University of Southern California Dissertations and Theses

Assessing the use of normalized difference chlorophyll index to estimate chlorophyll-A concentrations using Landsat 5 TM and Landsat 8 OLI imagery in the Salton Sea, California

(USC Thesis Other)

Assessing the use of normalized difference chlorophyll index to estimate chlorophyll-A concentrations using Landsat 5 TM and Landsat 8 OLI imagery in the Salton Sea, California

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

ASSESSING THE USE OF NORMALIZED DIFFERENCE CHLOROPHYLL INDEX TO
ESTIMATE CHLOROPHYLL-A CONCENTRATIONS USING LANDSAT 5 TM AND
LANDSAT 8 OLI IMAGERY IN THE SALTON SEA, CALIFORNIA

by

Alejandra G. Lopez

A Thesis Presented to the
FACULTY OF THE USC DORNSIFE COLLEGE OF LETTERS, ARTS AND SCIENCES
University of Southern California
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

May 2023

Copyright © 2023 Alejandra G. Lopez
ii

To my parents, siblings, and loving partner

iii
Acknowledgements
I would like to thank my advisor, Dr. Elisabeth Sedano, for her insight, encouragement, and
patience throughout the completion of this project. I would also like to thank my committee
members, Dr. Yi Qi and Dr. Diana Ter-Ghazaryan for their expertise, feedback, and guidance.

iv
Table of Contents
Dedication ...........................................................................................................................................ii
Acknowledgements ........................................................................................................................ iii
List of Tables ................................................................................................................................ vii
List of Figures .............................................................................................................................. viii
List of Abbreviations ...................................................................................................................... x
Abstract ......................................................................................................................................... xii
Chapter 1 Introduction .................................................................................................................... 1
1.1. Salton Sea Overview ...........................................................................................................2
1.2. Motivations .........................................................................................................................8
1.2.1. Human Health ............................................................................................................9
1.2.2. Lack of State Action ..................................................................................................9
1.2.3. Economic Development ...........................................................................................10
1.2.4. Gap in Current Literature .........................................................................................11
1.3. Methodological Overview ................................................................................................11
1.4. Thesis Overview ...............................................................................................................12
Chapter 2 Literature Review ......................................................................................................... 13
2.1. Traditional Water Quality Monitoring and Assessment Methods ....................................13
2.2. Water Quality Parameters .................................................................................................16
2.2.1. Suspended Sediments ...............................................................................................17
2.2.2. Chlorophyll and Algae .............................................................................................18
2.2.3. Temperature .............................................................................................................18
2.2.4. Salinity .....................................................................................................................19
2.3. Water Quality and Remote Sensing ..................................................................................19
2.3.1. Sensors and Satellites Used for Water Quality Assessments ...................................20
v
2.3.2. Remote Sensing Modeling Approaches ...................................................................22
2.3.3. RS and Water Quality Parameters ...........................................................................24
2.3.4. Remote Sensing Indices ...........................................................................................27
Chapter 3 Methods ........................................................................................................................ 33
3.1. Data Description ...............................................................................................................35
3.1.1. Landsat Imagery .......................................................................................................35
3.1.2. In-Situ Data ..............................................................................................................39
3.2. Data Preparation ................................................................................................................40
3.2.1. Landsat Imagery .......................................................................................................40
3.2.2. In-Situ Data ..............................................................................................................42
3.3. NDWI, NDCI, 2BDA and 3BDA .....................................................................................43
3.4. Global Regressions ...........................................................................................................47
3.4.1. Linear Regression Assumption ................................................................................48
3.4.2. Generalized Linear Regression ................................................................................54
Chapter 4 Results .......................................................................................................................... 56
4.1. Temporal Analysis of Chlorophyll-a Presence using NDCI .............................................56
4.1.1. Chlorophyll-a Presence in 2002 ...............................................................................57
4.1.2. Chlorophyll-a Presence in 2005 ...............................................................................59
4.1.3. Chlorophyll-a Presence in 2008 ...............................................................................61
4.1.4. Chlorophyll-a Presence in 2011 ...............................................................................63
4.1.5. Chlorophyll-a Presence in 2014 ...............................................................................65
4.1.6. Chlorophyll-a Presence in 2016 ...............................................................................67
4.1.7. Chlorophyll-a Presence in 2020 ...............................................................................69
4.2. Assessment of NDCI against 2BDA and 3BDA ...............................................................72
Chapter 5 Discussion .................................................................................................................... 75
vi
5.1. Discussion of Results ........................................................................................................75
5.2. Limitations and Challenges ...............................................................................................78
5.3. Future Work ......................................................................................................................80
References ..................................................................................................................................... 82
Appendix: A Distributions of NDCI Values ................................................................................. 97

vii
List of Tables
Table 1 Common Spaceborne Sensors and Satellites Used in Water Quality Assessments ........ 21
Table 2 McFeeter’s NDWI Pixel Range ....................................................................................... 28
Table 3 Datsets Overview ............................................................................................................. 36
Table 4 In-Situ and Remote Sensing Data Collection .................................................................. 40
Table 5 Summary of Statistics ...................................................................................................... 48
Table 6 NDCI Pixel Range ........................................................................................................... 57
Table 7 Chlorophyll-a and Trophic States .................................................................................... 57
Table 8 Summary of Linear Regression Results ........................................................................... 73
Table 9 Comparison of NDCI and In-Situ Measurements ............................................................ 74

viii
List of Figures
Figure 1 Salton Sea Vicinity .......................................................................................................... 3
Figure 2 Ancient Lake Cahuilla ...................................................................................................... 4
Figure 3 Colorado River Basin Map ............................................................................................... 7
Figure 4 Precipitation on the Salton Sea between 1914 to 1994 .................................................... 8
Figure 5 Sampling plan ................................................................................................................. 15
Figure 6 Modeling methods used in RS ........................................................................................ 23
Figure 7 Relationship between wavelength and reflectance for suspended sediment
concentrations ........................................................................................................................ 25
Figure 8 Relationship between wavelength and reflectance for cholorphyll concentrations ....... 26
Figure 9 Algorithms to asses chlorophyll-a concentrations .......................................................... 31
Figure 10 Algorithms used to assessed chlorophyll-a concentrations using Landsat 8 OLI ........ 32
Figure 11 Overview of Methodology ........................................................................................... 34
Figure 12 Mosaicked Band 2 (2002) ............................................................................................ 41
Figure 13 Sampling Stations ......................................................................................................... 43
Figure 14 Salton Sea Surface Area (2002 to 2020) ...................................................................... 44
Figure 15 Relationship Between Average Chlorophyll-a and NDCI ........................................... 49
Figure 16 Relationship Between Average Chlorophyll-a and NIR-R .......................................... 49
Figure 17 Relationship Between Average Chlorophyll-a and NIR-G .......................................... 50
Figure 18 Relationship Between Average Chlorophyll-a and NIR-B .......................................... 50
Figure 19 Relationship Between Average Chlorophyll-a and 3BDA ........................................... 51
Figure 20 Histograms of Variables ............................................................................................... 52
Figure 21 Quantile-Quantile Plots of Variables ............................................................................ 53
Figure 22 NDCI 2002 ................................................................................................................... 59
Figure 23 NDCI 2005 ................................................................................................................... 60
ix
Figure 24 NDCI Difference from 2002 to 2005 ........................................................................... 61
Figure 25 NDCI 2008 ................................................................................................................... 62
Figure 26 NDCI Difference from 2005 to 2008 ........................................................................... 63
Figure 27 NDCI 2011 ................................................................................................................... 64
Figure 28 NDCI Difference from 2008 to 2011 ........................................................................... 65
Figure 29 NDCI 2014 ................................................................................................................... 66
Figure 30 NDCI Difference from 2011 to 2014 ........................................................................... 67
Figure 31 NDCI 2016 ................................................................................................................... 68
Figure 32 NDCI Difference from 2014 to 2016 ........................................................................... 69
Figure 33 NDCI 2020 ................................................................................................................... 70
Figure 34 NDCI Difference from 2016 to 2020 ........................................................................... 71
Figure 35 NDCI Difference from 2002 to 2020 ........................................................................... 72

x
List of Abbreviations
2BDA Two-Band Algorithm
3BDA Three-Band Algorithm
AOI Area of Interest
AR Alamo River
BoR Bureau of Reclamation
FLH Florence Line Height
IOPs Inherent Optical Properties
IWRM Integrate Water Resource Management
MIR Middle Infrared
MNDWI Modified Normalized Difference Water Index
NDCI Normalized Difference Chlorophyll Index
NDVI Normalized Difference Vegetation Index
NDWI Normalized Difference Water Index
NIR Near Infrared
NR New River
NWI New Water Index
QSA Quantification Settlement Agreement
RS Remote Sensing
ST Salton Trough
US United States
USGS United States Geological Survey
WQ Water Quality
xi
WQMPs Water Quality Management Plan
WWR Whitewater River

xii
Abstract
The Salton Sea is the largest body of water in the State of California and has experienced a
decline in water quality within the last fifty years. This inland body of water serves as a reservoir
for agricultural runoff and maintains high concentrations of pesticides and nutrients that place
surrounding communities and ecological environments at risk. As a result of the degradation and
eutrophic state of the Salton Sea, it is important to identify historical trends and methodologies
that can be used for future water quality assessments. Traditional water quality assessments are
conducted onsite and require extensive financial and human resources. In order to mitigate some
of these costs while continuing to monitor water quality, more efficient assessment techniques
must be explored. This study explores one such technique by examining the use of remote
sensing techniques and the Normalized Difference Chlorophyll Index (NDCI) to assess
chlorophyll-a concentrations in the Salton Sea from 2002 to 2020 using Landsat 5 TM and
Landsat 8 OLI imagery. To assess the accuracy of this method, the NDCI is compared against
two-band and three-band algorithms proposed by literature. Results indicate that the NDCI has
largely underestimate chlorophyll-a concentrations within the Salton Sea and has incorrectly
suggested small variations across the temporal range. Linear regression results further reveal a
weak linear regression between NDCI, 2BDA and 3BDA values and in-situ measurements.

1

Chapter 1 Introduction
The Salton Sea is the largest body of waterbody in the State of California, spanning
approximately 340 square miles between the Riverside and Imperial counties (Cantor and Knuth
2019; Carpelan 1958). Deviations in weather conditions, increasing water demand and
unprecedented water drought episodes have contributed to the altering landscape of Southern
California (Cohen 2019; Doede and De Guzman 2020; Johnston et al. 2019). In addition to local
policies, environmental factors have contributed to the receding shoreline and poor water quality
(WQ) of the Salton Sea. Scholars and government agencies have categorized the Salton Sea as an
ecological disaster. This classification is attributed to its eutrophic state and uncommonly high
concentration of nutrients that have contributed to the collapse of multiple fish populations
(Bradley, Ajami, and Porter 2022; Forsman 2014; Riedel 2016). The Salton Sea faces an
uncertain future, as the termination of water deposits from the Colorado River, its increased
reliance for agricultural drainage and rising interest for lithium extraction within its lakebed.
Thus, the continued monitoring of the Salton Sea’s WQ is critical and calls for the adoption of
efficient monitoring methods. The aim of this study is to (1) use the Normalized Difference
Chlorophyll Index (NDCI) to assess historical chlorophyll-a concentrations within the Salton Sea
from 2002 to 2020, and (2) assess the accuracy and performance of this method against two-band
(2BDA) and three-band (3BDA) algorithms.
The remainder of this chapter is sectioned into four sections. Section 1.1 provides a
historical overview of the Salton Sea, including how the modern-day Salton Sea was formed.
Section 1.2 discusses the motivations behind this work, including the need for more extensive
data collection within this waterbody. Section 1.3 provides a general overview of the
2

methodology that was implemented within the study. Lastly, section 1.4 outlines how the
remainder of the remainder of the thesis is structured and formatted.
1.1. Salton Sea Overview
The Salton Sea is an inland body of water that is situated in Southern California,
approximately 35 miles north of the US-Mexico border and 32 miles south of the Coachella
Valley (Figure 1) (Cohen, Morrison, and Glen 1999; Doede and DeGuzman 2020). The
waterbody has a maximum depth of 51 feet and a surface area of 340 square miles (Cohn 2000;
King et al. 2011; Tompson 2016). The closed basin is bounded by multiple mountain ranges,
including the Santa Rosa, Chocolate, Peninsular and Orcopia mountains (Tompson 2016).
Geologically, the Salton Sea lies 200 feet below sea level and atop the Salton Trough
(ST). The ST is a rift valley formed along the San Andreas fault that is filled with high tectonic
activity (Cohen, Morrison, and Glenn 1999). The movement of the Pacific and North American
plates have formed local geothermal hotspots around the Salton Sea. These hotspots can exceed
680 degrees Fahrenheit in depths greater than 8,000 feet (Ajala et al. 2019; Tompson 2016).
Multiple earthquakes have been documented within the region, the largest occurring within El
Centro in 1940 and measuring 7.1 on the Richter scale (Cohen, Morrison, and Glenn 1999). The
ST also lays above a deeper underground sink that extends more than 20,000 feet and is
composed of alluvial deposits (Cohen, Morrison, and Glenn 1999). In 1997, the volume of
groundwater was estimated to range between 1.1 to 1.3 billion acre-feet (Tompson 2016). The
water remains inaccessible owing to its high depth and salinity levels (Tompson 2016).

3

Figure 1. Salton Sea vicinity
Historically, the Colorado River periodically inundated parts of the Salton Sink, also
called the Salton Basin, and creating a vast, intermittent lake. The ancient Lake Cahuilla had a
surface area of approximately 2,200 square miles and an estimated evaporation rate of 1.52 to
2.05 meters per year (Waters 1983). When left undisturbed, the lake would take approximately
47 to 64 years to evaporate completely (Laylander 1997; Rockwell, Meltzner and Haaker 2018).
Through the examination of historical records dating between 1774 and 1750, researchers
confirmed this timeframe as the last time the Colorado River had naturally streamed into the ST.
Expedition records dating between 1771 and 1776 confirmed the absence of the lake in later
4

years due to the high heat caused by the desert climate (Cohen 2019; Rockwell, Meltzner and
Haaker 2018). Despite this arid climate and limited supply of irrigated water, the Salton Basin
has a prolonged history of human settlement that dates back 12,000 years, including modern-day
Torres Martinez Desert Cahuilla Indians and the Cabazon Band of Mission Indians (Cohen,
Morrison, and Glenn 1999; Cohn 2000; Delfino 2006). Native American groups were able to
reside and prosper within this dry climate due to the availability of fish and respite the recurrent
bodies of water brought forth.

Figure 2. Ancient Lake Cahuilla
5

The expansion of irrigation within the region commenced in 1901 with the construction
of a massive canal system. The $150,000 investment by George Chaffey to the California
Development Company was used to build irrigation canals that would facilitate the transfer of
water deposits from the Colorado River to the Imperial Valley (Cannon 2022; Kershner 1953).
California Development Company had lain the framework of a canal system that required
additional enhancements, as the irrigation canals were frequently obstructed by silt and
sediments (Cohen, Morrison, and Glenn 1999). During the Winter of 1904 to 1905, the Colorado
River experienced heavy rain and snowmelt that resulted in three floods overflowing into the
Alamo Canal (Ross 2020). The modern-day Salton Sea was formed between 1905 and 1906
when floodwaters originating from the Colorado River ruptured an irrigation canal in the
Imperial County and proceeded to flow into the Salton Sink for 18 months (Cohen 2019;
Tompson 2016). At the height of the flooding, nearly 6 billion cubic feet of water was discharged
into the Salton Basin (Cohen, Morrison, and Glenn 1999; Kennan 1917).
Despite its rapid formation, the Salton Sea quickly became a critical habitat for migrating
birds and an area of interest (AOI) for real estate developers. The Colorado River floods
naturally populated the waterbody with fish. In the 1950s, the California Department of Fish and
Game introduced additional fish populations, including corina, sargo and croaker fish
(Associated Press 2015; Riedel 2016; Sheikh and Stern 2020; Taylor 2018; Tompson 2016). The
introduction of additional fish populations and continual water deposits from the Colorado River
cemented the Salton Sea as a recreational lake. At its tourism peak, the Salton Sea boasted more
visitors than the Yosemite National Park, including celebrity guests Frank Sinatra and the Beach
Boys (Clouse 2016; Gutierrez 2009; Picone 2021; Vizzo 2017). All these activities pointed to a
thriving tourist site that could serve as a sister location to the Palm Springs area. By the 1990s,
6

an evident decline in fish populations pointed toward a deteriorating environment caused by the
declining WQ in the Salton Sea.
Throughout most of the 20th century, Salton Sea water levels were maintained through
water deposits originating from the Colorado River and agricultural drainage from surrounding
agricultural communities. The modern-day Salton Sea would have followed the same
evaporation trajectory as Lake Cahuilla had it not been designated by President Coolidge as an
agricultural sump in 1924 (Cohen 2019). Toward the end of the 1990s, the State of California
was pressured by Wyoming, Utah, Colorado, Nevada, New Mexico Arizona, and the Federal
government to decrease water consumption originating from the Colorado River (Cohen 2019).
The Colorado River is the largest provider of water to California, as it supplies more than 60
percent of the water used by the Southern California region, see Figure 3 (Forsman 2014). The
Law of the River annually allocates California around 4.4 million acre-feet of water and this base
allotment was exceeded between 1983 and 1996 (Cohen, Morrison, and Glen 1999; Forsman
2014). The State of California had taken advantage of apportionments that went unused by other
states and was consequently instructed by the Secretary of Interior to develop a plan to reduce
Colorado River water consumption.
In 2003, Coachella Valley Water District, Imperial Irrigation District, San Diego County
Water Authority, and the Metropolitan Water District of Southern California mutually signed the
“Quantification Settlement Agreement” (QSA) that diverted Colorado River water resources
from the Salton Sea to the San Diego County (Tompson 2016; Forsman 2014; Cohen, Morrison,
and Glenn1999). Colorado River water deposits into the Salton Sea decreased by 10 percent and
completely halted in 2018 (Levers, Skaggs, and Schwabe 2019). Prior to the QSA, the Salton Sea
was already experiencing high levels of salinity due to the evaporation of water and
7

concentrations of salt. As a terminal lake, the Salton Sea does not have any physical outlets and
loses water volume through evaporation. An estimated 1.35 million acre-feet of water were
deposited annually into the Salton Sea to maintain water levels, 75 percent of which stems from
agricultural drainage (Cohen, Morrison, and Glenn 1999).

Figure 3. Colorado River basin
According to the Imperial Irrigation District, the Salton Sea basin can experience
temperatures exceeding 100 degrees Fahrenheit and generate conditions that allow for high
annual evaporation rates. Roughly 1,300,000 acre-feet of water evaporates annually from the
Salton Sea and scholars predict this amount will increase with ongoing climate changes,
8

primarily due to the effects of increasing drought episodes and decreasing precipitation rates
(Cohen, Morrison, and Glenn 1999; Marshall 2017). Figure 4 illustrates precipitation rates within
the Salton Sea between 1914 to 1994. The average annual precipitation rate of three inches per
year are too low to offset the annual evaporation rate of five feet per year (Cohen, Morrison, and
Glenn 1999; Hughes 2020). The reduction in contributing water resources and increased
dependence on agricultural runoff has led the waterbody to shrink in size and experience poor
WQ effects which impact wildlife and surrounding communities. In 2020, the State of California
ceased WQ monitoring in the Salton Sea, and it has since become the responsibility of local
agencies and non-profit organizations to continue assessing its WQ.

Figure 4. Precipitation on the Salton Sea between 1914 to 1994. Source: Cohen et al. 1999
1.2. Motivations
Interest and motivation for this research project stems from the unique circumstances that
have led to the decline in the overall WQ of the Salton Sea and the complex dynamics that stem
from the Coachella and Imperial valley regions.
9

1.2.1. Human Health
In combination with lower precipitation rates, the QSA has accelerated the shrinking of
the Salton Sea and increased exposure to lakebed sediment (Parajuli and Zender 2018). The
Salton Sea has served as a reservoir for agricultural runoff and the exposed lakebed sediment is
enriched with pesticides that date back to the 1930s, such as dichlorodiphenyltrichloroethane and
aldrin (Kim, Kabir, and Jahan 2017). The health and well-being of communities surrounding the
Salton Sea is of great importance, especially when they have historically been inhabited by
underserved population groups. In 1970, 57 percent of the total population in Imperial Valley
self-identified as White and 36 percent identified as Hispanic (Mead 2016). By 2010, the
Imperial Valley boasted a population of 180,000 and 74 percent of the population identified as
Hispanic.
Farzan et al. (2019) also determined that communities situated within close proximity to
the Salton Sea had a higher percentage of asthma prevalence among children (22.4%) than other
communities in the United States (US) (8.4%). This occurrence was largely associated to the
receding lakebed that resulted in approximately 40 to 80 tons of dust being released into
neighboring areas. The small particulates originating from the lakebeds are miniscule and easily
respirable by humans, consequently affecting the respiratory health of people in nearby
communities.
1.2.2. Lack of State Action
In recent years, the State of California has encountered a lot of criticism surrounding their
lack of organized efforts to address and improve the ongoing conditions of the Salton Sea. The
state has held multiple meetings and considered various proposals to identify a sustainable
management plan that will address rising concerns. In 2018, the State of California released a 10-
10

year management plan centered on aquatic habitat restoration and dust suppression around the
edge of the Salton Sea. Although this initiative is a step forward, the plan is centered around
smaller areas and does not address the overarching dilemma of poor WQ. In order to propose any
larger scale mitigation efforts, there is a need for WQ data to be accessible. At the moment, WQ
assessments are predominately performed at onsite locations and require a large workforce to
collect water samples. Since January 2020, the State of California, and Bureau of Reclamation
(BoR) have ceased WQ monitoring of the Salton Sea. Local entities and non-profit organizations
have taken it upon themselves to perform WQ assessments and fill the gap left by these
government agencies.
1.2.3. Economic Development
Since 2018, the State of California has held multiple discussions regarding the possibility
of extracting lithium mineral from the Salton Sea Basin. In March 2020, the California Energy
Commission released a report that outlined the proposed project and stated this venture would
produce nearly 600,000 tons of lithium per year that could result in $7.2 billion project (Ventura
2020). The demand for lithium-ion batteries has increased since 2008 due to their use within
electric vehicles and power grids (Agusdinata et al. 2018). The State of California recently
passed a law that would require all sales of passenger vehicles within the state to be electric
vehicles by 2035. Beyond California, it is estimated that the global demand of lithium will
steadily increase and exceed the resources available by 2025 (Wanger 2011).
At the time of writing this thesis, the California Energy Commission is requesting bid
proposals from companies to determine the best approach in extracting lithium using geothermal
technology which may reduce some environmental impact; however, many environmentalists
have already expressed major concerns (Gammon 2022). Most of the environmental impacts are
11

centered on polluting water sources, increasing carbon dioxide emissions, and creating a water
table reduction that may lead to a drier lakebed in the Salton Sea.
1.2.4. Gap in Current Literature
The study will help build upon progress made within the spatial sciences, as it explores
the application of NDCI in assessing chlorophyll-a concentration using Landsat 5 Thematic
Mapper (TM) and Landsat 8 Operational Land Imager (OLI) imagery. The NDCI has been
assessed using Landsat 8 OLI imagery, but studies have not incorporated Landsat 5 TM to
conduct a temporal analysis. The study also creates a framework that can be reproduced across
other bodies of water and across different temporal ranges. The research benefits local
community members by highlighting shrinkage rates and poor WQ in the Salton Sea. Moreover,
the study aims to address a scientific gap in the literature surrounding WQ within the Salton Sea.
The majority of the current literature in the Salton Sea is centered on salinity levels and the rate
in which the body of water is shrinking. Very few studies attempt to apply semi-empirical
indices to assess WQ parameters. Lastly, the results of this study and studies alike can aid local
agencies in assessing the Salton Sea and inform policy.
1.3. Methodological Overview
The methodology adopted in this study can be segmented into three different sections.
The first section consisted of evaluating peer-reviewed articles to identify common trends and
methodologies used to assess chlorophyll-a concentrations within inland bodies of water. Upon
analyzing the literature, semi-empirical modeling methods were deemed suitable for conducting
temporal analysis of chlorophyll-a concentrations in the Salton Sea. Semi-empirical indices, such
as the NDCI, are useful since they can be reproduced and compared across different bodies of
water and temporal ranges. In the second section, the study focused on the application of the
12

NDCI across the Salton Sea. This step entailed the use of Landsat 5 TM and Landsat 8 OLI
imagery to capture the selected temporal range (2002 to 2020). The third section assessed the
accuracy and performance of the NDCI by running multiple linear regressions against other
notable two-band (2BDA) and three-band (3BDA) algorithms. The 2BDA and 3BDA were
selected based on peer-reviewed articles that demonstrated comparable or better results than the
NDCI in assessing chlorophyll-a presence using Landsat imagery.
1.4. Thesis Overview
The thesis is comprised of five organized chapters. Chapter 2 provides a comprehensive
literature review of related work centered on WQ assessments, remote sensing (RS) modeling
approaches and RS indices for WQ assessments. Chapter 3 describes the data sources and
provides an overview of the methodology adopted within the study. Chapter 4 summarizes the
findings from the adopted methodology and applied analysis. Lastly, Chapter 5 discusses the
results, identifies limitations of the study and outlines future work.

13

Chapter 2 Literature Review
In this chapter, a survey of the literature related to traditional sampling methods, RS imagery in
WQ assessments, WQ parameters and RS data is presented. The findings of the literature were
used to inform the methodology and the direction of the overall study.
2.1. Traditional Water Quality Monitoring and Assessment Methods
Water shortages and low precipitation rates have historically been documented by early
civilizations and have contributed to the adoption of water management practices aimed at
preserving water resources (Behmel at al. 2016; Neary, Ice, and Jackson 2007). In the 20th
century, the concept of water management gained renewed interest owing to population growth
and limited access to natural resources. In 1972, the US enacted the Clean Water Act, which
regulated the discharge of pollutants into US waterbodies and established WQ standards (Keiser
and Shapiro 2019). In 1992, the United Nations Earth Summit introduced the Integrate Water
Resource Management (IWRM) process to participating nations (Saravanan, McDonald, and
Mollinga 2009; Savenije and Van der Zaag 2008). IWRM is a framework that promotes
equitable and sustainable approaches to managing water and land resources. Savenije and Van
der Zaag (2008) state that there is an emerging consensus on IWRM and water management
necessitating an integrated approach, but there remain some issues that are unresolved. Biswas
(2008) analyzes existing literature surrounding IWRM and identified 41 components that would
need to be integrated to water resource management, including WQ, water demand, economic
factors, ground water estimates, municipal water activities and land related issues. Although
Biswas (2008) disputes the ability for this approach to be truly implemented there has been an
increase in Water Quality Management Plans (WQMPs) within the past thirty years.
14

WQPMs are long-term programs designed to monitor aquatic environments, in order to
better respond to emerging problems and assess developing water trends (Behmel et al. 2016;
Bartram and Ballance 1996). WQPMs require standardized WQ assessments and continual
evaluation of all monitoring activity. Behmel at al. (2016) and Strobl and Robillard (2008)
outline eight different elements necessary for the implementation of WQMPs. These include (1)
identifying program objectives, (2) determining location of sampling sites, (3) selecting relevant
WQ parameters, (4) defining sampling frequencies, (5) calculating necessary human and fiscal
resources, (6) establishing logistics and quality control checks, (7) launching data distribution
platform and (8) assessment on use of public data distributed. Madrid and Zayas (2007) concur
on the importance of these elements, but also include sampling as an additional element.
Madrid and Zayas (2007) also highlight the importance of maintaining sampling
equipment, following post sampling procedure and maintaining clear record-keeping of samples
collected. Figure 5 illustrates the various processes and considerations associated with collecting
a water sample. WQ assessments are an integral part of WQPMs and consist of both data
collection and data analysis processes. Traditional WQ assessments predominately use water
sampling as part of their data collection process in combination with laboratories to conduct
measurements of the collected samples. According to Madrid and Zayas (2007), water samples
must abide by strict processing requirements to maintain the integrity of specimen and ought to
be representative of the environment being evaluated. There are two kinds of water sampling
approaches concerning the time a sample is collected, discrete and composite samples (Cassidy
et al. 2018; Matamoros 2012). Discrete or grab samples are single samples collected within a
single container. These types of samples represent the chemical composition of the waterbody at
a given time and place and are primarily used when temporal considerations are not of
15

importance. In contrast, composite samples consist of multiple samples collected within 24 hours
and combined within a single container.

Figure 5. Sampling plan. Source: Madrid and Zayas 2007
Traditional water sampling methods for chemical pollutants include spot, instrumental
and extraction sampling (Cassidy et al. 2018; Madrid and Zayas 2007). Spot or bottle sampling is
a peer-reviewed method that has been widely adopted for governing and legislative efforts
(Madrid and Zayas 2007). Spot sampling consists of individuals taking manual water samples at
onsite locations using a bottle to extract the water. This approach is predominately used within
surface waters, as individuals would be required to be onshore or on a boat. To ensure there is
minimal cross contamination, the bottle is rinsed multiple times with the surface water prior to
collecting the final sample (EPA 2017; Madrid and Zayas 2007). After the water sample has
been extracted, the water undergoes instrumental analysis within a laboratory or onsite location.
Water samples transported to laboratories are stored within a cooler at 4 to 6 degrees Celsius to
ensure the temperature does not breakdown the contaminants during transit (EPA 2017). To
16

extract water samples from waters that exceed 0.5 meters, the bottle is open upon entry and
lowered to the desired depth using a rope or cable (Madrid and Zayas 2007).
Although conventional sampling methods have become the norm within WQMPs, they
have discernable limitations. Spot sampling can be categorized as being discrete, as it consists of
taking a singular water sample at a given location and time (Cassidy et al. 2018; Matamoros
2012). This approach fails to reflect intermittent contamination and can be susceptible to quality
control errors (Gagnon et al. 2007; Madrid and Zayas 2007). Because traditional sampling
methods are conducted manually, they require high financial and human resources that limit the
frequency of testing and the number of sampling sites. There are also concerns on the type of
instruments that are used to measure water samples, given that they can vary across long-term
programs and between laboratories. Schaeffer et al. (2013) and Madrid and Zayas (2007) agree
that WQMPs should adopt novel approaches and emerging tools to conduct water monitoring.
Some of these new approaches include using onsite sensors, RS imagery and online monitoring
systems.
2.2. Water Quality Parameters
This section outlines different WQ parameters that are of importance to any WQ assessment.
WQ is a measurement that allows us to determine the chemical, biological and physical
characteristics of water in order to assess any degradation on WQ. These measurements allow
the assessment of degradation on WQ. WQ indicators can be further categorized into subgroups,
including chemical (e.g., dissolved oxygen, pH, organic compounds, and nutrients), biological
(e.g., algae and bacteria), physical (e.g., temperature, turbidity, clarity, color, and salinity), and
other (odor, color, and floating material). These subgroups are important to differentiate, as
17

screening tests will vary according to the selected WQ parameter and the overall objective of the
study (Gholizadeh, Melesse, and Reddi 2016; Gorde and Jadhav 2013; Usali and Ismail 2010).
Stakeholders and local agencies have an avid interest in identifying WQ parameters that
are best suited for a given body of water, as they are often adopted and standardized within local
WQMPs. The selection of WQ parameters is dependent on the conditions surrounding the AOI
and publications released. Multiple studies have been conducted around the Salton Sea to
determine its physical and chemical compositions. Local and state agencies have equally
conducted their own assessments, yet much of the data and findings remains unpublished
(Holdren and Montano 2002). The earliest publication conducted on the Salton Sea regarding
WQ dates back to the MacDougal publication of 1907. The journal examined major ion
compositions including sodium chloride, magnesium chloride, magnesium sulphate, potassium
sulphate, calcium sulphate, magnesium sulphate and calcium carbonate against other sources of
sea water and determined high levels of salinity within the Salton Sea (Holdren and Montano
2002; MacDougal 1907). Publications thereafter centered on the concentrations of nutrients,
metals and pesticides and have contributed to local WQMPs and WQ parameter selections.
2.2.1. Suspended Sediments
The US Environmental Protection Agency defines “suspended sediments” as fine
inorganic specks of silt and clay that measure less than 0.063 millimeters. Also included are fine
sands between 0.63 to 0.25 millimeters and other grainy matter found within the water column
(Droppo 2001; Spehar, Taylor, and Cormier 2002). According to Ritchie, Zimba and Everitt
(2003), suspended sediments are the most common water pollutant within the surface of
freshwater systems. Traditional water assessments measure concentrations of suspended
sediment by collecting water samples and transporting them to labs.
18

2.2.2. Chlorophyll and Algae
Chlorophyll is a green pigment and natural compound that allows plants to absorb
sunlight and undergo photosynthesis (EPA 2022; Ritchie, Zimba, and Everitt 2003). There are
two types of chlorophyll: chlorophyll-a and chlorophyll-b. Chlorophyll-a is the primary pigment
of photosynthesis and it absorbs orange and violet light between 430 to 660 nanometers (Martin
2019). Chlorophyll-b is an accessory pigment that absorbs light between 450 to 650 nanometers.
This pigment is not always present within the photosynthesis process and will pass absorbed
light to the primary pigment (Martin 2019). Chlorophyll-a is often used as an indicator of algae
growth and has been adopted in WQ assessments to manage eutrophic bodies of water.
Eutrophication refers to high concentrations of nutrients in a lake, river or other bodies of water
(Ansari et al. 2010; Dorgham 2014; Lin et al. 2021; Qin et al. 2013). Algae are native to
freshwater systems, but excessive amounts result in high plant-life density and occasionally dead
oxygen zones or loss of aquatic life (Ansari et al. 2010; Bhateria and Jain 2016; EPA 2022;
Dorgham 2014).
2.2.3. Temperature
Temperature is a physical WQ indicator that measures the thermal energy of a
constituent. Thermal energy is the movement of molecules and atoms and can be transmitted
between constituents as heat (Ling et al. 2017; Raman et al. 2017; Verones et al. 2010). The
transfer of heat to water bodies can occur by means of natural phenomena or human activity;
excessive amounts are thermal pollution (Bhateria and Jain 2016; Raman et al. 2017; Ritchie et
al. 2013). Outside of measuring thermal energy, high temperatures can decrease dissolved
oxygen levels and regulate the type of aquatic species that reside in a river or lake (Bhateria and
Jain 2016; Vasistha and Ganguly 2020).
19

2.2.4. Salinity
Salinity is a physical WQ indicator that measures dissolved salt concentrations in a
waterbody (EPA 2022b; Wang and Xu 2012). Salt originates from three main sources: (1)
evaporated ocean water, (2) landscape weathering and (3) other environmental settings, such as
drainage water and road salts (Dailey, Welch, and Lyons 2014; Obianyo 2019; Nielsen et al.
2003). Small quantities of water are suitable for aquatic life and plants, but higher amounts
negatively affect the viability of eggs and marine plant seeds (Nielsen et al. 2003). Compared to
oceans, inland waters have lower salinity levels and higher variations in salinity due to seasonal
water deposits, evaporation rates and precipitation frequency (Gholizadeh et al. 2016). Hua
(2017) performed a WQ assessment on the Malacca River in Malaysian and determined that high
salinity pollution resulted from pesticide usage in nearby rubber and oil plantations. This study
echoes numerous studies conducted around the Salton Sea, which conclude that salinity pollution
was largely attributed to pesticide usage and nearby agricultural practices (Bradley et al. 2022;
Cohen 2019; Gao et al. 2022).
2.3. Water Quality and Remote Sensing
Multiple academic studies have reported on the effectiveness of using RS techniques and
tools to better monitor WQ (Madrid and Zayas 2007; Schaeffer et al. 2013; Usali and Ismail
2010). RS as process for obtaining information on an object, AOI or through the use of satellites
and sensors that are able to measure reflected and emitted radiation (Gholizadeh, Melesse, and
Reddi 2016; Usali and Ismail 2010). In the past decades, there have been advancements in
sensors, satellites and modeling approaches used in WQ assessment. The following subsections
explore these topics further.
20

2.3.1. Sensors and Satellites Used for Water Quality Assessments
Technological advancements and public-facing NASA missions have made RS imagery
readily available. However, it has not been commonly incorporated into WQ assessments and
overarching WQMPs. The selection of RS imagery for water assessment is largely dependent
upon the satellites and their associate multispectral sensors; other factors to consider include
temporal range and spatial resolution. RS sensors measure the physical properties of an object or
environment through the emission of heat, radiation, sound, light or motion (Gholizadeh,
Melesse, and Reddi 2016; Sanderson 2010; Zhu et al. 2018).
Sensors can be categorized into two broad categories, airborne and spaceborne sensors.
Airborne sensors are mounted on aircrafts (i.e., helicopters, balloons, or aircrafts) and were first
introduced in 1859 when Gasper Felix Tourmachon captured the first aerial image of Paris on a
hot air balloon (Waghmare and Suryawanshi 2017; Zhu et al. 2018). Aerial sensors are flexible
instruments better suited for smaller waterbodies, as they require smaller pixel sizes (Gholizadeh,
Melesse and Reddi 2016). Costs associated with airborne data can drastically increase as the
surface area increases—standard cost of 350 dollars per square mile—and is one of the main
impediments within its broader application in WQ assessments (Chipman, Olmanson, and
Gitelson 2009; Gholizadeh, Melesse, and Reddi 2016).
Spaceborne or satellite sensors are aboard spacecrafts and satellites orbiting the earth’s
atmosphere (Gholizadeh, Melesse and Reddi 2016; Roy, Behera and Srivastav 2017). Satellite-
derived data have lower associated costs and are more readily available, given fixed revisiting
time frames. In terms of spatial resolution, satellite data tends to be coarse to moderate and
ranges between 30 to 120 meters. Table 1 contains common spaceborne sensors and satellites
used in WQ assessments, such as SPOT-5, MODIS, Landsat, Terra and Sentinel 2. Table 1 was
compiled based on data from Gholizadeh, Melesse and Reddi (2016) and this literature review.

21

Table 1. Common Spaceborne Satellites Used in Water Quality Assessments
Satellite Sensor

Launch
Year

Spectral Bands (mm) Spatial Resolution (m) Swath Width (km) Revisit Interval (Days)
Landsat 5 Thematic Mapper
(TM)
1984 Band 1: 0.45 - 0.52
Band 2: 0.52 - 0.60
Band 3: 0.63 - 0.69
Band 4: 0.77 - 0.90
Band 5: 1.55 - 1.75
Band 6: 10.40 - 12.50
Band 7: 2.09 - 2.35

30 m
30 m
185 km 16
Landsat 5 Multispectral
Scanner (MSS)
1984 Band 4: 0.5 - 0.6
Band 5: 0.6 - 0.7
Band 6: 0.7 - 0.8
Band 7: 0.8 - 1.1
60 m 185 km 18
Landsat 7 Enhanced Thematic
Mapper Plus
(ETM+)
1999 Band 1: 0.45 - 0.52
Band 2: 0.52 - 0.60
Band 3: 0.63 - 0.69
Band 4: 0.77 - 0.90

Band 5: 1.55 - 1.75
Band 6: 10.40 - 12.50
Band 7: 2.09 - 2.35
Band 8: 0.52 - 0.90

15 m
30 m
60 m
183 km 16
Landsat 8 Operational Land
Imager (OLI) and
Thermal Infrared
Sensor (TIRS)
2013 Band 1: 0.43 - 0.45
Band 2: 0.45 - 0.51
Band 3: 0.53 - 0.59
Band 4: 0.64 - 0.67
Band 5: 0.85 – 0.88
Band 6: 1.57 – 1.65
Band 7: 2.11 - 2.29
Band 8 : 0.50 - 0.68
Band 9: 1.36 - 1.38
Band 10: 10.60 - 11.19
Band 11: 11.50 -12.51

15 m
30 m
100 m
170 km 16
Sentinel 2 Multi-Spectral
Instrument (MSI)

Band 1: 0.43 - 0.45
Band 2: 0.45 - 0.52
Band 3: 0.54 - 0.57
Band 4: 0.65 - 0.68
Band 5: 0.69 - 0.71
Band 6: 0.73 - 0.74
Band 7: 0.77 - 2.29
Band 8: 0.78 - 1.38
Band 8A: 0.85 - 0.87
Band 9: 0.93 - 0.95
Band 10: 1.36 - 1.39
Band 11: 1.56 - 1.65
Band 12: 2.1 - 2.28

10 m
20 m
60 m
290 km 10
SPOT - 5 HRG 2005 Band 1: 0.50 - 0.59
Band 2: 0.61 - 0.68
Band 3: 0.78 - 0.89
Band 4: 1.58 – 1.75

2.5 m
5 m
10 m
20 m
60 km 2 - 3
Terra Moderate
Resolution Imaging
Spectroradiometer
(MODIS)
1999 Bands 1 to 19: 0.40 - 2.15
Bands 20 to 36: 3.66 - 14.28

250 m
500 m
1000 m
2330 km 1 - 2
22

2.3.2. Remote Sensing Modeling Approaches
The use of RS imagery in WQ dates to the 1970s with the work of Ritchie et al. (1974).
The study focused on the use of RS techniques to identify and estimate the presence of
suspended sediment within six Mississippi reservoirs (Ritchie et al. 1974; Usali and Ismail
2010). The study resulted in the following empirical equation:
𝑌 =𝐴+𝐵𝑋 or 𝑌 =𝐴𝐵
!
(1)

where Y is the RS measurement (i.e., reflectance, energy, radiance), X represents the WQ
parameter and A and B are empirically derived factors (Ritchie, Zimba, and Everitt 2003). This
empirical equation was then adopted by other researchers to examine various WQ parameters
such as suspended matter, algae, and temperature. Beyond the development of empirical
equations, RS techniques have continued to evolve and now include semi-empirical, analytical
and semi-analytical models. These models are bio-optical algorithms that determine the
concentration of nutrients using both optical qualities and in-situ WQ parameter measurements
(Topp et al. 2020).
Empirical models remain the most common RS approach, as it involves fitting a linear
regression between in-situ WQ measurements and spectral bands (Topp et al. 2020; Wang and
Yang 2019). These models are highly dependent on the availability of in-situ data to validate the
accuracy of the model and are rarely used across large temporal and spatial resolutions (Wang
and Yang 2019). Gholizadeh, Melesse and Reddi (2016) argue that empirical models are not
suited for all WQ parameters, especially those that rely on multispectral sensors that can yield
ambiguous results. For example, researchers have long debated on the use of empirical methods
to assess chlorophyll-a concentrations. Existing spectral satellites have difficulty differentiating
between suspended sediments and chlorophyll-a spectral signals when large amounts of
23

sediments are present within eutrophic bodies of water. Wang and Yang (2019) conducted a
literature review of current research using RS techniques for WQ assessments in China and
determined that chlorophyll-a was the highest WQ indicator examined using empirical models.
Figure 6 illustrates the number of publications for different indicators among different model
types.

Figure 6. Modeling methods used in remote sensing. Source: Wang and Yang 2019
In contrast, semi-empirical methods use spectral band index values rooted in physical
measurements to retrieve WQ parameters from RS imagery (Lednicka and Kubacka 2022; Mejia
Avila, Torres-Bejarano, and Martinez Lara 2022; Topp et al. 2022). These models are designed
to augment the spectral properties of the WQ indicators and reduce noise from other parameters.
Since semi-empirical models are based on physical properties, model users are expected to have
prior knowledge of sensors, spectral bands, and inherent optical properties (IOPs) to select
appropriate RS imagery (Topp et al. 2022). IOPs are the scattering and absorption characteristics
of water and suspended material (Lee 2006). Semi-empirical models are low cost and can be
easily reproduced across different spatial and temporal scales (Mejia Avila, Torres-Bejarano and
24

Martinez Lara 2022). Wang and Yang (2019) and Topp et al. (2022) identified chlorophyll-a,
suspended sediments, turbidity, and cyanobacteria as the most common WQ indicators used in
semi-empirical modelling.
Both analytical and semi-analytical models are physics-based. They determine WQ
concentrations by modeling surface water reflectance through IOPs of water and atmosphere
(Gholizadeh, Melesse, and Reddi 2016; Topp et al. 2022). Analytical approaches are best suited
for analyzing features that have distinct absorption values. Hence, they are not suited for
measuring dissolved oxygen or suspended sediments whose absorption values can vary. Given
that in-situ WQ measurement are used to parameterize semi-analytical models, they are often
more accurate and thus more commonly used for assessing in inland bodies of water (Morel and
Gordon 1980; Topp et al. 2020). However, due to their complex nature and high data
requirements relative to empirical models, analytical models have lower application within WQ
assessments (Wang and Yang 2019).
2.3.3. RS and Water Quality Parameters
In this section, an overview of RS techniques used to assess WQ parameters is provided.
WQ parameters discussed within this section include suspended sediments, chlorophyll,
temperature and salinity.
2.3.3.1. Suspended Sediments
There has been a growing consensus among researchers on the use of RS techniques to
monitor suspended sediments. Current sensors and satellites (e.g., Landsat, IRS, and SPOT) are
able to detect and quantify their optical signals, including Landsat, IRS, and SPOT (Sherman,
Houser and Baas 2013; Usali and Ismail 2010). Because the quantity of signal returned from a
sensor is correlated to the surface area of a particle, it can provide rough approximation of
25

suspended sediment concentrations (Sherman, Houser, and Baas 2013). To determine the optimal
wavelength for use, researchers need to monitor in-situ measurements. Many studies have
developed distinctive algorithms to establish a relationship between the concentration of
suspended sediments and reflectance using in-situ data. However, since the algorithms are rooted
in in-situ measurements, they cannot be repurposed for different bodies of water or even within
large temporal scales for the same water body. Despite variance between studies, wavelengths
between 700 and 800 nanometers have appeared to be the most effective in capturing suspended
sediments (Ritchie, Zimba, and Everitt 2003; Usali and Ismail 2010). Figure 7 illustrates the
relationship between reflectance and wavelength, as affected by the varying concentration levels
of suspended sediments (Ritchie, Zimba, and Everitt 2003).

Figure 7. Relationship between wavelength and reflectance for suspended sediment
concentrations. Source: Ritchie, Zimba, and Everitt 2003
2.3.3.2. Chlorophyll
Wang and Yang (2019) consider chlorophyll-a as the most studied WQ indicator in
inland bodies of water and claim it is widely assessed across all RS models. In contrast,
26

Brezonik, Menken and Bauer (2005) and Ritchie, Zimba and Everitt (2003) claim most RS
studies employ empirical models centered on the relationship between chlorophyll and
reflectance values. Chlorophyll-a absorptions occur at shorter wavelengths, between 450 to 475
nanometers and between 650 to 680 nanometers (Ritchie, Zimba, and Everitt 2003). Figure 8
illustrates the relationship between wavelength and reflectance for differing chlorophyll
concentrations. However, researchers are in agreement that existing satellites, such as SPOT and
occasionally Landsat, sometimes have limitations to assessing chlorophyll concentrations. Strong
signals emanating from suspended sediments can block chlorophyll-a reflectance. This
occurrence is more prominent in bodies of water with low water volume and small surface areas.

Figure 8. Relationship between wavelength and reflectance for chlorophyll concentrations.
Source: Ritchie, Zimba, and Everitt 2003
2.3.3.3. Temperature
In RS studies, researchers employ thermal sensors aboard planes or rely on satellite
imagery from MODIS and Landsat to measure temperature (Gholizadeh, Melesse, and Reddi
2016; Ritchie, Zimba, and Everitt 2013). Thermal infrared data incorporates middle wavelengths
27

ranging between 3 to 5 micrometers and long infrared wavelength ranging between 7 to 14
micrometers (Scafutto and de Souza Filho 2018; Wu et al. 2019).
2.3.3.4. Salinity
Similar to other WQ indicators, RS techniques can assist in monitoring water salinity
using microwave wavelengths. Salinity is correlated with water conductivity. Changes in
conductivity can be captured through microwave radiations (Lewis and Perkin 1978; Vinas
2012). Gholizadeh, Melesse and Reddi (2016), Ramadas and Samantaray (2018), Wang and Xu
(2012), and Vinas (2012) identified numerous RS satellites suited for salinity, including Landsat,
Aquarius, SEASAT and SMOS.
2.3.4. Remote Sensing Indices
In recent years, the use of semi-empirical methods and spatial indices to assess WQ
parameters has increased. A spectral index is an equation that combines pixel values from more
than one spectral band in multispectral images (Tran, Reef, and Zhu 2022). Spectral indices are
derived from simple band ratios that use adding and subtracting bands to highlight targeted
features and reduce environmental effects. Common spatial indices employ normalized
difference, in which the band ratio is standardized by the sum of the two bands selected as shown
in the following equation:
𝑁𝑜𝑟𝑚𝑎𝑙𝑖𝑧𝑒𝑑 𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 =
"
!
# "
"
"
!
% "
"
(2)
where Bx is the first selected band and By is the second selected band. This method ensures that
the spatial index layer ranges between -1 and 1. Band ratios without standardization can vary in
range and can be difficult to compare among one another.
28

There is an array of spatial indices that have been developed to highlight water bodies
and WQ parameters, while filtering out surrounding landcover. One of the most common is the
normalized difference water index (NDWI), which uses two separate equations to identify water
in RS imagery. The first equation employs shortwave infrared (SWIR) and near infrared
channels (NIR). It was first introduced by Gao (1996) and is expressed as follows:
𝑁𝐷𝑊𝐼 (𝐺𝑎𝑜 1996)=
&'(#)*'(
&'(%)*'(
(3)
Gao’s NDWI can assess water content within vegetation, as NIR and SWIR have a high
correlation with vegetation water content (Gao 1996; Gao et al. 2015; Xu 2006). The second
NDWI equation was introduced by McFeeters (1996). It employs NIR and green bands as shown
in the following expression:
𝑁𝐷𝑊𝐼 (𝑀𝑐𝐹𝑒𝑒𝑡𝑒𝑟𝑠 1996)=
+,--.#&'(
+,--.%&'(
(4)
McFeeter’s NDWI highlights waterbodies by amplifying the reflectance of water using the green
band and minimizing the reflectance of NIR seen in vegetation and other landcover types
(McFeeters 1996; Xu 2006). This expression produces an index in which water features have
high values and non-water feature have low or negative values, see Table 2. The NDWI has been
used to extract bodies of water and is especially useful when the water surface area fluctuates
through seasons or years.
Table 2. McFeeter’s NDWI Pixel Range
Range Land use/Landcover Type
-1.00 to -0.30 Drought, non-water surface
-0.30 to 0.00 Moderate Drought, non-water surface
0.00 to 0.20 Flooding
0.20 to 1.00 Water Surface
29

Xu (2006) argues McFeeter’s NDWI generates positive values for urban-made structures and
proposed the modified NDWI (MNDWI). The MNDWI equation uses middle infrared (MIR)
and green bands as shown in the following equation:
𝑀𝑁𝐷𝑊𝐼 =
+,--.#/'(
+,--.%/'(
(5)
According to Xu (2006), the MNDWI can generate higher values for waters of bodies, given the
MIR’s ability to absorb a higher percentage of light than NIR. The MNDWI should also produce
lower pixel values for urban areas and structures (Nugroho 2013; Xu 2006). Nugroho (2013)
assessed the NDWI, MNDWI and the new water index (NWI) to identify inundated areas within
Java Island. The NWI was introduced by Yang et al. (2011) as a method for identifying and
extracting bodies of water using the following equation:
𝑁𝑊𝐼 =
"01-# (&'(%/'(3%/'(4)
"01-%(&'(%/'(3%/'(4)
(6)
Nugroho (2013) concluded that McFeeter’s method provided more accurate results when
compared against MNDWI and NWI. Similarly, Ali et al (2019) determined the usefulness and
accuracy of using NDWI and MNDWI in extracting bodies of water, including those located in
urban areas. In contrast, a study conducted by Jiang et al. (2021) determined that NDWI may not
be suitable for identifying and extracting waterbodies that are frozen or are extremely large.
Despite some reservations, the use of NDWI to extract bodies of water is a continued practice
when assessing WQ parameters (Ali et al. 2019; Garg, Aggarwal, and Chauhan 2020; Haibo et
al. 2011; Yue et al. 2020).
Aside from water indices used to identify and extract bodies of water, there has been a
large development in RS algorithms used to assess WQ parameters, including chlorophyll-a.
The majority of RS algorithms for chlorophyll-a have employed red and NIR channels, given
that studies have shown blue and green band algorithms show inaccuracies when assessing
30

inland bodies of water with high turbidity (Mishra, Schaeffer, and Keith 2014). Mishra and
Mishra (2012) developed the NDCI using red-edge and red bands as shown in the following
equation:
𝑁𝐷𝐶𝐼 =
(-6 768-#(-6
(-6 768-%(-6
(7)
The spectral index was modeled after the normalized difference vegetation index
(NDVI), which was initially introduced by Tucker (1979). The NDCI uses a spectral range that
lies between 665 and 708 nm. This maximizes chlorophyll-a sensitivity and addresses several
limitations found when using the 748 to 778 nm spectral range (Kamerosky, Cho and Morris
2015; Mishra, Schaeffer, and Keith 2014). The use of band ratio standardization ensures that the
NDCI values not only range between -1 and 1, but also address seasonal variability. It was
initially assessed using the Medium Resolution Imaging Spectrometer sensor aboard the
ENVISAT and has since been tested across different sensors and satellites, including Landsat 8,
Sentinel-2 and Worldview-3 (Buma and Lee 2020; Watanabe et al. 2015). Given that not all
sensors have the capacity to capture RS imagery utilizing a red-edge band, the formula has been
amended in multiple studies. The red edge band has been replaced with the NIR band and
essentially models the NDVI as shown in the following equation:
𝐴𝑚𝑒𝑛𝑑𝑒𝑑 𝑁𝐷𝐶𝐼 =
&'(#(-6
&'(%(-6
(8)
In the RS community, there is a growing consensus that NDCI estimates chlorophyll-a
concentrations more accurately when using Sentinel-2 versus Landsat-8 and WorldView-3
imagery (Beck et al. 2016; Caballero et al. 2020; Karimi, Hashemi and Aghighi 2022;
Ogashawara and Moreno-Madrinan 2016). This consensus is largely attributed to the red-edge
band that is present within Sentinel-2. However, there have been studies that demonstrate the
usefulness of using Landsat imagery to apply NDCI. Buma and Lee (2020) assessed the use of
31

2BDA, 3BDA, florescence line height (FLH) and NDCI using Landsat 8, WorldView-3 and
Sentinel-2 imagery on Lake Chad, Africa. Figure 9 contains the algorithms used to derive
chlorophyll-a concentrations in Lake Chad. They determined that the four algorithms had an
average R
2
of 0.74 when using Seninel-2 data. The 2BDA had the lowest R
2
value at (R
2
= 0.60)
and the 3BDA had the highest (R
2
= 0.95). This value was in contrast with Landsat 8 that had an
overall R
2
of 0.78, a slightly higher R
2
than Sentinel-2 data. The Landsat 8 performance
diagnostics also revealed that 3BDA (R
2
= 0.89) and NDCI (R
2
= 0.75) performed better than
2BDA (R
2
= 0.71) and FLH (R
2
= 0.73).

Figure 9. Algorithms to assess chlorophyll-a concentrations. Source: Buma and Lee 2020
Watanabe et al. (2015) conducted a similar assessment using only Landsat 8 imagery to
determine chlorophyll-a concentration within a water reservoir in Brazil. The study examined
various two-band and three-band models proposed by literature, including NDCI and the 2BDA
performed within the Buma and Lee (2020) study (see Figure 10). Their findings revealed that
the NIR-Red, NIR-Green and NIR-Blue ratios generated high R2 values and demonstrated their
sensitivity towards detecting chlorophyll-a concentrations. The NDCI had one of the lowest R2
at 0.39, which contrasted greatly with the findings of Buma and Lee (2020). In terms of fitting
the models, Watanabe et al. (2015) concluded that all algorithms illustrated satisfactory fits,
32

however, they underestimated the actual concentrations and were unable to accurately capture
the correct trophic state. The NIR-Green ratio had the poorest results and NIR-Red (linear and
polynomial) yielded the best fit. Watanabe et al. (2015) states the algorithms may be affected by
the trophic state of the water feature and might display better results when used with other bodies
of water. Overall, the benefit of continuing to use Landsat imagery to assess chlorophyll-a
concentrations centers on its temporal range, given Sentinel-2 only launched in 2015.

Figure 10. Algorithms used to assess chlorophyll-a concentrations using Landsat 8 OLI. Source:
Watanabe et al. 2015

33

Chapter 3 Methods
The objective of this study is twofold: (1) use a semi-empirical RS index to perform a temporal
analysis of chlorophyll-a presence between 2002 and 2020, and (2) assess the efficacy of the
NDCI against 2BDA and 3BDA using linear regressions. Historically, WQMPs have been
implemented using traditional WQ assessment methods that require high financial and labor
resources. The use of NDCI to assess chlorophyll-a presence can help alleviate some of the
resource requirements in WQMPs. The application of NDCI, 2BDA and 3BDA for WQ
assessments have been examined in multiple studies and have proven to be comparable with
empirical methods (Buma and Lee 2020; Das, Kaur, and Jutla 2021; Mishra and Mishra 2012;
Xu et al. 2019a; Xu et al. 2019b).
This chapter summarizes the data and methodologies necessary for performing a
temporal analysis of the Salton Sea using the NDCI. Section 3.1 provides a detailed description
of the in-situ measurements and RS data sources. Section 3.2 expands on how the two datasets
were prepared for use, including data cleaning, coordinate system projections and mosaicking RS
imagery. Section 3.3 summarizes the application of NDWI to extract bodies of water and the
application of NDCI to view chlorophyll-a presence in the Salton Sea. Lastly, Section 3.4
explains the linear regression assumptions and the application of linear regressions to assess the
performance of NDCI against 2BDA and 3BDA. Figure 11 displays a general overview of the
methodology.

34

Figure 11. Overview of methodology
35

3.1. Data Description
The monitoring of WQ within inland waterbodies requires particular consideration to
satellite imagery, as cloud coverage and atmospheric radiation can be impediments to assessing
surface reflectance. Satellite imagery is the primary data type that is used within the study, while
the in-situ chlorophyll-a measurements will serve to assess the use of the NDCI against 2BDA
and 3BDA. Table 3 provides a brief overview of the imagery and in-situ data, including short
descriptions and source specifics.
3.1.1. Landsat Imagery
The Landsat collection was established in 1972 with the launch of Landsat 1 and was
predominantly inspired by the Apollo moon missions that resulted in early land surface images
of the Earth (Irons and Dwyer 2010; Williams, Goward, and Arvidson 2006). According to the
USGS, the main objective in establishing a RS satellite program was to allow civilians and
scientists the opportunity to further explore the natural resources found within the planet
(Goward et al. 2006; Loveland and Dwyer 2012). The satellite program was initially met with
contention by the Department of Defense, which had concerns regarding near-real time RS data
being readily available to the public. From a foreign relations perception, the Department of
Defense was apprehensive of openly monitoring other countries without their explicit consent.
Ultimately, the RS program was adopted, and it has since become the longest serving satellite
program.

36

Table 3. Datasets Overview reach the velocity necessary to obtain orbit. Amongst all satellites,
Type of Data Source

Dataset Description Data Format Spectral Resolution Spatial Resolution Temporal Resolution
Satellite
Imagery
United
States
Geological
Survey
(USGS)
Landsat 5 TM
C2 U.S.
Analysis
Ready Data
(ARD)
Dataset contains satellite
imagery spanning from
2002 to 2020 and covers the
extent of the Salton Sea.
The dataset also has a level
1 processing and provides
digital numbers (DN).
GeoTIFF Band 1: 0.45 – 0.52 mm
Band 2: 0.52 – 0.60 mm
Band 3: 0.63 – 0.69 mm
Band 4: 0.76 – 0.90 mm
Band 5: 1.55 – 1.75 mm
Band 6: 10.40 – 12.50 mm
Band 7: 2.08 – 2.35 mm

30 meters
120 meters
16 - Days
Satellite
Imagery
United
States
Geological
Survey
(USGS)
Landsat 8 OLI
C2 U.S.
Analysis
Ready Data
(ARD)
Dataset contains satellite
imagery spanning from
2002 to 2020 and covers the
extent of the Salton Sea.
The dataset also has a level
1 processing and provides
digital numbers (DN).
GeoTIFF Band 1: 0.43 – 0.45 mm
Band 2: 0.45 – 0.51 mm
Band 3: 0.53 – 0.59 mm
Band 4: 0.64 – 0.67 mm
Band 5: 0.85 – 0.88 mm
Band 6: 1.57 – 1.67 mm
Band 7: 2.11 – 2.29 mm
Band 8: 0.50 – 0.68 mm
Band 9: 1.36 – 1.38 mm
Band 10: 10.6 – 11.9 mm
Band 11: 11.50 – 12.51 mm
15 meters
30 meters
100 meters
120 meters
16- Days
In-Situ United
States
Bureau of
Reclamation
(USBR)
Salton Sea
Data (2004 to
2020)
Dataset contains satellite
imagery spanning from
2002 to 2020 and covers the
extent of the Salton Sea.
The dataset also has a level
1 processing and provides
digital numbers (DN).
.CSV N/A N/A Quarterly per Year
37

Since the successful launch of Landsat 1, the National Aeronautics and Space
Administration has launched seven successful Landsat satellites (i.e., Landsat 2, Landsat 3,
Landsat 4, Landsat 5, Landsat 7, Landsat 8, and Landsat 9) (Wulder et al. 2022). Landsat 6
launched on October 5, 1993, from the Vandenberg Air Force Base in California, but failed to
Landsat 5 is the longest operating RS satellite. Landsat 5 orbited for 28 years and captured 2.5
million images before being discontinued in 2013 (Wulder et al. 2022). Similar to is predecessor,
Landsat 5 carried aboard the Multispectral Scanner (MSS) and TM sensors. The MSS captured
four spectral bands between 0.5 to 1.1 micrometers, whereas the TM sensors added mid-range
infrared bands. In contrast, Landsat 7 carries the Enhanced Thematic Mapper Plus (ETM+) that
offers higher quality products and the addition of a panchromatic band with a 15-meter
resolution. In May 2003, the Scan Line Corrector, a mechanism responsible for the line of sight
of the satellite, malfunctioned and was unable to be repaired (Scaramuzza and Barsi 2005). As a
result, RS images in Landsat 7 are duplicated and follow a zig-zag pattern along the ground.
Landsat 8 was launched in February 2013, and it carries two sensors, the OLI and the Thermal
Infrared Sensor (TIRS) (Wulder et al. 2019; Wulder et al. 2022). The OLI contains nine spectral
bands, including SWIR and panchromatic bands. Similar to other Landsat satellites, Landsat 8
OLI has a 16-day repeat cycle and a 30-meter spatial resolution. Landsat 9 is the latest satellite to
be launched and it contains improved versions of the OLI and TIRS sensors.
The Landsat collection provides RS data to public users at different levels of processing
and resolution. WQ studies commonly use level-1, level-2 or modified versions of level-1 and
level-2 to assess surface reflectance and temperature values. Level-1 products consist of raw data
products that are derived from the satellite sensors and have not undergone any processing or
corrections. This minimal level of processing is common in the research community, given that it
38

allows users the flexibility to select the correction methodology best suited for their study. In
contrast, level-2 data products have undergone radiometric and atmospheric corrections. These
corrections often entail converting digital numbers to radiance and radiance to top of atmosphere
(TOA) reflectance.
Due to the spatial, spectral, and temporal resolution requirements, this study employs
Landsat Analysis Ready Data (ARD) that has been pre-processed, corrected and is ready to use.
ARD data was developed by Global Land Analysis and Discovery and is processed using the
following four steps for surface reflectance values: (1) conversion to TOA, (2) quality
assessment of observations, (3) reflectance normalization and (4) temporal aggregation of images
into 16-day composites (Potapov et al. 2012; Dwyer et al. 2018). Typical Landsat images have a
spectral reflectance range from 0 to 1, but this process rescales the range from 1 to 40,000. This
normalization of spectral values allows images originating from different Landsat satellites to be
compared against one another.
Landsat 5 TM and Landsat 8 OLI U.S. ARD spanning from 2002 to 2020 (i.e., 2002,
2005, 2008, 2011, 2014, 2016 and 2020) was acquired using the USGS EarthExplorer. Landsat 7
ETM+ data was not employed within the study, given visible scanline errors within the AOI
impeded the extraction and analysis of the waterbody. The selection of Landsat imagery was
dependent on the following criteria: (1) RS image needed to be collected less than 16 days from
when the in-situ measurements were collected, (2) RS image needed to fall between May and
August to align with high temperatures, and (3) cloud coverage could not exceed 10 percent of
image tile. Exceptions to the criteria were made in 2002, 2014 and 2020. In 2014, RS imagery
surrounding the day of in-situ measurement collection had excessive cloud coverage and the
selected RS imagery was collected 26 days prior to in-situ data. In 2020, in-situ measurements
39

were collected only in January of that year and were consequently incorporated in the study.
Lastly, note that in-situ measurements were not collected by the BoR until 2004 and as such, no
in-situ measurements were obtained for 2002. The selection for RS imagery for this year was
guided by the preset May to August range, the month of July was selected.
3.1.2. In-Situ Data
As part of the Salton Sea Reclamation Act of 1998, the BoR was directed by the
Department of Interior to investigate feasible options to manage salinity and water elevation at
the Salton Sea (Cohn 2000; Sheikh and Stern 2020; Vessey 2000). The intention behind the act
was to preserve fish and wildlife, increase recreational opportunities and aid in overall economic
development of surrounding areas. The legislation was approved with the understanding that the
Salton Sea would remain a reservoir for irrigation drainage. In collaboration with the Coachella
Valley Water District and the Salton Sea Authority, the BoR collects quarterly water samples
from preselected sampling stations located in and around the Salton Sea. As documented by the
BoR, the samples are manually collected at mid-water depths using a YSI meter and the spot
sampling method. The YSI meter can measure conductivity, water temperature, dissolved
oxygen and salinity in real-time. Selenium, nitrogen, phosphorous, chlorophyll-a and other
nutrients are measured at nearby laboratories.
In-situ measurements of the Salton Sea spanning between 2004 to 2020 were acquired
from the BoR site. The study uses in-situ values collected between May and August for the
following years: 2005, 2008, 2011, 2014, 2016 and 2020. The selected month range ensures that
in-situ measurements are being captured during the hottest months within the region, in order to
assess maximum values. Evaporation and temperature are correlated, and high temperatures
cause higher evaporation rates that result in higher nutrient concentrations (Roland et al. 2012).
40

Table 4 provides a summary of in-situ measurements and Landsat imagery data collection dates
and the date difference between them.
Table 4. In-Situ and Remote Sensing Data Collection
3.2. Data Preparation
Given that the AOI spans across multiple RS imagery tiles, there were a few steps that
needed to be implemented prior to applying NDWI and NDCI. Similarly, the in-situ data is
formatted as a .csv file with multiple sheets that needed to be condensed and formatted prior to
being imported into ArcGIS Pro 2.9.2. This section explains how the different datasets were
prepared for use.
3.2.1. Landsat Imagery
Landsat 5 TM and Landsat 8 OLI surface reflectance data spanning from 2002 to 2020
were acquired through USGS EarthExplorer. The EarthExplorer interface allows public users to
acquire RS imagery based on desired parameters, including satellites, sensors, temporal range,
and cloud coverage percentage. The selected RS data for this study was downloaded in .TIFF file
format via a compressed folder. The .TIFF files were unzipped upon download and imported into
ArcGIS Pro 2.9.2. Attributing to the extended temporal range of this study, RS imagery
Year Landsat Satellite
Satellite Collection
Date

In-Situ
Collection Date
Date Difference Known Limitations
2002 Landsat 5 TM 07/04/2002 N/A N/A In-situ data not
available prior to 2004
2005 Landsat 5 TM 06/26/2005 06/21/2005 05-Days Missing latitude and
longitude values
2008 Landsat 5 TM 08/21/2008 08/20/2008 01-Days
2011 Landsat 5 TM 06/11/2011 06/02/2011 09-Days
2014 Landsat 8 OLI 06/19/2014 05/28/2014 22-Days Exceeds 16-Day
criteria due to cloud
coverage
2016 Landsat 8 OLI 05/23/2016 06/06/2016 13-Days
2020 Landsat 8 OLI 01/11/2020 01/14/2020 03-Days In-situ data only
collected in January
41

originating from Landsat 8 OLI and Landsat 5 TM imagery were used to capture all the
preselected years. These two satellites have different sensors aboard and render RS bands with
differing spectral ranges. Accordingly, the bands imported into the geographic information
software are dependent on the year and associated satellite. For Landsat 5 TM imagery (i.e.,
2002, 2005, 2008 and 2011), Band 1, Band 2, Band 3, and Band 4 were introduced. For Landsat
8 OLI imagery (i.e., 2014, 2016 and 2020), Band 2, Band 3, Band 4, and Band 5 were imported.
Table 3 provides the spectral ranges of bands originating from Landsat 5 TM and Landsat 8 OLI.
At the time of import, the coordinate system for all raster band layers had not been
predefined and were consequently projected to the NAD 1983 (2011) State Plane California VI
(WKID: 0406) coordinate system. The Californian coordinate system is based on the North
American Datum of 1983 and the Lambert conformal projection. The coordinate system divides
the State of California and its counties into 6 different zones. Given that the Salton Sea extends
across both the Riverside and Imperial counties, Zone IV of the California State Plane was
selected for use.
After the RS bands had been projected, similar RS bands in a given year were combined
into a single raster using the Mosaic tool. Figure 12 illustrates the mosaicked process of
combining two raster layers into a singular raster layer.

Figure 12. Mosaicked band 2 (2002)
42

3.2.2. In-Situ Data
The data was obtained from the BoR Lower Colorado Region website and downloaded in
.csv file format. The original dataset contains multiple WQ parameter measurements taken at the
various sampling stations, including depth, temperature, dissolved oxygen, pH, turbidity,
chlorophyll-a, total dissolved oxygens, total suspended solids, calcium, magnesium, sodium,
alkalinity, nitrate, salinity and phosphorous. Water samples were collected at multiple sampling
stations, however, only six stations consistently provided data for the specified temporal range.
Among those six stations, only three stations (SS1, SS2 and SS3) were in the Salton Sea. The
other three stations were located at offsite locations, such as the Alamo River (AR), New River
(NR) and Whitewater River (WWR). Note that these are streams of water that drain into the
Salton Sea. Figure 13 illustrates the study area and the location of the different sampling stations.
The original in-situ dataset was modified to include only chlorophyll-a measurements
that originated from SS1, SS2 and SS3 and spanned between May and August (January 2020
being an exception). For sampling events that captured two chlorophyll-a measurements, the
average of the two measurements was recorded. This process allows us to capture a more
accurate representation of the chlorophyll-a concentration. Moreover, the selected 2005 sampling
event had missing latitude and longitude values that were replaced with values from the prior
sampling event. The latitude and longitude for sampling stations vary slightly within each
sampling event, less than one quarter of a mile.
After the in-situ dataset had been reformatted, it was uploaded as a single table into
ArcGIS Pro 2.9.2. The table was then converted to a point feature layer using the XY to Point
tool. This geoprocessing tool can generate a point layer based on latitude, longitude and z values
obtained from a .txt or .csv file. Once the points had been created, they were projected to the
43

NAD 1983 (2011) State Plane California VI (WKID: 0406) coordinate system to ensure that all
the new point features aligned with the raster data layers.

Figure 13. Sampling stations
3.3. NDWI, NDCI, 2BDA and 3BDA
This section introduces the application of NDWI, NDCI, 2BDA and 3BDA. After the RS
imagery data had been projected and mosaicked, the NDWI expression was applied to extract the
body of water. In recent years, the Salton Sea has experienced a decline surface area and the
NDWI method allows for the water feature to be accurately delineated, see Figure 14. Surface
area results will be discussed further in Chapter 4. The Raster Calculator tool was used to
perform the NDWI expression, see the following equations:
𝑁𝐷𝑊𝐼
9:.6;:< = >/
=
":.6 4#":.6 ?
":.6 4%":.6 ?
(9)
44

𝑁𝐷𝑊𝐼
9:.6;:< @ A9'
=
":.6 B#":.6 =
":.6 B%":.6 =
(10)
This process generated a new NDWI raster layer that enhanced water surfaces and ranged from -
1 to 1, see Table 2. To extract the extent of the body of water, the raster layer needed to be
converted to a binary layer prior to being converted to a polygon feature. Before proceeding with
these steps, the symbology of the raster layer needed to be altered from stretch to classify.
Classify symbology allocates a color to each group of values and allows users to determine the
number of classes and method to which the values will be grouped. This additional step was
necessary to facilitate the reclassification process that would need to occur thereafter. Through
the use of two classifications and the manual method, water features were classified into a
separate grouping from vegetation and other land surfaces.

Figure 14. Salton Sea surface area
45

After modifying the symbology, the NDWI values were reclassified to 1 and 2 values.
The reclassified layer was subsequently converted to a polygon layer, pixels with similar values
were grouped into individual polygons. This step facilitated the selection of the body of water
and its export as a new feature layer. The executed layer consisted of a water boundary for a
specific year. The new water boundary layers were then used to clip the initial mosaicked raster
layers to the extent of the body of water.
Once the mosaicked raster layers had been clipped, the next step consisted of assessing
the presence of chlorophyll-a in the Salton Sea. This was implemented using the following NDCI
formulas within the Raster Calculator tool:
𝑁𝐷𝐶𝐼
9:.6;:< = >/
=
":.6 ?#":.6 B
":.6 ?%":.6 B
(11)

𝑁𝐷𝐶𝐼
9:.6;:< @ A9'
=
":.6 =#":.6 ?
":.6 =%":.6 ?
(12)
The tool rendered a layer consisting of NDCI values that ranged from 1 to -1. To ensure that all
years were comparable, the symbology was altered from stretch to classify. The classifications
were determined based on the distribution of the NDCI values across the different years (see
Appendix A for distributions).
To assess NDCI differences between the years, NDCI differencing was applied using the
following expression:
∆𝑁𝐷𝐶𝐼 =𝑁𝐷𝐶𝐼
C3
−𝑁𝐷𝐶𝐼
D4
(13)
The subtraction of pixels generated a raster layer that ranged from -1 to 1, zero values indicated
no change. Positive values indicated negative change and negative values indicated positive
change in chlorophyll-a presence.
46

The NDCI was then compared against other 2BDA and 3BDA to assess its accuracy and
performance. Watanabe et al. (2015) and Buma and Lee (2020) compared the NDCI to various
2BDA and 3BDA and their findings revealed NIR-Red, NIR-Green, NIR-Blue and 3BDA had
generated higher adjusted R-squares than the NDCI. Higher adjusted R-square values are
indicative of good model fit and 2BDA and 3BDA were therefore incorporated within the study.
The NIR-Red ratios were calculated using the following equations:
𝑁𝐼𝑅−𝑅𝑒𝑑
9:.6;:< = >/
=
":.6 ?
":.6 B
(14)
𝑁𝐼𝑅−𝑅𝑒𝑑
9:.6;:< @ A9'
=
":.6 =
":.6 ?
(15)
The Raster Calculator tool was then used to calculate the following NIR-Green expressions:
𝑁𝐼𝑅−𝐺𝑟𝑒𝑒𝑛
9:.6;:< = >/
=
":.6 ?
":.6 4
(16)
𝑁𝐼𝑅−𝐺𝑟𝑒𝑒𝑛
9:.6;:< @ A9'
=
":.6 =
":.6 B
(17)

After the NIR-Green expression had been applied across all years, the NIR-Blue equation was
applied:
𝑁𝐼𝑅−𝐵𝑙𝑢𝑒
9:.6;:< = >/
=
":.6 ?
":.6 3
(18)
𝑁𝐼𝑅−𝐵𝑙𝑢𝑒
9:.6;:< @ A9'
=
":.6 =
":.6 4
(19)

Lastly, the Raster Calculator tool was used to calculate the following 3BDA suggested by Buma
and Lee (2020):
3𝐵𝐷𝐴
9:.6;:< = >/
=
":.6 3#":.6 B
":.6 4
(20)
3𝐵𝐷𝐴
9:.6;:< @ A9'
=
":.6 4#":.6 ?
":.6 B
(21)
47

3.4. Global Regressions
Regression analysis is a statistical method used to assess the relationship between
dependent and explanatory variables. Linear regressions were performed for NDCI, NIR-Red,
NIR-Green, NIR-Blue and 3BDA. The regression models will assess the accuracy in which
NDCI estimates chlorophyll-a concentrations in the Salton Sea and how it performs against the
other 2BDA and 3BDA. The linear regressions were performed in ArcGIS Pro 2.9.2 and the
program requires the dependent and independent variables to be within the same dataset.
Because raster values will need to be extracted and conjoined to the point layer, a neighborhood
operation was performed on the raster layers prior to the extraction of values. This additional
step will ensure that the extracted values represent the average value of 3 x 3 neighborhood cells
(rectangle shape), instead of a singular point. After this step had been completed, a spatial
analyst tool was used to extract the raster values to the sampling points and the raster values
were recorded in the attribute table of the output feature layer.
To perform a linear regression, it is recommended that the dataset contain at least ten data
observations for each independent variable. The in-situ dataset only captured chlorophyll-a
measurements from three sampling sites that were directly on the Salton Sea. This meant that
each year (excluding 2002) had three data points observations that could contribute to the linear
regression and would individually not meet the observation criteria. To address this limitation,
the data points across all years were stacked within a single table. The new table was created in
Microsoft Excel, reimported into ArcGIS 2.9.2, and projected accordingly to the NAD 1983
(2011) State Plane California VI (WKID: 0406) coordinate system.
48

3.4.1. Linear Regression Assumptions
There are four linear assumptions that can affect the output results of a model. The first
assumption is linearity, and it assumes the relationship between dependent and independent
values is linear. The second assumption is homoscedasticity, and it assumes equal variance in the
residuals. The third assumption is no multicollinearity, and it assumes independent values are not
correlated within one another. The last assumption is the Gaussian distribution of error terms,
and it assumes residuals are normally distributed.
In an attempt to assess whether the dataset meets the linear assumptions, explanatory data
analysis was conducted on all variables. The dataset was constructed around in-situ data
measurements and the algorithm ratios were extracted and subsequently adjoined to the dataset.
Table 5 provides a summary statistic of all the variables, and it includes their minimum,
maximum, mean, median, standard deviation and sample size. Average chlorophyll-a is the
independent variable for all the regression models, and it has the largest range of values from
4.15 to 256.56. The median and mean value are widely apart at 46.64 and 72.98 respectively and
suggest high variance. In contrast, all the dependent variables tend to have smaller ranges
between their minimum and maximum values. The sampling size of all variables is on the
smaller side with 18 samples; however, it meets the 10 observations per variable threshold.
Table 5. Summary of Statistics
Variable Type Minimum Maximum Mean Median Std. Deviation Sample Size
NDCI Dependent -0.05 -0.01 -0.02 -0.01 0.01 18
NIR-Red Dependent 0.88 0.97 0.95 0.96 0.02 18
NIR-Green Dependent 0.81 0.95 0.92 0.93 0.37 18
NIR-Blue Dependent 0.92 1.00 0.96 0.96 0.02 18
3BDA Dependent -0.03 0.97 0.31 -0.00 0.46 18
Avg. Chlorophyll Independent 4.15 265.56 72.98 45.64 72.98 18
49

Scatterplots are commonly used to evaluate the linearity assumption. They are able to
provide insight into the relationship between variables and can be used to identify the degree of
linearity and slope. Figure 15 displays the relationship between the NDCI and average
chlorophyll-a variables. The R
2
for these two variables is low at 0.01 and it indicates that the
relationship between the variables is not linear. Based on the distribution of points, the slopes
appear to be negative with strong gaps between the values.

Figure 15. Relationship between average chlorophyll-a and NDCI
When we examine the relationship between NIR-R and average chlorophyll-a, the
relationship appears to be non-linear, see Figure 16. The R
2
is 0.01 and the points distribution
suggests a negative slope with large gaps between values. There are also outliers that indicate
large and small average chlorophyll-a values.

Figure 16. Relationship between average chlorophyll-a and NIR-Red
50

Figure 17 displays the relationship between NIR-G and average chlorophyll-a. The R
2
in
this relationship is 0, indicating no linearity between the independent and dependent variables.
The distribution of points also suggests the slope is near 0. When compared to NIR-Red, both
2BDA demonstrate the same outliers within the independent variable.

Figure 17. Relationship between average chlorophyll-a and NIR-Green
Figure 18 illustrates the relationship between NIR-B and average chlorophyll-a. The R
2

in this relationship is 0 and it indicates there is no linearity between the variables. Compared to
the other 2BDA, NIR-Green has a more scattered distribution making it difficult to assess
whether it has a negative or positive association.

Figure 18. Relationship between average chlorophyll-a and NIR-Blue
Lastly, Figure 19 illustrates the relationship between 3BDA and average chlorophyll-a.
The R
2
in this relationship is 0.33 and the clustering of points at opposite ends suggests a
51

bimodal distribution. Compared to NDCI and 2BDA, 3BDA has a larger gap between values and
no clear association.

Figure 19. Relationship between average chlorophyll-a and 3BDA
The scatterplot findings align with the individual histograms of the variables shown in
Figure 20. In the figure, the NDCI, NIR-R and NIR-G values skew largely to the right. In
contrast, average chlorophyll-a and NIR-B tend to skew to the left. The 3BDA values are
bimodal, as data is clustered away from the mean and median. Non-linear relations are normally
addressed by transforming the independent or dependent variables using a linear transformation.
52

Figure 20: Histograms of variables
The distribution of error terms were evaluated using quantile-quantile plots. Figure 21
illustrates that the error terms are not normally distributed across the majority of the variables.
NDCI, NIR-R, NIR-B and NIR-G have larger gaps between values and are lightly tailed. They
also have a comparable number of outliers within third quadrant. In regard to 3BDA and average
chlorophyll-a, the error terms are heavily tailed and more rightly skewed, respectively.
53

Figure 21. Quantile-Quantile plots of variables
As previously described, an assumption of linear regression is no multicollinearity among
the independent variables. In this study, each linear regression was ran with one independent
variable and one dependent variable. This means that there was no correlation among the
54

independent variables, as there was only one. This assumption is normally assessed using the
variance inflation factor.
To address the issue of linearity and distribution, the independent and dependent
variables were transformed using linear transformations. Negative values and outliers were taken
into account when selecting the box-cox transformation for the dependent variables. For the
independent variable, a logarithm transformation was used to correct the skewness of the
variable values. Despite the transformations, the R
2
for majority of the variables remained low.
For example, NDCI and NIR-Red increased to an R
2
of 0.08. While NIR-Green rose to a value of
0.21 and 3BDA maintained an R
2
of 0.33. NIR-Blue maintained a low R
2
after transformation
with 0.02. Collectively, the R
2
are low and still demonstrate the low linearity between the
variables. To further offset low linearity, outliers were removed from the dataset. The alternative
solution would be to include additional data points to fill in the large value gaps, but this was not
readily available.
3.4.2. Linear Regression
The Generalized Linear Regression is a tool that can be used to fit continuous, binary and
count models. The linear performs an array of diagnostic tests that help us understand whether
the model is useful or if there is additional work to be done (Esri n.d.). The first diagnostic
examines the significance and robustness of explanatory variables. The second diagnostic
observes the coefficient of each explanatory variable to ensure they capture the relationship you
are expecting, whether it be positive or negative. The third diagnostic assesses whether the model
includes redundant variables through the computation of the variance inflation factor (VIF). The
fourth diagnostic considers the distribution of the model’s residuals through the Jarque-Bera test.
55

The fifth diagnostic evaluates the overall performance of the model through adjusted R
2
and
Akaike’s information criterion (AIC).
For the linear regression, the NDCI values were imputed as the dependent variable and
the in-situ chlorophyll-a measurements as the independent variable. Each data observation was
examined using a continuous model (ordinary least squares). Ordinary least squares is a
continuous model used estimate coefficients and model the relationship between dependent and
explanatory variables (Esri n.d.). The process was repeated for NIR-Red, NIR-Green, NIR-Blue
and the 3BDA. Linear regression results are summarized in Chapter 4.

56

Chapter 4 Results
This chapter presents chlorophyll-a patterns identified in the temporal analysis and the results of
the linear regression analysis. NDCI values, NDCI changes and linear regression results are
quantified, tabulated, and spatially presented in maps across the temporal range of the study.
Section 4.1 describes chlorophyll-a presence in the Salton Sea from 2002 to 2020 and evaluates
the NDCI differences across those years. Section 4.2 describes the linear regression results of
NDCI, 2BDA and 3BDA.
4.1. Temporal Analysis of Chlorophyll-a Presence using NDCI
This section explores how chlorophyll-a concentrations temporally fluctuated and
steadily increased throughout the years. The NDCI was used to assess chlorophyll-a
concentrations in the Salton Sea and the raster layers range between -1 and 1. Higher values
represent higher concentrations of chlorophyll-a in the waterbody. Table 6 illustrates the
correlation between NDCI values and in-situ chlorophyll-a (μg/L) concentrations. These pixel
ranges and associated concentrations were first introduced by Mishra and Mishra (2012) and
were used in this study to quantify chlorophyll-a concentrations within the Salton Sea.
While NDCI values range from -1 to 1, the symbology of each NDCI map was chosen to
highlight changes in pixel values. The symbology classifications were based on the distribution
of pixel values across all years. The symbology was then applied to all NDCI layers to ensure
they are comparable against one another. Moreover, Table 7 was adopted from Tas, Can and
Koloren (2011) to explain the correlation between chlorophyll-a concentrations (μg/L) and
trophic states.

57

Table 6. NDCI Pixel Range (Adopted from Mishra and Mishra 2012)

In addition to calculating NDCI values for all years, NDCI differences were calculated
between two selected years (e.g., 2002 and 2005). The NDCI difference layers also had their
symbology modified based on pixel value distribution. Pixel values depicted in shades of blue
experienced a decrease in chlorophyll-a concentrations from the first year to the second year.
Whereas pixel values depicted in shades of red experienced an increase in chlorophyll-a
concentration from the first year to the second year. Darker hues for either color translated to
higher pixel differences and lighter hues to minimal or no pixel difference.
Table 7. Chlorophyll-a and Trophic States (Adopted from Tas, Can and Koloren et al. 2011)
4.1.1. Chlorophyll-a Presence in 2002
In 2002, the Salton Sea extended 365.71 square miles. A total of 1,052,428 pixel values
were assessed for NDCI. The pixel values had a range of 0.24, a minimum of -0.18 and a
NDCI Range Chlorophyll-a Range (μg/L)
-1.00 to -0.10 <7.5
-0.10 to 0.00 7.5 to 16
0.00 to 0.10 16 to 25
0.10 to 0.20 25 to 33
0.20 to 0.40 33 to 50
0.40 to 0.50 > 50
0.50 to 1.00 Severe Algae Bloom
Trophic State Levels Average Max
Ultraoligotrophic <1 < 2.5
Oligotrophic <2.5 <8
Mesotrophic 2.5 – 8 8 - 25
Eutrophic 8 - 25 25 -75
Hypereutrophic > 25 >75
58

maximum of 0.05. The pixel values also had a mean of -0.31 and a median of -0.02. As shown in
Figure 22, most of the visible pixel values are less than -0.02. According to the pixel range
introduced by Mishra and Mishra (2012), the chlorophyll-a concentration should range between
7.5 to 16 μg/L. Chlorophyll-a concentrations that range between these values tend to indicate
eutrophic bodies of water, high nutrient concentrations and condensed plant populations.
Although, the NDCI values are minimal there is discernable differences at the mouths of the
WWR, AR and NR. NDCI values are noticeably higher in AR and NR and may be indicative of
higher agricultural runoff water and higher pesticide concentrations. In contrast, the WWR has
smaller NDCI values along its cove and the higher NDCI values are constrained to the shoreline.

Figure 22. NDCI 2002
59

4.1.2. Chlorophyll-a Presence in 2005
In 2005, there is a slight reduction in the Salton Sea with a surface area of 364.62 square
miles. A total of 1,049,292 pixel values were assessed in the NDCI. The range is 0.24, the
minimum is -0.17 and the maximum is 0.06. Compared to the 2002 pixel values, the range
remained the same and the overall NDCI range moved by 0.01 or less. The mean is -0.03 and the
median is -0.02, indicating a chlorophyll-a concentration that ranges between 7.5 to 16 μg/L. In
Figure 23, one can also detect a small plume of higher NDCI values located northwest of the AR
and these values range between 0.01 and 0. At the WWR mouth, values range between 0.02 to
0.06, indicating a chlorophyll-a concentration of 16 to 25 μg/L. The AR and NR also appear to
demonstrate higher concentrations, especially when compared the rest of pixel values in the
Salton Sea.
60

Figure 23. NDCI 2005
Figure 24 illustrates the NDCI difference amid 2002 and 2005 and 48 percent of all the
pixel values fall within the 0.00 to 0.05 array. This range indicates that there was an overall
decrease or no change in NDCI values between 2002 and 2005. The larger reduction of pixel
values occurred alongside the northern and western shorelines of the Salton Sea. Less than 4
percent of NDCI values had pixel changes that differed more than -0.05.
61

Figure 24. NDCI difference from 2002 to 2005
4.1.3. Chlorophyll-a Presence in 2008
By 2008, the Salton Sea had been further reduced in size with a surface area that
extended 356.61 square miles. For the 2008 NDCI assessment, a total of 1,026,255 values were
evaluated. The range of these values is 0.34, the maximum is 0.06 and the minimum is -0.28
(Figure 25). From a visual perspective, 2008 appears to have NDCI values that are mainly less
than -0.02 with smaller areas fluctuating between -0.02 and 0. Based on the NDCI pixel range,
chlorophyll-a concentrations still range between 7.5 to 16 μg/L. When we observe the rivers
mouths, a noticeable difference is their overall reduction in surface area and lower NDCI values.
There is however a visible area surrounding the perimeter of the Salton Sea that appear to have
elevated NDCI values compared to the rest of the extent.
62

Figure 25. NDCI 2008
Compared against 2005 values, the range has grown by 0.10 values and the maximum has
grown by 0.03 values (Figure 26). Whereas the differentiation between 2002 and 2005 had lower
NDCI values along the west shoreline of the Salton Sea, the 2005 and 2008 differentiation reveal
that there has been an increase in NDCI values within this year range.
63

Figure 26. NDCI difference from 2005 to 2008
4.1.4. Chlorophyll-a Presence in 2011
In 2011, the Salton Sea further reduced in size to 352.32 square miles. A total of
1,013,895 pixel values were processed for the NDCI, the range was 0.19, the minimum was -
0.12 and the maximum was 0.06. The mean and median pixel value for this year was -0.01.
Compared to previous year, 2011 has the shortest range at 0.19 and the smallest maximum at
0.19. The maximum value of 0.06 has remained consistent between 2008 and 2011. NDCI values
across the extent of the waterbody appear to have risen and broaden from <0.02 to a range
between -1 and 0 (Figure 27). When we examine the pixel count per classification, 44 percent of
all NDCI values fall within the -0.02 to -0.01 range and 17 percent fall between -0.01 to 0. In
64

terms of the WWR, AR and NR mouths, there is a notable decrease in NDCI values. This is
especially true when compared to the original 2002 values.

Figure 27. NDCI 2011
In Figure 28, the NDCI values from 2008 and 2011 were differenced and the majority of
the Salton Sea extent is shaded in light red and blue shades. This suggests that NDCI values
between 2008 and 2011 largely remained the same and if they deviated from the previous year, it
was by less than 0.05 pixel values. The dark red plume at the southern part of the Salton Sea
implies that there was a higher increase in NDCI values from 2008 to 2011.
65

Figure 28. NDCI difference from 2008 to 2011
4.1.5. Chlorophyll-a Presence in 2014
For 2014, a total of 997,611 pixel values were assessed for the NDCI, given that the
surface area of the Salton Sea extended only 346.66 square miles. The range for these pixels was
0.24, the minimum was -0.17 and the maximum was 0.06. These values deviate only by 0.01
from the original NDCI values in 2002. However, in 2014 there are three notable NDCI ranges: -
1 to -0.02, -0.02 to -0.01 and -0.01 to 0. The -0.02 to -0.01 had the largest surface area at 44
percent and this differentiates from previous years that had overall lower NDCI values (Figure
29). When we examine the river mouths, AR and NR continue to have lower NDCI values and
WWR continues to display more elevated ranges. Seen as most of the values remain below 0,
chlorophyll-a concentration should continue to range between 7.5 to 16 μg/L.
66

Figure 29. NDCI 2014
In comparing 2011 to 2014, it is evident that there has been a general decline in NDCI
values along the southern shoreline (see Figure 30). This may be linked to lower water deposits
or lower nutrient concentrations within the southern water streams, given that high nutrient
concentrations prompt the production of algae bloom and chlorophyll-a. The remainder of the
Salton Sea extent experienced smaller deviations from 2011 NDCI values.
67

Figure 30. NDCI difference from 2011 to 2014
4.1.6. Chlorophyll-a Presence in 2016
In comparison to the previous year, the Salton Sea in 2016 had a decreased surface area
of 341.10 square miles. A total of 981,626 pixels were assessed for NDCI and the mean and
median was -0.03 and -0.02 respectively. The range for this year increased to 0.31from 0.24 in
2014. The minimum value was -0.17 and the maximum value was 0.10, indicating that there
were areas within the Salton Sea that had chlorophyll-a concentration between 15 to 25 μg/L.
NDCI values largely ranged between -1 to 0, most of the -0.02 to -0.01 pixels were located at the
northern portion of the Salton Sea (Figure 31). NDCI values for AR and NR appear to remain the
same to 2014, whereas WWR illustrates an overall decrease in NDCI values. Along the lower
68

west shoreline, there also appears to be an increase in NDCI values from the -0.02 to -0.01 range
to the -0.01 to 0 range.

Figure 31. NDCI 2016
In Figure 32, we can examine NDCI difference between 2014 to 2016, the difference in
NDCI values does not appear to be too great. Although not visible to eye, the largest difference
occurred at the southern eastern area of the Salton Sea. In this area, there was a difference of
0.10 pixel values from the previous year.

69

Figure 32. NDCI differenced from 2014 to 2016
4.1.7. Chlorophyll-a Presence in 2020
By 2020 the Salton Sea had a surface area of 326.75 square miles, the majority of its
NDCI values range from -0.02 to -0.0. These values indicate a continual chlorophyll-a
concentration of 7.5 to 16 μg/L. A total of 940,317 pixel values were assessed in the NDCI, the
range is 0.27, the minimum is -0.19 and the maximum is 0.08. In contrast to early years, the AR
has NDCI values less than -0.02. Both the NR and WWR had higher NDCI values along their
shoreline (Figure 33). In the middle of the Salton Sea, we can also discern small areas with
higher chlorophyll-a concentrations.
70

Figure 33. NDCI 2020
The quantifiable difference between 2016 and 2020 reveals a large increase in
chlorophyll-a concentrations within the southern portion of the Salton Sea (Figure 34). Note that
this is the same area that experienced a decline in NDCI from 2014 to 2016.
71

Figure 34. NDCI differences from 2016 to 2020
NDCI difference values from 2002 to 2020 reveal that there has been a general increase
in chlorophyll-a concentrations within the Salton Sea (Figure 35). The majority of the variation
between these years is minimal, as it fluctuates between 0 to 0.05 values. However, these
findings indicate that the Salton Sea is largely eutrophic and that the overall chlorophyll-a
concentrations are slowly rising. The northern and western areas of the Salton Sea experienced
the greater change in water quality, value differentiations were as high as -0.10. We can also see
that the southern portion of the Salton Sea experienced a slighter decrease in NDCI values,
specifically around the AR mouth.
72

Figure 35. NDCI differences from 2002 to 2020
4.2. Assessment of NDCI against 2BDA and 3BDA
Table 8 provides a summary overview of the linear regression. The NDCI regression
model demonstrated a weak linear fit between the dependent and independent variable, meaning
that the independent variable was unable to add value to the model. This regression model had an
AIC
2
value of -104.40 and higher AIC
2
values are indicative of poor model performance;
regardless if the value is negative or positive. The NDCI regression also had a small R
2
of -
0.0005 that indicates a low level of correlation between variables. Because the p-value for
average chlorophyll-a was greater than the significance level of 0.05, we fail to reject the null
hypothesis that the coefficients are not zero.
73

Table 8. Summary of Linear Regression Results

For NIR-Red, the intercept is 0.9737 and the coefficient is -0.0021. This means that an
increase in average chlorophyll-a results in a 0.0021 decrease in NIR-Red values. The standard
error is 0.0021 and it suggests that the sample population does not deviate too largely from a true
mean population. The t-value is -0.9836 and the p-value for the coefficient of average
chlorophyll -a is 0.35. Since the p-value is greater than the significance level of 0.05, the variable
is not statistically significant, and we are unable to reject the null hypothesis that the coefficient
is not zero. The AIC
2
for NIR-Red is -87.25 and it indicates a relatively poorer model compared
to NIR-Blue model. The R
2
is low at -0.0043 and it suggests that the model is doing a poor job at
explaining the variance of the dependent variable.
Unexpectedly, out of all the regression models NIR-Green had the highest R
2
at 0.13. The
AIC
2
value remains larger than the NIR-Blue model, which had the lowest AIC
2
out of all the
models. In a study conducted by Watanabe et al. (2015), NIR-Green had the lowest R
2
compared
to the rest of the NIR-based 2BDA. The 3BDA model had an R
2
of 0.0036 and while it remains
low, it was the model with the second highest value. It was difficult to assess the performance of
the NDCI model against the rest of the regression models, when the R
2
were negative, and the
independent variable proved to statistically insignificant.
Table 9 provides a list of the in-situ measurements and their associated NDCI values. The
table also includes the expected NDCI range given the in-situ measurement and the correlated
Band Ratios Intercept Coefficient Std. Error t-Statistic P-Value R
2
AIC
2

NDCI 0.9866 -0.0007 0.0011 -0.9661 0.3547 -0.0055 -104.4077
NIR-Red (2BDA) 0.9737 -0.0021 0.0021 -0.9836 0.3511 -0.0043 -87.2594
NIR-Green
(2BDA)
1.9503 -0.0029 0.0017 -1.7151 0.1143 0.1392 -92.9739
NIR-Blue (2BDA) 0.9640 0.9640 0.0044 0.5193 0.6138 -0.0648 -68.2995
3BDA 0.0049 -0.0341 0.0003 -1.0250 0.3240 0.0036 -79.1229
74

trophic state based on the in-situ measurements. When we compare the NDCI and in-situ
measurements, it is evident that the NDCI underestimated the chlorophyll-a concentrations and
assigned incorrect NDCI values. Chlorophyll-a concentrations that varied more than 95μg/L
were assigned the same NDCI values. Between 2005 and 2008, we can see that Salton Sea had
the highest fluctuation in chlorophyll-a concentrations. Based on these limited in-situ
measurements, we can conclude that the Salton Sea is moving toward a hypereutrophic state
versus a eutrophic state.
Table 9. Comparison of NDCI and In-Situ Measurements
Year
In- Situ Chlorophyll-a
Concentrations (μg/L)
Measured NDCI
Value
Expected NDCI
Range
In-Situ Trophic
State
2005 229.44 -0.05 0.50 to 1.00 Hypereutrophic
2005 265.56 -0.02 0.50 to 1.00 Hypereutrophic
2005 171.26 -0.02 0.50 to 1.00 Hypereutrophic
2008 20.34 -0.04 0.00 to 0.10 Eutrophic
2008 4.16 -0.02 -1.00 to -0.10 Mesotrophic
2008 6.26 -0.06 -1.00 to -0.10 Mesotrophic
2011 10.56 -0.01 -1.00 to -0.10 Eutrophic
2011 17.57 -0.01 0.00 to 0.10 Eutrophic
2011 11.07 -0.01 0.00 to 0.10 Eutrophic
2014 112.70 -0.01 0.50 to 1.00 Hypereutrophic
2014 47.79 -0.02 0.20 to 0.40 Hypereutrophic
2014 42.91 -0.02 0.20 to 0.40 Hypereutrophic
2016 26.66 -0.02 0.10 to 0.20 Hypereutrophic
2016 46.87 -0.02 0.20 to 0.40 Hypereutrophic
2016 44.42 -0.02 0.20 to 0.40 Hypereutrophic
2020 71.56 -0.02 0.50 to 1.00 Hypereutrophic
2020 130.69 -0.02 0.50 to 1.00 Hypereutrophic
2020 53.96 -0.02 0.50 to 1.00 Hypereutrophic
75

Chapter 5 Discussion
This study conducted a temporal analysis of chlorophyll-a concentrations within the Salton Sea
from 2002 to 2020. The study also assessed the use of NDCI against 2BDA and 3BDA using
linear regression models. The contributions of this work lie within developing a more holistic
and efficient approach towards monitoring, predicting and safeguarding bodies of water. The use
of RS imagery to conduct WQ assessments, as opposed to traditional WQ methods, ensures
greater temporal and spatial coverage. The use of RS imagery can also aid policymakers and
community members in avoiding the high costs associated with conventional water sampling
methods while still making progress in monitoring and improving WQ. The results demonstrate
the potential of incorporating RS imagery in WQMPs, but they still show the value in the
continued collection of in-situ measurements. The linear regression findings revealed that the
NDCI underestimated chlorophyll-a concentrations by a wide margin. The significance of these
findings is that they highlighted the challenges and limitations of using RS imagery in WQ
assessment. The remainder of this chapter summarizes the main findings from the temporal
analysis and the linear regression models. A discussion of known limitations and challenges
within the study are also presented. Lastly, areas of potential future work and analysis
application are identified and discussed.
5.1. Discussion of Results
The temporal analysis evaluated chlorophyll-a concentrations in the Salton Sea. This
process generated NDCI values for all the years considered within the selected temporal range
(i.e., 2002, 2005, 2008, 2011, 2014, 2016 and 2020). The NDCI values primarily ranged between
-0.02 and 0.02. This NDCI range indicates that the Salton Sea has maintained concentrations
between 7.5 to 17 µg/L and 16 to 25 µg/L from 2002 to 2020. These chlorophyll-a
76

concentrations are in agreement with values measured in other eutrophic bodies of water and are
indicative of poor WQ within the Salton Sea, specifically in relation to eutrophication (Mishra
and Mishra 2012; Setmire et al. 2000). A eutrophic condition refers to an environment that is
enriched with nutrients and experiences high plant life density and dead oxygen zones (Cohen
2019; Setmire et al. 2000). The eutrophic state of the Salton Sea was an expected outcome, given
that it has served as an agricultural drainage since 1924 (Cohen 2019). Nevertheless, the extent
of eutrophication of the Salton Sea is alarming and has led to a large decrease in diversity of
aquatic life, increase in overall turbidity and has reduced the lifespan of the waterbody.
In terms of related work conducted on the Salton Sea, a study by Setmire et al. (2000)
examined inflow, nutrient, and chlorophyll-a concentrations in the Salton Sea from 1968-69 and
1999. The study findings revealed chlorophyll-a concentrations averaged 25 µg/L, indicating that
the Salton Sea had reached a eutrophic state as early as then. In 1999, the NR accounted for 46
percent of the inflow into the Salton Sea, the AR accounted for 32 percent and the WWR
accounted for 6 percent. Setmire et al. (2000) and Cohen (1999) recognized areas surrounding
the AR, NR and WWR experienced higher chlorophyll-a concentrations (an average of 70 µg/L)
compared to the rest of waterbody. The temporal trends found in this study illustrate minimal
variation of NDCI pixel values and indicate chlorophyll-a concentrations did not exceed 25
µg/L. Interestingly, when examining the difference in concentrations between 2002 and 2020, it
is evident that there has been a decrease in inflow from the AR and NR and increase in WWR.
Areas that were near the rivers experienced the highest change in in NDCI pixel values over the
years, but changes were less than 0.10 values. This means that the temporal analysis was unable
to identify areas that measured concentrations equal to or exceeding 70 µg/L.
77

Another goal of this study was to evaluate the performance of the NDCI in estimating
chlorophyll-a concentrations and assessing its performance against other band ratio algorithms.
This process generated statistical summaries for the NDCI, 2BDA and 3BDA models. None of
the linear regression results had p-values less than 0.05 to signify statistical significance of the
independent variable. The adjusted R2 for all the models was near 0 and it failed to indicate a
strong linear relationship between the independent and dependent variables. This finding
suggests Landsat imagery may not be suitable for estimating chlorophyll-a concentrations in the
Salton Sea using NDCI, 2BDA and 3BDA.
The conclusions of this study can build upon previous works in this space and more
insights can be gained when comparing these results against those found in similar studies. In the
RS community, Sentinel-2 imagery has long been favored to assess chlorophyll-a presence due
to the inclusion of the red-edge band. Buma and Lee (2020) evaluated the performance of NDCI,
2BDA, 3BDA and FLH using Sentinel-2 and Landsat 8 OLI imagery. Their findings revealed
that NDCI (R2 = 0.75) and 3BDA (R2 = 0.89) performed relatively better than FLH (R2 = 0.73)
and 2BDA (R2 = 0.71) when using Landsat 8 OLI imagery. When compared against Sentinel-2
imagery, NDCI and 3BDA had higher correlations with Worldview-3 measurements. Buma and
Lee (2020) findings deviate from those of this study, given that the adjusted R2 for NDCI was -
0.0055 and 0.0036 for 3BDA. The difference in results can be attributed to the use of
Worldview-3 values in Buma and Lee (2020), while this study relied on the use of in-situ
measurements.
Watanabe et al. (2015) conducted a study using Landsat OLI imagery and in-situ
measurements to examine multiple two-band and three-band models, including NIR-R, NIR-G
and NIR-B. Their findings revealed that the 2BDA (i.e., NIR-Red, NIR-Green and NIR-Blue)
78

performed better than the NDCI (R2 = 0.39). Most of the band ratios demonstrated acceptable
model fits, but the linear regression models underestimated chlorophyll-a concentrations and
incorrectly designated the associated tropic state. In this study, the NDCI values of the point
features were compared against the in-situ measurements. This evaluation revealed that the
NDCI also underestimated chlorophyll-a concentration in the Salton Sea by a wide margin.
Points that had high chlorophyll-a concentrations (>70 µg/L) were given the same NDCI values
as those that had low chlorophyll-a concentrations (> 5 µg/L). This finding shows limitations in
the use of RS imagery and emphasizes the importance of having a large enough dataset of in-situ
measurements to assess spatial indices. Moreover, the findings reveal that chlorophyll-a
concentrations have fluctuated greatly between 2002 and 2020. Since 2011, the sampling stations
have consistently recorded concentrations that exceed 25 µg/L and point toward a
hypereutrophic Salton Sea.
5.2. Limitations and Challenges
Several limitations and challenges exist within this study, which lead to uncertainty in the
results. A well-known constraint to performing traditional WQ assessment is the data collection
process and the high associated costs. The original in-situ dataset was constructed by the BoR,
which performed quarterly water sampling events at the Salton Sea from 2004 to 2020. In
WQMPs, the WQ data collection process is usually pre-planned as it requires assembling a
workforce and gathering water sensing equipment. The pre-planning process does not always
extend to the final product and a known limitation in this study was the data quality of the in-situ
dataset. The original dataset contained multiple data entry errors, including missing values,
duplicated items and unstructured data formats. The study had initially planned to evaluate the
Salton Sea every three years, beginning in 2002 to 2020. Missing values from the year 2017
79

made it difficult to maintain the intended temporal range. The original dataset also contained
entries for all the sampling events conducted since 2004. The location of the sampling events
varied across years and this inconsistency made it difficult to evaluate a single area temporally.
The second limitation of this study also results from issues in data-availability and is
centered on the limited in-situ sampling size. The study evaluated six years (i.e., 2005, 2008,
2011, 2014, 2016 and 2020) of in-situ measurements, which only had three consistent sampling
stations across all these years. This meant that between all six years, a total of only 18 data point
were acquired for the study. The large fluctuation in chlorophyll-a concentrations means that the
limited sampling size had a large range of values. When the data was being examined to meet
linear assumptions, the scatterplots would identify maximum and minimum values as outliers.
The values were removed in multiple iterations to increase the linearity between independent and
dependent variables, but new outliers would then emerge. This limitation could have been
mitigated had there been additional points to fill in the large gaps between in-situ values.
The third limitation of the study is the discrepancy between the days the RS imagery and
in-situ data were collected. The study used linear regressions to better understand the relationship
between band ratio values and in-situ measurements. It was important for the RS imagery to be
collected within a similar time window to when the in-situ data collection occurred to ensure
higher accuracy of assessment. However, when the in-situ measurements were collected, they
were not intended to be used in conjunction with satellite imagery and the data collection process
did not take into account cloud coverage, scanline errors or revisit intervals for Landsat satellites,
which are important considerations for RS imagery This meant that RS imagery selected for the
study were not always aligned with the days in which the in-situ measurements were collected,
which introduced large uncertainty in the final results.
80

The final limitation of this study was the temporal range of the analysis. The temporal
range spanned between 2002 to 2020 and required an RS program that was established prior to
2002. This limited the options in choice of RS programs that were publicly available and could
be used within the analysis. The only program that met all criteria was Landsat imagery,
however, this study showed the limitations of its use for WQ assessment. Sentinel-2 imagery
could have been an option because it incorporates a red-edge band, but it was not established
until 2014.
5.3. Future Work
Assessing chlorophyll-a concentrations is only one step in performing a complete WQ
assessments on a body of water. There is a large array of WQ parameters that are spectrally
active and can be assessed using RS imagery and RS techniques. Some of these spectrally active
WQ parameters include temperature, turbidity, salinity, and suspended sediments (Topp et al.
2022). This project can serve as a framework for future studies that can conduct WQ studies
using other parameters, as the analytical methods and workflows can be easily replicated. The
semi-empirical component of the NDCI means that the algorithm is not dependent on in-situ
measurements and can be reproduced in other bodies of water and across multiple temporal
ranges. The framework can also be implemented when using other spatial indices, as they will
require similar map algebra expressions. From a policy perspective, these analytical methods and
workflows can be beneficial to local government and non-profit organization that seek to create
more sustainable WQMPs.
Finally, future research can build upon this study by incorporating new satellite imagery
from other RS programs. For instance, the original NDCI expression contains a red-edge band
that captures the 748 to 778 nm spectral range. The red-edge band was replaced with a NIR band
81

when the NDCI was used with Landsat imagery. Thus, there is value in evaluating whether the
original NDCI expression can accurately assess chlorophyll-a presence within the Salton Sea.

82

References
“60 Years Ago First Satellite Image of Earth.” 60 Years Ago: First Satellite Image of Earth.
Kennedy Space Center, August 7, 2019.
Agusdinata, D.B., W. Liu, H. Eakin, and H. Romero. 2018. “Socio-Environmental Impacts of
Lithium Mineral Extraction: Towards a Research Agenda.” Environmental Research
Letters 13, no. 12: 16-20.
Ahn, J.M. and S. Lyu. 2020. “Selection of Priority Tributaries for Point and Non-Point Source
Pollution Management.” KSCE Journal of Civil Engineering 24, no. 4: 1060-1069..
Ajala, R., P. Persaud, J.M. Stock, G. S. Fuis, J.A. Hole, M. Goldman, and D. Scheirer. 2019.
“Three-Dimensional Basin and Fault Structure From a Detailed Seismic Velocity Model
of Coachella Valley, Southern California.” Journal of Geophysical Research: Solid Earth
124, no. 5: 4728-4750.
Ali, M.I., G.D Dirawan, A.H. Hasim and M.R. Abidin, 2019. “Detection of Changes in Surface
Water Bodies Urban Area With NDWI and MNDWI Methods. International Journal on
Advanced Science Engineering Information Technology 9, no. 3: 946-951.
Ansari, A.A., G.S. Singh, G.R. Lanza, and W. Rast, eds. 2010. Eutrophication: Causes,
Consequences and Control Vol. 1. Springer Science & Business Media
Armstrong, L.E. and E.C. Johnson. 2018. “Water Intake, Water Balance, and the Elusive Daily
Water Requirement.” Nutrients 10, no.12: 1-25.
Aulenbach, D.B. 1967. “Water--Our Second Most Important Natural Resource.” Boston College
Industrial and Commercial Law Review 9: 535-552.
Bartram, J. and R. Ballance, eds. 1996. 1996. Water Quality Monitoring: A Practical Guide to
the Design and Implementation of Freshwater Quality Studies and Monitoring
Programmes. CRC Press.
Beck, R., S. Zhan, H. Liu, S. Tong, B. Yang, M. Xu, Z. Ye, Y. Huang, S. Shu, Q. Wu, and S.
Wang. 2016. “Comparison of Satellite Reflectance Algorithms for Estimating
Chlorophyll-a in a Temperate Reservoir Using Coincident Hyperspectral Aircraft
Imagery and Dense Coincident Surface Observations.” Remote Sensing of Environment
178: 15–30.
Behmel, S., M. Damour, R. Ludwig, and M.J. Rodriguez. 2016. “Water Quality Monitoring
Strategies — A Review and Future Perspectives.” Science of The Total Environment 571:
1312–29.
Bhateria, R. and D. Jain. 2016. “Water Quality Assessment of Lake Water: A Review.”
Sustainable Water Resources Management 2, no. 2: 161–73.
83

Biswas, A.K. 2008. “Integrated Water Resources Management: Is It Working?” International
Journal of Water Resources Development 24, no. 1: 5–22.
Bradley, T., H. Ajami and W. Porter. 2022. “Ecological Transitions at the Salton Sea: Past,
Present and Future.” California Agriculture 76, no. 1: 8–15.
Brezonik, P., K.D. Menken, and M. Bauer. 2005. “Landsat-Based Remote Sensing of Lake
Water Quality Characteristics, Including Chlorophyll and Colored Dissolved Organic
Matter (CDOM).” Lake and Reservoir Management 21, no. 4: 373–82.
Brothers, D.S., N.W. Driscoll, G.M. Kent, R.L. Baskin, A.J. Harding, and A.M. Kell. 2022.
“Seismostratigraphic Analysis of Lake Cahuilla Sedimentation Cycles and Fault
Displacement History beneath the Salton Sea, California, USA.” Geosphere 18, no. 4:
1354–76.
Buma, W.G. and S.I. Lee. 2020. “Evaluation of Sentinel-2 and Landsat 8 Images for Estimating
Chlorophyll-a Concentrations in Lake Chad, Africa.” Remote Sensing 12, no. 15: 1-18.
Caballero, I., R. Fernández, O.M. Escalante, L. Mamán and G. Navarro. 2020. “New Capabilities
of Sentinel-2A/B Satellites Combined with in Situ Data for Monitoring Small Harmful
Algal Blooms in Complex Coastal Waters.” Scientific Reports 10, no. 1: 1-14.
Cannon, B.Q. 2022. “Irrigation, Reclamation, and Water Rights.” In A Companion to American
Agricultural History, 314–26. John Wiley & Sons, Ltd.
Cantor, A., and S. Knuth. 2019. “Speculations on the Postnatural: Restoration, Accumulation,
and Sacrifice at the Salton Sea.” Environment and Planning A: Economy and Space 51,
no. 2: 527–44.
Carpelan, L.H. 1958. “The Salton Sea. Physical and Chemical Characteristics.” Limnology and
Oceanography 3, no. 4: 373–386.
Cassidy, R., P. Jordan, M. Bechmann, B. Kronvang, K. Kyllmar, and M. Shore. 2018.
“Assessments of Composite and Discrete Sampling Approaches for Water Quality
Monitoring.” Water Resources Management 32, no. 9: 3103–3118.
Cheney, A.M., C. Newkirk, K. Rodriguez, and A. Montez. 2018. “Inequality and Health among
Foreign-Born Latinos in Rural Borderland Communities.” Social Science & Medicine
215: 115–22.
Chipman, J.W., L.G. Olmanson, and A.A. Gitelson, 2009. “Remote Sensing Methods for Lake
Management: A Guide for Resource Managers and Decision-Makers Developed By The
North American Lake Management Society In Collaboration With Dartmouth
College.” Madison, WI: University of Minnesota, University of Nebraska and University
of Wisconsin for the United States Environmental Protection Agency.
84

Clouse, L.E., 2016. “The History, Current State, and Future of Tourism at the Salton Sea,
California, USA.” Sustainability Of The Salton Sea: A Review of the Environmental,
Economic, And Social Climate Around California’s Largest Saline Lake, p.48.
Cohen, M.J., 2019. Past and future of the Salton Sea.
Cohen, M.J., J.I. Morrison and E.P. Glenn, 1999. Haven or hazard: The ecology and future of the
Salton Sea. Oakland, CA: Pacific Institute for Studies in Development, Environment and
Security.
Cohn, J.P., 2000. Saving the Salton Sea. Bioscience, 50(4), pp.295-301.
Crane, W.B. 1914. “The History of the Salton Sea.” Annual Publication of the Historical Society
of Southern California 9, no. 3: 215–24.
Dailey, K.R., K.A. Welch, and W.B. Lyons. 2014 “Evaluating the Influence of Road Salt on
Water Quality of Ohio Rivers over Time.” Applied Geochemistry 47: 25–35.
Damalas, C.A. 2009. Understanding Benefits and Risks of Pesticide Use. Scientific Research and
Essays 4, no. 10: 945-949.
Damalas, C.A. and I.G. Eleftherohorinos. 2011. “Pesticide Exposure, Safety Issues, and Risk
Assessment Indicators. International Journal of Environmental Research and Public
Health 8, no. 5: 1402-1419.
Das, S., S. Kaur, and A. Jutla. 2021. “Earth Observations Based Assessment of Impact of
COVID-19 Lockdown on Surface Water Quality of Buddha Nala, Punjab, India.” Water
13, no. 10: 1-15.
Delfino, K. 2006. “Salton Sea Restoration: Can There Be Salvation for the Sea.” Pacific
McGeorge Global Business & Development Law Journal 19, no. 1: 157-172.
Doede, A.L. and P.B. DeGuzman. 2020. “The Disappearing Lake: A Historical Analysis of
Drought and the Salton Sea in the Context of the GeoHealth Framework.” GeoHealth 4,
no. 9: 1-12.
Dorgham, M.M. 2014. “Effects of Eutrophication”. In Eutrophication: Causes, Consequences
and Control: Volume 2, 29-44. Springer Netherlands.
Droppo, I.G. 2001. “Rethinking What Constitutes Suspended Sediment.” Hydrological
Processes 15, no. 9: 1551–1564.
Dwyer, J.L., D.P. Roy, B. Sauer, C.B. Jenkerson, H.K. Zhang, and L. Lymburner. 2018.
“Analysis Ready Data: Enabling Analysis of the Landsat Archive.” Remote Sensing 10,
no. 9: 1-19.
85

Elahi, E., C. Weijun, H. Zhang, and M. Nazeer. 2019. “Agricultural Intensification and Damages
to Human Health in Relation to Agrochemicals: Application of Artificial Intelligence.”
Land Use Policy 83: 461–74.
EPA. 2017. Quick guide to drinking water sample collection - US EPA. Environmental
Protection Agency
EPA. 2022a. Indicators: Chlorophyll a. Environmental Protection Agency.

EPA. 2022b. Indicators: Chlorophyll a. Environmental Protection Agency.

Esri. n.d. What they don't tell you about regression analysis (no date) What they don't tell you
about regression analysis-ArcGIS Pro | Documentation.
Farzan, S.F., M. Razafy, S.P. Eckel, L. Olmedo, E. Bejarano, and J.E. Johnston. 2019.
“Assessment of Respiratory Health Symptoms and Asthma in Children near a Drying
Saline Lake.” International Journal of Environmental Research and Public Health 16,
no. 20: 1-15.
Forsman, T.N. 2014. “What the QSA Means for the Salton Sea: California’s Big Blank Check.”
Arizona State Law Journal 46: 365-387.
Gagnon, B., G. Marcoux, R. Leduc, M.F. Pouet, and O. Thomas. 2007. “Emerging Tools and
Sustainability of Water-Quality Monitoring.” TrAC Trends in Analytical Chemistry,
Emerging Tools as a New Approach for Water Monitoring 26, no.4: 308–314.
Gammon, K. 2022. As Lithium Drilling Advances at the Salton Sea, Researchers Work out the
Details. KCET.
Gao, B.C. 1996. “NDWI—A Normalized Difference Water Index for Remote Sensing of
Vegetation Liquid Water from Space.” Remote Sensing of Environment 58, no. 3: 257–
266.
Gao, J., J. Liu, R. Xu, S. Pandey, V.S.K.S Vankayala Siva, and D. Yu. 2022. “Environmental
Pollution Analysis and Impact Study—A Case Study for the Salton Sea in California.”
Atmosphere 13, no. 6: 1-23.
Gao, Y., J.P. Walker, M. Allahmoradi, A. Monerris, D. Ryu, and T.J. Jackson. 2015. “Optical
Sensing of Vegetation Water Content: A Synthesis Study.” IEEE Journal of Selected
Topics in Applied Earth Observations and Remote Sensing 8, no. 4: 1456–1464.
Garg, V., S.P. Aggarwal, and P. Chauhan. 2020. “Changes in Turbidity along Ganga River Using
Sentinel-2 Satellite Data during Lockdown Associated with COVID-19.” Geomatics,
Natural Hazards and Risk 11, no. 1: 1175–1195.
Garry, V. F. “Pesticides and Children.” Toxicology and Applied Pharmacology, Special Pediatric
Volume 198, no. 2: 152–63. https://doi.org/10.1016/j.taap.2003.11.027.
86

Gholizadeh, M.H., A.M. Melesse, and L. Reddi. 2016. “A Comprehensive Review on Water
Quality Parameters Estimation Using Remote Sensing Techniques.” Sensors 1, no. 8:
1298.
Glenn, E.P., M.J. Cohen, J.I. Morrison, C. Valdés-Casillas, and K. Fitzsimmons. 1999. “Science
and Policy Dilemmas in the Management of Agricultural Waste Waters: The Case of the
Salton Sea, CA, USA.” Environmental Science & Policy 2, no. 4: 413–423.
Gobalet, K.W. and T.A. Wake. 2000. “Archaeological and Paleontological Fish Remains from
the Salton Basin, Southern California.” The Southwestern Naturalist 45, no. 4: 514–520.
Goodchild, M.F. and L. Li. 2012. “Assuring the Quality of Volunteered Geographic
Information.” Spatial Statistics 1: 110–120.
Gorde, S.P. and M.V. Jadhav. 2013. “Assessment of Water Quality Parameters: A Review.” J
Eng Res Appl 3, no. 6: 2029-2035.
Goward, S., T. Arvidson, D. Williams, J. Faundeen, J. Irons, and S. Franks. 2006. “Historical
Record of Landsat Global Coverage.” Photogrammetric Engineering & Remote
Sensing 72, no. 10: 1155-1169.
Gutierrez, A., 2009. Dead Fish in the Desert: A Brief Photo-History of The Salton Sea. History
in the Making, 2(1), p.8.
Haftendorn, H.. 2000. “Water and International Conflict.” Third World Quarterly 21, no.1: 51–
68.
Haibo, Y., W. Zongmin, Z. Hongling, and G. Yu, 2011. “Water Body Extraction Methods Study
Based on RS and GIS.” Procedia Environmental Sciences, 2011 3rd International
Conference on Environmental Science and Information Application Technology ESIAT
2011, 10 (January): 2619–24.
Hasmadi, M., H.Z. Pakhriazad, and M.F. Shahrin, 2009. “Evaluating Supervised and
Unsupervised Techniques for Land Cover Mapping Using Remote Sensing Data.”
Geografia : Malaysian Journal of Society and Space 5, no.1: 1–10.
Harrold, Z., M.M. Arienzo, M. Collins, J.M. Davidson, X. Bai, S. Sukumaran and J. Umek,
2022. “A Peristaltic Pump and Filter-Based Method for Aqueous Microplastic Sampling
and Analysis.” ACS ES&T Water 2, no.2: 268–77.
Holdren, G.C. and A. Montaño. 2002. “Chemical and Physical Characteristics of the Salton Sea,
California.” Hydrobiologia 473, no. 1: 1–21.
Hua, A.K., 2017. “Land Use Land Cover Changes in Detection of Water Quality: A Study Based
on Remote Sensing and Multivariate Statistics.” Journal of Environmental and Public
Health 2017 (March): e7515130.
87

Hughes, B., 2020. “Turning Off the Tap: Will California Let the Salton Sea Go Down the
Drain?” California Western Law Review 56, no. 1: 245-281.
Irons, J.R. and J.L. Dwyer. 2010. “An Overview of the Landsat Data Continuity Mission.”
Remote Sensing of Environment 122: 11-21
Jiang, W., Y. Ni, Z. Pang, X. Li, H. Ju, G. He, J. Lv, K. Yang, J. Fu, and X. Qin, 2021. “An
Effective Water Body Extraction Method with New Water Index for Sentinel-2 Imagery.”
Water 13, no. 12: 1-17.
Johnston, J.E., M. Razafy, H. Lugo, L. Olmedo, and S.F. Farzan. 2019. “The Disappearing
Salton Sea: A Critical Reflection on the Emerging Environmental Threat of Disappearing
Saline Lakes and Potential Impacts on Children’s Health.” Science of The Total
Environment 663: 804–17.
Jones, B.A. and J. Fleck. 2020. “Shrinking Lakes, Air Pollution, and Human Health: Evidence
from California’s Salton Sea.” Science of The Total Environment 712 : 136490.
Kamerosky, A., H.J. Cho,. and L. Morris, 2015. “Monitoring of the 2011 Super Algal Bloom in
Indian River Lagoon, FL, USA, Using MERIS.” Remote Sensing 7 (2): 1441–60.
Karimi, B., S.H. Hashemi and H. Aghighi, 2022. “Performance of Sentinel-2 and Landsat-8
Satellites in Estimating Chlorophyll-a Concentration in a Shallow Freshwater Lake.”
Preprint. In Review.
Keiser, D.A. and J.S. Shapiro, 2019. “Consequences of the Clean Water Act and the Demand for
Water Quality.” The Quarterly Journal of Economics 134 (1): 349–96..
Kennan, G., 1917. The Salton Sea: An Account of Harriman’s Fight with the Colorado River.
Macmillan.
Kershner, F.D. 1953. “George Chaffey and the Irrigation Frontier.” Agricultural History 27, no.
4: 115–122.
Kim, K.H., E. Kabir, and S.A. Jahan. 2017. “Exposure to Pesticides and the Associated Human
Health Effects.” Science of The Total Environment 575: 525–535.
King, J., V. Etyemezian, M. Sweeney, B.J. Buck, and G. Nikolich. 2011. “Dust Emission
Variability at the Salton Sea, California, USA.” Aeolian Research, AGU Fall Meeting
session on Aeolian Dust: Transport Processes, Anthropogenic Forces and
Biogeochemical Cycling, 3 (1): 67–79.
“Landsat Science.” n.d. NASA. NASA. Accessed May 7, 2021.
Laylander, D. 1997. “The Last Days of Lake Cahuilla: The Elmore Site.” Pacific Coast
Archaeological Society Quarterly 33, no. 2: 1-138.
88

Lednicka, B. and M. Kubacka. 2022. “Semi-Empirical Model of Remote-Sensing Reflectance for
Chosen Areas of the Southern Baltic.” Sensors 22, no. 3: 1-23.
Lee, G.F., Jones, R.A. and Newbry, B.W. 1982. “Water Quality Standards and Water
Quality.” Journal (Water Pollution Control Federation): 1131-1138.
Lee, Z.P., 2006. Remote Sensing of Inherent Optical Properties: Fundamentals, Tests of
Algorithms, and Applications. International Ocean Colour Coordinating Group (IOCCG).
Lewis, E.L. and R.G. Perkin, 1978. “Salinity: Its Definition and Calculation.” Journal of
Geophysical Research: Oceans 83 (C1): 466–78.
https://doi.org/10.1029/JC083iC01p00466.
Levers, L.R., T.H. Skaggs and K.A. Schwabe, 2019. “Buying Water for the Environment: A
Hydro-Economic Analysis of Salton Sea Inflows.” Agricultural Water Management 213
(March): 554–67. https://doi.org/10.1016/j.agwat.2018.10.041.
Loveland, T.R. and J.L. Dwyer, 2012. “Landsat: Building a Strong Future.” Remote Sensing of
Environment, Landsat Legacy Special Issue, 122 (July): 22–29.
https://doi.org/10.1016/j.rse.2011.09.022.
Lichtenberg, E., 2002. “Chapter 23 Agriculture and the Environment.” In Handbook of
Agricultural Economics, 2:1249–1313. Agriculture and Its External Linkages. Elsevier.
https://doi.org/10.1016/S1574-0072(02)10005-3.
Lin, S.S., S.L. Shen, A. Zhou and H.M. Lyu, 2021. “Assessment and Management of Lake
Eutrophication: A Case Study in Lake Erhai, China.” Science of The Total Environment
751 (January): 141618. https://doi.org/10.1016/j.scitotenv.2020.141618.
Ling, F., G.M. Foody, H. Du, X. Ban, X. Li, Y. Zhang and Y. Du, 2017. “Monitoring Thermal
Pollution in Rivers Downstream of Dams with Landsat ETM+ Thermal Infrared Images.”
Remote Sensing 9 (11): 1175. https://doi.org/10.3390/rs9111175.MacDougal, D.T.,
1907. The Desert Basins of the Colorado Delta. American Geographical Society.
Madrid, Y. and Z.P. Zayas, 2007. “Water Sampling: Traditional Methods and New Approaches
in Water Sampling Strategy.” TrAC Trends in Analytical Chemistry, Emerging tools as a
new approach for water monitoring, 26 (4): 293–99.
https://doi.org/10.1016/j.trac.2007.01.002.
Malmqvist, B. and S. Rundle, 2002. “Threats to the Running Water Ecosystems of the World.”
Environmental Conservation 29 (2): 134–53.
https://doi.org/10.1017/S0376892902000097.
Marshall, J.R., 2017. “Why Emergency Physicians Should Care About the Salton Sea.” Western
Journal of Emergency Medicine 18 (6): 1008.
https://doi.org/10.5811/westjem.2017.8.36034
89

Marti-Cardona, B., T.E. Steissberg, S.G. Schladow and S.J. Hook, 2008. Relating fish kills to
upwellings and wind patterns in the Salton Sea. In The Salton Sea Centennial Symposium
(pp. 85-95). Springer, Dordrecht.
Martin, L., 2019. What Are the Roles of Chlorophyll A & B?. Sciencing.
Matamoros, V., 2012. Equipment for water sampling including sensors.
McFeeters, S.K., 1996. “The Use of the Normalized Difference Water Index (NDWI) in the
Delineation of Open Water Features.” International Journal of Remote Sensing 17 (7):
1425–32. https://doi.org/10.1080/01431169608948714.
Mead, S.J.F., 2016. A Green Oasis in the Desert: The History of Immigration in the Imperial
Valley. Sustainability Of The Salton Sea: A Review Of The Environmental, Economic,
And Social Climate Around California’s Largest Saline Lake, p.43.
Mejía Ávila, D., F. Torres-Bejarano and Z. Martínez Lara, 2022. “Spectral Indices for Estimating
Total Dissolved Solids in Freshwater Wetlands Using Semi-Empirical Models. A Case
Study of Guartinaja and Momil Wetlands.” International Journal of Remote Sensing 43
(6): 2156–84. https://doi.org/10.1080/01431161.2022.2057205.’
Miller, W. and A. Rango, 1984. “Using Heat Capacity Mapping Mission (Hcmm) Data to Assess
Lake Water Quality1.” JAWRA Journal of the American Water Resources Association
20 (4): 493–501. https://doi.org/10.1111/j.1752-1688.1984.tb02831.x.
Mishra, S. and D.R. Mishra, 2012. “Normalized Difference Chlorophyll Index: A Novel Model
for Remote Estimation of Chlorophyll-a Concentration in Turbid Productive Waters.”
Remote Sensing of Environment, Remote Sensing of Urban Environments, 117
(February): 394–406. https://doi.org/10.1016/j.rse.2011.10.016.
Mishra, D.R., B.A. Schaeffer and D. Keith, 2014. “Performance Evaluation of Normalized
Difference Chlorophyll Index in Northern Gulf of Mexico Estuaries Using the
Hyperspectral Imager for the Coastal Ocean.” GIScience & Remote Sensing 51 (2): 175–
98. https://doi.org/10.1080/15481603.2014.895581.
Morel, A.Y. and H.R. Gordon, 1980. “Report of the Working Group on Water Color.”
Boundary-Layer Meteorology 18 (3): 343–55. https://doi.org/10.1007/BF00122030.
Moses, M., E.S. Johnson, W.K. Anger, V.W. Burse, S.W. Horstman, R.J. Jackson, R.G. Lewis,
K.T Maddy, R. McConnell, W.J. Meggs and S.H. Zahm, 1993. “Environmental Equity
and Pesticide Exposure.” Toxicology and Industrial Health 9 (5): 913–59.
https://doi.org/10.1177/074823379300900512.
National Geographic Society, 2019. Freshwater Resources. National Geographic Society.
Accessed March 17, 2022.
90

Neary, D.G., G.G. Ice and C.R. Jackson, 2009. “Linkages between Forest Soils and Water
Quality and Quantity.” Forest Ecology and Management, 258 (10): 2269–81.
https://doi.org/10.1016/j.foreco.2009.05.027.
Nielsen, D.L., M.A. Brock, G.N. Rees and D.S. Baldwin, 2003. “Effects of Increasing Salinity
on Freshwater Ecosystems in Australia.” Australian Journal of Botany 51 (6): 655–65.
https://doi.org/10.1071/bt02115.
Niepytalska, M. and P. Warschau, 2019. The Salton Sea: An Eco-biography of California's
Largest Lake (Doctoral dissertation, Universitätsbibliothek der Ludwig-Maximilians-
Universität).
Nugroho, J.T., 2013. Identification of Inundated Area Using Normalized Difference Water Index
(NDWI) on Lowland Region of Java Island. International Journal of Remote Sensing and
Earth Sciences (IJReSES), 10(2).
Obianyo, J.I., 2019. “Effect of Salinity on Evaporation and the Water Cycle.” Emerging Science
Journal 3 (4): 255–62. https://doi.org/10.28991/esj-2019-01188.
Ogashawara, I., L. Li and M.J. Moreno-Madriñán, 2016. “Slope Algorithm to Map Algal Blooms
in Inland Waters for Landsat 8/Operational Land Imager Images.” Journal of Applied
Remote Sensing 11 (1): 012005. https://doi.org/10.1117/1.JRS.11.012005.
Omer, N.H., 2019. “Water Quality Parameters.” In Water Quality - Science, Assessments and
Policy, edited by Kevin Summers. IntechOpen. https://doi.org/10.5772/intechopen.89657.
Parajuli, S.P. and C.S. Zender, 2018. “Projected Changes in Dust Emissions and Regional Air
Quality Due to the Shrinking Salton Sea.” Aeolian Research 33 (August): 82–92.
https://doi.org/10.1016/j.aeolia.2018.05.004.
Payán-Rentería, R., G. Garibay-Chavez, R. Rangel-Ascencio, V. Preciado-Martinez, L. Munoz-
Islas, C. Beltrán-Miranda, S. Mena-Munguía, L. Jave-Suárez, A. Feria-Velasco and R. De
Celis, 2012. “Effect of Chronic Pesticide Exposure in Farm Workers of a Mexico
Community.” Archives of Environmental & Occupational Health 67 (1): 22–30.
https://doi.org/10.1080/19338244.2011.564230.
Picone, K., 2021. “Salton Sea: The Rise and Fall of a Toxic California Lake.” All That's
Interesting. All That's Interesting.
Pillsbury-Coyne, A., 2018. Policy Trials and Tribulations of the Salton Sea. UC San Diego:
Climate Science and Policy.
Poole, G.C., J.B. Dunham, D.M. Keenan, S.T. Sauter, S.T., D.A. McCULLOUGH, C. Mebane,
J.C. Lockwood, D.A. Essig, M.P Hicks, D.J. Sturdevant and E.J. Materna, 2004. “The
Case for Regime-Based Water Quality Standards.” BioScience 54 (2): 155–61.
https://doi.org/10.1641/0006-3568(2004)054[0155:TCFRWQ]2.0.CO;2
91

Potapov, P.V., S.A. Turubanova, M.C. Hansen, B. Adusei, M. Broich, A. Altstatt, L. Mane and
C.O. Justice, 2012. “Quantifying Forest Cover Loss in Democratic Republic of the
Congo, 2000–2010, with Landsat ETM+ Data.” Remote Sensing of Environment,
Landsat Legacy Special Issue, 122 (July): 106–16.
https://doi.org/10.1016/j.rse.2011.08.027..
Qin, B., G. Gao, G. Zhu, Y. Zhang, Y. Song, X. Tang, H. Xu and J. Deng. 2013. “Lake
Eutrophication and Its Ecosystem Response.” Chinese Science Bulletin 58 (9): 961–70.
https://doi.org/10.1007/s11434-012-5560-x.
Ramadas, M. and A.K. Samantaray, 2018. “Applications of Remote Sensing and GIS in Water
Quality Monitoring and Remediation: A State-of-the-Art Review.” In Water
Remediation, edited by Shantanu Bhattacharya, Akhilendra Bhushan Gupta, Ankur
Gupta, and Ashok Pandey, 225–46. Energy, Environment, and Sustainability. Singapore:
Springer. https://doi.org/10.1007/978-981-10-7551-3_13.
Råman Vinnå, L., A. Wüest and D. Bouffard, 2017. “Physical Effects of Thermal Pollution in
Lakes: Physical Effects of Thermal Pollution.” Water Resources Research 53 (5): 3968–
87. https://doi.org/10.1002/2016WR019686.
Reis, S. 2008. “Analyzing Land Use/Land Cover Changes Using Remote Sensing and GIS in
Rize, North-East Turkey.” Sensors 8, no. 10: 6188–6202.
https://doi.org/10.3390/s8106188.
Riedel, R., 2016. “Trends of Abundance of Salton Sea Fish: A Reversible Collapse or a
Permanent Condition?” Natural Resources 07, no.10: 535.
https://doi.org/10.4236/nr.2016.710045.
Richards, J.A., 2022. “Supervised Classification Techniques.” In Remote Sensing Digital Image
Analysis, edited by John A. Richards, 263–367. Cham: Springer International Publishing.
https://doi.org/10.1007/978-3-030-82327-6_8.
Ritchie J.C., F.R. Schiebe, and J.R. McHenry, 1976. “Remote Sensing of Suspended Sediments
in Surface Waters.” Photogramm. Eng. Remote Sens 42: 1539-1545.
Ritchie, J.C., P.V. Zimba and J.H. Everitt, 2003. “Remote Sensing Techniques to Assess Water
Quality.” Photogrammetric Engineering & Remote Sensing 69, no. 6: 695–704.
https://doi.org/10.14358/PERS.69.6.695.
Rockwell, T.K., A.J. Meltzner, and E.C. Haaker, 2018. Dates of the two most recent surface
ruptures on the southernmost San Andreas fault recalculated by precise dating of Lake
Cahuilla dry periods. Bulletin of the Seismological Society of America, 108(5A), pp.2634-
2649.
Ross, J.E., 2020. Formation of California’s Salton Sea in 1905–07 was not “accidental”.
In Changing Facies: The 2020 desert symposium field guide and proceedings (pp. 103-
116). Zzyzx, CA: Desert Symposium.
92

Rwanga, S.S. and J.M. Ndambuki. 2017.“Accuracy Assessment of Land Use/Land Cover
Classification Using Remote Sensing and GIS.” International Journal of Geosciences 8,
no.04: 611. https://doi.org/10.4236/ijg.2017.84033.
Rockwell, T.K., A.J. Meltzner and E.C. Haaker. 2018. “Dates of the Two Most Recent Surface
Ruptures on the Southernmost San Andreas Fault Recalculated by Precise Dating of Lake
Cahuilla Dry Periods.” Bulletin of the Seismological Society of America 108, no.5A:
2634–2649. https://doi.org/10.1785/0120170392.
Rode, E. and J. Wilson, 2022. “Lithium Valley: A Look at the Major Players near the Salton Sea
Seeking Billions in Funding.” The Desert Sun. Palm Springs Desert Sun, May 15, 2022.
Roland, F., V.L.M. Huszar, V.F. Farjalla, A. Enrich-Prast, A.M. Amado, and J.P.H.B. Ometto,
2012. “Climate Change in Brazil: Perspective on the Biogeochemistry of Inland Waters.”
Brazilian Journal of Biology 72: 709–22. https://doi.org/10.1590/S1519-
69842012000400009.
Roy, P.S., M.D. Behera, and S.K. Srivastav. 2017. “Satellite Remote Sensing: Sensors,
Applications and Techniques.” Proceedings of the National Academy of Sciences, India
Section A: Physical Sciences 87, no.4: 465–72. https://doi.org/10.1007/s40010-017-0428-
8.
Sanderson, R. 2010. “Introduction to Remote Sensing.” New Mexico State University: 25-26.
Saravanan, V.S., G.T. McDonald, and P.P. Mollinga. 2009. “Critical Review of Integrated Water
Resources Management: Moving beyond Polarised Discourse.” Natural Resources
Forum 33, no. 1: 76–86. https://doi.org/10.1111/j.1477-8947.2009.01210.x.
Savenije, H.H. and P. Van der Zaag. 2008. “Integrated Water Resources Management: Concepts
and Issues.” Physics and Chemistry of the Earth, Parts A/B/C, Integrated Water
Resources Management in a Changing World 33, no. 5: 290–97.
https://doi.org/10.1016/j.pce.2008.02.003.
Scafutto, R.D.P.M. and C.R. de Souza Filho, 2018. “Detection of Methane Plumes Using
Airborne Midwave Infrared (3–5 Μm) Hyperspectral Data.” Remote Sensing 10 (8):
1237. https://doi.org/10.3390/rs10081237.
Scaramuzza, P. and J. Barsi, 2005, October. Landsat 7 scan line corrector-off gap-filled product
development. In Proceeding of Pecora (Vol. 16, pp. 23-27).
Schaeffer, B.A., K.G. Schaeffer, D. Keith, R.S. Lunetta, R. Conmy and R.W. Gould. 2013.
“Barriers to Adopting Satellite Remote Sensing for Water Quality Management.”
International Journal of Remote Sensing 34, no. 21: 7534–44.
https://doi.org/10.1080/01431161.2013.823524.
Setmire, J., C. Holdren, D. Robertson, C. Amrhein, J. Elder, R. Schroeder, G. Schladow, H.
McKellar, and R. Gersberg, 2000, September. Eutrophic conditions at the Salton Sea.
93

In A topical paper from the Eutrophication Workshop convened at the University of
California at Riverside.
Sheikh, P.A. and C.V. Stern, 2020. Salton Sea Restoration. LIBRARY OF CONGRESS
WASHINGTON DC.
Sherman, D.J., C. Houser and A.C. Baas. 2013. “Electronic Measurement Techniques for Field
Experiments in Process Geomorphology.” Treatise on Geomorphology: 195–221.
https://doi.org/10.1016/B978-0-12-374739-6.00401-2.
Smith, V.H. and D.W. Schindler. 2009. “Eutrophication Science: Where Do We Go From
Here?” Trends in ecology & evolution 24, no. 4: 201-207.
Spehar, R., D. Taylor, and S. Cormier. 2022. EPA. Environmental Protection Agency.
Stanley, M. 2022. Point source and nonpoint sources of Pollution, National Geographic Society.
Strobl, R.O. and P.D. Robillard. 2008. “Network Design for Water Quality Monitoring of
Surface Freshwaters: A Review.” Journal of Environmental Management, Microbial and
Nutrient Contaminants of Fresh and Coastal Waters, 87 (4): 639–48.
https://doi.org/10.1016/j.jenvman.2007.03.001.
Tas, B., O. Can, and Z. Koloren. 2011. “Investigation on Photosynthetic Pigments Content of
Lotic Systems (Blacksea River Basin, Ordu-Turkey). Structure 1: 417-426.
Taylor, M., 2018. “The Salton Sea: A Status Update”. Sacramento, CA: Legislative Analysts.
The Associated Press. 2015. “History of the Salton Sea.” Tribune. San Diego Union-Tribune.
Tompson, A.F. 2016. “Born from a Flood: The Salton Sea and Its Story of Survival.” Journal of
Earth Science 27, no. 1: 89–97. https://doi.org/10.1007/s12583-016-0630-7.
Topp, S.N., T.M. Pavelsky, D. Jensen, M. Simard, and M.R. Ross. 2020. “Research Trends in the
Use of Remote Sensing for Inland Water Quality Science: Moving Towards
Multidisciplinary Applications.” Water 12, no. 1: 1-34.
https://doi.org/10.3390/w12010169.
Tran, T.V., R. Reef, and X. Zhu. 2022. “A Review of Spectral Indices for Mangrove Remote
Sensing”. Remote Sensing 14, no. 19: 1-29. https://doi.org/10.3390/rs14194868.
Tucker, C.J. 1979. “Red and Photographic Infrared Linear Combinations for Monitoring
Vegetation.” Remote Sensing of Environment 8, no. 2: 127–150.
https://doi.org/10.1016/0034-4257(79)90013-0.
Usali, N. and M.H. Ismail. 2010. “Use of Remote Sensing and GIS in Monitoring Water
Quality.” Journal of Sustainable Development 3, no. 3: 228-238.
https://doi.org/10.5539/jsd.v3n3p228.
94

U.S. Environmental Protection Agency. 2022. What are Water Quality Standards?, EPA.
Environmental Protection Agency.
U.S. Geological Survey, Water quality information by topic completed. Water Quality
Information by Topic | U.S. Geological Survey. Accessed March 21, 2022.
U.S. Geological Survey, What is remote sensing and what is it used for? What is remote sensing
and what is it used for? | U.S. Geological Survey. Accessed March 21, 2022.
Vasistha, P. and R. Ganguly. 2020. “Water Quality Assessment of Natural Lakes and Its
Importance: An Overview.” Materials Today: Proceedings, 3rd International Conference
on Innovative Technologies for Clean and Sustainable Development, 32: 544–52.
https://doi.org/10.1016/j.matpr.2020.02.092.
Ventura, S., 2020. “Selective Recovery of Lithium from Geothermal Brines.” Accessed June 20,
2022.
Verones, F., M.M. Hanafiah, S. Pfister, M.A. Huijbregts, G.J. Pelletier and A. Koehler, 2010.
Characterization factors for thermal pollution in freshwater aquatic
environments. Environmental Science & Technology, 44(24), pp.9364-9369.
Vessey, K.B. 2000. “Salton: A Sea of Controversy.” Journal of College Science Teaching 30, no.
1: 67-69.
Viñas, M.J., 2012. Notes from the field - measuring salinity from space, NASA. NASA.
Vizzo, E.C. 2017. “The Salton Sea: An Unfished and Quickly Loved Foray.” The Trumpeter 33,
no.1: 81-91.
Waghmare, B. and M. Suryawanshi. 2017. A Review-Remote Sensing. Int. J. Eng. Res. Appl 7,
no. 6: 52-54.
Wang, G., S. Mang, H. Cai, S. Liu, Z. Zhang, L. Wang, and J.L. Innes. 2016. “Integrated
Watershed Management: Evolution, Development and Emerging Trends.” Journal of
Forestry Research 27, no. 5: 967–94. https://doi.org/10.1007/s11676-016-0293-3.
Wang, F. and Y.J. Xu. 2012. “Remote Sensing to Predict Estuarine Water Salinity.”
Environmental Remote Sensing and Systems Analysis; 85-108.
Wang, X. and W. Yang. 2019. “Water Quality Monitoring and Evaluation Using Remote
Sensing Techniques in China: A Systematic Review. Ecosystem Health and
Sustainability 5, no. 1: 47-56. https://doi.org/10.1080/20964129.2019.1571443.
Wanger, T.C. 2011. “The Lithium Future—Resources, Recycling, and the Environment.”
Conservation Letters 4, no. 3: 202-206. https://doi.org/10.1111/j.1755-
263X.2011.00166.x.
95

Watanabe, F.S.Y., E. Alcântara, T.W.P. Rodrigues, N.N. Imai, C.C.F. Barbosa and L.H.D.S.
Rotta. 2015. “Estimation of Chlorophyll-a Concentration and the Trophic State of the
Barra Bonita Hydroelectric Reservoir Using OLI/Landsat-8 Images.” International
Journal of Environmental Research and Public Health 12, no. 9: 10391-10417.
Waters, M.R. 1983. “Late Holocene Lacustrine Chronology and Archaeology of Ancient Lake
Cahuilla, California.” Quaternary Research 19, no. 3: 373–387.
https://doi.org/10.1016/0033-5894(83)90042-X.
Williams, D.L., S. Goward, and T. Arvidson, 2006. “Landsat: Yesterday, Today, and
Tomorrow.” Photogrammetric Engineering & Remote Sensing 72, no.10: 1171-1178.
Wu, Z., G. Snyder, C. Vadnais, R. Arora, M. Babcock, G. Stensaas, P. Doucette, and T.
Newman, 2019. “User Needs for Future Landsat Missions.” Remote Sensing of
Environment 231 : 1-13. https://doi.org/10.1016/j.rse.2019.111214.
Wulder, M.A., T.R. Loveland, D.P. Roy, C.J. Crawford, J.G. Masek, C.E. Woodcock, R.G.
Allen, M.C. Anderson, A.S. Belward, W.B. Cohen, and J. Dwyer. 2019. “Current Status
of Landsat Program, Science, and Applications.” Remote Sensing of Environment 225:
127–47. https://doi.org/10.1016/j.rse.2019.02.015.
Wulder, M.A., D.P. Roy, V.C. Radeloff, T.R. Loveland, M.C. Anderson, D.M. Johnson, S.
Healey, Z. Zhu, T.A. Scambos, N. Pahlevan, and M. Hansen. 2022. “Fifty Years of
Landsat Science and Impacts.” Remote Sensing of Environment 280: 1-21.
https://doi.org/10.1016/j.rse.2022.113195.
Xu, H. 2006. “Modification of Normalised Difference Water Index (NDWI) to Enhance Open
Water Features in Remotely Sensed Imagery.” International Journal of Remote Sensing
27, no. 14: 3025–3033. https://doi.org/10.1080/01431160600589179.
Xu, M., H. Liu, R. Beck, J. Lekki, B. Yang, S. Shu, E.L. Kang, R. Anderson, E. Johansen, E.
Emery and M. Reif. 2019a. “A Spectral Space Partition Guided Ensemble Method for
Retrieving Chlorophyll-a Concentration in Inland Waters from Sentinel-2A Satellite
Imagery. Journal of Great Lakes Research 45, no. 3: 454-465.
Xu, M., H. Liu, R. Beck, J. Lekki, B. Yang, S. Shu, Y. Liu, T. Benko, R. Anderson, R. Tokars,
and R. Johansen. 2019b. “Regionally and Locally Adaptive Models for Retrieving
Chlorophyll-a Concentration in Inland Waters From Remotely Sensed Multispectral and
Hyperspectral Imagery.” IEEE Transactions on Geoscience and Remote Sensing 57, no.
7: 4758–4774. https://doi.org/10.1109/TGRS.2019.2892899
Yang, J. and X. Du. 2017. “An Enhanced Water Index in Extracting Water Bodies from Landsat
TM Imagery.” Annals of GIS 23, no. 3: 141–48.
https://doi.org/10.1080/19475683.2017.1340339.
Yang, W., S.L. Carmichael, E.M. Roberts, S.E. Kegley, A.M. Padula, P.B. English, and G.M.
Shaw. 2014. “Residential Agricultural Pesticide Exposures and Risk of Neural Tube
Defects and Orofacial Clefts Among Offspring in the San Joaquin Valley of California.”
96

American Journal of Epidemiology 179, no. 6: 740–748.
https://doi.org/10.1093/aje/kwt324.
Yue, H., Y. Li, J. Qian, and Y. Liu. 2020. “A New Accuracy Evaluation Method for Water Body
Extraction.” International Journal of Remote Sensing 41, no. 19: 7311–7342.
https://doi.org/10.1080/01431161.2020.1755740.
Zhu, L., J. Suomalainen, J. Liu, J. Hyyppä, H. Kaartinen and H. Haggren. 2018. “A Review:
Remote Sensing Sensors.” In Multi-Purposeful Application of Geospatial Data, edited by
Rustam B. Rustamov, Sabina Hasanova, and Mahfuza H. Zeynalova. InTech.
https://doi.org/10.5772/intechopen.71049.

97

Appendix: A Distributions of NDCI Values
98

99

Abstract (if available)

Abstract The Salton Sea is the largest body of water in the State of California and has experienced a decline in water quality within the last fifty years. This inland body of water serves as a reservoir for agricultural runoff and maintains high concentrations of pesticides and nutrients that place surrounding communities and ecological environments at risk. As a result of the degradation and eutrophic state of the Salton Sea, it is important to identify historical trends and methodologies that can be used for future water quality assessments. Traditional water quality assessments are conducted onsite and require extensive financial and human resources. In order to mitigate some of these costs while continuing to monitor water quality, more efficient assessment techniques must be explored. This study explores one such technique by examining the use of remote sensing techniques and the Normalized Difference Chlorophyll Index (NDCI) to assess chlorophyll-a concentrations in the Salton Sea from 2002 to 2020 using Landsat 5 TM and Landsat 8 OLI imagery. To assess the accuracy of this method, the NDCI is compared against two-band and three-band algorithms proposed by literature. Results indicate that the NDCI has largely underestimate chlorophyll-a concentrations within the Salton Sea and has incorrectly suggested small variations across the temporal range. Linear regression results further reveal a weak linear regression between NDCI, 2BDA and 3BDA values and in-situ measurements.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A spatial and temporal exploration of how satellite communication devices impact mountain search and rescue missions in California’s Sierra Nevada mountain range

PDF

Assessing modern conflict to monitor human rights with remote sensing: Russia's war in Ukraine

PDF

Temporal analysis of soil degradation in San Joaquin County, California: a close examination of soil erosion using RUSLE

PDF

Human suffering during wartime: a StoryMap of violations of international law during the Russo-Ukrainian War

PDF

A spatial analysis of beef production and its environmental and health impacts in Texas

PDF

That sinking feeling: predicting land subsidence in California’s San Joaquin Valley with a spatial regression model

PDF

A comparative study of ground and satellite evapotranspiration models for southern California

PDF

Water quality in the Los Angeles River: a remote sensing based analysis

PDF

Detection and accuracy assessment of mountain pine beetle infestations using Landsat 8 OLI and WorldView02 satellite imagery: Lake Tahoe Basin-Nevada and California

PDF

Satellite derived bathymetry in the Canadian Archipelago using multispectral and LiDAR space-based remote sensing

PDF

Development of a historical urban land use web application for the city of Hong Kong

PDF

Use of remotely sensed imagery to map sudden oak death (Phytophthora ramorum) in the Santa Cruz Mountains

PDF

From ruins to pixels: using remote sensing and GIS to analyze, document, and visualize archaeological sites and ancient roadways in Chaco Canyon, New Mexico

PDF

Projecting vulnerability: a combined analysis of sea-level rise, hurricane inundation, and social vulnerability in Houston-Galveston, Texas

PDF

Object detection and digitization from aerial imagery using neural networks

PDF

Climate conservation priorities: using MCDA to identify future refuge for the Joshua Tree Forest in California

PDF

Spatiotemporal analysis of the SLOSH and ADCIRC storm surge models: a case study of hurricane Ida

PDF

A comparison of weighted and fuzzy overlays in mapping landslide susceptibility, south-central front range, Colorado

PDF

The role of precision in spatial narratives: using a modified discourse quality index to measure the quality of deliberative spatial data

PDF

Using aerial imagery to assess tropical forest cover surrounding restoration sites in Costa Rica

Asset Metadata

Creator Lopez, Alejandra Garcia (author)

Core Title Assessing the use of normalized difference chlorophyll index to estimate chlorophyll-A concentrations using Landsat 5 TM and Landsat 8 OLI imagery in the Salton Sea, California

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Degree Conferral Date 2023-05

Publication Date 03/09/2023

Defense Date 01/11/2023

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag landsat,NDCI,normalized difference chlorophyll index,OAI-PMH Harvest,Salton Sea,satellite imagery

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sedano, Elisabeth (committee chair), Qi, Yi (committee member), Ter-Ghazaryan, Diana (committee member)

Creator Email alopez02@usc.edu,lopez.alejandra.garcia@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC112764295

Unique identifier UC112764295

Identifier etd-LopezAleja-11495.pdf (filename)

Legacy Identifier etd-LopezAleja-11495

Document Type Thesis

Format theses (aat)

Rights Lopez, Alejandra Garcia

Internet Media Type application/pdf

Type texts

Source 20230313-usctheses-batch-1009 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu