurement of diversity. Nature 163:688. Singh, A. 1989. Digital change detection techniques using remotely sensed data. International Journal of Remote Sensing 10:989–1003. Small T., 2004. CDMO NERR SWMP Data Management Manual Version 5.1. Georgetown, South Carolina: NOAA Centralized Data Management Office. 431p. Smith G., T. Spencer, A.L. Murray and J.R. French 1998. Assessing seasonal vegetation change in coastal wetlands with airborne remote sensing: an outline methodology. Mangroves and Salt Marshes 2:15-28. Somers G F. and D. Grant 1981. Influence of seed source upon phenology of flowering Spartina alterniflora Loisel. and the likelihood of cross pollination. American Journal of Botany 68: 6-9. Sparks, P.R. 1991. National Research Council (U.S>) and Committee on Natural Disasters. Hurricane Elena, Gulf Coast, August 29-Sept 2, 1985. Natural Disaster Studies, v.2. Washington D.C.: National Academy Press pp 104. Stauffer D and A. Aharony 1992. Introduction to Percolation Theory, 2nd ed. London: Taylor and Francis. Stefanov, W.L., and M. Netzband 2005. Assessment of ASTER land cover and MODIS NDVI data at multiple scales for ecological characterization of an arid urban center. Remote Sensing of Environment 99 (1-2), pp. 31-43. Stehman, S. V. 1996. Estimating the Kappa coefficient and its variance under stratified random sampling Photogrammetric Engineering and Remote Sensing 62(4):401-407. Steiger, J. and A.M. Gurnell 2003. Spatial hydro-geomorphological influences on riparian zone sedimentation: observations from the Garonne

  170  

River, France. Geomorphology 49:1-23. Steneck, R.S., M.H. Graham, B.J. Bourque, D. Corbett, J.M. Erlandson, J.A. Estes and M.J. Tegner 2002. Kelp forest ecosystems: biodiversity, stability, resilience and future. Environmental Conservation 29(4):436459. Stewart, S.R. 2004. Tropical Cyclone Report – Hurricane Ivan, NOAA Hurricane Center, Updated 27May2005. Available on the internet at http://www.nhc.noaa.gov/pdf/TCR-AL 092004_Ivan.pdf Stone, R. 2006. Ecology: A rescue effort for tsunami-ravaged mangrove forests. Science 314:404. Stout, J.P., M J. Lelong, H.M. Dowling, M.Y. Powers 1982. Wetland habitats of the Alabama coastal Zone, Part III. An inventory of wetland habitats of the Mobile-Tensaw River Delta. Alabama coastal Area Board. Technical Report 81-49A. 25pp. Stout, J.P. 1984. The ecology of irregularly flooded salt marshes of the northeastern Gulf of Mexico: a community profile. U.S. Fish and Wildlife Service Biological Report 85(7.1). 98 pp. Stovall, R.H. (editor). 1982. Proceedings of the 8th Annual Conference on Wetland Restoration and Creation. Hillsborough Community College. Tampa, FL. pp200. Straker, A. and S. Platt 2001. Living systems resource kit-biodiversity in property management. Department of Natural Resources and Environment DNRE, Melbourne. Stumpf, R.P. and J.W. Haines 1998. Variations in tidal level in the Gulf of Mexico and implications for tidal wetlands. Esturarine, Coastal and Shelf Science 46(2):165-173. Suding, K.N., K.L. Gross and G.R. Houseman 2004. Alternative states and positive feedbacks in restoration ecology. Trends in Ecology and Evolution 19:46-53. Suding, K.N. 2001. The effects of gap creation on competitive interactions: Separating changes in overall intensity from relative rankings. Oikos 94:219-27.

  171  

Sullivan, C.L. 1986. Hurricanes of the Mississippi gulf Coast, Biloxi, MS, Gulf Publishing. Sutcliffe, J.F. 1962. Mineral Salts Absorption in Plants. Pergamon Press. Oxford 194p. Swart, J.A.A., H.J. van der Windt, and J. Keulartz 2001. Valuation of nature in conservation and restoration. Restoration Ecology 9:230-238. Teal, J.M. 1962. Energy flow in the salt marsh ecosystem of Georgia. Ecology 43:614-624. Tedesco, L.P., H.R. Wanless, L.A. Scusa, J.A. Risi and S. Gelsanliter 1995. Impact of Hurricane Andrew on south Florida’s sandy coastlines. Journal of Coastal Research. Special Issue 21:59-82. Theodose, T.A. and J.B. Roths 1999. Variation in nutrient availability and plant species diversity across forb and graminoid zones of a Northern New England high marsh. Plant Ecology 143(2): 219-228. Thompson, S.K. 2002. Sampling, 2nd edition, Wiley, New York. Thorn, R.M. 1997. System-development matrix for adaptive management of coastal ecosystem restoration projects. Ecological Engineering 8:219-232. Tiner, R.W. 1993. Field Guide to Coastal Wetland Plants of the Southeastern United States. Amherst, The University of Massachusetts Press. Tischendorf, L. 2001. Can landscape indices predict ecological processes consistently? Landscape Ecology 16:235–254. Tou, J.L. and R.C. Gonzales 1974. Pattern Recognition Principles. AddisonWesley Publishing Company. Readings. MA. Tuxen K.A., L.M. Schile, M. Kelly and S.W. Siegel 2008. Vegetation colonization in a restoring tidal marsh: A remote sensing approach. Restoration Ecology 16(2):313-323. Twilley R.R. and J. Barras 2003. Formulation of the Louisiana Coastal Area Ecosystem Model, in Twilley, R.R. ed., LCA Ecosystem Restoration Plan, appendix E., p. 22-41, http://data.lca.gov/Ivan6/app/app_c_ch2.pdf Tyler, A.C. and J.C. Zieman 1999. Patterns of development in the creekbank

  172  

region of a barrier island Spartina alterniflora marsh. Marine Ecological Progress Series 180:161-177. Underwood, A.J., M.G. Chapman and S.D. Connell 2000. Observations in ecology: you can’t make progress on processes without understanding the patterns. Journal of Experimental Marine Biology and Ecology 250:97115. United State Army Corps of Engineers 1983. Low Coast Shore Protection. DACW 61-81-D-0012. Rogers, Golden & Harper, Philadelphia, PA. pp36. United States Geological Survey (USGS) 2010. U.S. Department of Interior, Seamless Data Internet Management System URL: http://seamless.usgs.gov/faq/general_faq.php United States Geological Survey (USGS) 2002. USGS Mapping Information-NMP Standards, United States National Map Accuracy Standards. http://rockyweb.cr.usgs.gov/nmpstds/nmas647.html United States and Weather Bureau 1986. Hurricanes and tropical storms of the Gulf of Mexico 1875-1956. New Orleans, LA, U.S. Weather Bureau. Valentine, J.F., K.L. Heck, M. Dardeau and H.Burch. In press. Ecosystembased management of Mobile Bay, Alabama. In: The Gulf of Mexico: Its Origins, Waters, Biota & Human Impacts. Volume V: The Gulf of Mexico: Ecosystem-Based Management. J. Day and A. Yanez-Arancibia. Texas A & M University Press. Valiela, I., J.M. Teal, S. Volkmann, D. Shafer and E.J. Carpenter 1978. Nutrient and particulate fluxes in a salt marsh ecosystem: Tidal exchanges and inputs by precipitation and groundwater. Limnology and Oceanography 23:798-812. Valta-Hulkkonen, K., P. Pellikka, H. Tanskanen, A. Ustinov and O. Sandman 2003. Digital false colour aerial photographs for discrimination of aquatic macrophyte species. Aquatic Botany 75: 71-88. Van de Koppel, J., D. van der Wal, J.P. Bakker and P.M.J. Herman 2005. Self-organization and vegetation collapse in salt marsh ecosystems. The American Naturalist 165(1):E1-12. Van der Valk, A.G. 1981. Succession in wetlands: a Gleasonian approach. Ecology 62:688-696.

  173  

Van der Valk, A.G. and C.B. Davis 1978. The role of seed banks in the vegetation dynamics of prairie glacial marshes. Ecology 59:322-335. Van Nes, E.H. and M. Scheffer 2007. Slow recovery from perturbations as a generic indicator of a nearby catastrophic shift. The American Naturalist 169:738-47. Virlag L.A. and J.E. Colwell 1987. An improved procedure for analysis of change in Thermatic Mapper image-pairs. International Symposium of Remote Sensing of Environment. 21st, Ann Arbor, MI, United States 2630 October. 1101-1110 pp. Walker, N.D. 2001. Tropical storm and hurricane wind effects on water level, salinity and sediment transport in the river-influenced AtchafalayaVermilion Bay System, Louisiana, USA. Estuaries 24(4):498-508. Walker, J. and P. Reddell 2007. Retrogressive succession and restoration on old landscapes. In Linking restoration and ecological succession, ed. L.R. Walker, J. Walker and R.J. Hobbs, 69-89. New York: Springer. Walker, L.R., J. Walker and R.J. Hobbs 2007. Linking restoration and ecological succession. New York: Springer. Walker B., and J.A. Meyers. 2004. Thresholds in ecological and socialecological systems: a developing database Ecology and Society 9(2):3 [online] URL: http://www.ecologyandsociety.org/vol9/iss2/art3. Wallington, T.J., R.J. Hobbs, and S.A. Moore 2005. Implications of current ecological thinking for biodiversity conservation: a review of the salient issues. Ecology and Society 10(1): 15. [online] URL: http://www.ecologyandsociety.org/vol10/iss1/art15/ Wang, J., D.M. Seliskar and J.L. Gallagher 2005. Tissue culture and plant regeneration of the salt marsh monocots Juncus roemerianus and Juncus gerardii . In Vitro Cellular and Developmental Biology-Plant 41:274-280. Wang, Y.P., R.S. Zhang and S. Gao 1990. Velocity variations in salt marsh creeks, Jiangsu, China. Journal of Coastal Restoration 15(2):471-477. Weinstein M.P. 2007. Linking restoration ecology and ecological restoration in estuarine landscapes. Estuaries and Coasts 30(2):365-370. Welch, R. 1992. Integration of GPS, remote sensing, and GIS techniques for

  174  

coastal resource management, Photogrammetric Engineering & Remote Sensing 58(11):1571-1578. Westmoreland, S. and D.A. Stow 1992. Category identification of changed land-use polygons in an integrated image processing/geographic information system Photogrammetric Engineering and Remote Sensing 58:1593–1599. Wheeler, D.J. 1993. Commentary: linking environmental models with geographic information systems for global change research. Photogrammetric Engineering and Remote Sensing 58, 1497-1501. Whisenant, S. 1999. Repairing Damaged Wildlands: A process-oriented, Landscape Scale Approach, Cambridge University, New York 328 pp. White, P.S. and S.T.A. Pickett 1985. Natural disturbance and patch dynamics: An introduction. San Diego, CA: Academic Press. Wiegert, R. G., L.R. Pomeroy and W.J. Wiebee 1981. Ecology of salt marshes: An Introduction, in, Pomeroy, L. R., and Wiegert, R. G., eds, The Ecology of Salt Marsh. New York: Springer-Verlag, p. 3-36. Wiegart, R.G., A.G. Chalmers and P.F. Randerson 1983. Productivity gradients in salt marshes: the response of Spartina alterniflora to experimentally manipulated soil water movement. Oikos 41:1-6. Wilcox, D.A. and J.E. Meeker 1995. Wetlands in regulated Great Lakes. In Our living resources: a report to the nation on the distribution, abundance, and health of U.S. plants, animals and ecosystems, (EDs.) E.T. LaRoe, G.S. Farris, C.E. Puckett, P.D. Doran, and M.J. Mac, pp. 247-249. Washington D.C. U.S. DOI, National Biological Service. Wilkie, D.S. and J.t. Finn 1996. Remote sensing imagery for natural resources monitoring: a guide for first-time users. New York: Columbia University Press. 295 p. Wilkinson, G.G. 1996. A review of current issues in the integration of GIS and remote sensing data. International Journal of Geographic Information Systems 10(1): 85-101. Williams, D.J., N. Rybicki, A. Lombana, T. O’Brien and R. Gomez 2003. Preliminary investigation of submerged aquatic vegetation mapping using hyperspectral remote sensing. Environmental Monitoring and Assessment 81:383-392, Kluwer Academic Publishers, Netherlands.

  175  

Williamson, T.N., P.E. Gessler, P.J. Shouse and R.C. Graham 2006. Pedogenesis-terrain links in zero-order watersheds after chaparral to grass vegetation conversion. Soil Science Society of America Journal 70:2065-74. Williams, D.J., N. Rybicki, A. Lombana, T. O’Brien and R. Gomez 2003. Preliminary investigation of submerged aquatic vegetation mapping using hyperspectral remote sensing. Environmental Monitoring and Assessment 81: 383-392, Kluwer Academic Publishers, Netherlands. With K.A. and A.W. King 1997. The use and misuse of neutral landscape models in ecology Oikos 79(2):219–229. Woerner, L.S. and C.T. Hackney 1997. Distribution of Juncus roemerianus in North Carolina tidal marshes: the importance of physical and biotic variables. Wetlands 17:284-291. Woodell, S.R.J. 1985. Salinity and seed germination patterns in coastal plants. Vegetation 61:223-229. Young, T.P., D.A. Petersen and J.J. Clary 2005. The ecology of restoration: Historical links, emerging issues and unexplored realms. 8:662-673. Zar, J.H. 1996. Biostatistal Analysis Printice Hall Sect 13.3 pp. 282-283. Zedler, J.B. 2000. Progress in wetland restoration ecology. Trends in Ecology and Evolution 15:402-07. Zedler, J.B. and J.C. Callaway 1999. Tracking wetland restoration: Do mitigation sites follow desired trajectories? Restoration Ecology 7:69-73. Zedler, J.B., J. Covin, C. Nordby, P. Williams and J. Boland 1986. Catastrophic events reveal the dynamic nature of salt-marsh vegetation in southern California. Estuaries 9:75-80. Zeff, M.L. 1999. Salt marsh tidal channel morphometry: Applications for wetland creation and restoration. Restoration Ecology 7(2): 205-211. Zhan, X., R.A. Sohlberg, J.R.G. Townshend, C. DiMiceli, M.L. Carroll, J.C. Eastman, M. C. Hansen, and R.S. DeFries 2002. Detection of land cover changes using MODIS 250 m data. Remote Sensing of Environment 83:336–350.

  176  

Zilkoski D.B., J.D. D’Onofrio and S. Frakes 1997. Guidelines for establishing GPS-Derived Ellipsoid Heights (Standards: 2cm – 5cm), Version 4.3. NOAA Technical Memorandum NOS NGS-85 National Geodetic Survey Information Center, Silver Springs, MA. Zilkoski, D.B. and L.D. Hothem 1989. GPS satellite survey and verticle control. Journal of Survey Engineering 115(2):262-282.

  177  

APPENDICES

Appendix I: Tables A-1 through A-3 Table A-1. Line transects geolocation for Kreola, Mississippi study site. Coordinates were collected using NAD 83 datum with Wide Area Augmentation System (WAAS) enabled Differential GPS (DGPS). Transect

Lat (Begin)

Lon (Begin)

Lat (End)

Lon (End)

1

30.41190

-88.40457

30.41184

-88.40482

2

30.41180

-88.40484

30.41159

-88.40492

3

30.41112

-88.40524

30.41106

-88.40497

4

30.41157

-88.40545

30.41139

-88.40560

5

30.41173

-88.40592

30.41171

-88.40618

6

30.41162

-88.40507

30.41187

-88.40534

  178  

Table A-2. Vegetation transects biodiversity statistics for Kreole, Mississippi (Grand Bay National Estuarine Research Reserve study site). The chart indicates the biodiversity of each transect (1-6) in number of individuals for each plant species (ni), total number of individuals for all plant species (N), the percent transect coverage for each plant species (ni/N), The Simpson Index for each plant species (ni(ni1)/(N(N-1))), and Shannon-Wiener Index for each plant species (ni)ln(ni/N). Transect

Transect

Transect

Transect

Transect

Transect

Grand Bay

1

2

3

4

5

6

Mean

179

Juncus roemerianus (ni)

50

36

36

25

27

38

35.3333

Spartina alterniflora (ni)

0

0

1

0

0

0

0.1667

Distichlis spicata (ni)

0

0

10

0

23

12

7.5000

Batis maritima (ni)

0

7

0

15

0

0

3.6667

0 1 50

0 2 43

3 4 50

0 2 40

0 2 50

0 2 50

0.5000 2.1667 47.1667

100.00%

83.72%

72.00%

62.50%

54.00%

76.00%

0.7470

Spartina alterniflora (ni/N)

0.00%

0.00%

2.00%

0.00%

0.00%

0.00%

0.0033

Distichlis spicata (ni/N)

0.00%

0.00%

20.00%

0.00%

46.00%

24.00%

0.1500

Batis maritima (ni/N)

0.00%

16.28%

0.00%

37.50%

0.00%

0.00%

0.0896

Borrichia frutescens (ni/N)

0.00%

0.00%

6.00%

0.00%

0.00%

0.00%

0.0100

Juncus roemerianus (ni(ni-1)/(N(N-1))) Spartina alterniflora (ni(ni-1)/(N(N-1))) Distichlis spicata (ni(ni-1)/(N(N-1))) Batis maritima (ni(ni-1)/(N(N-1))) Borrichia frutescens (ni(ni-1)/(N(N-1)))

1.0000 0.0000 0.0000 0.0000 0.0000

0.6977 0.0000 0.0000 0.0233 0.0000

0.5143 0.0000 0.0367 0.0000 0.0024

0.3846 0.0000 0.0000 0.1346 0.0000

0.2865 0.0000 0.2065 0.0000 0.0000

0.5739 0.0000 0.0539 0.0000 0.0000

0.5762 0.0000 0.0495 0.0263 0.0004

Borrichia frutescens (ni) Total Number of Species Total Number of Individuals (N) Juncus roemerianus (ni/N)

Table A-2. continued.

Simpson Index Simpson Index (adjusted)* Juncus roemerianus (ni)ln(ni/N) Spartina alterniflora (ni)ln(ni/N) Distichlis spicata (ni)ln(ni/N) Batis maritima (ni)ln(ni/N) Borrichia frutescens (ni)ln(ni/N) Shannon-Wiener Index Shannon-Wiener Index (adjusted)**

GB 001 1.0000 0.0000

GB 002 0.7209 0.2791

GB 003 0.5535 0.4465

GB 004 0.5192 0.4808

GB 005 0.4931 0.5069

GB 006 0.6278 0.3722

Grand Bay Mean 0.6524 0.3476

1.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

-0.1488 0.0000 0.0000 -1.2955 0.0000 0.4443 0.6409

-0.2365 -0.0782 -0.3219 0.0000 -0.1688 0.8055 0.5810

-0.2938 0.0000 0.0000 -0.3678 0.0000 0.6616 0.9544

-0.3327 0.0000 -0.3572 0.0000 0.0000 0.6899 0.9954

-0.2086 0.0000 -0.3425 0.0000 0.0000 0.5511 0.7950

-0.0367 -0.0130 -0.1703 -0.2772 -0.0281 0.5254 0.6611

180

Table A-3. Juncus roemerianus average stem count (m2) and height (cm) compiled based on line transect (25m) sampling (random quadrat) at Kreole, Mississippi.

Transect T1 T2 T3 T4 T5 T6

Stem Count 1 72 16 112 0 28 96

Stem Count 2 100 36 56 92 40 136

Stem Count 3 52 124 52 88 104 164

Mean Stem Count 75 59 73 60 57 132

STDEV Stem Count 24 57 33 52 4 34

Stem Height 1 55 54 45 53 46 51

Stem Height 2 53 49 51 48 49 60

Stem Height 3 51 54 50 59 51 45

Mean Stem Height 53.00 52.33 48.67 53.33 48.67 52.00

STDEV Stem Height 2.00 2.89 3.21 5.51 2.52 7.55

181

Appendix II: Figures A-1 and A-2

Figure A-1. Multispectral image showing the location of surface water wells and carbon and nitrogen tissue sampling sites at Kreole, Mississippi. Source: USDA Farm Service Agency image date 08August2007.

  182  

Figure A-2. Annual percent mean of carbon and nitrogen ratios for ten, Juncus roemerianus tissue, collection sites at the Kreola, Mississippi study area.

  183  

Appendix III: Table A-4 Table A-4. Percent nitrogen and carbon ratios recorded from ten Juncus roemerianus tissue samples within the study area at Kreola, Mississippi. Elevation data (meters) from the National Elevation Data (NED) considered the best public domain elevation raster data archived by Earth Resources Observation and Sciences (EROS) Center. The elevation data is a deliverable from the USGS Light Detection and Ranging (LIDAR) high resolution topography program and meets the remote sensing National Standards for Spatial Data Accuracy (NSSDA) ISO 19115. Source: http://nationalmap.gov Accessed September 8, 2009. Site Number

Latitude

Longitude

1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

30.41175 30.41191 30.41177 30.41168 30.41147 30.41091 30.41113 30.41176 30.41184 30.41192 30.41175 30.41191 30.41177 30.41168 30.41147 30.41091 30.41113 30.41176 30.41184 30.41192

88.40505 88.40537 88.40570 88.40623 88.40555 88.40514 88.40490 80.40472 88.40463 88.40457 88.40505 88.40537 88.40570 88.40623 88.40555 88.40514 88.40490 80.40472 88.40463 88.40457

Collection Date 2/12/08 2/12/08 2/12/08 2/12/08 2/12/08 2/12/08 2/12/08 2/12/08 2/12/08 2/12/08 4/18/08 4/18/08 4/18/08 4/18/08 4/18/08 4/18/08 4/18/08 4/18/08 4/18/08 4/18/08

  184  

C:N % .400448 .482455 .306023 .425290 .314804 .340519 .374812 .445683 .546114 .395938 .504581 .319712 .440480 .376918 .421732 .348370 .541229 .326198 .347076 .445646

Elevation (m) (1/9 Arc) 0.3840 0.5944 0.4846 0.6096 0.6553 0.3048 0.5486 0.8169 0.8443 0.1554 0.3840 0.5944 0.4846 0.6096 0.6553 0.3048 0.5486 0.8169 0.8443 0.1554

Table A-4. continued. Site Number 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Latitude 30.41175 30.41191 30.41177 30.41168 30.41147 30.41091 30.41113 30.41176 30.41184 30.41192 30.41175 30.41191 30.41177 30.41168 30.41147 30.41091 30.41113 30.41176 30.41184 30.41192

Longitude 88.40505 88.40537 88.40570 88.40623 88.40555 88.40514 88.40490 80.40472 88.40463 88.40457 88.40505 88.40537 88.40570 88.40623 88.40555 88.40514 88.40490 80.40472 88.40463 88.40457

Collection Date 6/27/07 6/27/07 6/27/07 6/27/07 6/27/07 6/27/07 6/27/07 6/27/07 6/27/07 11/1/07 11/1/07 11/1/07 11/1/07 11/1/07 11/1/07 11/1/07 11/1/07 11/1/07 11/1/07

  185  

C:N % .575530 .413712 .550542 .474401 .494096 .478285 .521356 .483002 .387074 .529110 .671325 .663683 .547908 .442981 .568128 .753394 .680517 .571444 .828432

Elevation (m) (1/9 Arc) 0.5944 0.4846 0.6096 0.6553 0.3048 0.5486 0.8169 0.8443 0.1554 0.3840 0.5944 0.4846 0.6096 0.6553 0.3048 0.5486 0.8169 0.8443 0.1554

Appendix IV: Figure A-3

Figure A-3. Ten plant tissue samples were collected at each study area during four seasonal timeframes. The tissue samples were collected from both mature and young leaf tissue and dried at ~ 60o C in a drying oven and then finely ground. Samples of ~1.5 -4.0 mg of ground tissue were placed into a pressed tin capsule (5 X 9 mm) (Costech Analytical Technologies, Valencia, CA). Five standard curve samples were prepared using approximately 2.5 mg of Acetanilide (71.09% C, 10.36% N). The carbon and nitrogen contents of the Juncus roemerianus bracts were ignited in an Elemental Combustion System (ECS) 4010 CHNS-O (Costech AnalyticalTechnologies, Valencia, CA) auto analyzer (Newell et al. 1996). The C:N ratios were measured over one year at four time periods (June 27, 2007, November 1, 2007, February 12, 2008, and April 18, 2008). The data bars represent mean carbon and nitrogen ratios ± standard deviation. An asterisk indicates difference between among months from a one factor ANOVA. Statistical test were performed on arcsine transformed datasets. Levine’s homogeneity of variance test on the transformed data resulted in F=2.376 and p(0.05)=0.086. All statistical test utilized p