Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Water quality in the Los Angeles River: a remote sensing based analysis
(USC Thesis Other)
Water quality in the Los Angeles River: a remote sensing based analysis
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
Water Quality in the Los Angeles River:
A Remote Sensing Based Analysis
by
Shane Daniel
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 Shane Daniel
ii
To my grandparents
iii
Acknowledgements
I am grateful to Dr. Ruddell for guidance in the first half of this endeavor. I would like to
acknowledge my classmates for giving great feedback during peer reviews in SCCI 594A. Thank
you to Drs. Vos and Wilson for their beneficial commentary during our committee meeting.
Lastly, I would like to thank Dr. Sedano for tireless guidance and continued support.
iv
Table of Contents
Dedication ...................................................................................................................................... ii
Acknowledgements ........................................................................................................................ iii
List of Tables ................................................................................................................................. vi
List of Figures ............................................................................................................................... vii
Abbreviations ............................................................................................................................... viii
Abstract .......................................................................................................................................... ix
Chapter 1 Introduction .................................................................................................................... 1
1.1. Study Area ..........................................................................................................................2
1.2. Background and Motivation ...............................................................................................4
1.3. Augmentation of Data through Remote Sensing ................................................................7
1.4. Thesis Structure ..................................................................................................................9
Chapter 2 Related Work ................................................................................................................ 11
2.1. Electromagnetic Radiation and Hydrospheric Science .....................................................11
2.2. Remote Sensing Based Hydrologic Observation ..............................................................13
2.3. Validity of Remote Sensing Derived Measurements ........................................................16
2.4. Municipal and Federal Reports Focusing on the Los Angeles River ...............................17
Chapter 3 Methods ........................................................................................................................ 19
3.1. Data Description ...............................................................................................................19
3.2. Sample Sites ......................................................................................................................21
3.2.1. Sepulveda Basin .......................................................................................................22
3.2.2. Glendale ...................................................................................................................24
3.2.3. Willow Street ...........................................................................................................27
3.3. Research Design................................................................................................................29
Chapter 4 Results .......................................................................................................................... 34
v
4.1. Sepulveda Basin ................................................................................................................34
4.1.1. NDTI-Turbidity........................................................................................................34
4.1.2. NDVI-Chlorophyll-a ................................................................................................35
4.2. Glendale ............................................................................................................................37
4.2.1. NDTI-Turbidity........................................................................................................37
4.2.2. NDVI-Chlorophyll-a ................................................................................................38
4.3. Willow Street ....................................................................................................................39
4.3.1. NDTI-Turbidity........................................................................................................39
4.3.2. NDVI-Chlorophyll-a ................................................................................................40
4.4. ANOVA ............................................................................................................................41
Chapter 5 Conclusions .................................................................................................................. 44
5.1. Discussion .........................................................................................................................44
5.2. Limitations ........................................................................................................................46
5.3. Future Work ......................................................................................................................47
References ..................................................................................................................................... 49
vi
List of Tables
Table 1 Correlation strength ......................................................................................................... 17
Table 2 Required data ................................................................................................................... 20
Table 3 Image acquisition dates .................................................................................................... 21
Table 4 Sepulveda Basin sample distribution ............................................................................... 23
Table 5 Glendale sample distribution ........................................................................................... 26
Table 6 Willow Street sample distribution ................................................................................... 28
Table 7 NDTI ANOVA results ..................................................................................................... 42
Table 8 NDVI ANOVA results ..................................................................................................... 43
vii
List of Figures
Figure 1 Los Angeles River extent and watershed ......................................................................... 3
Figure 2 Concrete and natural reaches of LAR .............................................................................. 5
Figure 3 Glendale natural riverbed ................................................................................................. 8
Figure 4 Light behavior in water bodies and atmosphere ............................................................. 12
Figure 5 Sepulveda Basin sample frame ....................................................................................... 24
Figure 6 Glendale sample frame ................................................................................................... 27
Figure 7 Willow Street sample frame ........................................................................................... 29
Figure 8 Image analysis workflow ................................................................................................ 30
Figure 9 Glendale sample frame raster clips ................................................................................. 33
Figure 10 Sepulveda Basin NDTI ................................................................................................. 35
Figure 11 Sepulveda Basin NDVI ................................................................................................ 36
Figure 12 Glendale NDTI ............................................................................................................. 38
Figure 13 Glendale NDVI ............................................................................................................. 39
Figure 14 Willow Street NDVI ..................................................................................................... 40
Figure 15 Willow Street NDVI ..................................................................................................... 41
viii
Abbreviations
ANOVA Analysis of Variance
BWRP Burbank Water Reclamation Plant
DTCWRP Donald C. Tillman Water Reclamation Plant
LAC Los Angeles County
LAPIER Los Angeles County Public Works Draft Program Environmental Impact Report
LAR Los Angeles River
NAIP National Agriculture Imagery Program
NDTI Normalized Difference Turbidity Index
NDVI Normalized Difference Vegetation Index
NDWI Normalized Difference Water Index
RS Remote Sensing
SSI Spatial Sciences Institute
UAVs Unmanned aerial vehicles
USACE United States Army Corps of Engineers
USC University of Southern California
WQ Water quality
WRP Water Reclamation Plant
ix
Abstract
Approximately 11 miles of concrete channelized Los Angeles River (LAR) are planned to be
restored to natural riverbed with engineered banks by the U.S. Army Corps of Engineers
(USACE) working with the City of Los Angeles. Recent studies centered on urban watersheds in
Southern California focus primarily on pollutants without addressing the effects of artificial
riverbed substrate on water quality (WQ). Departing from the field collection methodologies
employed by previous studies in the LAR, this project is a systematic remote sensing (RS) based
investigation of differences in WQ found between natural and concrete riverbed environments in
the LAR. Two RS indices are applied to aerial imagery to assess relative WQ in the LAR:
Normalized difference vegetation index (NDVI) and normalized difference turbidity index
(NDTI). National Agriculture Imagery Program imagery is integrated with NDVI to measure
chlorophyll-a concentration and NDTI to determine turbidity levels as parameters to gauge WQ.
Results indicate that natural riverbed sections typically contain lower NDTI values than concrete
sections, suggesting these environments abate turbidity. NDVI results detect statistically
significant differences in chlorophyll-a concentration between water in concrete and natural
sections that may be the product of several variables deserving further study. The RS
methodology herein provides a framework for monitoring of WQ parameters in the LAR.
Provision of a dense dataset of WQ parameters existing in concrete and natural riverbed zones
fills a critical knowledge gap regarding the poorly studied effects of the artificial substrate.
Examples of natural riverbed sections in the LAR with engineered banks harboring reduced
turbidity, while providing adequate flood protection for decades, supports the argument that
expanded restoration is both environmentally advantageous and safe for surrounding
communities.
1
Chapter 1 Introduction
The goal of this project was to use remote sensing (RS) based techniques to determine
fluctuations in water quality (WQ) as water travels through the highly modified Los Angeles
River (LAR). A large proportion of the 51-mile long LAR is lined with concrete, providing flood
protection for surrounding communities; however, some areas still contain natural riverbed
sections supporting limited habitat. Currently the LAR is in the process of being restored to a
more naturalized state possessing natural riverbed with engineered levees along an 11-mile
stretch. Since approximately 20% of the river is slated for restoration, and much of the river is
existing outside of the scope of the restoration effort, there is a need for better understanding of
the benefits of restoration and consequences of the river remaining in an artificial state.
This study examined fluxes in WQ parameters where water flows from fully concrete
lined sections of the LAR into natural riverbed sections. Two RS indices were chosen to carry
out analysis on two specific WQ parameters: the Normalized Difference Turbidity Index (NDTI)
to measure turbidity and Normalized Difference Vegetation Index (NDVI) to measure
chlorophyll-a, respectively. Turbidity provides a physical WQ characteristic while chlorophyll-a
provides a biological WQ characteristic. NDTI and NVDI were calculated using National
Agriculture Imagery Program (NAIP) aerial imagery and the results were processed herein.
Research questions in this study focus on analyzing spectral characteristics of remotely
sensed NAIP aerial imagery over each study frame in the LAR. Study frames were chosen with a
focus on areas where concrete riverbed and natural riverbed sections interface. Readily available
NAIP aerial imagery contains adequate spectral and spatial resolution to perform NDTI and
NDVI functions and produce information on WQ. The NAIP imagery used in this study had a
spatial resolution of 0.6m and 4 spectral bands: visible red, blue, green, and near infrared.
2
Previous investigations by the author revealed fully concrete-lined sections of the LAR
had higher levels of turbidity than natural riverbed sections, with concrete sections also
containing considerable variation in spectral distribution among different sample areas. These
results support the assumption that water quality is relatively better in natural riverbed sections
while the diversity of spectral radiances in concrete sections calls for further research and
indicates the resilient behavior of the system. Substantial restoration of the LAR channel stands
to increase environmental services and opportunities for recreation, while providing improved
environmental conditions and positive outcomes for the Los Angeles region.
1.1. Study Area
This study is focused on the LAR which is situated entirely within Los Angeles County
in California within a highly urbanized environment. The LAR as it currently exists is comprised
primarily of engineered concrete flood control channel with a few exceptions where natural
riverbeds persist. The modern-day LAR begins in the San Fernando Valley at the confluence of
Bell Creek and Arroyo Calabasas (U.S. Corps of Army Engineers 2015). From its headwaters in
the San Fernando Valley the LAR follows an easterly course along the north side of the Santa
Monica Mountains until reaching their easterly edge where the river bears south at an area called
the Glendale Narrows (Gumprecht 2001). The Glendale Narrows is formed by a gap between the
Santa Monica Mountains and the Verdugo Mountains. Once the LAR exits the Glendale Narrows
it flows into a vast coastal plain to the Pacific Ocean. The full course of the LAR is
approximately 51 miles long. The full extent of the LAR and watershed boundary can be seen in
Figure 1 for reference.
3
Figure 1. Los Angeles River extent and watershed
4
1.2. Background and Motivation
The reason the present-day LAR is functionally a flood control channel is due to a
combination of rapid urban development and local geology. The eastern portion of the San
Fernando Valley and the coastal plain that the LAR flows through is comprised of porous
decomposed minerals with origins in the surrounding mountain ranges (Gumprecht 2001).
Historically the LAR changed course often and unpredictably through the coastal plain as the
river would reduce down to a trickle in dry months and become a raging torrent during the wet
season. Extremely variable in nature, the LAR is prone to flooding during the rainy season from
the months of October to March and capable of moving around 14 times the volume of the flow
of the Hudson River in New York (U.S. Army Corps of Engineers Los Angeles District 2015).
As more people settled into Los Angeles the region shifted from an agricultural landscape
to an urbanized cityscape. As development encroached into the broad floodplain of the LAR in
the 19th and 20th centuries, storms produced massive flows that caused loss of life and millions
of dollars in property damage (U.S. Army Corps of Engineers Los Angeles District 2015). The
culminating event that spurred the mass construction project that produced the channelized
version of the LAR today was the Flood of March 1938. The peak flow of the LAR during the
1938 flood was greater than the average flow of the Mississippi River at St. Louis and the flood
water led to 45 deaths and $17 million dollars in damages (Gumprecht 2001). The first flood
control projects began shortly after 1938 flood and continued until 1970 (Gumprecht 2001). The
character of the river has changed dramatically with 94 percent of the LAR having concrete
banks and three quarters of the river’s course being fully concreted (Gumprecht 2001). Concrete
and natural riverbed reaches of the LAR are shown with locations of sample frames used in this
study in Figure 2.
5
Figure 2. Concrete and natural reaches of LAR
A feasibility report completed in 2015 compressed the original 32-mile long study area
into plans for an 11.5-mile long restoration effort (U.S. Army Corps of Engineers Los Angeles
6
District 2015). Being that the LAR is approximately 51 miles in length, a substantial portion of
river corridor is not covered by the USACE-led restoration effort. The USACE work to be
conducted is substantial; however, the example set by the partial restoration can serve as a proof
that further naturalization of the LAR is feasible. The City and County of Los Angeles can
observe the USACE restoration and consider expanding restoration to other portions of the river
channel in years to come. Proposed work over the 11.5 mile stretch in the study may return that
portion of river to relative normalcy although the remainder of the river will remain in a fully
channelized state. Moreover, communities downstream of the USACE restoration area will not
reap the same environmental benefits of restoration as their neighbors upstream. Large-scale
restoration of the LAR can revive the river system to a more natural ecological state while
improving the quality of life for people living in adjacent communities.
Past studies on the LAR and other urban watersheds in the vicinity of the study area
focused mainly on pollutants in the corridor (Ackerman 2003; Ackerman and Schiff 2003; Stein
and Tiefenthaler 2005; Stein and Ackerman 2007; Stein and Yoon 2007; Boroon and Coo 2015).
Alternatively, reports conducted by Los Angeles County focus on civil engineering perspectives
and leave out assessments of the benefits from semi-natural sections of the river. Environmental
consequences due to full channelization are accepted as the price of flood protection. Ultimately,
there is not much known about the effects of the LAR’s channel substrate on the hydrochemistry
of the water that flows though it (Boroon and Coo 2015). Moreover, previous studies fail to
evaluate the environmental benefits and costs of both the semi-natural and fully channelized
states of river sections. Observation of water quality can serve as a litmus test linking the
environmental effects imposed by a concrete riverbed versus a natural riverbed. A focused chain
of analysis using remotely sensed data in the LAR can provide data rich insights. Using ArcGIS
7
Pro to identify and analyze the spectral radiance of the environment at specific dates can provide
temporal snapshots of water quality in the LAR through time.
1.3. Augmentation of Data through Remote Sensing
Since imagery sources for this project encompass the natural riverbed and concrete
riverbed sections in the same temporal frame, comparison of the two environments is direct.
These analyses are sufficient to augment previous studies and provide a more complete
understanding of the river system at present. A remote sensing-based approach to analysis of the
LAR can capture the spectral characteristics of the river channel prior to, during, and after
restoration efforts have concluded. The temporal resolution of the aerial imagery provides for
monitoring and investigation of the LAR unlike other temporally sporadic methods of
investigation which require field work.
For the LAR downstream of the spatial limits of the USACE study and restoration area, a
Draft Program Environmental Impact Report (LAPIER) was prepared by ICF International Inc.
for Los Angeles County Public Works. This report concluded that a restoration alternative of
implementing a natural riverbed while maintaining the current engineered channel configuration
would increase flood risk and not improve water quality and therefore was eliminated from
project plans (ICF International Inc. 2021). This assertion appears to contradict the USACE-
conducted study upstream which is used sections of semi-natural LAR in Glendale as an example
for future restoration efforts. Portions of the Glendale reaches of the river are composed of a
natural riverbed within an engineered channel in which natural hydrologic processes persist and
flooding events are controlled. The USACE expects similarly designed future restoration areas to
have the same behavior (U.S. Army Corps of Engineers Los Angeles District 2015). An example
of this area can be seen in Figure 3.
8
Figure 3. Section of natural bottom riverbed adjacent to the Glendale Narrows Riverwalk.
Author January 2022.
The Glendale Narrows is a working example of a hybrid natural/engineered river channel
providing adequate flood protection while maintaining a functioning riparian habitat (U.S. Army
Corps of Engineers Los Angeles District 2015 ES-4). A remote sensing-based approach to
observe this area and several other study sites along the LAR can provide evidence to test the
conclusion found in the LAPIER report from 2021 regarding any naturalized channel restoration
efforts being unfeasible. Specifically, a remote sensing-based methodology which focuses on
comparing water quality in natural and concrete riverbed reaches of the LAR can illustrate the
benefits of environment services provided in the semi-natural stretches of the river.
Utilizing remote sensing technology to determine water quality in the LAR serves
significant purposes closely linked with restoration efforts. Critically, as stated previously, Los
9
Angeles County Public Works has arrived at the conclusion that restoration alternatives which
include naturalizing the riverbed while maintaining engineered levees do not improve water
quality. Remote sensing derived water quality data can provide evidence that natural riverbed
portions of the LAR contain better water quality than concrete lined sections. The observation of
better water quality in natural riverbed sections coupled with the USACE assertion that natural
riverbed-engineered levee sections are flood safe can serve as support for policy makers to
encourage a pro-restoration position on the LAR.
Furthermore, as restoration efforts are conducted observation of water quality can be
monitored at all stages of construction. USACE anticipates increased turbidity during restoration
therefore remote sensing based observation could provide an oversight function during the
construction process. Additionally, beyond the link to restoration efforts, the LAR is heavily
affected by effluent discharges (U.S. Army Corps of Engineers Los Angeles District 2015, 3-24).
The remote sensing workflow developed within this study can also be focused on areas where
effluent enters the LAR to provide monitoring of potential pollutant loads affecting water
quality. The dense datasets that imagery provides can serve myriad use cases. Lastly, the image-
based analysis of water quality in the LAR serves to bolster the scientific literature related to
remotely sensed water quality observation in urban environments.
1.4. Thesis Structure
The remainder of thesis consists of four chapters. In Chapter 2 focuses on related works
investigation and provides information on the behavior of electromagnetic radiation in relation to
hydrospheric science, characteristics of remote sensing-based hydrologic observations and the
corresponding validity, as well as information on current local restoration efforts on the LAR.
Chapter 3 describes research questions, the data, the sample sites, and the research design.
10
Chapter 4 summarizes the results of the study for all of the sample sites. Lastly, Chapter 5
discusses the utility and limitations of the current work and suggests future work.
11
Chapter 2 Related Work
This includes sections providing a foundation for a remote sensing-based proof of concept for
investigating water quality in the LAR. The first section provides background information on the
behavior of electromagnetic radiation and the integral aspects of hydrospheric science in relation
to the study. The second section provides a basis for exploring the hypothesis that water quality
is better in natural riverbed sections compared to fully concrete riverbed sections. This section
also provides background on remote sensing-based techniques for measuring chlorophyll-a and
turbidity levels in water bodies. Lastly, the third section focuses on municipal reports outlining
planned and potential restoration efforts in the LAR corridor.
2.1. Electromagnetic Radiation and Hydrospheric Science
Owing to the nuanced nature of applying remote sensing techniques to extract
information from water bodies, it is appropriate to provide some background related to the
methodology used in this study. Remote sensing-based observation of hydrologic features
fundamentally differs from observation of the Earth’s surface. In observing land remotely, the
electromagnetic radiation received at the sensor is the product of reflection from the surface.
Alternatively, in water body observations the majority of the electromagnetic radiation that
reaches the sensor is produced from volume reflection which is a composite of the spectral
properties of the full water column (Campbell 2011). As illustrated in Figure 4, additional
sources of electromagnetic radiation that reach the sensor include atmospheric scattering, surface
reflection and bottom reflection (Campbell 2011). These additional sources can distort the true
spectral profile of the observed water column and its constituents (Yu et al. 2020).
12
Figure 4. Electromagnetic radiation behavior in water bodies and atmosphere (Campbell
2011, 551 as modified from Alföldi 1982, 318)
A common phenomenon that alters the spectral profile of light entering a remote sensor is
atmospheric scattering. Atmospheric scattering can be defined as disruption of the original path
of electromagnetic radiation by the contents of the atmosphere, with this effect being exacerbated
by the distance traveled through the atmosphere (Campbell 2011). For remote sensors fixed to
aircraft and Unmanned Aerial Vehicles (UAVs), atmospheric effects are not as significant as
with satellite mounted sensors which may contend with light traveling through the full width of
Earth’s atmosphere (Campbell 2011). The metadata accompanying the 2020 NAIP imagery
applied in this study indicates that atmospheric effects were rectified by Leica XPro software
prior to being made available to the public for download (U.S. Geologic Survey 2021).
A key challenge to account for in remote sensing of water is surface reflection. Surface
reflection, also known as sun glint, is caused by water body surfaces being uneven and sun light
13
not entering the water medium but changing direction back into the atmosphere (McKim et al.
1984). Surface reflected electromagnetic energy collected by remote sensors skews data since the
spectral strength is higher than volume reflected light (McKim et al. 1984). Effective solutions
for removing sun glint from imagery have been developed. Notably, a study published in 2005
by Hedley provides an elegant regression equation based on manual pixel sampling and using
visible and NIR wavelengths to eliminate glint successfully.
Water depth in the LAR is relatively shallow in the concrete sections with the submerged
riverbed being visible in the NAIP aerial imagery making bottom reflectance a factor that must
be considered. Bottom reflectance is electromagnetic radiation reflected off the terrestrial bottom
of a water body and contributes to distortion of the observed spectral character of the water
column (Tolk, Han and Rundquist 2000; Katlane et al. 2020; Yu et al. 2020). It is possible the
bottom reflectance from the concrete riverbed sections is spectrally significant in shallow water.
To accurately measure chlorophyll-a and turbidity levels in the water column, imagery
containing the deepest possible water in each sample frame was selected to mitigate spectral
distortion from bottom reflectance prior to NDVI and NDTI function deployment.
2.2. Remote Sensing Based Hydrologic Observation
Anthropogenic caused upward nutrient fluxes impacting water systems can lead to
negative consequences on environments (Carpenter et al. 1998). It has been documented in the
past that the LAR has elevated nutrient (nitrogen) levels beyond the acceptable threshold as
determined by the Clean Water Act and that these nutrient loads can lead to algal growth (U.S.
Army Corps of Engineers Los Angeles District 2015). Algal blooms have the effect of removing
oxygen from the water body they occupy as they decompose, effectively producing an
environment in which many organisms cannot survive (Chislock et al. 2013). Algal growth in the
14
LAR beyond a typical baseline have the potential to cause harm to water quality and the
ecosystem as a whole.
Living plant matter contains pigments integral in the process of photosynthesis, with the
most common being chlorophyll-a (Campbell 2011). The specific spectral character of
chlorophyll-a provides a linkage for the detection of algae in water bodies (Richardson 1996).
Previous studies have successfully detected chlorophyll-a concentrations in aquatic environments
(Avdan et al. 2019; Kuhn et al. 2019). Vegetation indices such as NDVI are designed to identify
vegetation in digital imagery data by exploiting inherent spectral properties in plant tissue using
ratios of light wavelengths to produce a relative intensity value (Campbell 2011). NDVI is
attractive for use in this study since it requires visible red and near infrared wavelengths which
are included in available NAIP imagery. NDVI specifically has been shown to be effective in
measuring chlorophyll-a in streams when utilized with high spatial resolution imagery containing
visible blue, green, red, and near infrared bands (Kim et al. 2021).
Turbidity can be summarized as the level of sediment immersed in a body of water
(Campbell 2011). Measurement of turbidity has direct significance to the USACE restoration of
the LAR since construction activities will cause elevated sediment concentration in the river
during restoration efforts and afterwards as the water channel returns to normal suspended
sediment levels (U.S. Army Corps of Engineers Los Angeles District 2015
). Remotely sensed detection of turbidity has potential advantages of efficiency and low-
cost over analysis of field collected water samples (McKim et al. 1984).
It was hypothesized that natural riverbed portions of the LAR possess the behavior of
having lower turbidity levels than concrete-lined portions of the river. On a macroscopic scale, it
has been observed that natural watershed catchments contain lower levels of WQ parameters
15
such as nitrates and total suspended solids than developed watershed catchments (Stein and
Yoon 2007). By extension, on a smaller scale, it can be surmised that natural riverbed sections
contain better water quality than artificial sections. Natural riverbed sections contain riparian
vegetation whereas concrete reaches of the LAR are barren. Vegetation has the behavior of
reducing turbidity in waterbodies (Austin et al. 2017; U.S. Army Corps of Engineers Los
Angeles District 2015). In addition, higher velocity water allows for higher turbidity thresholds
and concrete lined sections of the LAR are designed to convey water through the channel as
efficiently as possible. Natural riverbed portions of the LAR contain heterogenous fluvial
morphology that has an effect of reducing flow velocity and by extension the turbidity threshold.
Lacaux et al. (2007) produced the Normalized Difference Turbidity Index (NDTI) for
measuring turbidity in ponds in West Africa. The NDTI requires the visible green and red
spectral bands to function and the available NAIP imagery includes these bands. This index was
applied in a study published by Chen et al. (2022) to measure the effects of the Covid-19
lockdowns on water quality in the Haihe River Basin in China, effectively documenting
significant drops in turbidity after the Covid-19 outbreak. The spatial resolution of the imagery
used in many remote sensing based-water quality studies is sourced from satellites and is coarse
when compared to the resolution available from aerial sensors. The studies by Lacaux et al.
(2007) and Chen et al. (2022) relied on imagery with a maximum spatial resolution of 10m
which is relatively coarse in comparison to the 0.6m resolution utilized in this study. The NAIP
imagery available in the study areas allows for analysis of water quality phenomena at a fine
spatial resolution.
16
2.3. Validity of Remote Sensing Derived Measurements
Another consideration is the strength of correlation between remote sensing-based
measurements with measurements derived from field collected samples. Numerous studies have
explored the efficacy of remote sensing as a means to determine various water quality
parameters. Within the literature, strong correlations have been demonstrated between remote
sensing-based measurements and measurements derived from in-situ water samples of turbidity
and chlorophyll-a concentration. Bid and Siddique (2019) derived NDTI values from LANDSAT
5 and 8 satellite images then validated them against laboratory tested water samples and yielded
an average correlation coefficient value (R²) of 0.90. Kim et al. (2021) achieved a strong
correlation between NDVI and chlorophyll-a of R² = 0.88 using an UAV to collect imagery and
unmanned surface vehicle (USV) equipped with a fluorometer to collect water samples. Studies
specific to validating remote sensing derived NDTI and NDVI values with in-situ water samples
tested for actual turbidity and chlorophyll-a are listed in Table 1. Water quality assessment by
remote means has been established as a reliable technique in lieu of in-situ water quality
measurements. For the scope of studying how the concrete lined and natural riverbed sections of
the LAR may impact water quality in distinct aspects, the remote sensing indices NDTI and
NDVI are suitable. Additionally, imagery offer the advantage of capturing the spectral
reflectance of a relatively large spatial area at one simultaneous moment in time whereas in-situ
samples are collected as data points. Any portion of an image can be analyzed. Alternatively,
each in-situ data point is confined to a minute extent.
17
Table 1 Correlation strength of RS indices and in-situ samples
Study RS Index RS Data In-Situ Data R
2
Kaplan et al.
2020
NDTI RapidEye-3 satellite
images (5m spatial
resolution)
Water samples laboratory
tested for actual turbidity
level
0.82
Elhag et al. 2019 NDTI Sentinel-2 satellite
images (10m spatial
resolution)
Water samples laboratory
tested for actual turbidity
level
0.94
Bid and
Siddique 2019
NDTI LANDSAT 5 and 8
satellite images (30m
spatial resolution)
Water samples laboratory
tested for actual turbidity
level
0.90
Rodríguez-
López et al.
2020
NDVI LANDSAT 5, 7 and 8
satellite images (30m
spatial resolution)
Water samples laboratory
tested for chlorophyll-a
using fluorometric technique
0.78
Kim et al. 2021 NDVI Parrot Sequoia UAV
images (11cm spatial
resolution at 120m
flight altitude) Pix4D.
2022.
Water samples collected by
an unmanned surface vehicle
and tested for chlorophyll-a
using an on board
fluorometer
0.88
2.4. Municipal and Federal Reports Focusing on the Los Angeles River
The Los Angeles County Public Works Draft Program Environmental Impact Report
contains the most recent released Environmental Impact Report and is critical to planning and
development of the LAR corridor within Los Angeles County (LAC) for the next 25 years (ICF
International Inc. 2021). This expansive text will serve as a foundational reference related to
municipal planning of the LAR in LAC. The 2015 USACE produced report (Feasibility Report)
centers on an approximately 11-mile stretch of the LAR and serves as the guiding document for
restoration efforts over that study area. This report has a wealth of information related to flood
18
control, local ecology, and the history of the LAR. Each of the reports conducted for local
municipalities offer a window into the future of policy making revolving around the LAR.
A remote sensing-based focus on water quality in the LAR corridor has the potential to
fill knowledge gaps that are either under studied or not studied at all. In addition, a richer
knowledge of the implications of natural riverbed versus concrete riverbed sections on WQ may
encourage LAC to revise its position that a natural channel will not improve water quality (ICF
International Inc. 2021). The conclusions reached in the reports produced by the City of Los
Angeles (working with USACE) and Los Angeles County are fundamentally different and
possibly contradictory. Whereas the USACE report has elected to go forward with restoration
efforts of the LAR that include unearthing natural riverbed while maintaining engineered levees,
the Los Angeles County draft report has decided that such alternatives do not meet their criteria.
Remote sensing focused on the true spectral character of the LAR has the potential to illuminate
the water quality dynamics along one of the most infamous stretches of river in the United
States.
19
Chapter 3 Methods
The following research questions inform the subsequent research within the LAR. What
is the extent of the differences in chlorophyll-a and turbidity concentrations when comparing
natural riverbed sections with concrete riverbed sections of the Los Angeles River? What
negative consequences on WQ, if any, can be attributed to sections of concrete river channel?
What benefits on WQ, if any, can be observed in natural riverbed sections? Two indices will be
used to determine different aspects of WQ. NDVI will provide reflectance values corresponding
to chlorophyll-a mostly associated with algae within the LAR channel. NDTI will provide
turbidity levels in the LAR. Each index delivers information correlated with WQ metrics that
will provide a deeper understanding of the environmental effect of river channelization.
The results of the image data processed through RS indices in this study serve to preserve
a window into baseline environmental behavior prior to river restoration. Availability of water
quality metrics available at adjustable locations along the full length of the LAR during specific
temporal windows offer benchmarks for post-restoration LAR water quality to be compared
against. In addition, with the availability of relatively low cost of UAVs the workflow presented
within this experiment can easily be adapted for concurrent and post-restoration observation of
aquatic environmental metrics. Furthermore, using the same methodology, effluent discharge
zones along the LAR can be readily monitored aerially.
3.1. Data Description
In order to carry out precise remote sensing-based water quality analysis in the LAR,
there are some specific resolution requirements. High spatial resolution is crucial since parts of
the river channel contract to relatively narrow widths. For example, in the Glendale concrete
sample area the width of channel containing flowing water is roughly 3.7m wide. The raster
20
imagery needs a spatial resolution capable of resolving the highly variable water surface extents
within the LAR. Within the vicinity of the LAR, NAIP aerial imagery is available from 2005 at a
lower resolution and available roughly biennially with some larger gaps in availability. Starting
in 2016 imagery was made available at a 0.6 meter spatial resolution.
Each of the indices employed in the methodology, NDVI and NDTI, need differing
combinations of electromagnetic wavelengths to function. NDVI measures relative vegetative
vigor and requires visible red and near infrared spectral bands. NDTI is used to measure turbidity
in water and requires visible red and visible green spectral bands. NAIP imagery is adequate for
the purposes of the study as it consists of the following four spectral band: visible red 619-
651nm, visible green 525-585 nm, visible blue 435-495 nm and near infrared 808-882 nm. Table
2 lists the resolution information for NAIP aerial imagery.
Table 2 Required data.
Required
Data
Spatial
Resolution
Spectral Resolution Temporal
Resolution
Data
Availability
NAIP aerial
imagery
*0.6 meter *Four spectral bands:
visible red 619-651 nm
visible green 525-585
nm
visible blue 435-495
nm
near infrared 808-882
nm
Imagery dataset
available annually
to biennially.
*Image tiles
captured 2018 and
2020.
Freely available
for download
from USGS Earth
Explorer.
Coverage is
available over full
extent of the LAR.
*Data resolutions are reported in metadata file of NAIP imagery download (U.S. Geologic
Survey Earth Explorer 2021)
The spatial scope of NAIP imagery covers the full extent of the approximately 51-mile
long LAR allowing for manually extracted samples to be taken anywhere along the channel. The
image tiles used in the study were collected in the spring summer months of 2018 and 2020.
21
Fortunately, NAIP image tiles are available at no cost to the public. Image collection dates for
each sample frame in each year are recorded in Table 3.
Table 3 Image acquisition dates
Sample Frame Date
Sepulveda Basin 2018 July 22, 2018
Sepulveda Basin 2020 April 25, 2020
Glendale 2018 July 23, 2018
Glendale 2020 May 05, 2020
Willow Street 2018 July 23, 2018
Willow Street 2020 May 05, 2020
3.2. Sample Sites
The central component under review within this study is identifying variations in water
quality between concrete and natural riverbed sections within the LAR system. Areas where
these two differing environments transition are a key requirement in sample site selection. The
sole examples of where concrete to natural riverbed transitions currently occur in the LAR are
the Sepulveda Basin, Glendale, and at West Willow Street in Long Beach. Each of these
locations was selected as sample locations and each location holds unique characteristics that set
them apart from each other.
Within each sample site, equally sized polygons were chosen from natural and concrete-
lined sections for study. These polygons were chosen visually to ensure an unobscured water
surface for study. Another method for extracting the water surface in an image relies of Equation
1:
22
Normalized Difference Water Index (NDWI) which is the result of the equation:
NDWI = (Visible Green - Near Infrared) / (Visible Green + Near Infrared) (1)
Theoretically, placing NDWI into the methodology would be advantageous over manually
creating polygons to extract areas of interest. However, shadows prevented accurate results with
an NDWI analysis, so this method was deemed insufficient.
3.2.1. Sepulveda Basin
The Sepulveda Basin sample frame lies within the San Fernando Valley in a flood control
basin enclosed by the Sepulveda Dam which was completed in 1941 by the USACE in response
to the extensive flooding in the Los Angeles region in 1938 (U.S. Army Corps of Engineers Los
Angeles District 2011). Being one of the last remaining natural corridors along the LAR, there is
a continuous stretch of natural riverbed through the basin of approximately 2.3 miles. This semi-
natural stretch of the LAR contains several confluences with other surface water sources that
have downstream proximity to sample sites. Approximate measurements taken in ArcGIS Pro
beginning from the concrete to natural riverbed transition marking the beginning of the
Sepulveda Basin indicate Bull Creek enters the river 4145 ft downstream, treated discharge
originating from the Donald C. Tillman Water Reclamation Plant (DTCWRP) enters 5180 ft
downstream, and Woodley Creek enters the channel 8875 ft downstream.
In the Sepulveda Basin sample area, a shallow trapezoidal concrete channel
approximately 14 feet wide containing water transitions into a natural riverbed containing
relatively deep slow moving water surrounded at its banks with thick vegetation. All sample sites
occur upstream of any major surface water confluences in the Sepulveda Basin; however, the
final natural sample site (SN4) lies 600 ft upstream of Bull Creek within a stretch of water
23
bounded by dam-like structures. Relative locations of all concrete and natural sample sites are
listed in Table 4. and Figure 5 specifies the complete Sepulveda Basin sample frame.
Table 4 Sepulveda Basin Sample Distribution
Sample
Polygon
and Key
Features
Distance
(ft)
SC1 0
SC2 84
SC3 166
SC4 250
Concrete
to Natural
Riverbed
Transition
615
SN1 1330
SN2 1703
SN3 3292
SN4 4128
24
Figure 5. Sepulveda Basin sample frame
3.2.2. Glendale
The Glendale sample frame is located on a stretch of the LAR situated along the
boundary of Griffith Park where the cities of Burbank, Glendale and Los Angeles converge. The
water table in portions of this zone exist too close to the land surface to allow for paving and
accordingly, the current iteration of the LAR here contains stretches of natural riverbed with
engineered concrete banks (Gumprecht 1999). The concrete stretches were sampled 2.6 miles
upstream from the samples collected in the natural riverbed area within this sample area due to
the physical characteristics of the channel in this location. There is an approximately 12 ft wide
rectangular concrete channel conveying relatively deep water ideal for RS analysis that abruptly
25
transitions into a roughly 105 ft wide concrete channel with water too shallow to permit capture
of genuine spectral characteristics.
The structure of the LAR maintains a wide concrete channel containing shallow water for
1.7 miles before reaching a confluence with another concrete lined waterway, known as the West
Burbank Channel, and then travels 550 ft to a concrete to natural riverbed transition. This stretch
of natural riverbed bed runs for 1700 ft, changes to a 330 ft concrete strip before another
transition to a 1511 ft run of natural riverbed where the natural samples were collected for this
sample area. Cobble stones and sediment with trees and vegetation have taken hold in the soft
bottom stretches of this zone. Also, another notable factor related to this sample frame is the
Burbank Water Reclamation Plant (BWRP), which discharges treated wastewater into the West
Burbank Channel before the confluence with the LAR. Specific locations of sample polygons
and key features for the Glendale sample area are listed in Table 5 and Figure 6 encapsulates the
extent of the sample frame.
26
Table 5 Glendale Sample Distribution
Sample
Polygon
and Key
Features
Distance
(ft)
GC1 0
GC2 119
GC3 242
GC4 355
West
Burbank
Channel
Confluence
9792
Concrete
to Natural
Riverbed
Transition
10342
Natural to
Concrete
Riverbed
Transition
12038
Concrete
to Natural
Riverbed
Transition
12368
GN1 13884
GN2 13948
GN3 14003
GN4 14011
27
Figure 6. Glendale sample frame
3.2.3. Willow Street
Located in Long Beach the Willow Street sample area occurs near the end of the course
of the LAR. Similar to Glendale the water table in this area occurs too close to the ground
surface for the bottom of the LAR to be paved. For approximately 2.5 miles the river consists of
an earthen bottom before reaching its mouth and exchanging water with the Pacific Ocean
(Gumprecht 1999). The river in this zone consists of a wide sediment laden concrete channel
with a roughly 25 foot wide rectangular depression situated at its centerline containing relatively
deep flowing water. Abruptly after the LAR passes under the Willow Street bridge concrete
channel transitions to a natural riverbed area with slow moving water.
28
In comparison to the other sample frames the natural samples sites at Willow Street occur
far closer to their upstream concrete sample counterparts. Natural sample sites were chosen close
to the upstream concrete sample sites in this case in an effort to avoid river water subject to tidal
mixing. The surface elevation of the LAR drops shortly after the concrete to natural riverbed
transition following the Willow Street bridge and becomes continuous with estuarine waters that
have a higher likelihood of being brackish. Freshwater in the LAR is the subject of this study
thus the natural riverbed sample site placements in this sample frame are chosen using this
additional selection criterion to avoid river water mixed with ocean water. Sample site locations
are indicated per Table 6, and Figure 7 details the complete Willow Street sample frame.
Table 6 Willow Street sample distribution
Sample
Polygon
and Key
Features
Distance
(ft)
WC1 0
WC2 59
WC3 117
WC4 172
Concrete
to Natural
Riverbed
Transition
686
WN1 746
WN2 789
WN3 810
WN4 861
29
Figure 7. Willow Street sample frame
3.3. Research Design
The image analysis workflow begins with acquisition of image tiles covering the sample
sites. The full workflow is summarized in Figure 8. Raster images are generally relatively large
files and NAIP tiles are quite expansive measuring approximately 6.27 km x 7.4 km, thus it is
necessary to remove the portions of image that are outside of the river channel to improve
processing speed in ArcGIS Pro. To make the image tiles containing the study areas of interest
more manageable a raster clip function was performed. The significance of isolating the river
channel from the remainder of the image tile was also central to the WQ analysis conducted as
30
part of this research project. The aim was to isolate spectrally consistent portions of water
surface that represent the spectral character of the water column.
Figure 8. Image analysis workflow
31
The clip function isolating the LAR channel is conducted by manually creating a polygon
feature that closely outlines the outer edge of the engineered banks of the river. This polygon is
then used to clip the preferred section of the raster image for further processing. Moving forward
in the chain of analysis, the best sample areas are determined in concrete riverbed and natural
riverbed sections using visual image interpretation. Criteria employed for selecting ideal sample
areas included deeper water devoid of surface debris, no shadows and minimal sun glint. All of
these factors are in keeping with the goal of collecting genuine volume reflectance of the water
column with little or no spectral noise.
In targeting deeper water, the assumption is bottom reflectance will be minimized, as it
has been observed in laboratory testing that bottom reflectance is minimal in turbid water at a
depth of 80 cm (Tolk, Han and Rundquist 2000). When water surfaces become uneven, they can
produce sun glint by reflecting the sun and producing brighter reflectance values than calm water
(Campbell 2011). The LAR is structurally dynamic containing areas that cause some relatively
significant water surface disturbance. Therefore, manual imagery interpretation will be
administered prior to imagery analysis to rule out pixels that are spectral outliers containing sun
glint. Acquiring a sample as close to spectral truth as possible from the LAR is crucial in
developing a clear understanding of the state of water quality. The surface debris and shadows
observable in the LAR also contain spectral characteristics that diverge from the volume
reflectance of smooth water.
When areas containing the aforementioned criteria were located, manually created
polygons were produced in ArcGIS Pro using the Create Features tool to isolate and clip the
raster image samples. Efforts were made to place sample polygons at uniform distances from
each other however placement is largely dictated by the availability of samples meeting the
32
criteria determined by manual image interpretation techniques. Each clip polygon measured 10 ft
by 40 ft. The reasoning for these specific sample polygon dimensions is threefold. First the 10 ft
polygon width is necessary since certain concrete stretches of the LAR containing water
condense to a width of approximately 12 ft wide. The second consideration is obtaining an
adequate number of pixels in each raster image sample to capture statistically archetypal
information which is spectrally consistent with the target samples. Third, in observing
phenomenon in a feature such as a river, the phenomena occur over a continuous field; however,
it is necessary to organize samples into discrete objects due to finite thresholds in sampling
precision (Montello and Sutton 2013). Breaking samples into standardized discrete objects
within the study is especially significant owing to the multiple criteria used a viable image raster
to calculate the RS indices.
It is suggested to use 10 x n pixels for developing training sites for supervised
classification with n representing the number of raster image bands (Green, Congalton, and
Tukman 2017). This method has been extrapolated here to set a minimum pixel count for input
rasters being processed to produce the RS indices used in this study. The imagery used in the
workflow contains four spectral bands and a pixel resolution of 0.6 m thus the clip polygons
exceed the minimum pixel counts per the 10 x n pixels benchmark while retrieving spectrally
precise samples.
At each sample frame, four image samples are taken in the concrete section and four
samples are taken in the natural section downstream of the concrete to natural riverbed interface.
After image raster samples have been extracted for each riverbed type in each study area for all
image tile years, NDVI and NDTI functions are applied to them using the Band Arithmetic tool
in ArcGIS Pro:
33
NDVI = (Near Infrared - Visible Red) / (Near Infrared + Visible Red) (2)
NDTI = (Visible Red- Visible Green) / (Visible Red + Visible Green) (3)
Raster images clipped with uniform polygons can be seen below in Figure 9.
Figure 9. Glendale sample frame. Left: concrete riverbed sample. Right: natural riverbed sample.
NAIP Imagery (U.S. Geologic Survey 2021). Sample area rasters produced in ArcGIS Pro.
Following NDVI and NDTI outputs for each sample area, histograms were produced for
initial statistical visualization. Histograms are effective in identifying patterns in NDVI and
NDTI value distributions as well as the overall means. Next, mean index values produced by the
histograms from each sample site index result were organized into .csv files for further analysis.
The .csv files containing mean RS index values are then inputted into Python and charts are
created with Matplotlib in order to better visualize the chlorophyll-a and turbidity levels as they
pass downstream through the LAR and crucially as water transitions from concrete to natural
riverbed environments.
34
Chapter 4 Results
This chapter documents the NDTI and NDVI results for each sample site in each sample frame
within the LAR. The mean index values derived from pixels confined in each image raster clip
are represented in charts produced for each sample frame. RS index result charts are produced
using Pandas and Matplotlib library commands in a Python environment. For each sample frame
location, charts represent results in the direction of flow in the LAR traveling downstream
beginning with concrete lined sample sites and then transitioning to natural riverbed sites. As
noted in Chapter 2, the distances between each sample site within each sample location was not
uniform and placement was dictated by availability of samples meeting various criteria.
4.1. Sepulveda Basin
This section summarizes the results for the Sepulveda Basin sample frame, the NDTI-
turbidity results, and finishing with the NDVI-chlorophyll-a results.
4.1.1. NDTI-Turbidity
Within the Sepulveda Basin turbidity was observed to be of lower concentration in
natural sample sites as opposed to concrete sample sites for all data points collected. In
comparing physical characteristics of the other sample frames in the study, Sepulveda Basin
contains the longest stretch of natural riverbed, is heavily vegetated, and contains slow moving
water. Vegetation has an attenuating effect on turbidity intensity in waterbodies. Moreover, there
are no WRPs upstream of the sample sites at this location that would influence hydrochemistry.
The DTCWRP discharges treated water into the LAR downstream of the sample sites and is
separated by a rudimentary dam. The features of the Sepulveda Basin make it the most ideal
sample frame in this study for comparing turbidity in concrete and natural riverbed stretches.
35
The final sample site (SN4 in Figure 10) in the natural riverbed section is close to the
confluence with Bull Creek. SN4 is isolated on a stretch of the LAR between two dam like
structures and Bull Creek discharges into the same stretch. It is possible the creek could
influence the NDTI levels near the SN4 sample site. Bull Creek has been observed to exceed
water quality objectives for turbidity set by the California Department of Water Quality Control
Board in wet and dry seasons (Weston Solutions, Inc. 2005). In the 2020 NDTI results the
highest natural riverbed turbidity value was located at the SN4 sample site whereas the 2018
NDTI result at the same site recorded the lowest turbidity which is indicative of the highly
dynamic nature of turbidity within a river system.
Figure 10 Sepulveda Basin NDTI: (a) 2018 and (b) 2020.
4.1.2. NDVI-Chlorophyll-a
NDVI results in the Sepulveda Basin for the 2018 imagery indicate a rise in chlorophyll-a
levels as water travels from the concrete to natural riverbed. In the 2020 imagery, the
chlorophyll-a levels decline as water transitioned from concrete to natural riverbed (Figure 11).
The 2018 results pointing to increased chlorophyll-a levels in the natural stretch of the LAR in
the Sepulveda Basin conflict with the original hypothesis that chlorophyll-a levels should
36
decrease as water flows through the natural riverbed section. Nonetheless, chlorophyll-a levels
are a biological indicator driven by environmental inputs such as hydrochemistry and
seasonality.
Figure 11 Sepulveda Basin NDVI: (a) 2018 and (b) 2020.
One significant factor when interpreting NDVI results in this study is the relationship
between algal growth and nitrate levels in the LAR. Elevated nitrate levels increase algal growth
which can lead to eutrophication and poor water quality (Boroon and Coo 2015). It has been
observed in the LAR that nitrate levels are highest during the dry periods and mainly originate
from runoff in the Sepulveda Basin and runoff from surface streets in highly urbanized areas of
Los Angeles (Boroon and Coo 2015). Within the Sepulveda Basin sample frame, the natural
riverbed sample sites are surrounded by several sports fields with maintained grass. The concrete
sample sites in Sepulveda Basin are upstream of any runoff that would flow from these
manicured sports fields thus less nitrate would be expected to enter the concrete section
ultimately leading to reduced algal growth and lower chlorophyll-a levels detected by NDVI.
Another factor to consider is the time of year each image tile was collected. The 2018
images for Sepulveda Basin were collected in late July while the 2020 images were collected in
37
late April. A spring season algal spike is not apparent in the Sepulveda Basin sample frame
although elevated NDVI values were observed in the 2020 results at the Glendale and Willow
Street sample frames. Each of the natural riverbed sample locations are isolated by concrete
sections of river and have specific physical characteristics that may produce site-specific results.
4.2. Glendale
This section summarizes the results for the Glendale sample site, starting with the NDTI-
turbidity results, and finishing with the NDVI-chlorophyll-a results.
4.2.1. NDTI-Turbidity
The NDTI results for the Glendale sample frame were not as uniform as expected. The
2018 NDTI results shown in Figure 12 show that the turbidity levels in the natural riverbed
sample sites were all higher than any of the turbidity levels in the upstream concrete sample
sites. The elevated turbidity values in the natural riverbed samples in 2018 could be attributed to
the unique characteristics the Glendale sample frame. The BWRP discharges treated wastewater
into the West Burbank Channel roughly two miles upstream, and constituents discharged by the
BWRP may contaminate the river by adding partially treated effluent into the system (Boroon &
Coo 2015). The concrete sample sites in the Glendale sample frame were collected upstream of
the West Burbank Channel and LAR confluence and therefore the hydrochemistry was affected
by this source. The 2020 NDTI results for Glendale indicate the turbidity was lowest in the
natural riverbed sample sites (Figure 12b).
38
Figure 12. Glendale NDTI: (a) 2018 and (b) 2020.
4.2.2. NDVI-Chlorophyll-a
Similar to the NDVI results in Sepulveda Basin, the results for the Glendale sample
frame are split. At Glendale chlorophyll-a concentrations declined in the concrete to natural
riverbed transition in 2018 but increased in 2020. Unlike the Sepulveda Basin sample frame, the
natural riverbed section sampled at Glendale is influenced by an upstream WRP. The highest
nutrient levels in the LAR have been observed downstream of WRPs (Stein and Ackerman
2007). Since the concrete sample sites at Glendale occur upstream of the inlet channel connected
to the BWRP, they are not influenced by the same constituents as the natural rived bed sites. In
addition, the intermittent nature of WRP discharges creates fluctuating hydrochemistry
downstream of any outlet. In any circumstance however, mean NDVI values derived from
individual pixel values in each sample polygon and charted in Figure 13 indicate a clear shift in
chlorophyll-a concentration at concrete to natural riverbed transitions.
39
Figure 13. Glendale NDVI: (a) 2018 and (b) 2020.
4.3. Willow Street
This section summarizes the results for the Willow Street sample site, starting with the
NDTI-turbidity results, and finishing with the NDVI-chlorophyll-a results.
4.3.1. NDTI-Turbidity
The Willow Street sample site signifies the end of the freshwater run of the LAR before
meeting and mixing with the saline water in the Pacific Ocean. River water at this site passes the
Willow Street bridge and abruptly shifts from flowing over a concrete channel into a wide soft
bottom pool where water velocity was observed to be reduced before traveling down grade
through a series of small rapids into the broader soft bottom estuarine stretch of the LAR. The
concrete sample sites are the closest in proximity to the natural sample sites of any of the sample
frames in this study. Additionally, there are no major surface water inlets that mix with the LAR
in the vicinity of the sample sites at Willow Street which means this site may prove to provide a
true view of how turbidity varies in transitions between concrete lined and natural riverbed lined
LAR sections. The 2018 and 2020 NDTI results both indicate lower turbidity in natural sample
sites than concrete sample sits at Willow Street (Figure 14).
40
Figure 14. Willow Street NDTI: (a) 2018 and (b) 2020.
4.3.2. NDVI-Chlorophyll-a
Mirroring the Glendale sample frame, the NDVI results at Willow Street pointed to lower
algal concentrations in the natural section for 2018 while algal concentrations were higher for the
natural section in 2020 (Figure 15). It should also be noted that the highest observed NDVI
values in the study occurred in the Glendale and Willow Street 2020 imagery. Glendale is
downstream of the DTCWRP and BWRP while Willow Street is downstream of all WRPs on the
river and nearly all urban runoff inlets contributing to the LAR. Different from Sepulveda Basin
and Glendale, the concrete and natural riverbed sections at Willow Street have the same inputs
meaning there are no waterway confluences affecting the sample sites, and similar to other
sample frames, the natural riverbed section at Willow Street contains lower velocity water and a
higher residency time than the concrete section.
41
Figure 15. Willow Street NDVI: (a) 2018 and (b) 2020
4.4. ANOVA
The methodology utilized in this study consists of uniform polygon clip boundaries used
to created raster images of the same size for calculation and analysis of RS indices. Index results
within this study therefore are the outcome of a deliberate, balanced experimental design. The
RS index histogram value distributions yielded a mean index value for each sample site and
represent continuous values which are dependent on the execution of each RS index ratio. In
order to determine whether the variation the results of the RS indices used in this study are
statistically significant analysis of variance (ANOVA) was executed. RS index results for
concrete and natural sample sites are compared using a one-way ANOVA. This approach is
appropriate for comparing two or more datasets beholden to one dependent variable (Ross and
Willson 2017). Microsoft Excel was used to apply one-way ANOVA to .csv files containing
NDTI and NDVI index means for each sample site in each sample frame in the study. A check
was also performed running ANOVA on index values in polygon sample sites at the individual
pixel level and it was determined that using the mean index values at the concrete to natural
transition yields conservative p-values in comparison.
42
ANOVA results for four of six NDTI results indicate p-values below the 0.05 threshold
which is widely used as a benchmark for determining statistical significance. The two p-values
above the 0.05 figure were both from 2018 imagery and occurred at Sepulveda Basin and
Glendale with values of 0.08 and 0.16, respectively. Both Sepulveda and Glendale are sites
influenced by several unique factors that may influence the natural riverbed section water
chemistry in ways that are different from the preceding concrete riverbed sections. Constituents
present in the LAR are variable and unpredictable and fluctuate on time scales from hours to
months due to several sources (Stein and Ackerman 2007). The two NDTI ANOVA results
above the p-value threshold could have been the product of turbidity fluctuations local to the
natural riverbed sections that skewed the variance.
Table 7 NDTI ANOVA p-values
Sample Frame p-value 2018 p-value 2020
Sepulveda Basin 0.079202 0.000325
Glendale 0.1575 0.01581
Willow Street 2.01 × 10
-7
0.000891
The ANOVA results for NDVI produced statistically p-values for all outputs considered
in terms of 0.05 standard (Table 8). The least statistically significant NDVI p-value output (0.04)
was recorded at the Sepulveda 2018 sample frame. Relying solely on the 0.05 p-value for
determining whether a claim is correct or false can be reckless when study design, measurement
accuracy and external variables are not also considered (Wasserstein and Lazar 2016). Despite
NDVI results being split, indicating three sample frames having higher algal concentrations in
concrete sections and three sample frames having higher algal concentrations in natural sections,
43
all the ANOVA outputs represent significant variation in RS index values through the concrete to
natural riverbed transitions.
Table 8 NDVI ANOVA p-values
Sample Frame 2018 2020
Sepulveda Basin 0.042524 6.02 × 10
-5
Glendale 8.56 × 10
-6
2.78 × 10
-6
Willow Street 0.005859 0.000485
44
Chapter 5 Conclusions
The main objective of this study was to determine the effects of concrete and natural riverbed
substrates on WQ as water flows through the LAR. More precisely, the extent in variation of
turbidity and chlorophyll-a concentration within the artificial and semi-natural channels of the
LAR were examined by processing imagery data with RS indices. Lastly, the validity of benefits
of natural riverbed sections on WQ was investigated by confirming statistical significance with
ANOVA.
This chapter considers the implications of the results produced in this project as well as
their utility and the limitations associated with the applied methodology. Finally, an alternative
methodology is proposed which expands upon the findings in this study may be used in real
world use cases.
5.1. Discussion
A pattern of turbidity level decrease was observed in five out of six sample sites at
concrete to natural riverbed transitions despite the variations in the physical conditions present at
the three sample frames. The one NDTI result for which there was not a decrease in turbidity in
the natural riverbed zone occurred at Glendale in 2018. The natural section of LAR at the
Glendale sample frame is influenced by discharge from the BWRP, and this may have caused a
turbidity spike in the natural section in 2018. The NDTI result for the 2018 Sepulveda Basin
sample frame showed a decrease in turbidity in the natural sample sites; however, ANOVA
produced a p-value of 0.08 which exceeded the 0.05 p-value benchmark used in this study to
denote statistical significance. Nonetheless, when compared to the NDTI results of imagery
acquired in different years and other sample frames, while bearing in mind the variable nature of
45
constituents present within the LAR at brief temporal scales, it can be observed that turbidity in
the LAR reliably decreases as water flows from the artificial reaches into natural reaches.
With respect to chlorophyll-a concentrations in the LAR, results were split with half the
low chlorophyll-a levels occurring in the concrete sections and half in the natural sections.
Chlorophyll-a in the LAR is a biological indicator subject to delayed response times given
several factors such as water velocity, water temperature and nutrient concentration. All of the
chlorophyll-a results were statistically significant, but it is likely that the prediction of
chlorophyll-a levels in the LAR would be better if additional variables beyond riverbed type
were included.
The cumulative results of each RS index in this study serves as an impetus for continued
attention, study, and action on the LAR. Policy makers in the Los Angeles area can use the NDTI
results in this study as a basis to advocate for expanded restoration of the LAR outside the scope
of the ongoing USACE project. This study suggests naturalized sections of the LAR provide
environmental services that reduce the turbidity of water traveling through them. Additionally,
the NDVI results indicate significant fluxes in chlorophyll-a concentration at concrete to natural
riverbed transition zones which have dangerous potential when associated with algal blooms and
the associated die-offs that cause anoxic conditions. Comprehension of the factors leading up to
sharp shifts in WQ are possible since each of the RS indices used in the study have proven
themselves as promising detection and monitoring tools for turbidity and chlorophyll-a
concentration levels in the LAR. Modifying the methodology in this study to increase temporal
resolution and field validation of the turbidity and chlorophyll-a concentrations will further
increase utility.
46
The current purpose of the LAR is mainly to transport water as efficiently as possible
through the system to protect surrounding areas from floods. Environmental services that a
natural system would typically perform are traded for flood protection. Study frame locations in
this study such as Glendale and Willow Street are working examples of natural riverbed sections
with engineered banks providing flood protection and environmental services simultaneously.
The observed reduction in turbidity in natural sections of river in this study is consistent with the
2015 USACE feasibility report and undermines the assumption by the 2021 LAC draft
environmental impact report that natural sections of the LAR do not contribute to improved WQ.
With knowledge that sections of the LAR possessing natural riverbeds with engineered levees
have increased WQ over fully concrete sections, a more aggressive might be taken by USACE to
restore longer stretches of this river moving forward.
5.2. Limitations
One factor that would improve the practicality of this study would be to increase the
temporal resolution of imagery acquisition. In the workflow of the current study the temporal
resolution of image capture is once every two years, which represents a key limitation. In a
methodology where images are captured at a higher temporal frequency WQ fluctuations could
be more closely studied and better understood.
Another data limitation to consider in this study is related to bottom reflectance in the
LAR. There was considerable effort exerted in searching for ideal areas of the LAR to sample
which contained minimal bottom reflectance. Some areas of the LAR, particularly concrete
sections containing shallow water which may have spectral characteristics influenced by the
bottom reflectance of the concrete riverbed. In order to truly validate the quality of imagery data,
in-situ sampling of WQ should be conducted simultaneously with image capture.
47
Moreover, the present RS methodology provides a qualitative measure of the WQ
parameters of turbidity and chlorophyll-a in the LAR. The results derived from this project are
valuable in the sense that they detect a reduction in turbidity in natural sections of river while
effectively detecting differences in chlorophyll-a in concrete and natural portions of the LAR.
While the results provided by this study constitute relative intensities of turbidity and
chlorophyll-a in the LAR they are not interchangeable with field collected samples of these WQ
parameters. Future work is necessary in order to reliably correlate RS index values with field
collected WQ samples.
5.3. Future Work
In reviewing the results and limitations of the study there are some key components in the
methodology that can be modified or augmented to increase reliability and practicality. Looking
forward and expanding upon the findings in this study, increasing the temporal resolution of
imagery and collecting in-situ WQ samples for validation are of high value. In order to develop a
dependable template that relates RS index values to physical parameters, in-situ WQ sampling
can be conducted concurrently with UAV flight missions. UAVs with adequate spatial and
spectral resolution to collect imagery are available for purchase by the public. Utilization of
UAVs increases the temporal resolution of imagery capture manyfold over the biennially
captured NAIP aerial imagery used in this project. With greatly increased temporal resolution,
WQ patterns can be monitored on a daily or weekly basis.
In-situ field collected WQ results can be compared with the RS index results derived
from simultaneously collected UAV imagery to determine correlation strength. Once correlation
is firmly established between field collected WQ data and RS index results, WQ in-situ sampling
can be relegated to occasional data quality checks of RS data captured by UAVs. This proposed
48
future work is essential in enabling remote monitoring of WQ that can be compared against local
water quality standards for compliance. WQ parameters can be determined at daily intervals and
used to alert local agencies WQ in the LAR in near real time. This work could provide great
benefit to the residents of Los Angeles and the communities that reside in the Los Angeles River
Basin.
49
References
Ackerman, D. and Schiff, K. 2003. “Modeling storm water mass emissions to the Southern
California Bight.” Journal of Environmental Engineering, Vol. 129 No. 4. 308-317.
Ackerman, D., Schiff, K., Trim, H., and Mullin, M. 2003. "Characterization of water quality in
the Los Angeles River." Bulletin of the Southern California Academy of Sciences 102,
No. 1
Alföldi, T.T. 1982. “Remote sensing for water quality monitoring”. In Remote Sensing for
Resource Management, Johannsen, C.J. & Sanders, J.L., (Eds), Soil Conservation Society
of America, 317-328.
Avdan, Z.Y., Kaplan, G., Goncu, S., and Avdan, U. 2019. "Monitoring the water quality of small
water bodies using high-resolution remote sensing data." ISPRS International Journal of
Geo-Information 8, no. 12: 553.
Austin, Å. N., Hansen, J. P., Donadi, S., and Eklöf, J. S. 2017. "Relationships between aquatic
vegetation and water turbidity: A field survey across seasons and spatial scales." PLOS
one 12, no. 8: e0181419.
Bid, S., Siddique, G. 2019. Identification of seasonal variation of water turbidity using NDTI
method in Panchet Hill Dam, India. Model. Earth Syst. Environ. 5, 1179–1200.
https://doi.org/10.1007/s40808-019-00609-8
Boroon, M. and Coo, C. 2015. “Water quality assessment of the Los Angeles river watershed,
California, USA in wet and dry weather periods.” Journal of Geography, Environment
and Earth Science International 4(3) 1–17.
Campbell, J.B. 2011. Introduction to Remote Sensing. 5th edition. New York: Guilford Press.
Weston Solutions, Inc. 2005. California Department of Water Quality Control Board, Final
Report.
Carpenter, S., Caraco, N.F., Correll, D.L., Howarth, R., Sharpley, A.N., and Smith, V.H. 1998.
“Non-Point Pollution of Surface Waters with Phosphorus and Nitrogen.” Ecological
Applications. Vol. 8, No. 3 559-68
County of Los Angeles. 2022. Accessed April 4, 2022.
https://urldefense.com/v3/__https://geohub.lacity.org/datasets/lacounty::streams-and-
rivers/explore?location=33.824213*2C-
118.226184*2C8.99__;JSU!!LIr3w8kk_Xxm!9pWOVH3D4AxGhL1K-
eD4S97LlBUSxu0ZFs9IhP6cbnwrqDol1Bs2eV6fzdofapQ$
Chen, X., Chen, W., Bai, Y., and Wen, X. 2022. "Changes in turbidity and human activities
along Haihe River Basin during lockdown of COVID-19 using satellite data."
Environmental Science and Pollution Research 29, no. 3: 3702-3717.
50
Chislock, M. F., Doster, E., Zitomer, R. A. & Wilson, A. E. 2013. “Eutrophication: Causes,
Consequences, and Controls in Aquatic Ecosystems.” Nature Education Knowledge
4(4):10
Elhag, M., Gitas, I., Othman, A., Bahrawi, J., and Gikas P. 2019. "Assessment of Water Quality
Parameters Using Temporal Remote Sensing Spectral Reflectance in Arid Environments,
Saudi Arabia" Water 11, no. 3: 556. https://doi.org/10.3390/w11030556
Gumprecht, B. 1999. The Los Angeles River: Its Life, Death and Possible Rebirth. Baltimore:
The Johns Hopkins University Press.
Congalton, R. G. and Green. K. 1999. Assessing the accuracy of remotely sensed data: principles
and practices. Boca Raton, FL: Lewis Publishers.
Hedley, J. D., Harborne, A. R., and Mumby, P. J. 2005. "Simple and robust removal of sun glint
for mapping shallow ‐water benthos." International Journal of Remote Sensing 26, no.
10: 2107-2112.
ICF International Inc. 2021. “2020 LA River Master Plan Draft Program Environmental Impact
Report.” Los Angeles County Public Works.
Kaplan, G., Avdan, Z. Y., Goncu, S., and Avdan, U. 2020. "Evaluation of RapidEye-3 Satellite
Data for Assessing Water Turbidity of Lake Borabey" Proceedings 48, no. 1: 14.
https://doi.org/10.3390/ECWS-4-06424
Katlane, R., Dupouy, C., El Kilani, B., Berges, J. C. 2020. “Estimation of chlorophyll and
turbidity using sentinel 2A and EO1 data in Kneiss Archipelago Gulf of Gabes, Tunisia.”
International Journal of Geosciences, Scientific Research 11, p. 708-728.
Kim, E., Sook-Hyun, N., Koo, J., and Hwang, T. 2021. "Hybrid Approach of Unmanned Aerial
Vehicle and Unmanned Surface Vehicle for Assessment of Chlorophyll-a Imagery Using
Spectral Indices in Stream, South Korea." Water 13, no. 14: 1930.
Kuhn, C., Valerio, A., Ward, N., Loken, L., Sawakuchi, H. O., Kampel, M., Richey, J., et al.
2019. "Performance of Landsat-8 and Sentinel-2 surface reflectance products for river
remote sensing retrievals of chlorophyll-a and turbidity." Remote Sensing of
Environment 224.104-118.
Lacaux, J. P., Tourre, Y. M., Vignolles, C., Ndione, J. A., and Lafaye, M. 2007. "Classification
of ponds from high-spatial resolution remote sensing: Application to Rift Valley Fever
epidemics in Senegal." Remote Sensing of Environment 106, no. 1: 66-74.
McKim, H. L., Merry, C. J., and Layman, R.W. 1984. "Water quality monitoring using an
airborne spectroradiometer." Photogrammetric Engineering and Remote Sensing 50, no.
3: 353-360.
Montello, D. and Sutton, P. 2012. An Introduction io Scientific Research Methods in Geography
and Environmental Studies. Vol. 1. Sage.
51
Pix4D. 2022. “Learn more about Sequoia+.” Accessed September 11, 2022.
https://www.pix4d.com/product/sequoia/faq.
Rodríguez-López, L., Duran-Llacer, I., González-Rodríguez, L., Abarca-del-Rio, R., Cárdenas,
R., Parra, O., Martínez-Retureta, R., and Urrutia, R. 2020. "Spectral analysis using
LANDSAT images to monitor the chlorophyll-a concentration in Lake Laja in Chile."
Ecological Informatics 60: 101183.
Ross, A. and Wilson, V. 2017. Basic and Advanced Statistical Tests Writing Results Sections and
Creating Tables and Figures. Sense Publishers.
Saberioon, M., Brom, J., Nedbal, V., Souc ̆ ek, P., and Císa r ̆ , P. 2020. "Chlorophyll-a and total
suspended solids retrieval and mapping using Sentinel-2A and machine learning for
inland waters." Ecological Indicators 113: 106236.
Stein, E. D., and Ackerman, D. 2007. “Dry weather water quality loadings in arid, urban
watersheds of the Los Angeles Basin, California, USA.” JAWRA Journal of the American
Water Resources Association. Vol. 43 No. 2. 398– 413.
Stein, E.D. and Yoon, V.K. 2007. “Dry weather flow contribution of metals, nutrients, and solids
from natural catchments.” Water, Air, and Soil Pollution. Vol. 190 No1.183-195.
Tolk, B. L., Han, L. and Rundquist. D. C. 2000. "The impact of bottom brightness on spectral
reflectance of suspended sediments." International Journal of Remote Sensing 21, no. 11:
2259-2268
U.S. Army Corps of Engineers Los Angeles District. 2011. Sepulveda Dam Basin Master Plan
and Draft Environmental Assessment.
U.S. Army Corps of Engineers Los Angeles District. 2015. Los Angeles River Ecosystem
Restoration Integrated Feasibility Report. FINAL Feasibility Report and Environmental
Impact Statement/Environmental Impact Report. Volume 1: Integrated Feasibility Report.
U.S. Geologic Survey. 2021. Accessed November 7, 2021. https://earthexplorer.usgs.gov/
Wasserstein R. L. and Lazar N. A. 2016. “The ASA’s statement on p-values: Context, process,
and purpose.” The American Statistician. 70. No. 2: 129-133.
Yu, Y., Chen, S., Qin, W., Lu, T., Li, J. and Cao, Y. 2020. “A Semi-Empirical Chlorophyll-a
Retrieval Algorithm Considering the Effects of Sun Glint, Bottom Reflectance, and Non-
Algal Particles in the Optically Shallow Water Zones of Sanya Bay Using SPOT6 Data."
Remote Sensing 12, no. 17: 2765.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Finding the green in greenspace: an examination of geospatial measures of greenspace for use in exposure studies
PDF
Classifying the alpines: developing a methodology to track environmental changes in the alpines utilizing remote sensing and ecozone vegetation patterns
PDF
Assessing modern conflict to monitor human rights with remote sensing: Russia's war in Ukraine
PDF
Exploring the relationship between urban forests and ethnicity in West Covina using GIS and object-based image analysis
PDF
Utilizing GIS and remote sensing to determine sheep grazing patterns for best practices in land management protocols
PDF
Classifying ash cloud attributes of Eyjafjallajökull volcano, Iceland, using satellite remote sensing
PDF
A scenario based fire susceptibility approach for remote sensing platform comparison: Los Angeles county area, southern California, USA
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
A comparison of pixel-based, object-based, deep learning, and data fusion image classification approaches in delineating urban tree canopy in City Terrace
PDF
Impacts of vegetation management on wildfire severity: a study of the 2021 Caldor fire
PDF
Comparing Landsat7 ETM+ and NAIP imagery for precision agriculture application in small scale farming: a case study in the south eastern part of Pittsylvania County, VA
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
Measuring seasonal variation in food access: a case study of Everett, Washington
PDF
A spatial analysis of veteran healthcare accessibility
PDF
The role of precision in spatial narratives: using a modified discourse quality index to measure the quality of deliberative spatial data
PDF
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
Utilizing online data sources to improve existing military aircraft systems
PDF
Integrating Landsat and California pesticide exposure estimation at aggregated analysis scales: accuracy assessment of rurality
PDF
Exploring remote sensing and geographic information systems technologies to understand vegetation changes in response to land management practices at Finke Gorge National Park, Australia Between ...
PDF
An analysis of the North Rainier Elk Herd area, Washington: change detection and habitat modeling with remote sensing and GIS
Asset Metadata
Creator
Daniel, Shane Reese
(author)
Core Title
Water quality in the Los Angeles River: a remote sensing based analysis
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
01/24/2023
Defense Date
01/13/2023
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
Los Angeles River,NDTI,NDVI,OAI-PMH Harvest,remote sensing,water quality
Format
theses
(aat)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Sedano , Elizabeth J. (
committee chair
), Vos, Robert O. (
committee member
), Wilson, John P. (
committee member
)
Creator Email
shanedan@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC112718705
Unique identifier
UC112718705
Identifier
etd-DanielShan-11435.pdf (filename)
Legacy Identifier
etd-DanielShan-11435
Document Type
Thesis
Format
theses (aat)
Rights
Daniel, Shane Reese
Internet Media Type
application/pdf
Type
texts
Source
20230126-usctheses-batch-1003
(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
Tags
NDTI
NDVI
remote sensing
water quality