remote sensing studies for the assessment of geohazards
TRANSCRIPT
Western Michigan University Western Michigan University
ScholarWorks at WMU ScholarWorks at WMU
Dissertations Graduate College
8-2008
Remote Sensing Studies for the Assessment of Geohazards: Remote Sensing Studies for the Assessment of Geohazards:
Toxic Algal Blooms in the Lower Great Lakes, and the Land Toxic Algal Blooms in the Lower Great Lakes, and the Land
Subsidence in the Nile Delta Subsidence in the Nile Delta
Richard H. Becker Western Michigan University
Follow this and additional works at: https://scholarworks.wmich.edu/dissertations
Part of the Geology Commons
Recommended Citation Recommended Citation Becker, Richard H., "Remote Sensing Studies for the Assessment of Geohazards: Toxic Algal Blooms in the Lower Great Lakes, and the Land Subsidence in the Nile Delta" (2008). Dissertations. 3382. https://scholarworks.wmich.edu/dissertations/3382
This Dissertation-Open Access is brought to you for free and open access by the Graduate College at ScholarWorks at WMU. It has been accepted for inclusion in Dissertations by an authorized administrator of ScholarWorks at WMU. For more information, please contact [email protected].
REMOTE SENSING STUDIES FOR THE ASSESSMENT OF GEOHAZARDS: TOXIC ALGAL BLOOMS IN THE LOWER
GREAT LAKES, AND THE LAND SUBSIDENCE IN THE NILE DELTA
by
Richard H. Becker
A Dissertation Submitted to the
Faculty of The Graduate College in partial fulfillment of the
requirements for the Degree of Doctor of Philosophy
Department of Geosciences Dr. Mohamed Sultan, Advisor
Western Michigan University Kalamazoo, Michigan
August 2008
Copyright by
Richard H. Becker
2008
ACKNOWLEDGMENTS
I would like to thank several people for helping and encouraging me
through the course of my doctoral work. I especially thank my advisor Dr.
Mohamed Sultan for his encouragement in pursuing this degree and for
mentoring me throughout the research process. Our discussions have been
very fruitful, and have helped me become a better researcher.
I would like to thank my dissertation committee, Dr. Joseph Atkinson,
Dr. Carla Koretsky, Dr. Barry Lesht, and Dr. William Sauck for their time
and effort in reviewing my proposal and dissertation.
Additional thanks go to Dr. Gregory Boyer, and Elizabeth Konopko at
SUNY ESF, as well as Dr. Michael Twiss at Clarkson University for access to
the data necessary to validate my satellite models for the Great Lakes.
I would also like to thank all of the researchers who have worked in
the ESRS lab, who provided input on these projects.
Finally, thanks go to my wife Dee, who has been very patient and
supporting as I have been pursuing this degree, and to our children, Anna,
James, and William, who have been great sources of joy and energy when it
was needed.
Richard H. Becker
11
REMOTE SENSING STUDIES FOR THE ASSESSMENT OF
GEOHAZARDS: TOXIC ALGAL BLOOMS IN THE LOWER
GREAT LAKES, AND LAND SUBSIDENCE
IN THE NILE DELTA
Richard H. Becker, Ph.D.
Western Michigan University, 2008
Remote sensing techniques provide valuable tools for assessing a
wide variety of environmental phenomena. They have been used for
monitoring and assessment of various types of geologic and environmental
hazards occurring on land, in the air, or in oceans. I present results from
two studies, the first of which examines the spatial and temporal
distribution of algal blooms in the Great Lakes; the second measures
subsidence in the Nile Delta.
In the first study, methodologies to investigate the extent and
distribution (temporally and spatially) of algal blooms in Lake Erie and
Lake Ontario are studied. Millions of people in the U.S and Canada rely
on the Great Lakes for drinking water, food, work, and recreation. Toxic
algal blooms present a hazard to the substantial number of communities
that draw water from the Great Lakes. Visible and infra red MODIS
satellite data are used to map the extent of algal blooms in these lakes.
Existing algorithms to retrieve chlorophyll concentrations are successfully
tested against in situ measurements from sampling cruises. Algorithms
are developed to identify the potentially toxic cyanobacterial blooms.
The second study examines subsidence in the Nile Delta. The
modern Nile Delta is the major agricultural production area for Egypt and
was formed from sediments supplied by at least 10 distinct distributary
channels throughout the Holocene. With an average elevation around a
meter above sea level and with a predicted rise in sea level of 1.8-5.9
mm/year the subsidence of the northern 30 km of the delta is a topic of
major concern to the Egyptian population and government. Ongoing
subsidence rates in the northeastern Nile Delta were estimated using
persistent scatterer radar interferometry techniques. The highest rates
( ~8 mm/yr; twice average Holocene rates) correlate with the distribution
of the youngest deposition, with older depositional centers subsiding at
slower rates of 2-6 mm/yr. Results are interpreted to indicate that: (1)
modern subsidence in the Delta is heavily influenced by the compaction of
the most recent sediments, and (2) the highly threatened areas are at the
terminus of the Damietta, where the most recent deposition has occurred.
TABLE OF CONTENTS
·ACKNOWLED GMENTS .......................................................................... 11
LIST OF TABLES ····················································································· Vl
LIST OF FIGURES . . . . . . . . .. . . . . . . . . .. . . . . . . . .. . . . . .. . . . . . . . . . . . . . . . . .. . . . .. . . . . . . . . . . . . . . . ... . . . . . . . Vll
CHAPTER
1. INTRODUCTION ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......... ...... 1
1.1 Remot e Sensing of Geologic Haza rds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Diss ert a tion Ov erview....................................................... 6
2. AQUATIC REMOTE SENSING .............................. ....................... 8
2.1 Ov erview............................................................................ 8
2.2 Light a t the Sens or............................................................ 10
2.3 At mos pheric Mod eling....................................................... 12
2.4 Op tica l Components of Wa t er........................................... 14
3. SPATIAL AND TEMPORAL VARIATIONS OF ALGAL
BLOOMS IN THE LOWER GREAT LAKES ............................... .. 16
3.1 Abs t ra ct .............................................................................. 16
3.2 Int roduction .................................................... ................. .. 18
3.3 Ima ge Acquisition a nd Processing.................................... 20
3.4 Field Measurements .......................................................... 21
3.5 Sea WiF S vs. MOD IS Measurements of Chlorophyll-a..... 24
lll
CHAPTER
Table of Contents - Continued
3.6 Satellite vs. In Situ Measurements of Chlorophyll-a....... 27
3. 7 Identification and Mapping of Algal Blooms .. .............. ... 34
3.8 Locating Phycocyanin Containing Blooms ....................... 36
4. MAPPING CYANOBACTERIAL BLOOMS IN THE GREATLAKES USING MODIS................................................................... 39
4.1 Abstract.............................................................................. 39
4.2 Problem Description .......................................................... 40
4.3 Methodology....................................................................... 45
4.4 Field Measurements .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .... .. .. .. .. .. .... .. .. .. 4 7
4.5 Algorithm Development and Processing ofMODIS Data.................................................................................... 49
4.6 Results................................................................................ 52
5. RADAR INTERFEROMETRY........................................................ 62
5.1 Overview............................................................................ 62
5.2 Two-Pass Method............................................................... 67
5.3 Three-Pass Method............................................................ 67
5.4 Four-Pass Method.............................................................. 68
5.5 Multi Temporal Methods................................................... 69
5.6 Summary............................................................................ 69
IV
Table of Contents - Continued
CHAPTER
6. LAND SUBSIDENCE IN THE NILE DELTA: INFERENCES
FROM RADAR INTERFEROMETRY............................................ 72
6.1 Abstract.............................................................................. 72
6.2 Introduction . . . . .. . . . . . . . .. . . . ... . . . . . . .. . . . . . . . ... . . . . .. . . . . . . .. . . . . . .. . . . . . . . . . . . 73
6.3 Geology of the Nile Delta................................................... 76
6.4 Radar Interferometry . . . . . .. . . . . . . .. . . . . . .. ....... .. . . . . . . . .. . . . ... . . . . .. . . . . 79
6.5 Discussion and Results...................................................... 82
7. CONCLUSIONS.............................................................................. 88
REFERENCES .. ................................................. ........................................ 91
APPENDIX
SELECTED INTERMEDIATE INSAR PRODUCTS .......................... 112
V
LIST OF TABLES
4.1 Cruise dates and locations on which fluorometer horizontal
profiles were acquired....................................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 7
6.1 ERS-1 and ERS-2 scenes used in permanent scatterers study.
Temporal and perpendicular baselines are shown relative to
the scene acquired on 8/24/1995........................................................ 82
Vl
LIST OF FIGURES
2.1 Case I and case II waters................................................................... 9
2.2 Schematic showing components contributing to radiance measured at the sensor...................................................................... 11
3.1 Density scatter plot comparing chlorophyll-a concentration derived from MO DIS and Sea WiFS data.......................................... 25
3.2 MODIS Band 16 reflectance images showing relative atmospheric
contributions .. . . . . . . . . . . . . . .. . . . . . . . ... . . . . .. . . . .. . . . ... . . . .. . . . ... . . . ... . . . ... . . . . . . . .. . . . . . . .. . . 26
3.3 Satellite-derived chlorophyll-a concentrations plotted against in situ-:r:neasured chlorophyll-a concentrations..................................... 29
3.4 Comparison between satellite-derived chlorophyll-a concentration using case I (OC4v4, blue diamonds) and case II (Carder, pink squares) algorithms and in situ measured chlorophyll-a
concentrations . . . .. .. . . . . .. . . . . . . . . . . .. .. . . . . . .. . ... . . . .. . . . . .. . . ...... ... . . . . . . . . . . . . ......... .. . 30
3.5 SeaWiFS case I (OC4v4, blue diamonds) and case II (Carder, pink squares) derived concentrations vs. in situ results overlain over Sea WiFS validation results (gray squares) from http://seabass.gsfc.nasa.gov/matchup_results.html ........................ 32
3.6 In situ chlorophyll-a concentration (Parsons, blue diamonds, Welchmeyer, pink squares) versus depth ......................................... 33
3. 7 Progression of an algal bloom in the Oswego Harbor using a timeseries generated from Sea WiFS and MOD IS images acquired in August of 2004 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.8 Chlorophyll-a concentrations extracted from Sea WiFS images acquired in September of 2004.......................................................... 35
Vll
List of Figures - Continued
3.9 Chlorophyll-a concentrations extracted from a processed SeaWiFS image acquired on June 23, 2003 (left), and June 23, 2004 (right) . 36
3.10 Relative abundance of phycocyanin from in situ fluorescence measurements (July 13-16, 2004) overlain on top of relative
abundance of phycocyanin derived from Sea WiFS imagery (July 13 and 15, 2004) ................................. :...................................... 37
4.1 Landsat Thematic Mapper false color composite showing the location of Lake Erie, Lake Ontario and surrounding metropolitan areas............................................................................. 41
4.2 Absorption curves for green and blue green algae ........................... 43
4.3 Relative absorption curves for CDOM, water, green and blue green algae . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.4 Cyanobacterial abundance from FluoroProbe profile (open blue circles) plotted vs. time since cruise start......................................... 54
4.5 Cyanobacterial abundance from FluoroProbe from September
2005 IFYLE cruise vs satellite-derived relative cyanobacterial concentrations from the MODIS scene of September 12, 2005 ..... 55
4.6 Abundances of cyanobacterial blooms in Lake Erie during Sept 2005 derived from MO DIS data ............... ........................................ 56
4. 7 Relative abundances of cyanobacterial blooms in Lake Erie
during August 2005 derived from MODIS data .............................. 57
4.8 Relative abundances of cyanobacterial blooms in Lake Erie (Aug 2004) derived from MODIS data (left) and contoured map generated from cell counts at 15 stations in the western basin of Lake Erie from Dyble (2008) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.1 Geometry of repeat pass interferometry (after Sandwell and
Price, 1998)......................................................................................... 64
5.2 Sample interferogram generated from a pair of radar images acquired on 9-2-92 and 10-31-92 ....................................................... 66
Vlll
List of Figures - Continued
6.1 False color Landsat TM mosaic showing the Nile Delta, the Damietta and Rosetta branches, and urbanized centers (cities and villages) that appear in shades of purple ....................... 7 4
6.2(a) Mean velocities extracted from 14 ascending ERS-1 and ERS-2
scenes acquired over the period 1992-1999 and expressed as
vertical motion of permanent scatterers (negative velocities indicate downward motion) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6.2(b) Average Holocene subsidence (Stanley and Warne, 1993a), and
Holocene thicknesses from individual borings (colored squares)
(Stanley et al., 1996) .. . . . . .. . . . . . . . ... . . . . . . .. . . . . . ... . . . . ... . . . . . ... . . . . . . .. . . . . .. . . . . .. . . . 83
IX
CHAPTER 1
INTRODUCTION
1.1 Remote Sensing of Geologic Hazards
Remote sensing techniques provide valuable tools for assessing a wide
variety of environmental phenomena. They have been used for monitoring
and assessment of various types of geologic and environmental hazards.
Examples of such applications on land include the use of interferometric
synthetic aperture radar (INSAR) studies to investigate motion associated
with earthquakes (e.g., Sandwell et al., 2002), subsidence related to mining
(e.g., Carnec and Delacourt, 2000) and to groundwater extraction (e.g.,
Hoffmann et al., 2001; Watson et al., 2002), and surface rise due to aquifer
recharge (Lu and Danskin, 2001). Landslides and slope stability have been
investigated using INSAR and visible techniques and data sets (e.g.,
Berardino et al., 2003; Hilley et al., 2004; Mason and Rosenbaum, 2002; Ostir
et al., 2003). Visible satellite data and satellite derived digital elevation
models have been used for assessment of mudflows from volcanic eruptions
(Ker le et al., 2003). Morphologic analyses of these data sets were used to
assess potential hazards form volcanic eruptions (Aydar et al., 2003).
Remote sensing applications have been quite useful in monitoring
1
geohazards in the atmosphere as well. The most obvious would be weather
predictions and related hazards (Lynch, 2008) and monitoring spatial and
temporal variations of the ozone layer in the middle stratosphere (Stolarski
et al., 2006). Recent advances in remote extraction of precipitation data are
now enabling early warning applications in flood prediction (Pultz and
Scofield, 2002) and landslide risk assessment (Yang et al., 2006).
Remote sensing datasets and techniques have been used to address
geohazards in oceans and inland lakes. Remote sensing data sets have been
used to monitor the ongoing changes in the ocean circulation, including El
Nino events (Fu and Smith, 1996). They have also been used to identify the
overall abundances of phytoplankton blooms in the oceans and to estimate
primary productivity (Campbell et al., 2002). Additional applications include
examining the role phytoplankton plays in carbon cycling and its potential
impacts on the global climate (Schneider et al., 2008).
Remote sensing techniques have also been used to examine geohazards
at the interface of the land, atmosphere, and ocean systems. For example,
remote sensing data sets were used to examine the impacts of land use and
land cover change on water quality in the adjoining lakes and ocean systems
(Ucuncuoglu et al., 2006). They have also been used to assess the impacts of
tsunami on coastlines and vegetation. Examples include studies
investigating the impacts of the 2004 Indian Ocean tsunami (Belward et al.,
2007; Curran et al., 2007; Yang et al., 2007).
2
International institutions have recognized the importance of the potential
of remote sensing techniques in assessing geohazards. UNESCO developed a
Geohazards theme as a part of their Integrated Global Observation Strategy
(IGOS). One of the key components of this theme is the strengthening of their
existing Geosciences Application in Remote Sensing (GARS) Programme
(IGOS, 2003).
A significant advantage of using remote sensing data techniques is
that they can provide simultaneous measurements of observational
parameters over large areas, and can make repeat measurements on regular
schedules. With the deployment of more satellites and airborne instruments
over the past two decades, the availability of these data sets is becoming far
more widespread. Missions such as the National Aeronautic and Space
Administration's (NASA) Earth Observation System (EOS) provide global
coverage using multiple sensors on 30 different operational or planned
platforms (e.g., visible, infra red optical sensors, passive microwave sensors,
gravity, radar scatterometer) (NASA, 2006). Platforms such as NASA's
ASTER and the commercial SPOT satellite series provide higher resolution
(3-15 m) multispectral images and panchromatic stereo images. The
European Space Agency and the Canadian MDA corporation have several
operational radar satellites (ERS-2, ENVISAT, RADARSAT-2), as well as
archived data from previous missions starting in the 1990s (ERS-1,
RADARSAT-1). These and other platforms provide an extensive library of
ongoing acquisitions as well as archival data which could be used to
investigate a wide range of geohazards on land, in ocean water and in the
atmosphere.
3
I have chosen to examine how remote sensing techniques can be used
to examine two specific geohazards which have the potential to affect millions
of people. The first geohazard examines the spatial and temporal distribution
of algal blooms in the Great Lakes and the second examines subsidence in the
Nile Delta. Both exercises are applications that heavily rely on the use of
remote sensing datasets and techniques. The studies provide ways to better
understand and monitor ongoing processes (algal bloom development and
subsidence in the Delta) and have the potential for providing predictive tools
to be used by community leaders and policy makers to address the
investigated hazards.
In the first study, methodologies to temporally and spatially
investigate the extent and distribution of algal blooms in Lake Erie and Lake
Ontario were evaluated. Millions of people in the U.S. and Canada rely on the
Great Lakes for drinking water, food, work, and recreation. Toxic algal
blooms present a hazard to the substantial number of communities which
draw water from the Great Lakes, or use them for recreation (Dyble et al.,
2008). Visible and near infra red MOD IS and Sea WiFS satellite data were
used to map the distribution of algal blooms in these lakes. Existing
algorithms were used to extract chlorophyll concentrations from MODIS
scenes (Carder et al., 1999; O'Reilly et al., 1998) and results were compared
to in situ measurements collected during sampling cruises conducted during
the periods of acquisition of satellite measurements. The successful
algorithms were defined as that best reproduced the in situ measurements.
Algorithms were developed to identify the potentially toxic cyanobacterial
blooms. Again, cyanobacterial concentrations that were derived using my
4
algorithms were successfully tested against in situ measurements.
The second study examines subsidence in the Nile Delta. The modern
Nile Delta is the major agricultural production area for Egypt. It is formed
from sediments supplied throughout the Holocene by at least 10 distinct
distributary channels (Stanley and Warne, 1993b). With an average elevation
of a meter or so above sea level and a predicted rise in sea level of 1.8-5.9
mm/year (IPCC, 2007) the subsidence of the northern 30 km of the delta is a
topic of major concern to the Egyptian population and government.
Ongoing subsidence rates in the northeastern Nile Delta were
measured using persistent scatterer radar interferometry techniques (Hooper
et al., 2007; Hooper et al., 2004). Fourteen ERS-1 and ERS-2 satellite radar
images covering the period of 1992-1999 were used. It was found that the
highest rates (~8 mm/yr; twice average Holocene rates) correlate with the
distribution of the youngest deposits, whereas the older depositional centers
were found to subside at slower rates of 2-6 mm/yr. Results were interpreted
to indicate that: (1) modern subsidence in the Delta is heavily influenced by
the compaction of the most recent sediments, and (2) the highly threatened
areas are at the terminus of the Damietta branch, where the most recent
deposition has occurred.
These two problems are both good candidates for the use of remote
sensing datasets and techniques. Both cover large areas, hundreds of square
kilometers with observational parameters acquired over long periods (years).
Predictions made from the remote sensing data over the study areas have the
potential for providing guidelines to the policy makers when it comes to
measures to address the geohazards of concern. For example, in the case of
the algal bloom study, the success in monitoring the propagation of blooms
spatially and temporally could potentially affect decisions on the operations
at water intake facilities, recreational usage of inland lakes and even
government regulation regarding development plans. Similarly, results on
the Nile subsidence could potentially be used to orchestrate plans for
protecting the threatened areas in the Delta and in developing educated
plans for urban development of these areas.
1.2 Dissertation Overview
Chapters two through four of the dissertation cover topics related to
aquatic remote sensing, and summarize the results of the investigations that
were conducted in Lake Erie and Lake Ontario. Chapter two of this
dissertation provides a general overview of aquatic remote sensing principles,
techniques, and applications. Chapter three is modified from two articles
which have already been published and chapter four has been submitted for
consideration for publication.
The content of chapter three is modified from two studies. The first is a
study that was published in the proceedings of the 8th International
Conference on Remote Sensing for Marine and Coastal Environments (Becker
et al., 2005), and the second a study that was published in the Great Lakes
Research Reviews (Becker et al., 2006). The study utilizes existing
chlorophyll extraction algorithms to extract chlorophyll concentrations in the
optically complex waters of Lake Erie and Ontario. Comparisons between
algorithms were conducted; the ability to integrate data series from MODIS
and Sea WiFS was examined through comparisons of observations extracted
from contemporaneous MOD IS and Sea WiFS acquisitions. The study
evaluates the effects of varying atmospheric conditions and new
methodologies for extracting phycocyanin concentrations from reflectance
data were developed and applied to Lakes Erie and Ontario to assess the
distribution of blue-green algae blooms.
Chapter four addresses the same problem of extracting blue-green
algae concentrations from satellite imagery, but uses absorption spectra
instead, and has been submitted for consideration to the Journal of Great
Lakes Research.
Chapter five provides a general overview of radar interferometry
techniques. Chapter six covers the application of persistent scatterer radar
interferometry techniques for measuring subsidence in the Nile Delta. The
contents of chapter seven have been submitted to the journal Holocene for
consideration for publication.
Chapter seven provides a summary of the work, and suggestion for
future work in these two areas.
7
CHAPTER2
AQUATIC REMOTE SENSING
2.1 Overview
Monitoring water quality in large water bodies (e.g., lake, sea, and
ocean) spatially and temporally using traditional in situ based methods of
sampling at discreet stations is costly due to the high costs for ship time and
for analysis of collected samples. Moreover, in situ methods do not provide
detailed analyses on regional scales given the large extent and the spatial
and temporal variability of such dynamic systems. Remote sensing
techniques have long held great promise for making synoptic measurements
of the near surface properties for such extensive systems. By virtue of their
capabilities for repeatedly conducting simultaneous measurements with the
same observational parameters, remote sensing sensors could potentially
provide consistent measurements of selected chemical and physical
properties for surface water systems on regional scales. Examples of remotely
sensed properties that could be readily measured include temperatures,
concentrations for chlorophyll, suspended sediments, and total dissolved
solids (Kirk, 1994).
8
For the majority of the published literature on this topic (remote
sensing-based water composition), a distinction is being made between case I
and case II waters (Morel and Prieur, 1977) as shown in figure 2.1.
Case I
Case II
2 Suspended Mate� Figure 2.1 Case I and case II waters. Case I waters are dominated by
phytoplankton. Case II waters have a significant portion of colored dissolved
organic matter (CDOM) and suspended material.
Case I waters, the oceanic waters, are those in which phytoplankton is
the dominant optically active component (Fig. 2.1). For case I waters, the
9
optical properties are dominated by the pigment concentration. Most of the
existing algorithms for extracting water quality parameters were originally
derived for these optically simpler case I waters (Muller-Karger et al., 2005).
For case II or coastal and inland waters, the concentrations of these optically
active components do not co-vary. The optically active components of
phytoplankton, suspended organic and inorganic particulates (8PM), and
colored dissolved organic matter (CDOM or gelbstofl) vary independently.
The rivers discharging in lakes and coastal systems play an important role
delivering these components in the form of sediments and nutrients.
2.2 Light at the Sensor
In using space-borne observation, the property that is being measured
is the amount of light received by the sensor in discrete wavelength regions
(bands). The intensity oflight measured at the satellite is a function of the
wavelength dependant solar irradiance, and interactions with the
atmosphere, the surface of the water, and with the components dissolved and
suspended in the water (Fig. 2.2).
10
SUN
Figure 2.2 Schematic showing components contributing to radiance measured
at the sensor.
The upwelling radiance received at the sensor at the top of the
atmosphere (TOA), L 7 (-l), can be broken down into multiple components
(Hooker and McClain, 2000):
The first two components are atmospheric Rayleigh scattering from air
and scattering from aerosols, including aerosol-Rayleigh interactions. The
terms L ..... c and L9are radiances due to whitecaps and sun glint respectively.
Lw
is the upwelling radiance directly above the water surface and t(J) is the
diffuse transmittance of the atmospheric column.
11
To remove the variations in radiances that are related to varying
illumination angles and viewing geometries, L..,..
is converted to normalized
water-leaving radiance (nLw
) (Gordon and Clark, 1981), which is the water
leaving radiance corrected for the varying illumination angles, viewing
angles and sun-earth distance.
2.3 Atmospheric Modeling
Atmospheric contributions dominate (up to 90%) the signal received at
the satellite in aquatic remote sensing applications (Martin, 2004). All the
data that are included in this dissertation were corrected for atmospheric
contributions using algorithms included in the SeaDAS (Sea WiFS Data
Analysis System) software package (Patt et al., 2003). To enable the
implementation of these atmospheric corrections, SeaDAS uses ancillary
meteorological data (e.g., magnitude and direction of surface winds, water
vapor, and humidity) generated by the National Center for Environmental
Prediction (NCEP), and ozone concentrations from the Earth Probe Total
Ozone Mapping Spectrometer (EPTOMS). Wind speed and direction data are
used as input for algorithms determining whitecap radiance. Other
parameters from the meteorological data such as humidity are used as inputs
to determine the appropriate atmospheric model (Gordon and Wang, 1994).
12
The ozone data are used to calculate diffuse transmittance through the ozone
layer (Gordon, 1997).
The atmospheric contributions to the signal received by the satellite
are affected by Rayleigh scattering, whitecaps, interactions between Rayleigh
scattering and aerosols, and water in the atmosphere. For the MODIS
(Moderate Resolution Imaging Spectrometer) data that are being used in this
work, atmospheric corrections are processed using iterative near infra red
correction techniques (Gordon and Wang, 1994; Stumpf et al., 2003). This
technique first estimates and removes Rayleigh and whitecap atmospheric
contributions, leaving only the remaining atmospheric contributions from
aerosols and water (Gordon and Wang, 1994). Rayleigh and whitecap
components are solved for in SeaDAS iteratively using the near infra red
(NIR) bands (7 48 and 869 nm) and using over 25000 radiative transfer
simulations that are based on multiple MODTRAN (Moderate resolution
atmospheric Transmission) atmosphere models (Patt et al., 2003). Because
water heavily absorbs in the NIR wavelength region, L w is assumed to be
close to zero and L 7 (.-l) will consist only of L R (A.)+ LA (A) in this wavelength
region. Since L R(il) has already been modeled and removed, what is left
consists only of contributions from aerosols and Rayleigh-aerosol interactions.
The aerosol contribution to the NIR bands is iteratively fit to an aerosol
model, and LA
(l) is calculated for all bands (Stumpf et al., 2003).
13
2.4 Optical Components of Water
A large number of previous studies have been carried out to
characterize the optical properties of optically active components in water
bodies in order to determine the relative abundance of these components in
the water (such as Bukata et al., 1981; Clarke et al., 1970; Kahru and
Mitchell, 1998; Morel and Prieur, 1977 among many). These components
include the water, suspended sediments (suspended particulate matter, SPM
and suspended particulate inorganic matter, SPIM), colored dissolved organic
material (CDOM or gelbstofl), particulate organics (POM), and pigmented
materials (i.e., chlorophyll-a, phycocyanin) (Davies-Colley et al., 2003).
Two basic approaches have been used to address this issue. The first
involves extracting concentrations of the examined components from the
apparent optical properties (AOPs) such as the radiance or reflectance at the
sensor (e.g. O'Reilly et al., 1998; 2000; Vincent et al., 2004). Knowing the
reflectances of various components, the atmospherically corrected water
leaving reflectance can be used to provide correlations with the
concentrations of the various components in the examined waters. Most of
the common chlorophyll algorithms used in case I waters are of this type (e.g.
O'Reilly et al., 1998). The drawback of this type of method in optically
complex (case II) waters is that the individual components concentrations are
14
non-linearly related to the measured apparent optical properties, requiring
significantly more data to successfully model.
The second approach relies on extracting theses concentrations from
the inherent optical properties (IOPs) (e.g. Bukata et al., 1998; Pegau et al.,
1999). IOPs refer to properties characterizing how a light field is modified by
physical processes of absorption and scattering independent of the geometric
properties of light field. Examples of the IOP include the total backscatter
and absorption (Bukata et al., 1998; Carder et al., 1999; Lee et al., 2002).
Several methods have been applied to retrieve inherent optical properties
from the apparent optical properties. Examples of these methods are those of
Maritorena (Maritorena et al., 2002), Carder (Carder et al., 1999) and Lee
(Lee et al., 2002). The application of these methods that allows the extraction
of total backscatter and absorption from the apparent optical properties is
enabled in the SeaDAS domain. Knowing the absorption and backscatter
characteristics of the individual components, total backscatter or absorption
can then be deconvolved to calculate the concentrations of individual
components contributing to the total absorption and backscatter.
15
CHAPTER 3
SPATIAL AND TEMPORAL VARIATIONS OF ALGAL BLOOMS IN THE LOWER GREAT LAKES
3.1 Abstract
As part of the MERHAB (Monitoring and Event Response for Harmful
Algal Blooms) Lower Great Lakes project, we are exploring the utility of
chlorophyll identification algorithms, and their use in mapping the spatial
and temporal distribution of algal blooms in Lake Erie and Lake Ontario. We
have processed Sea WiFS and MOD IS data acquired in the summers of 2003,
2004, and 2005 over the investigated lakes. The remotely-sensed data were
used to derive chlorophyll-a concentrations. The chlorophyll data were
compared with in situ measurements acquired largely from mooring stations,
buoys, and from cruise measurements. Comparisons were made between case
I and case II water satellite-based estimates of chlorophyll-a concentrations
and in situ cruise measurements for the 2003-2005 field seasons (July-August
2003, 2004, May-Sept 2005). We used algorithms which were developed for
case I waters (where chlorophyll content and other dissolved and suspended
materials are assumed to co-vary), and applied these algorithms to the more
optically complex waters of Lake Erie and Lake Ontario (Bukata et al., 1981;
O'Reilly et al., 2000). This approach has been successfully applied to Lake
16
Michigan waters (Lesht et al., 2002). We demonstrate similar applications
and successes in Lakes Erie and Ontario, where comparisons of satellite
based chlorophyll-a concentrations with in situ chlorophyll-a measurements
indicate that these algorithms provide reasonably good estimates of
chlorophyll-a. We also demonstrate an even better correlation with
chlorophyll-a concentrations derived from the case II water algorithm. Using
a series of Sea WiFS and MOD IS scenes with minimal cloud cover, we
mapped annual re-occurrence (same location and time) of chlorophyll
distribution patterns as well as unique algal bloom events in Lakes Erie and
Ontario in the summers of 2003-2005. Our results indicate: (1) a significant
correlation between chlorophyll concentrations extracted from Sea WiFS and
from MODIS scenes acquired on the same day, especially for days with
minimal atmospheric contributions, (2) a general correspondence between
Sea WiFS-based chlorophyll-a concentrations and in situ chlorophyll
measurements, and (3) annual recurrence (same location and time) of
chlorophyll distribution patterns as well as unique algal bloom events in
Lakes Erie and Ontario.
We used the unique spectral characteristics of the phycocyanin
pigment to discriminate between phycocyanin containing blooms and other
blooms. These results have been validated has using phycocyanin
fluorescence data showing relative phycocyanin abundance.
17
3.2 Introduction
We evaluated the utility of SeaWiFS (Sea-viewing Wide Field-of-view
Sensor) and MODIS (Moderate Resolution Imaging Spectrometer) platforms
for mapping algal blooms as a first step towards the spectral identification of
the toxic blooms in the lower Great Lakes. Comparisons were made between
satellite-based estimates of chlorophyll-a concentrations and in situ
measurements for the 2003-2004 field seasons (July and August 2003, 2004).
We used algorithms which were developed for case I waters where
chlorophyll content and other dissolved and suspended materials are
assumed to co-vary, and applied these algorithms to the more optically
complex waters of Lake Erie and Lake Ontario (Bukata et al., 1981; O'Reilly
et al., 2000). This approach has been successfully applied to Lake Michigan
waters (Lesht et al., 2002). We demonstrate similar applications and
successes in Lakes Erie and Ontario, where comparisons of satellite-based
chlorophyll-a concentrations with in situ chlorophyll-a measurements
indicate that these algorithms provide reasonably good estimates of
chlorophyll-a in these lakes. Current and future work is focusing on
evaluating the case I chlorophyll algorithms over a wider array of
concentrations and time spans of the year to further examine the reliability
and accuracy of these methodologies. A case II model (Carder, 2003) is also
tested, and the results are significantly better than the case I models.
18
Monitoring the distribution of algal blooms in the Great Lakes area
using satellite data is often hindered by the cloud cover. The acquisition of a
continuous series of images with minimal cloud cover over an algal bloom of
interest is needed for proper mapping of the temporal and spatial distribution
of the bloom. The acquisition of a series of Sea WiFS or a series of MO DIS
images of such quality over the study area has proven to be difficult. To begin
addressing this problem, we will explore the possibility of integrating the
information (derived chlorophyll-a) contained in Sea WiFS and MO DIS
images. Chlorophyll-a concentrations extracted from MODIS and SeaWiFS
are compatible but are not identical, given the variations in number of, and
the wavelength region covered by, the bands used in extracting the respective
chlorophyll-a values. Nevertheless, we show that using a series of SeaWiFS
and MODIS scenes that are acquired at approximately the same time and
which show minimal cloud cover, the spatial extent of algal blooms of interest
could be readily measured through time. Finally, we use this methodology to
map the temporal and spatial patterns of algal blooms in Lakes Erie and
Ontario in the summers of 2003 and 2004, and to identify the individual and
annually occurring algal bloom events.
19
3.3 Image Acquisition and Processing
We examined SeaWiFS and MODIS (raw and chlorophyll-a
concentrations) data for the summers of 2003 and 2004 over Lakes Erie and
Ontario. Approximately 120 Sea WiFS scenes (2003, 2004) and 20 MOD IS
scenes (2004) were processed using SeaDAS software and applying the
OC4V4 (SeaWiFS), and the OC3 (MODIS) algorithms (O'Reilly et al., 2000)
for the calculation of water leaving radiance and chlorophyll-a
concentrations. The OC3 and OC4V4 are techniques which extract empirical
relationships between ratios of reflectances from three (OC3) or four (OC4v4)
wavelength regions in the visible portion of the spectra and observed
chlorophyll-a concentrations (O'Reilly et al., 2000).
Chlorophyll-a concentrations were also derived using Carder case II
semi-analytic algorithm in SeaDAS (Carder et al., 1999) and results were
compared to those extracted using the OC3 and OC4V4 techniques. The
default iterative Near IR atmospheric models were utilized. Images were
brought to a common projection using commercial image processing software
(ENVI) to allow comparisons to be made between scenes and between remote
sensing and in situ observations.
20
3.4 Field Measurements
3.4.1 ln Situ Sampling and Measurements
Samples were collected and processed for station cruises on the RV
Guardian and RV Limnos during the summers of 2003, 2004 and 2005 by Dr.
Boyer's research group at the State University at New York School of
Environmental Science and Forestry (SUNY ESF). Fluorescence data were
acquired by Dr. Boyer's group at SUNY ESF during 2004 and 2005 and by
Dr. Michael Twiss at the University of Binghamton in 2005.
Discrete water samples (lL) were collected at 1 m depth and filter onto
4 7-mm Whatman 934-AH glass fiber filters for phycocyanin and chlorophyll
determinations. The concentration of chlorophyll-a was determined by
fluorescence using a Turner Designs TD-700 fluorometer equipped with a
chlorophyll excitation (436 nm) and emission (680 nm) filter set and a
mercury blue light source (Welshmeyer, 1994). The concentration of
chlorophyll-a was also computed using spectrophotogrammetry by analysis of
chlorophyll pigments by the trichromatic technique described by Parsons
(Parsons, 1963).
Phycocyanin concentrations were estimated fluorometrically using a
modification of the method of Abalde et al (1998) and Siegelman and Kycia
21
(1983) as detailed in Konopko (2007). Phycocyanin was extracted from these
filters by freezing the samples at -21 deg C and thawing at 4 deg C three
times in 10 mM phosphate buffer (pH 6.8) under dim light. The extract was
clarified by centrifugation at 22,000 x g for 15 min, and the phycocyanin
concentrations in the supernatant determined by fluorescence using a Turner
Designs 10-AU fluorometer equipped with a 577 nm band pass excitation
filter and a 660 nm cutoff emission filter with a cool while light source.
3.4.2 Fluorometer Measurements
Sampling surveys were conducted in Lake Erie and Lake Ontario in
2004 and 2005. Horizontal fluorometer profiles were obtained throughout the
surveys and water samples were occasionally collected for analysis and
calibration of the fluorometer data.
A 45 kg steel "fish" was towed by the RIV Lake Guardian at a speed of
8-11 knots at a depth of 1 m. Water was pumped through a Teflon™-lined
polyethylene tubing into a 15 L polyvinylchloride "ferry box" that provided an
integrated water sample at a horizontal transect resolution of 0.5 km.
During the Sept 2005 cruise a FluoroProbe (bbe Moldeanke, GmbH),
was submersed inside the ferry box and measured temperature and
phytoplankton concentrations. The FluoroProbe uses five different excitation
22
wavelength ranges (450, 525, 570, 590 and 610 nm) sequentially to illuminate
a sample. It detects fluorescence at 680 nm. Based on the different
fluorescence responses to light at the individual excitation wavelengths, the
concentration of different algal groups is calculated based on a gaussian fit.
The FluoroProbe distinguishes phytoplankton into four groups: 1)
Chlorophyta and Euglenophyta, 2) phycocyanin (PC)-rich Cyanobacteria, 3)
Heterokontophyta and Dinophyta, and 4) Cryptophyta and phycoerythrin
(PE)-rich Cyanobacteria based on fluorescence peaks specific to the
phytoplankton types (Beutler et al., 2003; 2002). All data were time stamped
and archived on a field computer (Panasonic, model CF-29). Time-stamped
geographic coordinates were obtained simultaneously from the ship
navigational system.
During the remaining 5 cruises, water was drawn by gravity from the
ferry box to two Turner Designs 10-AU fluorometers, equipped with flow cells
and either a chlorophyll or phycocyanin filter sets (340 to 500 nm excitation,
>665 nm emission) and phycocyanin (excitation 577 nm, emission 660 nm).
Water flow rate was controlled by the head pressure generated by a 5.5 cm x
1.3 m plastic tube. Air bubbles introduced during rough waters were also
removed by this tube. Readings from the fluorometers were collected at one
minute intervals and linked to the ship's GPS system. This corresponded to
approximately 300 m intervals when the ship was in motion at its normal
cruising speed of 10 knots.
23
3.5 Sea WiFS vs. MOD IS Measurements of Chlorophyll-a
Sea WiFS and MO DIS images are routinely acquired approximately an
hour apart over the study area (SeaWiFS around 12:00 pm and MODIS Aqua
at 13:00 local time). A comparison between chlorophyll-a concentration
extracted from SeaWiFS and MODIS Aqua was conducted for 20 pairs of
scenes with each pair acquired approximately one hour apart in the summer
of 2004. We found a correlation between chlorophyll-a concentrations derived
from SeaWiFS and MODIS data (Fig. 3.1). The figure displays the range of
observed variations in the degree of correspondence between the Sea WiFS
and MODIS data sets. The highest correlation between SeaWiFS and MODIS
data was observed for the pair of scenes acquired on 8/2/04 (correlation
coefficient of 0.65) and the lowest correlation was for the pair of images
acquired on the previous day (on 8/1/04) (correlation coefficient of 0.48). In
addition, a least squares linear fit between the OC3 and OC4 satellite derived
data for 8/2/04 had a slope of 0.78 (perfect fit slope=l), where the least
squares linear fit for the data from 8/1/04 had a slope of 0.28, much further
from the ideal 1:1 relationship. The correlation coefficient for the 6/23/04 data
was .55, with a slope of 0.33, so again there is a correlation, but one where a
bias is introduced, and the OC4 is underestimated relative to the OC3. The
9/23/04 had a correlation coefficient of .55, and a slope of the linear fit of 0.74,
much closer to an ideal fit. Because such differences occurred in a relatively
24
short time period (1 day), we suspect that these variations largely reflect
temporal variations in atmospheric contributions and are less likely to be
related to variations in water characteristics. The variations cause a general
decorrelation between the images, and also a bias towards higher OC3 values
relative to OC4v4 derived values.
u 0
>.c a.
0
08/02/04
11.25
7.50
3.75 -'.l�f.:,_. r?t:_-· __
0.00-+-....;.,..����������-+ 0,00 3.75 7.50 11.25 15.0
09/23/04
11.25
7.50
3.75
0.00 3.75 7.50 11.25 15.00
0
.c
u Vl LL
15.00 -f--L......L-'---'-L......L-'---'-L......L----'---'-L......L----'---+ 15.00 -f--L......L-'---'-L......L-'---'-L......L----'---'-L......L----'---+
ro Q)
Vl
06/23/04 08/01/04
11.25 11.25
7.50 . :_ ·�\\t:.- :7.50
3.75 3.75
o.oo-+--�����������--+- o.oo-+--�����������--+-
o.oo 3.75 7.50 11.25 15.00 0.00 3.75 7.50 11.25
MODIS Chlorophyll-a (µg/L)(OC3)
Figure 3.1 Density scatter plot comparing chlorophyll-a concentration
derived from MODIS and SeaWiFS data. Areas in shades of red and blue
represent the largest and the smallest densities of picture elements,
respectively.
25
15.00
08/02/04 09/23/04
06/23/04 08/01/04
Figure 3.2 MODIS Band 16 (846 nm) reflectance images showing relative
atmospheric contributions. Brighter intensities show increased atmospheric
contribution (e.g., increased cloud cover).
Atmospheric contributions will cause reflectance values for clear water
to exceed zero in the wavelength regions: 745-785 nm and 845-885 nm. These
spectral regions are covered by MODIS bands 15 and 16 and SeaWiFS bands
7 and 8 (Land and Haigh, 1996; Sturm and Zibordi, 2002). Reflectance over
clear water ranged from 0.010 to 0.014 on 08/02/04 and varied from 0.010 to
0.030 on 08/01/04. Figure 3.2 shows MODIS band 16 images acquired over
Lake Erie on 06/23/04, 08/01/04, 08/02/04, and 09/23/04. The brightest image,
the one displaying the highest reflectance, and the least correspondence
between the derived SeaWiFS and MODIS chlorophyll-a concentration is the
image acquired on 8/01/04. The scene that was acquired on 8/02/04 is among
the darkest images (lowest reflectance), and shows the highest
26
correspondence between the derived chlorophyll-a concentrations (Fig. 3.1).
Thus application of improved atmospheric models (Ruddick et al., 2000) over
inland waters such as Lakes Erie and Ontario should enhance the observed
correlations.
Our results indicate that chlorophyll-a concentration extracted from
MO DIS and Sea WiFS scene pairs acquired around the same time and under
favorable atmospheric conditions can be integrated together to aid analyses.
3.6 Satellite vs. In Situ Measurements of Chlorophyll-a
In situ chlorophyll-a measurements using multiple measurement
techniques were compared with chlorophyll-a concentrations extracted from
Sea WiFS data (Fig. 3.3); the remote sensing and in situ measurements were
acquired around the same approximate date (within 48 hours). Inspection of
figure 3.3 suggests that chlorophyll concentrations from Sea WiFS correlate
well with in situ measurements (pink: Welchmeyer; blue: Parson) and shows
that the Sea WiFS concentrations are more or less similar to the in situ
measurements in magnitude. In situ chlorophyll-a measurements were also
compared with chlorophyll-a concentrations extracted from case I and case II
algorithms from SeaWiFS (Figs. 3.4 and 3.5); the remote sensing and in situ
measurements were acquired around the same approximate date (within 48
hours). Figure 3.4 shows that the Carder case II algorithm does provide a
27
better fit to the in situ data (R2 of 0.78 for Carder method vs R2 of 0.551 for
the OC4v4 method, and slope of the relationship closer to 1). Figure 3.5
shows the data (spectral and in situ) that were used to calibrate and derive
the case I water algorithms for open oceans; the inner diagonal solid lines
mark± 35% agreement relative to the 1:1 central solid line and the dashed
lines are the 1:5 and 5:1 lines encompassing the data set
(http://seabass.gsfc.nasa.gov/matchup_results.html). For comparison, we
plotted our in situ chlorophyll-a data and the chlorophyll-a data derived from
Sea WiFS for the same locations and using case I and case II algorithms.
Inspection of figure 3.5 suggests that all of our case I samples plot within the
envelope defined by the 1:5 and 5:1 lines and for case II samples, the majority
of them plot within the envelope defined by the ± 35% lines. These results
suggest that the application of case I water algorithms provide reasonable
approximations for the chlorophyll-a content in Lakes Erie and Ontario and
that case II algorithms provide more precise estimates.
28
===
1
·c
C,
-�
25 • El<tracted Chi-a Parsons
• Extracted Chi-a Welcluneyer
20 --1:1
----- Linear (Extracted Chi-a Parsons)
-- - -- Linear (E;.ttacted Chi-aWelchmeyer)
15
■
• ♦
10
••
•
,,
y = 0.673x -t 1.534 R2 = 0.780 _.·
•
■ -✓
_,... /.-</= 0.52x+ 1.182
,'
/ 2 =-
o------�----�-----�-----------
0 5 10 15 20
In situ Chi-a (1t9/L)
Figure 3.3 Satellite-derived chlorophyll-a concentrations plotted
against in situ-measured chlorophyll-a concentrations. Blue diamonds
represent measurements made using Parsons technique (Parsons,
1963), whereas those measured using Welschmeyer technique
(Welschmeyer, 1994) are represented by pink squares. Linear fits for
both relationships are displayed, showing a better correlation with
measurements made using the Welschmeyer technique.
29
Differences between the Sea WiFS-based and in situ chlorophyll-a
concentrations extracted using case II algorithms could be attributed to: (1)
20 ,-------------------------------------.
18
16
14
-�12
<1> 10 .::::
:t:: 8
6
4
2
0
♦ Swa-V\Ms - OC4v4
■ Sea-\'Vifs Ca1der
----- Lin war (Sea-VVifs - OC4v4)
- - - Linear (Sea-\'\/ifs Ca1der)
--1:1 iperf9ctMatch)
•
0 2 4 6
♦
'I= 0.814, +0.323.,.. R' = 0 901 ;
. ;, 6'
6' ,
, ..... _.,.,, .., ,;: 0.599� + 1.807
, - R"= 0.648
✓ ,,, ,,
.,.. ,,
. .,, ..
8 10 12 In situ Chi-a (llg!L)
14 15 18
Figure 3.4 Comparison between satellite-derived chlorophyll-a
concentration using case I (OC4v4, blue diamonds) and case II (Carder,
pink squares) algorithms and in situ measured chlorophyll-a
concentrations. Linear fits for both relationships are displayed showing
a better correlation with the Carder-derived chlorophyll-a
concentrations.
30
20
differences in "sample size" (1 km.2 for satellite vs. point samples for ground
measurements), (2) variations in chlorophyll concentrations with depth that
are reflected in the depth-integrated satellite measurements, but not in the
in situ measurements, and (3) differences in the time of acquisition. Satellite
measurements are not always acquired at the same time during which the
ground sampling campaign is being conducted.
In figure 3.6, the chlorophyll-a concentrations obtained for seven water
samples from Lake Erie, acquired on July 13, 2004 (location: Lat: 41.816 N;
Lon: 82.983 E), at depths ranging from Oto 7 m, and analyzed using the
Parsons (blue symbol) and Welchmeyer (pink symbol) protocols are compared
to concentrations extracted from Sea WiFS data (yellow line). The figure
shows the large variations of in situ measurements with depth and
demonstrates how the Sea WiFS chlorophyll-a concentration could represent a
depth integrated measurement of the in situ chlorophyll-a concentrations.
31
l lzlc1r,-11 hyll
101 ◊ Sea-Wifs - OC4v4
□ Sea-Wifs Carder •
• • • 1:1
•
•
• •
-
10-� ------------------Lo-:: 10
° i0
1 10-
ln 1 u
Figure 3.5 SeaWiFS case I (OC4v4, blue diamonds) and case II (Carder, pink
squares) derived concentrations vs. in situ results overlain over Sea WiFS
validation results (gray squares) from
http://seabass.gsfc.nasa.gov/matchup results.html. Data from both case I and
case II algorithms fall within validation data set ranges.
32
-
-
-
■ •
■ •
- •
■
■
■ •
I I
• Extracted Chl-a Parsons(ug/L)
• Extracted Chl-a (ug/L)
Sea-Wifs
•
•
■ •
Figure 3.6 In situ chlorophyll-a concentration (Parsons, blue
diamonds, Welchmeyer, pink squares) versus depth. Figure also
shows the Sea WiFS-derived chlorophyll-a (yellow line).
33
3. 7 Identification and Mapping of Algal Blooms
Examination of chlorophyll-a concentration images (extracted from
Sea WiFS and MO DIS) acquired in summers of 2003 and 2004 over Lakes
Erie and Ontario revealed several algal blooms. Blooms were noted as early
as early June in Lake Ontario, and through mid-October in Lake Erie. As
described earlier, we integrated inferences from Sea WiFS and MO DIS data
to provide the most complete cloud free coverage. For example, mapping of a
bloom which was noted in the Oswego Harbor at the beginning of August
2004 was enabled using a time series of chlorophyll-a images extracted from
a combined set of Sea WiFS and MO DIS data sets. The assembled time series
provided a temporal coverage with minimal cloud cover (Fig. 3.7).
>6
µg/L
0.0
µg/L
Figure 3. 7 Progression of an algal bloom in the Oswego Harbor using a time
series generated from Sea WiFS and MOD IS images acquired in August of 2004.
34
Generally the blooms in Lake Erie originate in the western basin and
propagate into the central portion of the Lake (Fig. 3.8). Although these
blooms usually cover much of the central and western basins of Lake Erie,
they do not propagate into the eastern basin (Fig. 3.8). An algal bloom in the
western basin of Lake Erie coincided with a reported cyanobacterial bloom in
the same area (Rinta-Kanto et al., 2005). The only bloom that was observed
in the eastern basin of Lake Erie throughout the investigated period is a
bloom that occurred in the same approximate location in years 2003 and
2004, around the end of June (Fig. 3.9).
Figure 3.8 Chlorophyll-a concentrations extracted from Sea WiFS images acquired
in September of 2004. These are indicative of the propagation of an algal bloom in
the Western and Central Basins of Lake Erie; the inferred bloom (from satellite
data) coincided with a reported cyanobacterial bloom.
35
Figure 3.9 Chlorophyll-a concentrations extracted from a
processed SeaWiFS image acquired on June 23, 2003 (left), and
June 23, 2004 (right). Image shows blooms in the eastern parts
of Lake Erie. This late June bloom is the only bloom identified in
the Eastern Basin of the lake.
3.8 Locating Phycocyanin Containing Blooms
>6
µg/L
0.0
µg/L
One of the primary goals of the MERHAB project is to be able to
discriminate between potentially harmful and non-harmful algal blooms in
the lower Great Lakes. Microcystis species and other toxin-producing
cyanobacteria have been observed in Lake Erie (Brittain et al., 2000; Rinta
Kanto et al., 2005). As Microcystis is a cyanobacteria containing the
phycocyanin pigment it should be possible to rule out the possible toxicity of
many blooms by the absence of phycocyanin. We developed an algorithm to
allow the identification of the location of potentially toxic cyanobacterial
blooms from Sea WiFS satellite data, based on spectral characteristics of
phycocyanin pigment.
To accomplish this goal, Sea WiFS images were processed to remote
sensing reflectance using SeaDAS 4.8. A mixing model was adopted based on
the following end member components: chlorophyll-a, phycocyanin and water
36
(Vincent et al., 2004), chlorophyll and water (extracted from Sea WiFS images
in areas of >8 µg/L chlorophyll-a and low phycocyanin as determined from
fluorescence data), and water with chlorophyll-a levels below 1 µg/L.
Figure 3.10 Relative abundance of phycocyanin from in situ fluorescence
measurements (July 13-16, 2004) overlain on top of relative abundance of
phycocyanin derived from Sea WiFS imagery (July 13 and 15, 2004).
This phycocyanin abundance extracted from Sea WiFS was compared
with relative abundances of phycocyanin based on fluorescence data acquired
during July 2004 (Fig. 3.10). Our success in spectrally distinguishing
phycocyanin is consistent with the earlier findings in the western Lake Erie
using remote sensing techniques (Vincent et al., 2004). Despite the broad
wavelength regions covered by the TM Landsat bands, and the absence of TM
band(s) in the wavelength region affected by the phycocyanin (PC) (refer to
37
section 4.2 for further details), Vincent et al. (2004) successfully developed
algorithms to detect PC from Landsat TM data for mapping cyanobacterial
blooms in Lake Erie using ratios of reflectance data. We are currently
refining our mixing model to include better end members for high turbidity
and high CDOM areas in an attempt to remove some false positive
identification and increase the areas where the model resolves.
38
CHAPTER4
MAPPING CYANOBACTERIAL BLOOMS IN THE GREAT LAKES USING MODIS
4.1 Abstract
Microcystis and other toxin-producing cyanobacteria have been
documented in Lake Erie and Ontario in the last several years. We developed
algorithms to discriminate potentially toxic cyanobacterial blooms from other
harmless blooms and to extract relative phycocyanin abundances from
Moderate Resolution Imaging Spectrometer (MODIS) satellite data. Lee's
Quasi-Analytical Algorithm is used to calculate total absorption and
backscatter from the 250 m, 500 m and 1 km bands of MOD IS scenes. A
nonnegative least square algorithm was then utilized to unmix relative
concentrations of green algae, blue-green algae, and CDOM and sediments
combined in lake waters using published absorption spectra for these
components. MODIS-derived cyanobacterial concentrations and/or bloom
distributions from 10 scenes acquired in the summers of 2004 and 2005 were
successfully verified against contemporaneous calibrated measurements of
pigments that were acquired from along track fluorometer measurements
from six cruises, and three additional cyanobacterial blooms reported in the
scientific literature between 2002 and 2006. The developed methodologies
39
could potentially be used to develop cost-effective practical screening methods
for rapid detection of, and warning against, cyanobacterial blooms in the
lower Great Lakes.
4.2 Problem Description
Lake Erie (surface area: >25,000 km2; average depth: 20 m) and Lake
Ontario (surface area: 19,000 km2; average depth: 86 m) are two of the
largest 20 lakes in the world (Fig. 4.1). The Niagara River carries the outflow
from Lake Erie into Lake Ontario and accounts for over 85% of the total
inflow to Lake Ontario (Shear et al., 1995). The outflow of Lake Ontario is
carried by the St. Lawrence River which flows into the Atlantic Ocean.
Millions of citizens of the U.S and Canada rely on Lakes Erie and Ontario
and the remaining Laurentian lakes for drinking and recreation purposes.
Increasing reports over the past years of occurrences of harmful algae in the
Great Lakes demonstrated the vulnerability of these resources and the need
for developing methodologies for expedited detection of the spatial
distribution and propagation of these blooms on regional scales (Anderson et
al., 2002; Hallegraeff, 1993).
40
44°
N
42°
N
84°
W
Detroit �
· Windsor-1,
Toledo
80°
W 76°
W
Lake Erie
Erie·
Cleveland
Figure 4.1 Landsat Thematic Mapper false color composite showing the location of Lake Erie, Lake Ontario and surrounding metropolitan areas.
Conditions favoring the formation of the Harmful Algal Blooms (HABs)
are not completely understood. It is believed that a number of factors (e.g.,
sunlight, nutrients, grazing pressure, water temperature, water depth, water
flow, filter feeders) could collectively promote their development (Downing et
al., 2001; Havens, 2008; Paerl, 2008). Cyanobacterial blooms occur in warm
stable water masses where the species can maximize light capture and
nutrient uptake through the formation of gas vacuoles to regulate buoyancy
(Reynolds et al., 1987). Filter feeders (e.g., invasive zebra mussels; Dreissena
polymorpha) discriminate against toxic cyanobacteria in feeding, giving them
a competitive advantage (Vanderploeg et al., 2001).
In the Lower Great lakes, the toxin producing cyanobacterial species
which have been observed include Microcystis, Anabaena and Planktothrix
41
species (Boyer, 2006; Rinta-Kanto et al., 2005). Cyanobacterial blooms have
been reported in the western basin of Lake Erie in late summer to fall of 2000
through 2007. The most significant levels of microcystin toxicity (14.3 to 20.0
µg L-1) were associated with high chlorophyll-a (4 to 40 µg L-1) blooms and
was detected at the mouth of Maumee River in the summers of 2003 and
2004 (Rinta-Kanto et al., 2005) at stations in the western basin and at the
mouth of the Sandusky River. In Lake Ontario a toxic bloom of Microcystis
approached the World Health Organization (WHO) guideline value of 1 µg
microcystin-LR L-1 in 2003 (Boyer et al., 2004; Makarewicz et al., 2006).
Given the potential health risks associated with the blooms and the large
areal extent of these occurrences throughout Lakes Erie and Ontario,
regional and cost effective monitoring solutions should be developed to
provide adequate warning for the impacted populations. In this manuscript
we describe one such approach that takes advantage of the spectral
properties of these blooms as observed in the readily available, frequently
acquired remote sensing Moderate Resolution Imaging Spectrometer
(MODIS) data sets.
The toxic cyanobacteria in the Great Lakes not only possess the
chlorophyll-a pigments like all other phytoplankton, but also have the
accessory pigment phycocyanin (Fay, 1983) that strongly absorbs in the 620
to 630 µm wavelength region (Bryant, 1981; Sathyendranath et al., 1987),
giving the blue-green algae in Lakes Erie and Ontario their characteristic
42
C 0
0 (/)
.D.
Ctl
Q) >
� a> 0::
-Microcy
-Gree1
450 500 55:J 6JO
Wavelength (nm)
650 700
Landsat TM
MERIS
MODIS 500m
MODIS 1km
MODIS 250m
Figure 4.2 Absorption curves for green and blue green algae. Also shown are positions of MOD IS, MERIS and Landsat TM bands.
color. Because all of the toxin-producing species in the Great Lakes are
cyanobacterial, the accessory pigments associated with these blooms can
potentially be used to identify toxic blooms. Earlier attempts to use satellite
data to remotely map the distribution of toxic algae in the Great Lakes and
elsewhere taking advantage of the spectral signature introduced by the
phycocyanin pigment (PC) are limited. One of these methods utilized band
ratios of reflectance values extracted from Landsat TM images (Vincent et
al., 2004). A second method, a semi-analytical algorithm, was applied to
Medium Resolution Imaging Spectrometer (MERIS) reflectance images
43
(Simis et al., 2005). Vincent's approach takes advantage of the high spatial
resolution provided by the Landsat TM imagery, but is disadvantaged by the
infrequent visits (at best every 7 days) of the Landsat system over the study
area and by the absence of bands of sufficient spectral resolution over the
critical wavelength regions to distinguish the phycocyanin absorption feature
from the chlorophyll-a absorption feature. Specifically, there are no Landsat
bands that are positioned over the maximum phycocyanin absorptions in the
620-630 nm wavelength region (Fig. 4.2). The figure shows that this is not the
case with the MERIS data; the phycocyanin absorption features are covered
by MERIS band 6. However, MERIS's temporal availability (three day
repeat, but only acquired when ordered) and the relatively high commercial
cost of the data makes it a less practical system for routine screening and
monitoring purposes.
Monitoring systems that are based on airborne or space-borne (e.g.,
Hyperion) hyperspectral reflectance algorithms (Dekker et al., 1991; Gons et
al., 1992; Jupp et al., 1994) are costly and provide limited spatial coverage
(USGS, 2008). Previous attempts using the 1 km MODIS bands were
discounted as viable methods for discriminating cyanobacterial blooms
(Kutser et al., 2006a), as these bands do not adequately sample the
phycocyanin absorption feature.
In this study, we report the first MODIS-based algorithms which use
absorption spectra to identify potentially toxic cyanobacteria by
44
distinguishing the phycocyanin absorption feature from the chlorophyll
absorption features. Unlike previous attempts, that solely used the 1 km or
250 m bands (Kutser et al., 2006a; Kutser et al., 2006b), our approach
integrates the spectrally wider 250 m MODIS band 1 with the 1 km and 500
m bands enabling the discrimination between the green and cyanobacterial
blooms (Fig. 4.2). A MODIS-based algorithm is advantageous for the
following reasons: (1) the data are free and rapidly available, making the
developed methodologies more appealing to users; (2) reliable atmospheric
models are in place enabling comparisons to be made between MODIS
derived water compositions extracted from scenes acquired under varying
atmospheric conditions; (3) the revisit time is short (daily coverage from
Aqua) allowing monitoring of spatial and temporal propagation of blooms
with confidence; (4) archival MODIS data are available since July of 2002
facilitating the identification of recurrent seasonal development and
propagation of blooms; and (5) the MODIS bands are placed over critical
wavelength regions enabling the discrimination between the phycocyanin and
chlorophyll absorption (Fig. 4.2).
4.3 Methodology
In this study we developed an algorithm that is based on a non
negative linear least squares algorithm to unmix the prevailing component
45
concentrations in the waters of Lakes Erie and Ontario both spatially and
temporally. The investigated components include: Colored Dissolved Organic
Matter (CDOM), suspended sediments, green algae, and blue-green algae.
The adopted unmixing procedure was based on the individual components of
the examined waters and on inherent optical property (IOP) of the system.
The !OP-based algorithm provides linear relationships between components
concentrations and the observed properties (Bukata et al., 1998). The
developed algorithm was applied to selected MODIS data sets that were
acquired over the study areas in July and August, 2004, and July, August,
and September, 2005. The selected scenes were acquired around the time
periods during which our field campaigns were conducted (within one week of
the cruise) to enable comparisons to be made between relative cyanobacterial
and other phytoplankton concentrations extracted from space-borne and field
observations. The fluorescence data derived from along track horizontal
profiles acquired during cruises were calibrated against individual on-station
measurements. We further tested our algorithms by comparing space-borne
observations from additional MODIS scenes (late August 2005) to field
observations acquired in the study area (e.g. Dyble et al., 2008) for the
western basin of Lake Erie.
46
4.4 Field Measurements
Sampling surveys were conducted in Lake Erie and Lake Ontario in
2004-2005 (Table 4.1). Four offshore Lake Erie surveys and two Lake Ontario
surveys are reported here and compared with observations made from
MODIS data. Horizontal fluorometer profiles were obtained throughout the
surveys and water samples were occasionally collected for analysis and
calibration of the fluorometer data.
Cruise
Name
MELEE-VIII
Taste &Odor
MELEE-IX
MELEE-X
Taste &Odor
IFYLE
Lake
Erie
Ontario
Erie
Erie
Ontario
Erie
Departure Port and Date
Colborne, ON 7/12/2004
Burlington, ON
8/30/2004
Colborne, ON 7/11/2005
Amherstburg, ON 8/22/2005
Burlington, ON
8/29/2005
Cleveland, OH
9/7/2005
Return Port
and Date
Colborne, ON 7/16/2004
Burlington, ON
9/3/2004
Colborne, ON 7/15/2005
Burlington, ON 8/26/2005
Burlington, ON
9/2/2005
Cleveland, OH
9/11/2005
Instruments
used to measure PC
10-AU
10-AU
10-AU
10-AU &Hydrolab
10-AU &Hydrolab
Fluoroprobe
Table 4.1 Cruise dates and locations on which fluorometer horizontal profiles were acquired.
A 45 kg steel "fish" was towed by the RIV Lake Guardian at a speed of
8-11 knots at a depth of 1 m. Water was pumped through a Teflon™-lined
polyethylene tubing into a 15 L polyvinylchloride "ferry box" that provided an
47
integrated water sample at a horizontal transect resolution of 0.5 km. During
the September 2005 cruise a FluoroProbe (bbe Moldeanke, GmbH), was
submersed inside the ferry box and measured temperature and
phytoplankton concentrations. The FluoroProbe distinguishes phytoplankton
into four groups: 1) Chlorophyta and Euglenophyta, 2) phycocyanin (PC)-rich
Cyanobacteria, 3) Heterokontophyta and Dinophyta, and 4) Cryptophyta and
phycoerythrin (PE)-rich Cyanobacteria (Beutler et al., 2003; 2002). All data
were time stamped and archived on a field computer (Panasonic, model CF-
29). Time-stamped geographic coordinates were obtained simultaneously
from the ship navigational system. During the remaining 5 cruises, water
was drawn by gravity from the ferry box to a fluorometer (Turner Designs,
model 10-AU) which measured chlorophyll and phycocyanin abundance
(Konopko, 2007).
Discrete water samples (1 L) were collected at 1 m depth and filtered
onto 47-mm Whatman 934-AH glass fiber filters for phycocyanin and
chlorophyll determinations and calibration of the fluorometer data.
Chlorophyll-a was determined fluorometrically using the method of
Welschmeyer (1994). Phycocyanin concentrations were estimated
fluorometrically using a modification of the method of Abalde et al. (1998)
and Siegelman and Kycia (1978). Phycocyanin was extracted from these
filters and the phycocyanin concentrations determined by fluorescence as
described in Konopko (2007).
48
4.5 Algorithm Development and Processing of MODIS Data
MODIS images were selected to provide the best spatial coverage of
the lakes and with minimum cloud coverage. Preference was given to scenes
where the lakes were viewed closer to nadir (in the center of the MO DIS
swath) to maximize spatial resolution and minimize atmospheric
contributions. MODIS data scenes were processed in SeaDAS 5.2 software
package. Atmospheric corrections were conducted using the Iterative Near
Infra Red Atmosphere model (Patt et al., 2003). Atmospherically corrected
scenes were then processed using the developed algorithms to extract
compositional information (e.g., cyanobacterial and other phytoplankton
abundances).
The quasi analytical methods of Lee et al. (2002) (Lee QAA method)
was selected to solve for the total absorption (aT) and total back scatter at
each location from the reflectance data (Gordon et al., 1975) (equation 4.1):
(4.1)
Where R(0-,A) is the below surface reflectance, bb is the total backscatter, aT is
the total absorption, and f is a constant that depends on the nature of the
incident light field and volume scattering function.
The Lee QAA method was selected because it makes no a priori
assumptions of phytoplankton or CDOM absorption either in shape or
magnitude, has limited empirical relationships, has been successfully tested
49
with non case I datasets, and works with multi-spectral data. When the
water leaving radiance in any bands fell below a threshold value of 0.01
mW• cm-2 • µ.m- 1 • sr-1, the absorption calculations from the QAA algorithm
proved unreliable, producing artificially high very noisy absorption values.
These locations were masked in the results as unusable.
The absorption component is directly related to the concentration of
the individual colored components, through the following relationship
(Bukata et al., 1998):
ar(A,) = I at (,,l)ci + aw
(4.2)
i=l
Where aT is the total absorption, ai and Ci are the absorption and
concentration of each component, and aw is the absorption of pure water.
Knowing the absorption spectra of these components, this system of linear
equations was then solved for component concentrations using a non-negative
least squares technique (Lawson and Hanson, 1974).
The individual absorption spectra for the components in Lake Erie and
Ontario were extracted from published data: (1) green algae from Kirk (1994),
(2) blue-green Microcystis from Kuster (2006b), (3) water from Pope and Fry
(1997), and (4) CDOM and sediment are represented by an average spectra
generated from the spectra of these two endmembers (reported in Binding et
al., 2008) because the spectra of CDOM and sediment are quite similar.
50
Binding (2008) found that acnoM and asediment in the western basin of Lake
Erie are best represented by fitting an exponential function:
a= e<-s(,l-440)) (4.3)
where Sis best represented by a value of 0.011 nm-1 for suspended sediments,
and by 0.0161 nm-1 for CDOM.
The endmember spectra for CDOM/sediment, blue-green algae, green
algae, and water were resampled to MODIS bands using the MODIS relative
. spectral response functions (Fig. 4.3). The figure shows that the blue-green
algae possess absorption spectra that can be distinguished from those of the
C 0
:;::; Cl.
0 (/) .0 ro
(I) >
� Q) IY
450 500 550 600
Wavelength (nm)
-.-water
-CDOM+Sediment
..,.... Blue-Green
-Green
650 700
Figure 4.3 Relative absorption curves for CDOM, water, green and blue green
algae. Absorption curves resampled to MODIS bands
51
green algae using the MODIS bands, largely due to absorption differences
observed in MODIS bands centered at 645 and 555 nm.
Knowing the total absorption for each image pixel and the individual
absorption spectra for the four selected endmembers, we then solved for the
concentrations of these endmembers using all MO DIS bands centered
between 468 and 678 nm and applying a non-negative linear least squares
technique implemented in MATLAB. This method was adopted to avoid
physically meaningless negative solutions (i.e., negative concentrations).
Solutions for each pixel were evaluated by calculating the root mean squared
error (RMSE) between the modeled total absorption values to the satellite
derived total absorption values. Those with an RMSE value greater than
0.08m-1 (approximately 20% of the aT at 645 nm in the western basin) were
considered invalid and rejected. The solution for each endmember is a
relative concentration of that endmember.
4.6 Results
Results (relative cyanobacterial concentrations) were re-projected to
UTM zone 17, and compared to along-track fluorometer results from six
cruises (four in Lake Erie and two in Lake Ontario) conducted in 2004-2005
(Table 4.1). We first demonstrate the correspondence between the patterns of
blue-green algae abundances extracted from the September 7-11 2005 cruises
52
and the relative MODIS-derived cyanobacterial concentrations. We then
extract a linear relationship between the FlouroProbe derived cyanobacterial
abundances and the MODIS derived cyanobacterial abundances and use it to
represent the relative MODIS cyanobacterial concentrations as
concentrations. The calibrated algae abundance data from the FluroProbe
used for this relationship is only available in the September 7-11 2005 cruise.
The horizontal profile extracted from the FluoroProbe fluorimeter
measurements from the September 2005 cruise was compared with the
satellite-based relative cyanobacteria concentrations extracted from the
temporally closest satellite image with good coverage of the lake (Fig. 4.4).
The scene was acquired on September 12, 2005, one day following the end of
the five day cruise. The figure displays data as a function of time (first 48
hours of the cruise) and covers the western basin and the first section of the
central basin. The selected sections of the cruise encompass the full range of
measured variations in FluoroProbe phycocyanin concentrations.
53
1.40 - l 1 /
15 1.20
� 2 13 �
1.00
fl
::i
11 u...
5 O.EO
E
....,
� o.w
■ Cv<1nubdckria Aburdance from MODIS
O Cv<1nubacleriil Aburdance Frum FluoruP·ut;e
•.)
"'
� ;:;
7 C
E
� 0.40
0 "'
-�
,.,,
,:i
C
x
F �
3 C
::, 0.70 1
-1
0 1:J 15 20 25 30 35 40 45 Time of aquisition (hours f-om start of cruise)
Figure 4.4 Cyanobacterial abundance from FluoroProbe profile (open blue circles)
plotted vs. time since cruise start. Corresponding MODIS derived cyanobacterial
concentrations extracted for each the FluoroProbe location from the September
12, 2005 image are shown with red squares.
Inspection of figure 4.4 shows the agreement between the MODIS
derived relative cyanobacteria concentrations and the fluorometer
concentrations for the September 2005 bloom events. This correlation is also
observed on a plot of cyanobacterial abundance from FluoroProbe vs. MODIS
derived relative cyanobacterial concentrations (Fig 4.5). A correlation coefficient of
R2: 0.658 was observed despite the differences in acquisition time between
the two sets (up to 5 days). Variations are to be expected between the field
and satellite-based algal distributions given that blooms will have moved
54
1.0
0.9
0.8
CJ) 0.7
/{= 0.032x 1- 0.084
+---------------.1�-----/--�-�·�'•8�---•
0.6
0.5
0.4
0.3
0.2 ,!;;
.E 0.1
0.0
0 5 10 15 20 25 30
Cyanobacterial Abundance from Flouroprobe (as JL9/L of chlorophyll)
Figure 4.5 Cyanobacterial abundance from FluoroProbe from September 2005
IFYLE cruise vs satellite-derived relative cyanobacterial concentrations from the
MODIS scene of September 12, 2005.
between the time the satellite and the FluoroProbe measurements were made.
The linear relationship was used to represent the relative MODIS
cyanobacterial concentrations in units of µg/L. The results are displayed in
figure 4.6.
In this figure the MODIS-derived cyanobacterial concentrations for the entire
lake are expressed in the same units as the fluorometer derived value, in
units of µg/L as chlorophyll-a. Inspection of figure 4.6 shows the agreement
between the MODIS derived relative cyanobacteria concentrations and the
fluorometer concentrations for the September 2005 bloom events. Figure 4.4
demonstrates this correlation for a limited number of cruise line sections in
55
the western and central Basins, whereas Figure 4.6 shows that the
correlation applies to the remaining sections for the September 2005 cruise
as well.
82°
W 80°
W
0 12.525 50 75 100
Kilometers
42°
N
41°
N
Figure 4.6 Abundances of cyanobacterial blooms in Lake Erie during Sept
2005 derived from MODIS data. Overlain on satellite images are along track fluorescence data from temporally overlapping cruises.
MODIS derived relative cyanobacterial abundances were also
compared with phycocyanin fluorescence values from the remaining cruises
in the summer of 2004 and 2005 (Table 4.1). Because the fluorometer data
for these cruises were uncalibrated and reported in raw flouresence, we chose
to present our unmixing results as relative cyanobacterial abundances.
Another approach that we could have adopted is to apply the linear
relationship extracted from the September cruise data and report results in
units of µg/L as chlorophyll-a. However, such an approach could be inaccurate
given the reported variations in phycocyanin abundance (relative to algae
56
concentration) over the course of a season (Simis et al., 2007). Figure 4. 7
shows along track relative phycocyanin fluorescence data overlain on relative
cyanobacterial abundance derived from MODIS data. Again, the agreement
in bloom distribution between the two data sets is evident. Although not
shown, this is also true for the July 2004 sampling event, where the patterns
seen in the fluorometer data are evident in the satellite data. In one case, the
July 2005 cruise, the MODIS-derived concentrations showed a possible low
level bloom throughout the entire western basin of Lake Erie, whereas the
fluorescence data showed more variations in concentrations: higher towards
the mouth of the Maumee River, lower towards the north-central portion of
the western basin. Simis (2007) showed that the absorption spectra of
cyanobacteria can change over the course of a season as the amount of
82°
W 80°
W
0 12.525 50 75 100
KIiometers
Figure 4. 7 Relative abundances of cyanobacterial blooms in Lake Erie during August 2005 derived from MODIS data. Overlain on satellite images are along track fluorescence data from temporally overlapping cruises.
57
42°
N
41°
N
phycocyanin changes. This may provide one explanation for why the
technique worked well in August and September of 2004 and 2005, but not as
well in July of 2004.
Applications of our algorithms did not produce false positives when
compared with fluorometer data. No cyanobacterial phycocyanin-bearing
blooms were noted from the MODIS data acquired during two cruises over
Lake Ontario, consistent with the fluorometer measurements. The latter
showed negligible phycocyanin concentrations across the entire lake with one
exception, the bay of Quinte, a long and narrow bay (length: >100 km; width:
1-3 km on average) that cannot be resolved from the coarse MODIS data.
The developed algorithms were also successful in identifying
phycocyanin blooms in the western basin of Lake Erie that were reported by
other researchers. The MODIS-derived relative phycocyanin patterns
matched the field-based contoured cell count patterns reported by Dyble
(2008) (Fig. 4.8). Although not shown, the MODIS-derived phycocyanin
concentration patterns also matched those reported by Vincent (2004) for the
September 2002 bloom in the western basin, the blooms in August of 2003
(Rinta-Kanto et al., 2005), and others in 2006 (Bridgeman, personal
communication).
58
83°
W
42°
N
�4 _..:\_ Kilometers f
Figure 4.8 Relative abundances of cyanobacterial blooms in Lake Erie (August
2004) derived from MODIS data (left) and contoured map generated from cell
counts at 15 stations in the western basin of Lake Erie from Dyble (2008).
Despite the reported success identifying toxic algal blooms, one has to
be aware of the potential uncertainties that could be introduced given the
complexities of the investigated systems, the encountered technical
difficulties, and the adopted assumptions and approximations. These factors
could introduce uncertainties in the bloom patterns and concentrations as
attempts are made to extend the developed methodologies spatially (beyond
Lake Erie and Ontario) and temporally (2002 through 2006). Next, we
provide examples of such complexities, uncertainties, and approximations.
There are difficulties in correlating and calibrating MODIS data to
field data. Simultaneous acquisition of in situ and satellite data is not always
possible. Fluorometer measurements are acquired over a period of 4-6 days,
whereas the satellite data represent a single point in time. The sample size
for MO DIS (pixel size at nadir: 250 m to 1 km) exceeds that of the samples
acquired in the field. The sample depth differs as well, MODIS integrates
measurements across a water column of varying depth depending on the
59
concentration of the different absorbers and scatterers and the wavelength of
the light (Bukata et al., 1998) whereas field samples are generally acquired
at a particular depth (in this case 1.5 m below the surface for fluorometer
data). These difficulties are compounded by the fact that wind pattern can
significantly affect the movement of cyanobacterial blooms laterally
(Kanoshina et al., 2003) and the ability of cyanobacteria to regulate their
vertical position in the water column (Reynolds et al., 1987).
Compositionally distinct endmembers can display similar absorption
spectra as is the case with CDOM and sediments. In such cases, it is best to
represent the spectrally similar endmembers as a single endmember, making
it difficult to estimate the proportions of these two components. One or more
of the endmembers could experience natural variations, such as the
variations observed amongst individual cyanobacterial species making the
representation of such similar, yet spectrally variable population, by a single
spectrum, an oversimplification. There are conceivable potential spatial and
temporal variations in endmember compositions and associated spectral
signatures. One such potential variation was described above, the seasonal
variation in phycocyanin composition. Similarly, the CDOM and sediment
spectra can vary with the value of S varying from .007 to .018 nm-1 (refer to
equation 3), variations which can lead to significant spectral differences in
the shorter (412 nm and 443 nm) wavelengths. Locally, one or more
60
endmembers can go undetected resulting in fictitious solutions given the
unsatisfactory selection of endmembers for a particular location.
Unmixing of spectral components works best when extensive field data
are available for correlating and calibrating the sensor-derived component
concentrations, when all of the endmembers are well characterized spatially
and temporally and are spectrally orthogonal (distinct). The developed
methodologies could potentially be used to develop cost-effective practical
screening method for rapid detection of, and warning against, cyanobacterial
blooms in the lower Great Lakes.
61
CHAPTERS
RADAR INTERFEROMETRY
5.1 Overview
Radar interferometry is a technique that uses multiple radar images to
infer topography, and to detect subtle temporal changes in topographic relief.
It has been demonstrated that changes on the order of several cm/yr and as
small as 0.1 mm/yr (under optimum conditions) could be readily measured
using this technique (Massonnet and Feigl, 1998). This technique has been
used to map deformation and fault slip from earthquakes (Sandwell et al.,
2002), mine subsidence (Carnec and Delacourt, 2000), aquifer compaction
from pumping (Burbey, 2003), and landslides (Amelung and Day, 2002), as
well as seasonal changes in surface topography due to groundwater
extraction and recharge (Hoffmann et al., 2001). The ideal setting for
applying these techniques is in arid areas, where interferences from
vegetation and atmospheric effects are minimal. Over the past decade,
advances in radar interferometry applications and techniques have led to
successful applications in temperate regions and over a wider range of
climatic conditions and landscapes.
62
The Synthetic Aperture Radar (SAR) Interferometry technique exploits
the information contained in the phase of two or more acquired over the same
location. The technique makes use of the difference in phase (interferometric
phase) between two radar scenes to determine the exact differences in range
from the satellite, and subsequently determines the precise x, y, and z
location of the reflector, enabling the extraction of topography or subtle
changes in topography.
Figure 5.1 shows the basic configuration of a pair of images used in
repeat pass interferometry. p is the range to a target from the satellite
reference position, and p +b p is the range to the same target acquired in the
second pass. B is the baseline, or physical distance between the location of
the satellite in the first and second pass. 0 is the look angle, and a is the
angle between the baseline vector and the tangent plane. It is then possible
to define bp as a function of B, 0, a, p, and A, the wavelength of the radar
beam. bp is proportional to the phase difference component of the radar
return <j>, measured during the two radar acquisitions:
<j>=(4n/A)*bp (5.1)
The common terminology for the reference scene is the master scene, and the
repeat scene is the slave scene.
63
ref
erence
spheroid
Figure 5.1 Geometry of repeat pass interferometry (after Sandwell and Price, 1998).
The factors which contribute to· phase differences between two radar
scenes include topography, deformation, and atmospheric effects.
<j>= <l>':t'opo + $def + <l>atm+ <!>noise, (5.2)
The phase <j> is recorded cyclically from -n<<j><n, so there is by default
an ambiguity in determining p from <j>.
64
This process is described extensively by Gabriel and Massonnet
elsewhere (Gabriel et al., 1989; Massonnet and Feigl, 1998) .
There are three basic families of radar interferometry techniques
currently in use and under development. These are the basic 2-4 pass
differential INSAR (DINSAR) techniques, as well as two classes of multi
temporal techniques which use numbers of scenes ranging from tens to
hundreds. The multi-temporal techniques are expansions and refinements of
the basic 2-4 pass techniques. They repeat many of the same steps, and then
extract usable information from results which are ambiguous in the 2-4 pass
techniques.
The basis for all of these techniques is the generation of an
interferogram. To generate an interferogram, the two scenes need to be co
registered in radar-space. This means that the slave scene (or a subset
thereof) has to be co-registered to the master scene (or a subset of the master
scene). This processing is done in the Delft object-oriented radar
interferometric software (DORIS), a public domain radar interferometry
software application (Kampes et al., 2003). With DORIS, the orbits are used
to provide an initial estimate of the registration, and then the images are
iteratively correlated using the cross correlation amplitude of the radar
signal in individual subsets of the radar images. The radar images can be
filtered (optional) to improve the registration. The slave image is then re-
65
sampled to the master image. The interferogram is then calculated from the
co-registered images as the dot-product of the complex images (Fig. 6.2). This
step is repeated for every interferogram that is generated. Any interferogram
generated like this will have a phase component related to the curvature of
the earth's surface. The curvature is then calculated and removed before any
further processing is done.
"'#1 Phase (Band 1 :9-6·92·10·11 ·92·1nterfero c,nt) temp7
. �
Figure 5.2 Sample interferogram generated from a pair of radar images acquired on
9-2-92 and 10-31-92.
The second necessary component of these techniques is phase
unwrapping. In the above interferogram (Fig. 5.2), a repeating ripple pattern
is evident; the ripples trend from the lower left to the upper right. This is
caused by the phase ranging from -n<q><n, cyclically. In order for this to be
turned into a measure of range, the cycles have to be added together, so that
66
the phase numbers then extend from Oto ~20n, instead of the original cyclical
distribution (-n«j><n). This is called unwrapping, and is a major challenge in
interferometry. The method we use (snaphu) is described in full in Chane and
Zebker (2002). If correlation between scenes is low or if coherence in the
interferogram is low due to changes in the scene through time or increased
perpendicular baseline, then this step becomes almost impossible.
5.2 Two-Pass Method
In the two-pass INSAR technique an interferogram is generated from
two scenes which span a deformation event (the master is acquired before,
the slave after). <l>atm and <!>noise are assumed to be negligible. <)>Topo is calculated
from a digital elevation model (DEM) which has been registered with the
master scene and subtracted from <j> to yield <!>def. This method is fairly simple,
but it relies heavily on the availability of a high quality DEM and excellent
registration between the DEM and the master. Any error in the DEM or in
registration will cause ambiguities in detecting and mapping deformation.
5.3 Three-Pass Method
In three-pass interferometry, instead of a DEM being used, an
unwrapped interferogram is used to remove the <)>Topo component. Noise and
67
atmospheric contributions are again considered to be negligible. An
interferogram from a co-registered master and slave with a very small
temporal and spatial baseline (i.e., 1 day) is generated. An additional slave
image on the other side of the deformation (also with a small spatial baseline)
is co-registered to the same master, and an interferogram is generated. The
interferogram from the first pair is unwrapped, scaled to match the second
baseline, and re-wrapped. The second interferogram is subtracted from the
first, removing the topographic phase, leaving <!>def.
5.4 Four-Pass Method
In the four-pass interferometry, a master-slave pair with very small
temporal and spatial baselines is acquired before and after the deformation
event. Each slave is co-registered to the appropriate master. An
interferogram is generated for each pair. One interferogram is unwrapped
and re-sampled to match the radar coordinates of the other pair. It is then
scaled and re-wrapped. The second interferogram is subtracted from the first,
yielding <!>def. This method has several distinct advantages over the three-pass
method. The <!>noise due to de-correlation is significantly reduced. All four
scenes do not need to share the same small baseline range, but pairs can be
selected to minimize spatial and temporal baselines. This significantly
increases the detection of the resultant deformation. 68
5.5 Multi Temporal Methods
The multi-temporal methods generate a far higher number of
interferogram pairs, throughout a deformation event. By making a large
number of interferograms, and with educated assumptions about the nature
of the deformation (i.e. linear deformation) and of the atmospheric
contributions, the errors associated with the solution can be minimized. In
the Small Baseline Subset (SBAS) approach, patches of coherent data are
processed. In the Persistent/Permanent scatterer techniques, individual
objects (single rooftops, etc.) which are exceptionally good scatterers that stay
coherent over long times and spatial baselines are used instead.
The selection of the family of techniques to be used depends on data
quality, environment, deformation type, availability of scenes, and processing
time. The technique which we have focused on in the Nile Delta exercise is
the Persistent Scatterer techniques as outlined by Hooper (Hooper et al.,
2007; Hooper et al., 2004) and Kampes (Kampes, 2006). The following
summarizes the main characteristics of each of these techniques, as well as
the advantages and disadvantages of each method.
5.6 Summary
• Basic 2, 3 or 4 Pass DINSAR
o 2 to 4 scenes required
69
o Good coherence between scenes mandatory
o Atmospheric effects assumed negligible
o Two-pass
• Needs supplemental DEM information
• Subject to inaccuracies in DEM
• Pair needs to bracket deformation event
• DEM needs to be accurately radar coded
o Three-pass
• Subject to atmospheric effects
• One pair needs to be very close together in time (1 day) on
one side of the deformation event, the third scene has to
be on the other side of the event
• Difficult to maintain correlation over long times (years)
• Scenes registered to common master
o Four-pass
• 4 scenes, pairs of scenes from before and after the
deformation event
• Each pair needs small baseline, good correlation
• Allows for longer time for deformation to occur
• Scenes co-registered as pairs, pairs co-registered to each
other.
• Multi-temporal techniques:
o Small Baseline Techniques (SBAS)
• 10-20 scenes and more are used
• Assumes areas of good coherence in interferogram
• Entire scenes need not be coherent
• Scenes spatially resampled to one common scene, directly
or through cascading sequence (Refice et al., 2003)
• Only useful for gradual deformations (i.e., subsidence)
70
• Examples can be found in (Lanari et al., 2004a; Lanari et
al., 2004b)
• High number of interferograms generated
• Processing time intensive
o Permanent Scatterer Techniques
• >6 (Persistent Scatterer) to >40 (Permanent Scatterer)
scenes
• Good coherence at individual points
(Permanent/Persistent Scatterers)
• Deformation can be gradual (subsidence) or sharp
(faulting)
• Non-linear estimate of deformation
• Scenes spatially resampled to one common scene, directly
or through cascading sequence
• Very high number of interferograms generated
• Processing time intensive
• Examples include: (Ferretti et al., 2000, 2001; Hooper et
al., 2007; Hooper et al., 2004)
71
CHAPTER6
LAND SUBSIDENCE IN THE NILE DELTA: INFERENCES FROM RADAR INTERFEROMETRY
6.1 Abstract
The Nile Delta has formed by progression of a complex system of deltaic
fans throughout the Pleistocene, with the modern delta being formed from
sediments supplied by at least 10 distinct distributary channels throughout
the Holocene. With an average elevation of a meter or so above sea level and
a predicted rise in sea level of 1.8-5.9 mm/yr (IPCC, 2007) the subsidence of
the northern delta is becoming a topic of major concern both to the Egyptian
population and the government. The Nile Delta is home to more than 50
million people and the major agricultural production area for Egypt. We
evaluated the modern rates of subsidence of sections of the northeastern Nile
Delta (a total length of 110 km, up to 50 km from the coastline) using
persistent scatterer radar interferometry techniques applied to 14 ERS-1 and
ERS-2 scenes. The area covered includes the present active depocenter of the
Damietta promontories, and the nearby Mendesian depocenter that was
active up to recent times (up to 2500 years ago). The highest subsidence rates
( ~8 mm/yr; twice the average Holocene rates) do not correlate with the
distribution of the thickest Holocene sediments, but rather with the
72
distribution of the youngest depositional centers (major deposition occurred
between ~3500 years bp and present) at the terminus of the Damietta
branch. The adjacent slightly older (8000- 2500 years bp) Mendesian branch
depositional center is subsiding at slower rates of 2-6 mm/yr. Results are
interpreted to indicate that: (1) modern subsidence in the Delta is heavily
influenced by the compaction of the most recent sediments, and (2) the highly
threatened areas are at the terminus of the Damietta and possibly the
Rosetta branches, where active deposition is occurring.
6.2 Introduction
Deltas worldwide witness phases of progradation and destruction, phases
that are largely related to the interplay between fluvial processes, sea level
changes and coastal erosion at the mouth of the river. Deltas prograde when
fluvial processes dominate and retreat when coastal erosion intensifies. Many
of the world's large deltas are subsiding, in large part due to compaction and
isostatic response to loading by thick depositional sequences (Stanley et al.,
1996). Over the past 7000 years (Holocene) many deltas all over the world
have been witnessing an overall constructional phase. Recently (over the past
tens of years), there have been indications that this trend is being reversed in
a number of these deltas world-wide.
73
Modification of free-flowing rivers for energy generation or irrigation
projects can accelerate changes in delta plains and coasts including delta
subsidence due to the deprivation of the coastal systems of the silt and clay
which accumulate in deltas. Increased coastal erosion, and accelerated net
subsidence of deltas under the weight of the thick delta deposits combined
with rising sea level also exacerbates deltaic changes (Kay, 1993).
The Nile Delta (Fig. 6.1), the area of this study is witnessing a
30°
E
.,.. Alexandria
I .. ,,.Rosetta --
Branch Mahala
,..•.
0 25 50 75 100 N
-c:::=-K■il■o■mi=e =te=r=s=--• -i-
32°
E
31.5°
N
31°
N
30.5°
N
Figure 6.1 False color Landsat TM mosaic showing the Nile Delta, the Damietta and Rosetta branches, and urbanized centers (cities and villages) that appear in shades of purple. The thickest Holocene deposits are located north of the black dashed line (Stanley and Warne, 1993a). The red box
outlines the area covered in Figure 6.2. Inset (lower left corner) shows the location of Nile Delta, Nile River, and the Aswan High Dam.
74
destructional phase largely related to anthropogenic activities (damming,
water diversion, etc.) by the nations of the Nile Basin throughout the last
several centuries, with the most drastic changes occurring in the last 100
years. Egypt increased its water storage capacity by constructing dams (e.g.,
Aswan Dam, Aswan High Dam) and impounding water behind dams in
extensive artificial reservoirs (e.g., Lake Nasser: total capacity: 1.6 x 1011 m3)
(Said, 1993). Egypt is embarking on diversion projects (e.g., Toshka project,
El Salam Canal) to channel large volumes of Nile River water to reclaim
lands in the Western Desert and Sinai. As a result, water discharged at the
Rosetta and Damietta channels (Fig. 6.1) has been dramatically reduced
(1950s: 40 x 109 m3/yr; 2000s: < 5 x 109 m3/yr) (Frihy and Lawrence, 2004)
altering the balance that existed between sediment influx, sea level rise,
coastal erosion, and subsidence. For example, since 1972 the retreat of the
Damietta promontory has increased to alarming rates of up to 52 m/yr (Frihy
and Lawrence, 2004).
The pioneering work by Stanley, Warne, and co-workers (Krom et al.,
2002; Stanley and Warne, 1993b, 1998; Stanley, 2001; Stanley et al., 2004;
Warne and Stanley, 1993) on the subsidence of the Nile Delta throughout the
Holocene, and the accelerated subsidence of the area following the erection of
the Aswan High Dam and subsequent cut off of sediment supply, has received
a great deal of attention from the scientific community as well as extensive
media coverage. Eighty-seven radiocarbon-dated cores across the northern
75
delta were used to interpret the interactions between sea-level changes,
climatic oscillations, subsidence, and sediment transport. Average subsidence
rates (0.5 to 5 mm/yr for the past 7500 years) were calculated from
radiocarbon dated sediments from wells in the northern delta plain.
There are several uncertainties associated with using the average
Holocene subsidence rates to infer precise estimates of modern subsidence:
(1) subsidence rates for deltas change with time (Meckel et al., 2007) and
thus, average Holocene values do not necessarily represent shorter temporal
scales (modern) rates, (2) spatial heterogeneities in subsidence rates might
not be adequately captured using a limited number of samples collected from
widely spaced wells drilled in the northern delta, (3) uncertainties associated
with the 14C ages may be large as the position of dated samples in the
stratigraphic sequence may have been altered in many cases resulting in
underestimating the subsidence rates calculated from these samples (Said,
1993; Stanley and Hait, 2000; Stanley, 2001). Persistent scatterer radar
interferometry techniques were applied over large sections of the
northeastern sections of the Delta to overcome these problems.
6.3 Geology of the Nile Delta
The modern Nile Delta sits on the Nile Cone, a depositional feature
resulting from more than 5 million years of discharge and deposition of 3500
76
m of sediment by the Nile and Paleonile drainage systems (Sestini, 1989).
The Pleistocene section consists of up to 800 m of deltaic sands with minor
clay layers (RIGW, 1992). During the late Pleistocene, from roughly 35 to 18
ka, the region of the modern Nile Delta was a seasonally active alluvial plain
with braided stream channels. This was followed between 15 and 8 ka by a
period of rapid sea level rise, which reworked much of these sediments. Some
8000 years ago, a slowing down in sea level rise prompted the development of
the modern Nile Delta (Said, 1990; 1993). The Delta prograded up to 10 m/yr
into the Mediterranean through accretion of silts and clays (1-7mm/yr) giving
rise to thicknesses of up to 60 m of Holocene deltaic fluvial/marine deposits
(Stanley and Warne, 1993a).
Throughout the Holocene the centers of deposition continued to shift
from one spot to another along the delta coastline depending on which
channel(s) were active at any particular time (Stanley et al., 2004). For
example, from 8000 years ago, the eastern sections of the Nile delta were
building out along the Tanitic and Mendesian branches (Fig. 6.2a). The
Tanitic branch deposited material underneath what is now the Manzala
Lagoon, and prograded across to present day Port Said (Fig. 6.2). The
Mendesian branch formed a depositional center under the western portion of
the Manzala Lagoon and prograded to the northeast to merge with the
depositional center of the Tani tic branch. The supply of sediments by these
two branches gave rise to up to 40 m thick sections of Holocene deposits,
77
currently underlying the Manzala Lagoon. Around 1500 to 2500 years ago
these two branches silted up allowing the Damietta branch to become the
largest contributor of Nile sediment flux (Stanley and Warne, 1993a).
In addition to natural phenomena, and as early as 5000 bp, human
intervention has played an increasing role in modifying the natural flow
system and flood patterns. By the middle of the 19th century, earlier smaller
individual projects (e.g., raised levees) quickly gave way to larger irrigation
canal systems to irrigate large sectors of the Nile Delta (Sestini, 1989). Up to
that time, the annual floods still delivered considerable silt to the delta, a
situation that rapidly changed beginning in the late 19th century and
continued throughout the 20th century. Major engineering projects were
implemented including dams and barrages culminating in the construction of
the Aswan High Dam in 1964. Prior to the construction of the latter,
approximately 9.5x106 metric tons of sediment were deposited onto the Nile
Delta floodplain each year, an amount equivalent to a ~ 1 mm layer of silt
across the Delta. This added thickness partially balanced continued
subsidence of the Delta. Following the construction of the Aswan High Dam,
the sediment load no longer reaches the delta, instead it settles in Lake
Nasser behind the dam (Said, 1993).
78
6.4 Radar Interferometry
Radar interferometry has been successfully used to map small (as
small as 0.1 mm/yr) (Massonnet and Feigl, 1998) deformation and fault slip
from earthquakes (Sandwell et al., 2002), mine subsidence (Carnec and
Delacourt, 2000), aquifer compaction from pumping (Burbey, 2003), and
landslides (Amelung and Day, 2002), as well as seasonal changes due to
groundwater extraction and recharge (Hoffmann et al., 2001).
Interferometric synthetic aperture radar (InSAR) processing makes
use of the difference in phase between two radar scenes to determine precise
differences in range to a target and to subsequently determine the exact
surface location, and subtle changes in topography. In conventional three
pass InSAR interferometry, the deformation is calculated from the difference
between the unwrapped phase of a reference interferogram (no deformation
between scene acquisitions) and a pair spanning the deformation (Massonnet,
1996). Successful implementation of this method relies on maintaining a high
correlation for pairs of scenes over the time period of the deformation. If
changes in scatterers occur between passes, such as those introduced by
growing vegetation, phase will vary randomly and the process of extracting
phase differences indicative of subsidence will be jeopardized. Because the
Nile Delta is highly vegetated, and the variations we are examining are long
term variations compared to those introduced by growing vegetation, the
79
conventional three-pass interferometry technique was inadequate for the
study area (Aly et al., 2005).
We took advantage of recent refinements in persistent scatterer radar
interferometry techniques (Hooper et al., 2007; Hooper et al., 2004) that are
now enabling successful applications over a wider-range of physiographic and
atmospheric conditions. To date, applications of the related permanent
scatterer techniques in the Nile Delta have been restricted to two major
cities, Mahala and Mansura, in central Nile Delta (Fig. 6.1) (Aly et al., 2005).
We applied the persistent scatterer technique (Hooper et al., 2007; Hooper et
al., 2004) to investigate the spatial variations in subsidence rates across large
sections (area: ~2400 km2) of the heavily vegetated northeastern delta (Fig.
6.2). The adopted persistent scatterer method generates multiple three-pass
interferograms, but restricts the phase unwrapping and analysis to pixels
containing individual scatterers which dominate individual pixels, and
remain stable over the time period of interest. Because the signal from these
scatterers is much larger than the random noise from a large number of
small scatterers, the variance in phase in these pixels reflects the underlying
deformation (Ferretti et al., 2000; Hooper et al., 2004). In the northern delta,
these scatterers include buildings, individual outcrops, rocks, or utility poles.
In applying this technique, we took advantage of the extensive distribution of
small buildings and scattered dwellings, present largely in villages and cities
(areas appearing in shades of violet in Fig. 6.1) across the delta. The Stanford
80
Method for Persistent Scatterers (StAMPS) algorithm was adopted; this
method uses the spatial correlation of the interferogram phase to identify
such pixels (Hooper et al., 2007). For the majority of the investigated area,
the density of the permanent scatters ranged from 4 to 30 permanent
scatterers/km2 enabling the estimation of atmospheric contributions
(Colesanti et al., 2003).
Synthetic aperture radar (SAR) scenes covering the period 1992-1999
were acquired from the European Remote Sensing satellites (ERS-1 and
ERS-2; Table 6.1). In the selection of scenes, attempts were made to pick
scenes with minimal differences in their spatial baselines (perpendicular
baseline <500m) and which are temporally as evenly distributed as possible
throughout the investigated period. Fourteen scenes in track 436 were
processed, and mean line of sight velocities were first calculated and then
converted to equivalent vertical motions.
ERS precise orbit information was obtained from Delft Institute for Earth
Oriented Space Research (Scharroo and Visser, 1998). All of the
interferometric processing was conducted using: (1) the Delft Object-oriented
Radar Interferometric Software (DORIS) (Kampes et al., 2003), (2) the
Repeat Orbit Interferometry Package (ROI_PAC) from JPL, and (3) StAMPS
(Hooper et al., 2004).
81
Temporal Acquisition Perpendicular baseline Date Satellite Baseline (m) (days) Track
19920601 ERS-1 331 -1179 436
19930517 ERS-1 60 -829 436
19931004 ERS-1 77 -689 436
19931108 ERS-1 300 -654 436
19950720 ERS-2 -35 169 436
19950823 ERS-1 100 -1 436
19950824 ERS-2 0 0 436
19960111 ERS-2 129 140 436
19960530 ERS-2 -341 280 436
19970515 ERS-2 150 630 436
19980709 ERS-2 -333 1050 436
19980813 ERS-2 186 1085 436
19990311 ERS-2 64 1295 436
19990520 ERS-2 -61 1365 436
19991007 ERS-2 -190 1505 436
Table 6.1 ERS-1 and ERS-2 scenes used in permanent scatterers study. Temporal and perpendicular baselines are shown relative to the scene acquired on 8/24/1995.
6.5 Discussion and Results
Results of permanent scatterer analysis applied to the 13 ascending
scenes are shown in figure 6.2a. A large number (>10,000) of persistent
scatterers were identified and used to constrain modern subsidence rates in
the study area. Subsidence rates represented on this figure are relative to the
average motion of the permanent scatterers in the southwestern portion of
82
31.5°
E
Total Holocene Thickness (After Stanley, 1996)
• • •
Kilorneters -- -
0 4 8 12 16
Modern Relatlve Motion (mm/yr) Persistent Scatterer
■ ■ D
32°
E
Average Holocene Subsidence (After Stanley, 1993a)
Elevation (m)
SRTM 3 Arc Second
31.5°
N
31°
N
Figure 6.2 (a) Mean velocities extracted from 14 ascending ERS-1 and ERS-2 scenes acquired over the period 1992-1999 and expressed as vertical motion of permanent scatterers (negative velocities indicate downward motion). Areas of highest modern subsidence include areas at the terminus of the Damietta branch that are underlain by significant (>5 m) amounts of young (<3500 years) sediments (areas shaded in white). Also shown in white solid lines are the current Damietta branch, the paleo Mendesian and Tanitic branches, and shown in black dashed lines, the paleo-coastlines positions at 4000, 3500, 2000 and 1500 years bp (Coutellier and Stanley, 1987; Stanley and
Warne, 1993a). (b) Average Holocene subsidence (Stanley and Warne, 1993a), and Holocene thicknesses from individual borings (colored squares) (Stanley et al., 1996). High modern subsidence rates correlate with the distribution of younger (<3500 yr) Holocene sediments rather than thick Holocene sediments. Radar data provided by the European Space Agency.
83
the study area, an area that experienced minimal subsidence (<0.5mm/yr)
(Stanley and Warne, 1993a) throughout the Holocene. Previous estimates of
average Holocene subsidence rates were based on data from 87 wells across
the northern delta, 22 of which were drilled in the study area (Fig. 6.2b). The
subsidence rates that we report and their spatial distribution differ
considerably from those previously reported for the Holocene (Emery et al.,
1988; Stanley and Warne, 1993a). Modern rates are generally high (up to 8
mm/yr) compared to previously calculated average Holocene rates (0.5 to 4.5
mm/yr). The highest modern rates are observed over the youngest (<3500
year old) sediments under the terminus of the Damietta branch (Fig. 6.2a).
Moderate rates (4-6 mm/yr) are calculated around the Manzala Lagoon (Fig.
6.2), a depocenter that was active more than 3500 years ago.
The highest subsidence rates were calculated at the terminus of the
Damietta branch suggesting a causal relationship between rapid subsidence
and the distribution of young sediments that were probably deposited during
the recent progradation of the Damietta branch and have been undergoing
compaction and isostatic subsidence since then. This hypothesis is supported
by the general spatial correspondence between rapidly subsiding areas and
those underlain by thick, young ( <3500 years old) sediments. Figure 6.2
shows a correspondence between the area subtended by a contour defining
domains of 5 or more meter thick young deposits and the areas undergoing
rapid subsidence (clusters of orange, red, and yellow dots on Fig. 6.2a). The
84
contour was defined using data extracted from radiocarbon dated cores and
sediment thicknesses from 22 drilled wells in the study area (Stanley et al.,
1996). Only cores where the 14C age sequence increases sequentially with
depth were used. The locations of these cores are represented as squares color
coded to total Holocene depth in figure 6.2b.
The proposed hypothesis is consistent with the documented evolution of
the depositional environments over the past 3500 years in the northeast
delta. The Damietta branch has prograded some 40 km to the northeast over
the past 3500 years as the coastline has migrated to the north (Fig. 6.2a). The
highest modern subsidence rates (6-8 mm/yr) correspond to areas of this most
recent deposition along the Damietta branch between the paloeshoreline of
3500 years bp (Coutellier and Stanley, 1987) and the modern shoreline (Fig.
6.2a).
The proposed hypothesis is also consistent with the documented evolution
of the depositional environments for the time period preceding the past 3500
years. If modern subsidence is largely controlled by compaction of younger
sediments, one should expect to observe moderate subsidence rates in areas
of slightly older (>3500 years old) sediments that were transported by the
earlier Mendesian and Tanitic branches and deposited in the area occupied
by Manzala Lagoon. In these areas, only a thin cover of younger sediments
(1-4 meters) were deposited by the earlier Mendesian and Tanitic branches
85
(Stanley et al., 1996). Our results indicate that this is indeed the case;
moderate (2-6 mm/yr) subsidence rates are observed in the area occupied by
the Mendesian depocenter (area outlined by dashed white line: Fig. 6.2a).
Unfortunately the examined radar scene does not cover the Tanitic
depocenter. We suggest that these large thicknesses of the Holocene
sediments in the Manzala Lagoon must have witnessed early periods of rapid
compaction and subsidence followed by a progressive decline in subsidence
rates with time as sediments became less prone to compaction. Such an
interpretation could explain why the highest average Holocene rates are
reported for areas with the thickest Holocene sediment in the Manzala
Lagoon (Stanley and Warne, 1993a), whereas the modern subsidence rates
we report for these areas are moderate.
It is unlikely that the observed modern subsidence was caused by tectonic
movements, as there have not been any reported earthquakes (ISC, 2001)
along the previously identified faults (Frihy, 2003; Stanley, 1988) in the area
or in any other location within the study area throughout the investigated
period (1992-1999). Our findings do not preclude seismic activities occurring
on larger timescales. The lack of any recent (1992-1996) seismic activities
also argues against the observed displacements being related to horizontal
motion along faults. The observed modern displacements are interpreted here
to be vertical in nature and are largely attributed to compaction and isostatic
adjustment.
86
Our findings in the Nile Delta are consistent with those reported from
deltas elsewhere. Modern subsidence rates in the Mississippi are interpreted
to indicate that compaction of sediments constitutes a substantial source of
subsidence in deltaic systems, especially during the earlier periods following
deposition, and that average Holocene subsidence rates do not necessarily
predict current subsidence rates. Compaction rates in the Mississippi Delta
system within the first tens of years to centuries can range from 5 mm/yr to
values as high as 10 mm/yr in organic rich peat systems, considerably higher
than the estimated average Holocene subsidence rates of <3 mm/yr
(Tornqvist et al., 2008). Mekel et al. (2007) modeled the compaction of
Holocene sediments in the Mississippi Delta; his simulations predicted rapid
compaction of deposits shortly after deposition, with progressive decline in
compaction with increasing age of sediments.
In the northern delta, it has been assumed that the areas underlain by
the highest sediment thickness are currently subsiding the fastest, and would
be impacted most severely by projected sea level rises. This study shows that
the sediment age rather than the thickness might be the dominant factor
controlling subsidence in the Nile Delta. Results indicate that areas around
the city and harbor of Damietta may be the most vulnerable to inundation.
The conditions described for the Nile Delta are typical of many of the world's
deltas.
87
CHAPTER7
CONCLUSIONS
In this manuscript, I demonstrated methodologies and applications
addressing two specific problems pertaining to mapping the abundance of
algal blooms in the Great Lakes, and land subsidence in the Nile Delta.
Specifically, I investigated methodologies to temporally and spatially
investigate the extent and distribution of algal blooms in Lake Erie and Lake
Ontario, and proposed a method for discriminating between potentially toxic
and non-toxic blooms. Visible and near infra red MO DIS and Sea WiFS
satellite data were used to map the distribution of algal blooms in these
lakes. Existing algorithms for mapping blooms were applied and results
were compared. Moreover, new methodologies were developed to distinguish
between the non-toxic blooms and the potentially toxic cyanobacterial blooms.
Cyanobacterial concentrations that were derived using my algorithms were
successfully tested against in situ measurements.
These results show great promise for establishing cost-effective,
remotely-based monitoring systems which could be used by various
municipalities to secure water intake systems and the recreational areas in
Lakes Erie and Ontario. To achieve this goal, we now have in place an
automated system which produces chlorophyll images (using case II Carder
algorithms applied to MODIS data) and posts the results on the web
(http://www.esrs.wmich.edu/MERHAB-LGL) for all interested parties
Future steps will include automation of the cyanobacterial
88
identification method to provide rapid access to results and to make these
available to scientists and city managers. Better refinement of the spectral
properties of the endmembers is also planned. Currently, Microcystis is the
only cyanobacteria included in the model. Plans are underway to better
represent the known variations in toxic species in the lakes, including
Planktothrix, and Anabaena. The model should be also tested and refined in
other locations in the Great Lakes (e.g., Saginaw Bay) and in other settings
where cyanobacterial blooms were reported (e.g., Baltic Sea).
In the second study I examined subsidence in the northeastern portion
of the Nile Delta. Because of its low average elevation, and a predicted rise in
sea level of 1.8-5.9 mm/year (IPCC, 2007) the potential for inundation of this
region of the delta is high. This provides a significant concern to the Egyptian
population and government. I calculated modern subsidence rates in the
northeastern Nile Delta using persistent scatterer radar interferometry
techniques, and found that the modern subsidence rates are as high as 8
mm/year. The highest rates are twice the average Holocene rates, and
correlate with the distribution of the youngest deposits. Slower rates of 2-6
mm/yr. were found under the older depositional centers. These results were
interpreted to indicate that the modern subsidence in the Delta is heavily
influenced by the compaction of the most recent sediments, and that the
highly threatened areas are at the terminus of the Damietta branch, where
the most recent deposition has occurred.
This information about the ongoing modern distribution of subsidence
rates is useful for Egyptian officials in the Damietta region. Prior to my work
it had been assumed that the maximum subsidence rates were about 5
89
mm/yr, approximately half of the maximum rate we report in this study.
Moreover, the previous studies predicted that the highest subsidence rates
were underneath the Manzala Lagoon, whereas our study shows that it is
around the Damietta region. Naturally, the high subsidence rates under
Damietta city and its surroundings raises the concern of inundation in this
area.
Future work in this area should include verification of these
calculations. Geodetic stations could be established in or near Damietta city
and the harbor. The radar interferometry study should be extended both
spatially to cover larger portions of the northern Nile Delta region, and
temporally, to include data later than 1999. Future studies could take
advantage of the L-band radar instrumentation on the ALOS satellite. The
later provides better penetration of vegetation canopy, and could potentially
provide more coherent applications compared to those based on the C band of
ERS-1 and ERS-2. As longer acquisition periods of radar data are achieved
better precision in results will be accomplished.
90
REFERENCES
Abalde, J., Betanocourt, L., Torres, E., Cid, A., and Barwell, C., 1998,
Purification and characterization of phycocyanin from the marine
cyanobacterium Synechococcus: Plant Science, v. 16, p. 109-120.
Aly, M.H., Giardino, J.R., and Klein, A.G., 2005,-Local Subsidence in Greater
Mahala, Egypt, Detected by Radar Interferometry: Eos Trans. AGU,
Fall Meet. Suppl., v. 86.
Amelung, F., and Day, S., 2002, InSAR observations of the 1995 Fogo, Cape
Verde, eruption: Implications for the effects of collapse events upon
island volcanoes: Geophysical Research Letters, v. 29.
Anderson, D.M., Glibert, P.M., and Burkholder, J.M., 2002, Harmful algal
blooms and eutrophication: Nutrient sources, composition, and
consequences: Estuaries, v. 25, p. 704-726.
Aydar, E., Gourgaud, A., Ulusoy, I., Digonnet, F., Labazuy, P., Sen, E.,
Bayhan, H., Kurttas, T., and Tolluoglu, A.U., 2003, Morphological
analysis of active Mount Nemrut stratovolcano, eastern Turkey:
evidences and possible impact areas of future eruption: Journal of
Volcanology and Geothermal Research, v. 123, p. 301-312.
Becker, R., Sultan, M., Atkinson, J., Boyer, G.L., and Konopko, E., 2005,
Spatial and Temporal Variations of Algal Blooms in the Lower Great
91
Lakes, 8th International Conference on Remote Sensing for Marine
and Coastal Environment: Halifax, Nova Scotia, Altarim.
Becker, R., Sultan, M., Boyer, G., and Konopko, E., 2006, Mapping Variations
Of Algal Blooms In The Lower Great Lakes: Great Lakes Research
Reviews, v. 7, p. 14-17.
Belward, A.S., Stibig, H.J., Eva, H., Rembold, F., Bucha, T., Hartley, A.,
Beuchle, R., Khudhairy, D., Michielon, M., and Mollicone, D., 2007,
Mapping severe damage to land cover following the 2004 Indian Ocean
tsunami using moderate spatial resolution satellite imagery:
International Journal of Remote Sensing, v. 28, p. 2977-2994.
Berardino, P., Costantini, M., Franceschetti, G., Iodice, A., Pietranera, L.,
and Rizzo, V., 2003, Use of differential SAR interferometry in
monitoring and modelling large slope instability at Maratea
(Basilicata, Italy): Engineering Geology, v. 68, p. 31-51.
Beutler, M., Wilstshire, K.H., Arp, M., Kruse, J., Reineke, C., Moldaenke, C.,
and Hansen, U.-P., 2003, A reduced model of the fluorescence from the
cyanobacterial photosynthetic apparatus designed for the in situ
detection of cyanobacteria: Bioch. Bioph. Acta v. 1604, p. 33-46.
Beutler, M., Wilstshire, K.H., Meyer, B., Moldaenke, C., Li.iring, C.,
Meyerhofer, M., Hansen, U.-P., and Dau, H., 2002, A fluorometric
method for the differentiation of algal populations in vivo and in situ:
Photosyn. Res, v. 72, p. 39-53.
92
Binding, C.E., Jerome, J.H., Bukata, R.P., and Booty, W.G., 2008, Spectral
absorption properties of dissolved and particulate matter in Lake Erie:
Remote Sensing of Environment, v. 112, p. 1702-1711.
Boyer, G., 2006, Toxic Cyanobacteria in the Great Lakes: More than Just the
Western Basin of Lake Erie: Great Lakes Research Reviews, v. 7, p. 2-
7.
Boyer, G.L., Satchwell, M.F., Hotto, Y., A., X., Lewis, T., Oles, M.,
Makarewicz, J.C., and Watson, S.B., 2004, Occurrence of
Cyanobacterial Toxins in Lake Ontario, USA, 6th Internat Conf. Toxic
Cyanobacteria: Bergen Norway.
Brittain, S.M., Wang, J., Babcock-Jackson, L., Carmichael, W.W., Rinehart,
K.L., and Culver, D.A., 2000, Isolation and characterization of
microcystins, cyclic heptapeptide hepatotoxins from a Lake Erie stain
of Microcystis aeruginosa: J. Great Lakes Res, v. 26, p. 241-249.
Bryant, D.A., 1981, The Photoregulated Expression of Multiple Phycocyanin
Species - A General Mechanism for the Control of Phycocyanin
Synthesis in Chromatically Adapting Cyanobacteria: European
Journal of Biochemistry, v. 119, p. 425-429.
Bukata, R.P., Jerome, J.H., Bruton, J.E., Jain, S.C., and Zwick, H.H., 1981,
Optical Water-Quality Model of Lake-Ontario .1. Determination of the
Optical-Cross-Sections of Organic and Inorganic Particulates in Lake
Ontario: Applied Optics, v. 20, p. 1696-1703.
93
Bukata, R.P., Jerome, J.H., Kondratyev, K.Y., and Pozdnyakov, D., 1998,
Optical properties and remote sensing of inland and coastal waters:
Boca Raton Fla, C. R. C. Press.
Burbey, T.J., 2003, Use of time-subsidence data during pumping to
characterize specific storage and hydraulic conductivity of semi
confining units: Journal of Hydrology, v. 281, p. 3-22.
Campbell, J., Antoine, D., Armstrong, R., Arrigo, K., Balch, W., Barber, R.,
Behrenfeld, M., Bidigare, R., Bishop, J., Carr, M.E., Esaias, W.,
Falkowski, P., Hoepffner, N., Iverson, R., Kiefer, D., Lohrenz, S.,
Marra, J., Morel, A., Ryan, J., Vedernikov, V., Waters, K., Yentsch, C.,
and Yoder, J., 2002, Comparison of algorithms for estimating ocean
primary production from surface chlorophyll, temperature, and
irradiance: Global Biogeochemical Cycles, v. 16.
Carder, K.L., Chen, F.R., Lee, Z.P., Hawes, S.K., and Kamykowski, D., 1999,
Semianalytic Moderate-Resolution Imaging Spectrometer algorithms
for chlorophyll a and absorption with bio-optical domains based on
nitrate-depletion temperatures: Journal of Geophysical Research
Oceans, v. 104, p. 5403-5421.
Carnec, C., and Delacourt, C., 2000, Three years of mining subsidence
monitored by SAR interferometry, near Gardanne, France: Journal of
Applied Geophysics, v. 43, p. 43-54.
94
Chen, C.W., and Zebker, H.A., 2002, Phase unwrapping for large SAR
interferograms: Statistical segmentation and generalized network
models: IEEE Transactions on Geoscience and Remote Sensing, v. 40,
p. 1709-1719.
Clarke, G.L., Ewing, G.C., and Lorenzen, C.J., 1970, Spectra of backscattered
light from the sea from aircraft as a measure of chlorophyll
concentration.: Science, v. 167, p. 1119-1121.
Colesanti, C., Ferretti, A., Novali, F., Prati, C., and Rocca, F., 2003, SAR
monitoring of progressive and seasonal ground deformation using the
permanent scatterers technique: IEEE Transactions on Geoscience and
Remote Sensing, v. 41, p. 1685-1701.
Coutellier, V., and Stanley, D.J., 1987, Late Quaternary Stratigraphy and
Paleogeography of the Eastern Nile Delta, Egypt: Marine Geology, v.
77, p. 257-275.
Curran, P.J., Dash, J., and Llewellyn, G.M., 2007, Indian Ocean tsunami:
The use of MERIS (MTCI) data to infer salt stress in coastal
vegetation: International Journal of Remote Sensing, v. 28, p. 729-735.
Davies-Colley, R., Vant, W.N., and Smith, D.G., 2003, Colour and Clarity of
Natural Waters: Science and Management of Optical Water Quality:
Caldwell, NJ, Blackburn Press.
95
Dekker, A.G., Malthus, T.J., and Seyhan, E., 1991, Quantitative modeling of
inland water-quality for highresolution mss systems: IEEE
Transactions on Geoscience and Remote Sensing, v. 29, p. 89-95.
Downing, J.A., Watson, S.B., and McCauley, E., 2001, Predicting
cyanobacteria dominance in lakes: Canadian Journal of Fisheries and
Aquatic Sciences, v. 58, p. 1905-1908.
Dyble, J., Fahnenstiel, G.L., Litaker, R.W., Millie, D.F., and Tester, P.A.,
2008, Microcystin concentrations and genetic diversity of Microcystis in
the lower Great Lakes: Environmental Toxicology, v. 9999, p. NA.
Emery, K.O., Aubrey, D.G., and Goldsmith, V., 1988, Coastal Neo-Tectonics
of the Mediterranean from Tide-Gauge Records: Marine Geology, v. 81,
p. 41-52.
Fay, P., 1983, The Blue-Greens: London, Arnold.
Ferretti, A., Prati, C., and Rocca, F., 2000, Nonlinear subsidence rate
estimation using permanent scatterers in differential SAR
interferometry: Ieee Transactions on Geoscience and Remote Sensing,
v. 38, p. 2202-2212.
-, 2001, Permanent scatterers in SAR interferometry: IEEE Transactions on
Geoscience and Remote Sensing, v. 39, p. 8-20.
Frihy, 0., 2003, The Nile Delta-Alexandria Coast: Vulnerability to Sea-Level
Rise, Consequences and Adaptation: Mitigation and Adaptation
Studies for Global Change, v. 8, p. 115-138.
96
Frihy, 0., and Lawrence, D., 2004, Evolution of the modern Nile delta
promontories: development of accretional features during shoreline
retreat: Environmental Geology, v. 46, p. 914-931.
Fu, L.L., and Smith, R.D., 1996, Global ocean circulation from satellite
altimetry and high-resolution computer simulation: Bulletin of the
American Meteorological Society, v. 77, p. 2625-2636.
Gabriel, A.K., Goldstein, R.M., and Zebker, H.A., 1989, Mapping Small
Elevation Changes over Large Areas - Differential Radar
Interferometry: Journal of Geophysical Research-Solid Earth and
Planets, v. 94, p. 9183-9191.
Gons, H.J., Kromkamp, J., Rijkeboer, M., and Schofield, 0., 1992,
Characterization of the Light-Field in Laboratory Scale Enclosures of
Eutrophic Lake Water (Lake Loosdrecht, the Netherlands):
Hydrobiologia, v. 238, p. 99-109.
Gordon, H.R., 1997, Atmospheric correction of ocean color imagery in the
Earth Observing System era: Journal of Geophysical Research
Atmospheres, v. 102, p. 17081-17106.
Gordon, H.R., Brown, O.B., and Jacobs, M.M., 1975, Computed relationships
between the inherent apparent optical properties of a flat
homogeneous ocean: Applied Optics, v. 14, p. 417-427.
97
Gordon, H.R., and Clark, D.K., 1981, Clear Water Radiances for Atmospheric
Correction of Coastal Zone Color Scanner Imagery: Applied Optics, v.
20, p. 4175-4180.
Gordon, H.R., and Wang, M.H., 1994, Retrieval of Water-Leaving Radiance
and Aerosol Optical-Thickness over the Oceans with Seawifs - a
Preliminary Algorithm: Applied Optics, v. 33, p. 443-452.
Hallegraeff, G.M., 1993, A Review of Harmful Algal Blooms and their
Apparent Global Increase: Phycologia, v. 32, p. 79-99.
Havens, K.E., 2008, Cyanobacteria blooms: effects on aquatic ecosystems, in
Hudnell, H.K., ed., Cyanobacterial Harmful Algal Blooms: State of the
Science and Research Needs: Advances in Experimental Medicine and
Biology: New York, NY, Springer, p. 950.
Hilley, G.E., Burgmann, R., Ferretti, A., Novali, F., and Rocca, F., 2004,
Dynamics of slow-moving landslides from permanent scatterer
analysis: Science, v. 304, p. 1952-1955.
Hoffmann, J., Zebker, H.A., Galloway, D.L., and Amelung, F., 2001, Seasonal
subsidence and rebound in Las Vegas Valley, Nevada, observed by
synthetic aperture radar interferometry: Water Resources Research, v.
37, p. 1551-1566.
Hooker, S.B., and McClain, C.R., 2000, The calibration and validation of
SeaWiFS data: Progress in Oceanography, v. 45, p. 427-465.
98
Hooper, A., Segall, P., and Zebker, H., 2007, Persistent scatterer
interferometric synthetic aperture radar for crustal deformation
analysis, with application to Volcan Alcedo, Galapagos: Journal of
Geophysical Research-Solid Earth, v. 112.
Hooper, A., Zebker, H., Segall, P., and Kampes, B., 2004, A new method for
measuring deformation on volcanoes and other natural terrains using
InSAR persistent scatterers: Geophysical Research Letters, v. 31.
IGOS, 2003, IGOS Geohazards Theme Report.
IPCC, 2007, Climate Change 2007: The Physical Science Basis. Contribution
of Working Group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change in Solomon, S., Qin, D.,
Manning, M., Chen, Z., Marquis, M., Averyt, K.B., M.Tignor, and
Miller, H.L., eds.: Cambridge, United Kingdom and New York, NY,
USA., Cambridge University Press.
ISC, 2001, International Seismological Centre On-line Bulletin,
http://www.isc.ac.uk, Volume 2008: Thatcham, United Kingdom,
Internatl. Seis. Cent.
Jupp, D.L.B., Kirk, J.T.O., and Harris, G.P., 1994, Detection, Identification
and Mapping of Cyanobacteria - Using Remote-Sensing to Measure the
Optical-Quality of Turbid Inland Waters: Australian Journal of Marine
and Freshwater Research, v. 45, p. 801-828.
99
Kahru, M., and Mitchell, B.G., 1998, Spectral reflectance and absorption of a
massive red tide off southern California: Journal of Geophysical
Research-Oceans, v. 103, p. 21601-21609.
Kampes, B., 2006, Radar Interferometry: Persistent Scatterer Technique,
Springer.
Kampes, B.M., Hanssen, R., and Perski, Z., 2003, Radar Interferometry with
Public Domain Tools, Fringe 2003: Frascati, Italy.
Kanoshina, i., Lips, U., and Leppanen, J., 2003, The influence of weather
conditions (temperature and wind) on cyanobacterial bloom
development in the Gulf of Finland (Baltic Sea): Harmful Algae, v. 2, p.
29-41.
Kay, R., 1993, Deltas of the World: New York, American Society of Civil
Engineers, p. 138.
Kerle, N., Froger, J.L., Oppenheimer, C., and De Vries, B.V., 2003, Remote
sensing of the 1998 mudflow at Casi ta volcano, Nicaragua:
International Journal of Remote Sensing, v. 24, p. 4791-4816.
Kirk, J.T.O., 1994, Light & Photosynthesis in Aquatic Ecosystems:
Cambridge, Cambridge University Press.
Konopko, E., 2007, Development of a Flow-through Fluorometric System for
the Detection of Phycocyanin in the Lower Great Lakes: Syracuse, NY,
State University of New York College of Environmental Science and
Forestry.
100
Krom, M.D., Stanley, J.D., Cliff, R.A., and Woodward, J.C., 2002, Nile River
sediment fluctuations over the past 7000 yr and their key role in
sapropel development: Geology, v. 30, p. 71-74.
Kutser, T., Metsamaa, L., Strombeck, N., and Vahtmae, E., 2006a,
Monitoring cyanobacterial blooms by satellite remote sensing:
Estuarine Coastal and Shelf Science, v. 67, p. 303-312.
Kutser, T., Metsamaa, L., Vahtmae, E., and Strombeck, N., 2006b, Suitability
of MOD IS 250 m resolution band data for quantitative mapping of
cyanobacterial blooms: Proceedings of the Estonian Academy of
Sciences Biology Ecology, v. 55, p. 318-328.
Lanari, R., Mora, 0., Manunta, M., Mallorqui, J.J., Berardino, P., and
Sansosti, E., 2004a, A small-baseline approach for investigating
deformations on full-resolution differential SAR interferograms: IEEE
Transactions on Geoscience and Remote Sensing, v. 42, p. 1377-1386.
Lanari, R., Zeni, G., Manunta, M., Guarino, S., Berardino, P., and Sansosti,
E., 2004b, An integrated SAR/GIS approach for investigating urban
deformation phenomena: a case study of the city of Naples, Italy:
International Journal of Remote Sensing, v. 25, p. 2855-2862.
Land, P.E., and Haigh, J.D., 1996, Atmospheric correction over case 2 waters
with an iterative fitting algorithm: Applied Optics, v. 35, p. 5443-5451.
Lawson, C.L., and Hanson, R.J., 1974, Solving Least Squares Problems:
Englewood Cliffs, NJ, Prentice Hall.
101
Lee, Z.P., Carder, K.L., and Arnone, R.A., 2002, Deriving inherent optical
properties from water color: a multiband quasi-analytical algorithm for
optically deep waters: Applied Optics, v. 41, p. 5755-5772.
Lesht, B.M., Stroud, J.R., McCormick, M.J., Fahnenstiel, G.L., Stein, M.L.,
Welty, L.J., and Leshkevich, G.A., 2002, An event-driven
phytoplankton bloom in southern Lake Michigan observed by satellite:
Geophysical Research Letters, v. 29.
Lu, Z., and Danskin, W.R., 2001, InSAR analysis of natural recharge to
define structure of a ground-water basin, San Bernardino, California:
Geophysical Research Letters, v. 28, p. 2661-2664.
Lynch, P., 2008, The origins of computer weather prediction and climate
modeling: Journal of Computational Physics, v. 227, p. 3431-3444.
Makarewicz, J.C., Boyer, G.L., Guenther, W., Arnold, M., and Lewis., T.W.,
2006, The occurrence of cyanotoxins in the nearshore and coastal
embayments of Lake Ontario: Great Lakes Research Reviews, v. 7, p.
25-29.
Maritorena, S., Siegel, D.A., and Peterson, A.R., 2002, Optimization of a
semianalytical ocean color model for global-scale applications: Applied
Optics, v. 41, p. 2705-2714.
Martin, S., 2004, An introduction to Ocean Remote Sensing: Cambridge, UK,
Cambridge University Press, 426 p.
102
Mason, P.J., and Rosenbaum, M.S., 2002, Geohazard mapping for predicting
landslides: an example from the Langhe Hills in Piemonte, NW Italy:
Quarterly Journal of Engineering Geology and Hydrogeology, v. 35, p.
317-326.
Massonnet, D., 1996, Tracking the Earth's surface at the centimetre level: an
introduction to radar interferometry: Nature & Resources, v. 32, p. 20-
29.
Massonnet, D., and Feigl, K.L., 1998, Radar interferometry and its
application to changes in the earth's surface: Reviews of Geophysics, v.
36, p. 441-500.
Meckel, T.A., Ten Brink, U.S., and Williams, S.J., 2007, Sediment compaction
rates and subsidence in deltaic plains: numerical constraints and
stratigraphic influences: Basin Research, v. 19, p. 19-31.
Morel, A., and Prieur, L., 1977, Analysis of variations in ocean color:
Limnology and Oceanography, v. 22, p. 709-722.
Muller-Karger, F.E., Chuanmin, H., Andrefouet, S., Varela, R., and Thunell,
R., 2005, The Color of the Coastal Ocean and Applications in the
Solution of Research and Management Problems, in Miller, R.L., Del
Castillo, C., and McKee, B.A., eds., Remote Sensing of Coastal Aquatic
Environments, Volume 7: Remote Sensing and Digital Image
Processing: Dordrecht, The Netherlands, Springer, p. 345.
103
NASA, 2006, Earth Science Reference Handbook: A guide to NASA's Earth
Science Program and Earth Observing Satellite Mission, in Parkinson,
C.L., Ward, A., and King, M.D., eds.: Washington D.C., National
Aeronautics and Space Administration.
O'Reilly, J.E., Maritorena, S., Mitchell, B.G., Siegel, D.A., Carder, K.L.,
Garver, S.A., Kahru, M., and McClain, C., 1998, Ocean color
chlorophyll algorithms for SeaWiFS: Journal of Geophysical Research
Oceans, v. 103, p. 24937-24953.
O'Reilly, J.E., Maritorena, S., Siegel, D.A., O'Brien, M.C., Toole, D., Chavez,
F.P., Strutton, P., Cota, G.F., Hooker, S.B., McClain, C.R., Carder,
K.L., Muller-Karger, F., Harding, L., Magnuson, A., Phinney, D.,
Moore, G.F., Aiken, J., Arrigo, K.R., Letelier, R., and Culver, M., 2000,
Sea WiFS Postlaunch Calibration and Validation Analyses, Part 3.
Ostir, K., Veljanovski, T., Podobnikar, T., and Stancic, Z., 2003, Application
of satellite remote sensing in natural hazard management: the Mount
Mangart landslide case study: International Journal of Remote
Sensing, v. 24, p. 3983-4002.
Paerl, H.W., 2008, Nutrient and other environmental controls of harmful
cyanobacterial blooms along the freshwater-marine continuum, in
Hudnell, H.K., ed., Cyanobacterial Harmful Algal Blooms: State of the
Science and Research Needs: Advances in Experimental Medicine and
Biology: New York, NY, Springer, p. 950.
104
Parsons, T.T.a.J.D.H.S., 1963, Discussion of spectrophotometric
determination of marine-plant pigments, with revised equations for
ascertaining chlorophylls and carotenoids.: J. Mar. Res, v. 21, p. 155-
163.
Patt, F.S., Barnes, R.A., Eplee, R.E.J., Franz, B.A., Robinson, W.D., Feldman,
G., Bailey, S.W., Gales, P.J., Werdell, P.J., Wang, M., Frouin, R.,
Stumpf, R.P., Arnone, R., Gould, R.W., Martinolich, P.M.,
Ransibrahmanakul, V., O'Reilly, J.E., and Yoder, J., 2003, Algorithm
Updates for the Fourth SeaWiFS Data Reprocessing, NASA Tech.
Memo. 2003--206892, Volume 22, NASA Goddard Space Flight Center,
Greenbelt, Maryland, p. 7 4.
Pegau, W.S., Zaneveld, J.R.V., Barnard, A.H., Maske, H., Alvarez-Borrego, S.,
Lara-Lara, R., and Cervantes-Duarte, R., 1999, Inherent optical
properties in the Gulf of California: Ciencias Marinas, v. 25, p. 469-
485.
Pope, R.M., and Fry, E.S., 1997, Absorption spectrum (380-700 nm) of pure
water .2. Integrating cavity measurements: Applied Optics, v. 36, p.
8710-8723.
Pultz, T.J., and Scofield, R.A., 2002, Applications of remotely sensed data in
flood prediction and monitoring: report of the CEOS Disaster
Management Support Group flood team: 2002 IEEE International
Geoscience and Remote Sensing Symposium. 24th Canadian
105
Symposium on Remote Sensing. Proceedings (Cat.
No.02CH37380) I 2002 IEEE International Geoscience and Remote
Sensing Symposium. 24th Canadian Symposium on Remote Sensing.
Proceedings (Cat. No.02CH37380), p. 768-70 vol.2 I 6 vol.xcvi+3694.
Refice, A., Bovenga, F., and Nutricato, R., 2003, Stepwise Approach to INSAR
Processing of Multi temporal Datasets, Fringe 20003, Volume SP-550:
Frascati, Italy, ESA.
Reynolds, C.S., Oliver, R.L., and Walsby, A.E., 1987, Cyanobacterial
dominance: the role of buoyancy regulation in dynamic lake
environments: New Zealand J ournalof Marine and Freshwater
Research, v. 21, p. 379-390.
RIGW, 1992, Nile Delta, Hydrogeological Map of Egypt: El Kanater el
Khariya, Research Institute for Groundwater (RIGW), Ministry of
Public Works and Water Resources (MPWWR).
Rinta-Kanto, J.M., Oullette, A.J.A., Twiss, M.R., Boyer, G.L., Bridgeman,
T.B., and Wilhelm, S.W., 2005, Quantification of Toxic Microcystis spp.
during the 2003 and 2004 blooms in Western Lake Erie:
Environmental Science and Technology.
Ruddick, K.G., Ovidio, F., and Rijkeboer, M., 2000, Atmospheric correction of
Sea WiFS imagery for turbid coastal and inland waters: Applied Optics,
V. 39, p. 897-912.
106
Said, R., 1990, The Geology of Egypt: Rotterdam, Netherlands, A. A.
Balkema, 734 p.
-, 1993, The River Nile: Geology, Hydrology and Utilization, Pergamon
Press.
Sandwell, D.T., and Price, E.J., 1998, Phase gradient approach to stacking
interferograms: Journal of Geophysical Research-Solid Earth, v. 103, p.
30183-30204.
Sandwell, D.T., Sichoix, L., and Smith, B., 2002, The 1999 Hector Mine
earthquake, southern California: Vector near-field displacements from
ERS InSAR: Bulletin of the Seismological Society of America, v. 92, p.
1341-1354.
Sathyendranath, S., Lazzara, L., and Prieur, L., 1987, Variations in the
Spectral Values of Specific Absorption of Phytoplankton: Limnology
and Oceanography, v. 32, p. 403-415.
Scharroo, R., and Visser, P., 1998, Precise orbit determination and gravity
field improvement for the ERS satellites: Journal of Geophysical
Research-Oceans, v. 103, p. 8113-8127.
Schneider, B., Bopp, L., Gehlen, M., Segschneider, J., Frolicher, T.L., Cadule,
P., Friedlingstein, P., Doney, S.C., Behrenfeld, M.J., and Joos, F., 2008,
Climate-induced interannual variability of marine primary and export
production in three global coupled climate carbon cycle models:
Biogeosciences, v. 5, p. 597-614.
107
Sestini, G., 1989, The Implications of Climate Changes for the Nile Delta:
Nairobi, Kenya, UNEP/OCA.
Shear, H., Fuller, K., and Wittig, J., 1995, The Great Lakes An
Environmental Atlas and Resource Book/Les Grands Lacs, atlas
ecologique et manuel des ressources: Toronto and Chicago, United
States Environmental Protection Agency and Government of Canada.
Siegelman, H.W., and Kycia, J.H., 1978, Algal biliproteins, in Hellebrust, J.,
and Craigie, J.S., eds., Handbook of Phycological Methods:
Physiological and Biochemical Methods: Cambridge, UK, Cambride
University Press, p. 71-80.
Simis, S.G.H., Peters, S.W.M., and Gons, H.J., 2005, Remote sensing of the
cyanobacterial pigment phycocyanin in turbid inland water: Limnology
and Oceanography, v. 50, p. 237-245.
Simis, S.G.H., Ruiz-Verdu, A., Dominguez-Gomez, J.A., Pena-Martinez, R.,
Peters, S.W.M., and Gons, H.J., 2007, Influence of phytoplankton
pigment composition on remote sensing of cyanobacterial biomass:
Remote Sensing of Environment, v. 106, p. 414-427.
Stanley, D.J., 1988, Subsidence in the Northeastern Nile Delta - Rapid Rates,
Possible Causes, and Consequences: Science, v. 240, p. 497-500.
Stanley, D.J., and Hait, A.K., 2000, Deltas, radiocarbon dating, and
measurements of sediment storage and subsidence: Geology, v. 28, p.
295-298.
108
Stanley, D.J., McRea, J.E., Jr., and Waldron, J.C., 1996, Nile delta drill core
and sample database for 1985-1994: Mediterranean basin (MEDIBA)
program, Smithsonian Contributions to the Marine Sciences, v. 37, p.
1-428.
Stanley, D.J., and Warne, A.G., 1993a, Nile Delta: Recent Geological
Evolution and Human Impact: Science, v. 260, p. 628-634.
-, 1993b, Sea-Level and Initiation of Predynastic Culture in the Nile Delta:
Nature, v. 363, p. 435-438.
-, 1998, Nile delta in its destruction phase: Journal of Coastal Research, v.
14, p. 794-825.
Stanley, J.D., 2001, Dating modern deltas: Progress, problems, and
prognostics: Annual Review of Earth and Planetary Sciences, v. 29, p.
257-294.
Stanley, J.D., Warne, A.G., and Schnepp, G., 2004, Geoarchaeological
interpretation of the canopic, largest of the relict Nile Delta
distributaries, Egypt: Journal of Coastal Research, v. 20, p. 920-930.
Stolarski, R.S., Douglass, A.R., Gupta, M., Newman, P.A., Pawson, S.,
Schoeberl, M.R., and Nielsen, J.E., 2006, An ozone increase in the
Antarctic summer stratosphere: a dynamical response to the ozone
hole: Geophysical Research Letters I Geophysical Research Letters, p.
1-4.
109
Stumpf, R.P., Arnone, R., Gould, R.W., Martinolich, P.M., and
Ransibrahmanakul, V., 2003, A partially coupled ocean-atmosphere
model for retrieval of water-leaving radiance from seawifs in coastal
waters, NASA Tech. Memo. 206892, National Aeronautics and Space
Administration, Goddard Space Flight Center, Greenbelt, MD.
Sturm, B., and Zibordi, G., 2002, SeaWiFS atmospheric correction by an
approximate model and vicarious calibration: International Journal of
Remote Sensing, v. 23, p. 489-501.
Tornqvist, T.E., Wallace, D.J., Storms, J.E.A., Wallinga, J., van Dam, R.L.,
Blaauw, M., Derksen, M.S., Klerks, C.J.W., Meijneken, C., and
Snijders, E.M.A., 2008, Mississippi Delta subsidence primarily caused
by compaction of Holocene strata: Nature Geosciences, v. 1, p. 173-176.
Ucuncuoglu, E., Arli, 0., and Eronat, A.H., 2006, Evaluating the impact of
coastal land uses on water-clarity conditions from Landsat TMIETM +
imagery: Candarli Bay, Aegean Sea: International Journal of Remote
Sensing, v. 27, p. 3627-3643.
Vanderploeg, H.A., Liebig, J.R., Carmichael, W.W., Agy, M.A., Johengen,
T.H., Fahnenstiel, GL., and Nalepa, T.F., 2001, Zebra mussel
(Dreissena polymorpha) selective filtration promoted toxic Microcystis
blooms in Saginaw Bay (Lake Huron) and Lake Erie: Canadian
Journal of Fisheries and Aquatic Sciences, v. 58, p. 1208-1221.
110
Vincent, R.K., Qin, X.M., McKay, R.M.L., Miner, J., Czajkowski, K., Savino,
J., and Bridgeman, T., 2004, Phycocyanin detection from LANDSAT
TM data for mapping cyanobacterial blooms in Lake Erie: Remote
Sensing of Environment, v. 89, p. 381-392.
Warne, A.G., and Stanley, D.J., 1993, Archaeology to Refine Holocene
Subsidence Rates Along the Nile Delta Margin, Egypt: Geology, v. 21,
p. 715-718.
Watson, K.M., Bock, Y., and Sandwell, D.T., 2002, Satellite interferometric
observations of displacements associated with seasonal groundwater in
the Los Angeles basin: Journal of Geophysical Research-Solid Earth, v.
107.
Welschmeyer, N.A., 1994, Fluorometric analysis of chlorophyll a in the
presence of chlorophyll b and pheopigments: Limnology and
Oceanography, v. 39, p. 1985-1992.
Yang, H., Adler, R., and Huffman, G., 2006, Evaluation of the potential of
NASA multi-satellite precipitation analysis in global landslide hazard
assessment: Geophysical Research Letters I Geophysical Research
Letters, p. 1-5.
Yang, M.D., Su, T.C., Hsu, C.H., Chang, K.C., and Wu, A.M., 2007, Mapping
of the 26 December 2004 tsunami disaster by using FORMOSAT-2
images: International Journal of Remote Sensing, v. 28, p. 3071-3091.
111
APPENDIX
SELECTED INTERMEDIATE INSAR PRODUCTS
112
Figure Al Three Pass interferogram generated from scenes acquired 8-24-1995
and 8-23-1995 prior to persistent scatterer selection. Good coherence is seen due
to small spatial and temporal baselines
113
Figure A.2 Three Pass interferogram generated from scenes acquired 8-24-1995
and 1-11-1996. Coherence is maintained in areas around cities and buildings.
These areas are good candidates for persistent scatterers.
114
t •
Stanley et al 1993
Figure A.3 Sample subsidence rate calculation for location in Damietta City
before atmospheric corrections are applied. Error bars show 1 standard deviation
from average of all nearby points within 1 km 2. Y axis shows vertical position
real relative to entire scene average using the 5-15-1997 scene as the reference.
115