remote sensing

Article
Assessing the Impacts of Rising Sea Level on Coastal
Morpho-Dynamics with Automated High-Frequency Shoreline
Mapping Using Multi-Sensor Optical Satellites
Naheem Adebisi 1 , Abdul-Lateef Balogun 1, * , Masoud Mahdianpari 2 and Teh Hee Min 3

1 Geospatial Analysis and Modelling (GAM) Research Group, Department of Civil and Environmental
Engineering, Universiti Teknologi PETRONAS (UTP), Seri Iskandar 32610, Malaysia;
naheem_18003002@utp.edu.my
2 C-CORE and Department of Electrical Engineering, Memorial University of Newfoundland,
St. John’s, NL A1B 3X5, Canada; masoud.mahdianpari@c-core.ca
3 Centre of Urban Resource Sustainability, Institute of Self-Sustainable Building, Universiti Teknologi
PETRONAS (UTP), Seri Iskandar 32610, Malaysia; heemin.teh@utp.edu.my
* Correspondence: alateef.babatunde@utp.edu.my

Abstract: Rising sea level is generally assumed and widely reported to be the significant driver of
coastal erosion of most low-lying sandy beaches globally. However, there is limited data-driven
evidence of this relationship due to the challenges in quantifying shoreline dynamics at the same



temporal scale as sea-level records. Using a Google Earth Engine (GEE)-enabled Python toolkit,


this study conducted shoreline dynamic analysis using high-frequency data sampling to analyze
Citation: Adebisi, N.; Balogun, A.-L.;
the impact of sea-level rise on the Malaysian coastline between 1993 and 2019. Instantaneous
Mahdianpari, M.; Min, T.H. Assessing
shorelines were extracted from a test site on Teluk Nipah Island and 21 tide gauge sites from the
the Impacts of Rising Sea Level on
combined Landsat 5–8 and Sentinel 2 images using an automated shoreline-detection method,
Coastal Morpho-Dynamics with
Automated High-Frequency
which was based on supervised image classification and sub-pixel border segmentation. The results
Shoreline Mapping Using indicated that rising sea level is contributing to shoreline erosion in the study area, but is not the only
Multi-Sensor Optical Satellites. driver of shoreline displacement. The impacts of high population density, anthropogenic activities,
Remote Sens. 2021, 13, 3587. https:// and longshore sediment transportation on shoreline displacement were observed in some of the
doi.org/10.3390/rs13183587 beaches. The conclusions of this study highlight that the synergistic use of multi-sensor remote-
sensing data improves temporal resolution of shoreline detection, removes short-term variability,
Academic Editors: Robin Teigland, and reduces uncertainties in satellite-derived shoreline analysis compared to the low-frequency
Torsten Linders, Sergio Leandro and sampling approach.
Ivan Masmitja

Keywords: optical imaging satellites; subpixel shoreline detection; GEE; sea-level rise
Received: 1 June 2021
Accepted: 26 August 2021
Published: 9 September 2021

1. Introduction
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
Coastal cities are spread around different regions of the world, with more than
published maps and institutional affil- two-thirds of the global population living within 100 km of the sea [1,2]. Coastal areas are
iations. essential regions of socio-economic and environmental benefits [3], densely populated, and
highly urbanized, and are usually highly morpho-dynamic [4]. The shoreline, an intersec-
tion of the marine environment and the mobile coastal system, is one of the most dynamic
features of coasts [3]. About 31% of the 1.63 million km of world coastlines are sandy coasts
Copyright: © 2021 by the authors.
that are highly susceptible to rapid advances/retreats and long-term accretion/erosion
Licensee MDPI, Basel, Switzerland.
due to changing environmental conditions and coastal processes [2]. Techniques that
This article is an open access article
couple physical landscape processes associated with climate change with human behavior
distributed under the terms and have been employed in many coastal susceptible and vulnerability assessment studies [5].
conditions of the Creative Commons Anfuso et al. [6] presented a comprehensive overview of these techniques, outlining data
Attribution (CC BY) license (https:// requirements and vulnerability indices used in adaptation planning. Generally, assessing
creativecommons.org/licenses/by/ the causes of coastal vulnerability, especially shoreline erosion and coastal flooding, is a
4.0/). difficult task due to the complexity of the processes acting at varying spatial and temporal

Remote Sens. 2021, 13, 3587. https://doi.org/10.3390/rs13183587 https://www.mdpi.com/journal/remotesensing

Remote Sens. 2021, 13, 3587 2 of 22

scales [6,7]. The processes could be anthropogenic activities, such as mineral extraction
and sediment removal; or natural forces, such as tectonics, wind, and waves. It could also
be due to rising sea level, which is a consequence of both anthropogenic and natural forces.
As such, assessing the coastal response to rising sea levels in particular is very challenging.
Readers are referred to [8,9] for various factors influencing shoreline dynamics.

1.1. Sea Level Rise (SLR) and Shoreline Dynamics

Globally, there has been an unprecedented rise in sea level [10], due mainly to human-
induced climate change, which is expanding the world oceans and melting the ice sheets
and glaciers [11,12]. Since the beginning of the 20th century, the global mean sea level
(GMSL) has risen by about 16–21 cm, with more than 7 cm of this occurring since 1993 [13].
Sea level has been projected to rise by 0.6–1.5 m by 2100 [14], or even reach 2 m under
extreme situations [15]. There will also be significant variation in regional sea-level changes
due to differences in gravitational processes, ground subsidence, current dynamics, salinity,
and heating effects [16,17]. Local sea-level changes will depend on the sum of eustatic,
steric, isostatic, and tectonic forces [8,18]; for example, relative sea-level falls in areas such
as the northern Baltic and Hudson Bay, where land is uplifting and rises more rapidly than
usual climate-induced trends on subsiding coasts [18,19].
The impact of current and future sea-level rise would be devastating, particularly
to coastal environments [10]. Rising frequency and intensity of flooding, coastal erosion,
and changes in coastal aquifer quality and groundwater are some of the physical impacts
of SLR on coastal communities [16,20–23]. These negative impacts of a rising sea level
on the coastal ecosystem present a global environmental challenge [13,24]. With a large
population inhabiting coastal areas and the present rate of SLR at 3.2 mm/y [13], analyzing
the impacts of sea level on coastal zones has become an important practice among coastal
scientists and planners.
As some studies have observed, estimating the contribution of sea-level rise to shore-
line change with great certainty is challenging due to two major problems [7,25]. On the
one hand, shoreline data is usually lacking or inadequately temporally resolved to establish
a correlation with sea levels, which are monitored at high sampling frequency—on hourly,
daily, and/or monthly basis [7]. On the other hand, records of local relative sea-level change
are only available at limited locations where tide gauges are installed and functional [7,26].
Consequently, data-driven studies of relationships between sea level and shoreline change
are limited [27]. The majority of the existing studies are either logic-based or based on
a simple examination of estimates of sea-level rise and shoreline changes. However, the
situation evolves with the growing availability in sea-level and shoreline-change data [7].
There is thus a critical need to develop an improved capability for assessing the impact of
sea-level rise on shoreline dynamics [25].

1.2. Shoreline Detection

Shoreline positions can be monitored with in-situ or site surveys using instruments
such as a Differential Global Positioning System (DGPS), Lidar, and total station. In situ
approaches provide very accurate shoreline position, but are costly, labor-intensive, and
have limited spatio-temporal coverages [28,29]. Shoreline positions can also be deter-
mined through remote-sensing techniques. Traditionally, coastlines are extracted using
photogrammetry methods such as video-imaging techniques, airborne, and UAVs [30].
Shore-based video systems are composed of optical equipment mounted on an elevated po-
sition to monitor the nearshore area [31]. Video-imaging techniques provide high-frequency,
top-quality, and continuous images of coasts [32]. They also offer a low-cost alternative
to the established airborne methods [30]. However, video-imaging techniques also have
limited spatial coverage since they are fixed, site-specific ground-based technologies [30].
While airborne methods and UAVs are more flexible in their potential coverage, they
are also limited due to dependence on human operators, which discourages continuous
observation [33], whereas understanding coastal dynamics requires continuous and fre-
Remote Sens. 2021, 13, 3587 3 of 22

quent monitoring of shorelines. Consequently, long-term, uninterrupted, and updated

information of shoreline displacement that is vital for coastal engineering and sustainable
adaptation plans is usually scarce in many coastal areas [34].
Crucially, the launch of satellites with sensors for the optical and microwave parts of
the electromagnetic spectrum provides an alternative approach to photogrammetry. Satel-
lites offer contemporaneous, repetitive, and quasi-global coverage to facilitate shoreline
monitoring and change analysis [30]. With the contemporary observation by multiple
optical satellites, it is possible to obtain hundreds of satellite-derived shorelines (SDS)
throughout a year over a coast with a global average revisit of about three days [35].
However, the technique for shoreline detection, data management, and analysis must be
addressed in shoreline monitoring using remote-sensing data.

1.3. Satellite Derived Shorelines

Different techniques have been employed for extracting instantaneous shoreline posi-
tions from satellite images. The manual method involves visual inspection and digitizing
of shorelines from imagery [36]. The semi-automatic method includes single band, band
ratio [36], image classification [37,38], and principal component [39] analyses. The Normal-
ized Difference Water Index (NDWI) [40], Modified Normalized Difference Water Index
(MNDWI) [41], and Automated Water Extraction Index (AWEI) [42] are the commonly
used indices for shoreline delineation.
With the advancements in machine learning, supervised and unsupervised classifica-
tion techniques have been also used for shoreline detection. Iterative Self-Organizing Data
Analysis (ISODATA) is a common unsupervised classification algorithm for extracting
shoreline features from satellite imagery [43]. In particular, ISODATA classification is an
improved version of K-means [28]. For example, García-Rubio, et al. [44] assessed the
various land–sea classification techniques employed in shoreline studies and confirmed
the great performance of ISODATA shoreline classification. Among supervised classifica-
tion methods, support vector machine is a well-known method for land–sea classification.
Elnabwy, et al. [45] classified Landsat images using SVM to extract instantaneous shorelines
along the coasts of Nile Delta. Studies have also employed neural network classification to
discriminate sand and water features [34]. Feed forward networks such as the multi-layer
perceptron (MLP), radial basis function (RBF), probabilistic neural network (PNN), and
generalized regression neural network (GRNN) are the commonly used ANN models in
remote-sensing applications [46]. This is because neural networks have a strong capability
to handle complex phenomena and can improve land–sea segmentation accuracy [47,48].
However, developing a more robust technique for precise and accurate extraction of
shorelines from the available satellite data remains a challenging task [38,49]. Moreover,
shoreline detection is a complex phenomenon that demands the use of several techniques,
rather than a single image-processing technique [28]. As for data management, the file size
of Landsat 8 images can be as high as 1 GB per scene, meaning that high-frequency shore-
line sampling for long-term studies requires high data storage capabilities and processing
power that is difficult to achieve with a conventional Geographic Information System (GIS)
approach. Based on the foregoing, it is evident that to understand shoreline dynamics
within the context of sea-level rise, there is a need for an innovative approach to manage
the large volume of data, a robust detection technique for accurate shoreline delineation,
and appropriate analysis techniques to accurately quantify the spatiotemporal evolution of
a beach.

1.4. Malaysia: Sea-Level Rise and Coastal Vulnerability

A rising sea level has been reported along the Malaysian coastline [50,51], and a
corresponding increase in coastal hazards such as flooding and erosion have been ob-
served in many coastal areas [52]. The coasts of Selangor and Batu Pahat have experienced
severe coastal erosion, leading to the loss of 18.785 km2 and 4.155 km2 of land, respec-
tively [53]. Similarly, coastal flooding in Johor has caused an economic loss of an estimated
Remote Sens. 2021, 13, 3587 4 of 22

RM 2.4 billion and the destruction of RM 0.35 billion worth of amenities [53]. Existing
studies of sea levels are fragmental in terms of space and time coverage. In addition, the
lack of accurate local sea-level information, in addition to high uncertainties in shoreline
monitoring, makes it difficult to quantify the impact of sea-level rise on shoreline dynamics.
A nationwide study of sea-level change and shoreline dynamic is lacking. To address
this gap, this study leveraged advances in geospatial technology to conduct a holistic
evaluation of long-term shoreline dynamics and sea-level change along the Malaysian
coastline. For an effective management of the shoreline data, Google Earth Engine (GEE)
was used to acquire and process multi-mission satellite images of Landsat series 4–8
and Sentinel 2. In particular, GEE is a free-to-use, cloud-based geospatial platform for
large-scale analysis of big EO data [54]. It provides a petabyte of satellite data, high-level
API, and machine-learning algorithms for processing and analyzing big earth observation
data, thereby overcoming the limitations of the conventional digital image-processing
workflows [55]. For shoreline extraction, a robust shoreline detection method based on a
supervised classification and sub-pixel resolution segmentation technique developed by
Vos, Harley, Splinter, Simmons and Turner [34] was employed to improve the detection
accuracy of the sand and land interface.
The rest of this paper is arranged as follows: descriptions of the study area and data
used are provided in Section 2; the methodology employed to estimate sea level and
shoreline displacement is provided in Section 3; and Section 4 presents the results of the
study. A discussion of the results is then presented in Section 5, followed by the conclusions
of the study in Section 6.

2. Study Area and Data Used

This work was performed around the Malaysian coastline, which measures about
4675 km, making it the 29th-longest coastline in the world. The country has a large popula-
tion inhabiting lowlands in coastal areas; more than 13% of its land area is within 5 km
of the coast, and thereby vulnerable to the harmful impacts of the sea-level rise [56]. The
aesthetic natural environment of most islands in Malaysia draws millions of tourists to the
country every year. Therefore, the development of tourist attractions in the natural-settings
area results in a large impact on the coastal zone.
The data employed to accomplish the aim of this study were acquired from different
sources. These data had different formats and dimensions, as presented in Table 1 below.
The table also provides information on the source of the data and their specific use in
this study. To quantify sea-level change over the study area, monthly averaged sea-
level data were acquired from the Permanent Service of Mean Sea Level (PSMSL) and
Copernicus Marine Environment Service (CMEMS), respectively. PSMSL provides high-
frequency sea-level data measured by tide gauges, while CMEMS provides multi-mission
altimeter satellite gridded sea-level anomalies (SLAs). A total of 21 tide-gauge stations
cover the Malaysian coastline—12 of these stations are along the coast of Peninsular
Malaysia (West Malaysia), and 9 tidal stations are along the coast of Sabah and Sarawak
(East Malaysia) (Figure 1).
Optical satellites were employed to evaluate shoreline dynamics at regions of extreme
SLR. These datasets, including Landsat (4,5,7,8) and Sentinel-2 imagery, were acquired from
GEE. Landsat imagery of the free and open satellite data were organized in two tiers. Tier 1
images had well-characterized geometry and valid geometric accuracy, and were thus suit-
able for time series analysis, while Tier 2 images were not too fit for this purpose [57]. Land-
sat Tier 1 images have three products, namely Raw scenes, Top-of-Atmosphere (TOA), and
Surface Reflectance. Raw scenes contain just the digital number (DN) as measured by the
sensor; the DN values from the sensor are converted to reflectance for Top-of-Atmosphere
(TOA) reflectance images [58]; while the Surface Reflectance image is atmospherically
corrected. TOA provides a standardized comparison between images acquired on different
dates. The Level-1C product of Sentinel-2 has similar standardized TOA reflectance im-
ages that are appropriate for time-series analysis. The Shuttle Radar Topography Mission
 Remote Sens. 2021, 13, 3587 5 of 22
Remote Sens. 2021, 13, x FOR PEER REVIEW 5 of 22

(SRTM) digital elevation dataset and the 2014 Finite Element Solution (FES2014) ocean tide
Table 1. model, which
Formats, are required
sources, for tidal
and purposes correction
of the of instantaneous
data employed shorelines, were acquired
in this study.
from GEE and AVISO, respectively. The next section describes the methodology employed
Category Dataand analyze the data.
to process Format Source Purpose
Tide gauge csv PSMSL For estimating sea-
Sea Level Table 1. Formats, sources, and purposes of the data employed in this study.
Satellite altimetry Netcdf CMEMS level trend
Category Data
Optical satellites: Landsat, Sen-Format Source ForPurpose
assessing shoreline
Shoreline Tide gauge csv GeoTIFF PSMSL GEE
Sea Level tinel dynamics
For estimating sea-level trend
Satellite altimetry Netcdf CMEMS
For estimating the
Optical satellites: For assessing
Elevation
Shoreline SRTMLandsat,
digitalSentinel
elevation modelGeoTIFF
GeoTIFF GEE GEE beachdynamics
shoreline slope required
SRTM digital for tidal
For estimating the beachcorrection
slope
Elevation GeoTIFF GEE
elevation model Archiving, Validation required for tidal correction
Archiving, For estimating tide
andValidation and
Interpretation of
Oceanic TideTide
Oceanic Model
Model FES2014
FES2014 Netcdf Netcdf
Interpretation of Satellite For level required
estimating tide levelfor tidal
Satellite Oceanographic
Oceanographic required for tidal correction
data (AVISO) correction
data (AVISO)

Figure 1. Map
Figure ofofstudy
1. Map studyarea
areashowing thetide
showing the tidegauge
gauge stations.
stations.

Optical satellites were employed to evaluate shoreline dynamics at regions of ex-

treme SLR. These datasets, including Landsat (4,5,7,8) and Sentinel-2 imagery, were ac-
quired from GEE. Landsat imagery of the free and open satellite data were organized in
two tiers. Tier 1 images had well-characterized geometry and valid geometric accuracy,
and were thus suitable for time series analysis, while Tier 2 images were not too fit for this
purpose [57]. Landsat Tier 1 images have three products, namely Raw scenes, Top-of-At-
mosphere (TOA), and Surface Reflectance. Raw scenes contain just the digital number
 image is atmospherically corrected. TOA provides a standardized comparison between
images acquired on different dates. The Level-1C product of Sentinel2 has similar stand-
ardized TOA reflectance images that are appropriate for time-series analysis. The Shuttle
Radar Topography Mission (SRTM) digital elevation dataset and the 2014 Finite Element
Remote Sens. 2021, 13, 3587
Solution (FES2014) ocean tide model, which are required for tidal correction of instanta-
6 of 22
neous shorelines, were acquired from GEE and AVISO, respectively. The next section de-
scribes the methodology employed to process and analyze the data.

3.
3. Methodology
Methodology
The approach
The approach to accomplish
accomplish the goals
goals of
of this
this study,
study, including estimation of sea-level
trend, extraction
trend, extraction of
ofshorelines,
shorelines,and
andthe
theanalysis of of
analysis displacement pattern
displacement is presented
pattern in this
is presented in
section.
this Figure
section. 2 below
Figure summarizes
2 below the study’s
summarizes methodology.
the study’s methodology.

Figure
Figure 2.
2. Workflow
Workflow of
of the
the study’s
study’s methodology.
methodology.

3.1. Estimation
3.1. Estimation ofof Sea-Level
Sea-Level Trend
Trend
The monthly mean sea-level
The monthly mean sea-level record of the
record of 21thetide
21 gauge stations
tide gauge acquired
stations from PSMSL
acquired from
was preprocessed,
PSMSL and the and
was preprocessed, meanthe seamean
levelsea
waslevel
subtracted to deriveto
was subtracted thederive
relative
thesea-level
relative
anomaly.anomaly.
sea-level Sea-level Sea-level
anomaly anomaly
values were alsowere
values extracted from gridded
also extracted fromaltimetry satellites
gridded altimetry
at the locations
satellites of the tide-gauge
at the locations stations
of the tide-gauge to estimate
stations the corresponding
to estimate the corresponding absolute sea-
absolute
level trend.
sea-level TheThe
trend. trend in absolute
trend andand
in absolute relative sea levels
relative was was
sea levels determined
determined for all
forstations.
all sta-
The trend
tions. was estimated
The trend usingusing
was estimated robust fit regression
robust techniques,
fit regression since
techniques, robust
since fit analysis
robust fit anal-
ysis optimizes solution determination and outlier detection [59]. This approach was
optimizes solution determination and outlier detection [59]. This approach was used
used
to fit
to fit aa linear
linear trend
trend atat each
each station with an
station with an iteratively re-weighted least
iteratively re-weighted least squares
squares (IRLS)
(IRLS)
technique, wherein
technique, whereinweights
weightsofofobservations
observations werewereadjusted
adjustedaccording to the
according todeviations
the deviationsfrom
the trendline to re-fit the trendline repeatedly until the solution converged
from the trendline to re-fit the trendline repeatedly until the solution converged [60]. [60].

3.2. Shoreline Extraction

3.2. Shoreline Extraction
Mapping of the coastline at sites of extreme sea-level change was conducted using
Mapping of the coastline at sites of extreme sea-level change was conducted using
CoastSat [4], a GEE-enabled Python package. The toolkit exploits the capabilities of Google
CoastSat [4], a GEE-enabled Python package. The toolkit exploits the capabilities of
Earth Engine, machine learning, and other image-processing packages, such as scikit-learn
Google Earth Engine, machine learning, and other image-processing packages, such as
and scikit-image, to facilitate access, preprocessing, detection, and automatic extraction
scikit-learn and scikit-image, to facilitate access, preprocessing, detection, and automatic
of instantaneous shoreline positions from multi-spectral optical imagery. The steps are
extraction of instantaneous shoreline positions from multi-spectral optical imagery. The
further described below; to support open reproducible science, the code written for this
process is available as a supplemental document.

3.2.1. Data Acquisition

Using CoastSat, Top-of-Atmosphere reflectance images from the Landsat 5, 7, and 8
Tier 1 collection, as well as Level-1C products of Sentinel-2 at regions of interest (ROI),
were retrieved from GEE. The four satellite missions were chosen to maximize the frequency
of sampling the coastline. To reduce the size of files to download, the toolkit cropped the
image to the ROI and downloaded only the bands necessary for shoreline detection. Details
Remote Sens. 2021, 13, 3587 7 of 22

of satellite data retrieved by CoastSat, such as pixel size of required bands, revisit period of
mission, and the GEE collections for the image storage, are provided in Table 2 below.

Table 2. Description of satellite datasets on GEE that were accessible with CoastSat.

Periodicity
Satellite Mission Pixel Size of Required Bands GEE Collections
(Days)
30 m R, G, B, NIR,
Landsat 5 (TM) 16 LANDSAT/LT05/C01/T1_TOA
SWIR1 bands
30 m R, G, B, NIR, SWIR1
Landsat 7 (ETM+) bands + 15 m 16 LANDSAT/LE07/C01/T1_RT_TOA
panchromatic band
30 m R, G, B, NIR, SWIR1
Landsat 8 (OLI) bands + 15 m 16 LANDSAT/LC08/C01/T1_RT_TOA
panchromatic band
10 m R, G, B, NIR +
Sentinel-2 5 COPERNICUS/S2
20 m SWIR1

3.2.2. Data Preprocessing

The acquired images were orthorectified by GEE. Sharpening of Landsat 7 and
Landsat 8 images with higher resolution panchromatic bands was performed by Coast-
Sat to enhance the spatial resolution of the images from 30 m to 15 m, with the aim of
achieving an optimal shoreline detection. Landsat 5 scenes were down-sampled from 30 m
to 15 m, as they do not include a panchromatic band. Sentinel-2 images have a higher
spatial resolution: 10 m visible and NIR bands, and 20 m SWIR1. The SWIR band was also
down-sampled to 10 m using bilinear interpolation to maintain equal resolution among
the S2 bands. For cloud-masking, we chose a threshold of 0.2 to discard all images with
more than 20% cloud cover in our ROI.

3.2.3. Shoreline Detection

The instantaneous boundary between water and sand captured at the time of im-
age acquisition was taken as the indicator of the shoreline here. The CoastSat toolkit
applied a robust shoreline detection algorithm to the pre-processed satellite images [4]. The
technique involves two major steps—image classification and sub-pixel segmentation. A
multi-layer perceptron, a neural-network-supervised classifier algorithm, was pretrained,
with 99% accuracy based on cross-validation, to categorize the pixels in the image into
four classes—sand, water, white water, and other features (containing features such as
vegetation, rock, buildings, and coastal-defense structures). A sample classification of a
Landsat 5 image for Kukup Beach is shown in Figure 3 below. The sand/water boundary
for each classified image was then delineated using the Modified Normalized Difference
Water Index (MNDWI).

3.2.4. Tidal Correction

The satellite images were acquired at different stages of tides. So, a linear tidal
correction [61] was applied to the extracted shorelines using Equation (1) below to remove
the displacement due to tidal fluctuation.

Zre f − Zlocal
∆x = (1)
tanβ

where ∆x is the cross-shore horizontal shift, Zre f is the reference tidal datum, Zlocal
is the measured (or modelled) tide level at the time of image acquisition, and tanβ is a
characteristic beach face slope of the site. The slope of the beach at each site was derived
from the SRTM elevation dataset, and the tide level was acquired from the FES2014 ocean
tide model [62].
Remote
Remote Sens. 2021, 13,
Sens. 2021, 13, 3587
x FOR PEER REVIEW 88 of
of 22
22

Figure 3. A sample
sample processed
processedLandsat
Landsat55image
imageofofKukup
KukupBeach:
Beach:(a)(a)ananRGB
RGB composite;
composite; (b)(b) output
output of of
thethe image
image classifica-
classification
tion with
with labelslabels forfour
for the theclasses;
four classes; (c) single-colour-band
(c) single-colour-band image
image of of the MNDWI
the MNDWI pixel (d)
pixel values; values; (d) thresholding
refined refined thresholding
peculiar
peculiar
to to the
the sand sand and/water
and/water interface.interface.

3.3. Shoreline
3.2.4. Analysis
Tidal Correction
Aftersatellite
The the extraction
imagesofwere the shorelines,
acquired at baseline andstages
different shore-normal
of tides. transects were
So, a linear created.
tidal cor-
Using the Digital Shoreline Analysis System (DSAS) [63], significant
rection [61] was applied to the extracted shorelines using Equation (1) below to remove shoreline change
statistics
the such as the
displacement dueshoreline change envelope (SCE), net shoreline movement (NSM), end
to tidal fluctuation.
point rate (EPR), linear regression rate (LRR), and weighted linear regression (WLR) were
calculated. The principle behind the calculation 𝑍𝑟𝑒𝑓 − was𝑍𝑙𝑜𝑐𝑎𝑙
based on the measured differences(1) in
∆𝑥 =
shoreline position over time. SCE is the range𝑡𝑎𝑛𝛽 of all the shorelines that intersect a given
transect, and NSM is the distance between the oldest and the youngest shorelines for each
where ∆𝑥 is the cross-shore horizontal shift, 𝑍𝑟𝑒𝑓 is the reference tidal datum, 𝑍𝑙𝑜𝑐𝑎𝑙 is
transect. The end point rate (EPR) was calculated by dividing the distance of shoreline
the measured (or modelled) tide level at the time of image acquisition, and 𝑡𝑎𝑛𝛽 is a char-
movement by the time elapsed between the oldest and the most recent shoreline. The
acteristic beach face slope of the site. The slope of the beach at each site was derived from
major advantages of the EPR are the ease of computation and the minimal requirement of
the
onlySRTM elevationdates.
two shoreline dataset, and the
A linear tide level
regression was acquiredstatistic
rate-of-change from the can FES2014 ocean tide
be determined by
model
fitting a[62].
least-squares regression line to all shoreline points for a transect. The regression
line is placed so that the sum of the squared residuals (determined by squaring the offset
3.3. Shoreline
distance Analysis
of each data point from the regression line and adding the squared residuals
Afteristhe
together) extraction The
minimized. of the shorelines,
linear regressionbaseline and
rate is theshore-normal
slope of the line.transects were cre-
In a weighted
ated.
linearUsing the Digital
regression, the more Shoreline
reliable Analysis System
data are given (DSAS)
greater [63], significant
emphasis or weightshoreline
towards
change statistics such as the shoreline change envelope (SCE), net shoreline
determining a best-fit line. In the computation of rate-of-change statistics for shorelines, movement
(NSM), end pointisrate
greater emphasis (EPR),
placed on linear regression
data points rate (LRR),
for which and weighted
the position uncertainty linear regression
is smaller. The
(WLR)
weight were
(w) is calculated.
defined as aThe principle
function of thebehind theincalculation
variance was based
the uncertainty on the measured
of the measurement (e):
differences in shoreline position over time. SCE is the range of all the shorelines that in-
tersect a given transect, and NSM is thewdistance 1 between the oldest and the youngest
= 2 (2)
shorelines for each transect. The end point ratee (EPR) was calculated by dividing the dis-
tance
whereofe is shoreline movement
the shoreline by the value.
uncertainty time elapsed between the oldest and the most recent
shoreline.
NSMThe andmajor
SCE are advantages of the EPR arestatistics
distance-measurement the ease of computation
calculated and the
in meters, minimal
while EPR,
requirement of only two shoreline dates.
LRR, and WLR are rates of change calculated in m/y. A linear regression rate-of-change statistic can
be determined by fitting a least-squares regression line to all shoreline
The relationship between sea-level change and shoreline dynamics at all tide-gauge points for a tran-
sect.
sites The
wasregression
determined line is placed
based so that correlation
on Pearson the sum of the squared
analysis. residuals
Pearson’s (determined
correlation by
coeffi-
squaring the offset distance of each data point from the regression line
cient evaluates the statistical relationship, or association, between two continuous variables. and adding the
squared
It is based residuals together)
on the concept of is minimized.
covariance andThe linear regression
recognized rate isapproach
as an optimal the slopeforof the line.
measur-
In a weighted linear regression, the more reliable data are given greater emphasis or
Remote Sens. 2021, 13, 3587 9 of 22

ing the association between variables of interest [64]. The correlation formula is as shown
in Equation (3) below:
∑ ( xi − χ)(yi − γ)
r= q (3)
2 2
∑ ( xi − χ ) ∑ ( yi − γ )
where r is the Pearson correlation coefficient, the value of which ranges between −1
and 1, where +1 indicates a perfect positive relationship, −1 indicates a perfect negative
relationship, and 0 denotes the absence of a relationship; xi is the yearly averaged sea-level
change values; yi is the yearly average shoreline displacement values; χ is the mean of the
yearly average sea-level change values; and γ is the mean of the yearly averaged shoreline
change values.

4. Results
Relative and absolute sea-level trends estimated for the Malaysian coastline are pre-
sented in Figures 4 and 5, respectively.
Shorelines were extracted from the multi-sensor satellites at all 21 locations with tide
gauges using the shoreline-detection method described above. The application of the
automated shoreline-detection method was first validated at a test site prior to its use
at other tide gauge sites. Teluk Nipah was selected as the test site because a previous
study [52] had been conducted at this location using the conventional approach. The
change statistics for the shorelines extracted at the test sites using the robust shoreline-
detection algorithm employed in this study are presented here. A total of 428 shorelines
were correctly extracted from the 825 images acquired over Teluk Nipah Beach. The
inability to extract shorelines from some images was due to excessive cloud cover. The
result of shoreline analysis at the test sites presented in Table 3 indicated both an erosion
trend (negative values) and an accretion trend (positive values) occurred. NSM, EPR, LRR,
and WLR statistics showed that erosion occurred at 12 of the 14 transects. The average
erosion trend was −26.05 m, −0.86 m/y, −0.41 m/y, and −0.51 m/y based on NSM, EPR,
LRR, and WLR, respectively. Accretion occurred only at the two extreme ends of the
transect: Transect IDs 1 and 14. The average accretion trend was 5.56 m, 0.19 m/y, 0.86 m/y,
and 0.96 m/y based on NSM, EPR, LRR, and WLR respectively. The map and plot of the
shoreline change rate for EPR, LRR, and WLR are shown in Figure 6. The mean shoreline
change envelope was 59.14 m. Figure 7 shows the map and plot of the shoreline change
distance for SCE and NSM. The mean of erosion/accretion values at all transects showed
that erosion was more dominant across the coastline of Teluk Nipah Beach at a rate of
−0.711 m/y, −0.23 m/y, and −0.30 m/y based on EPR, LRR, and WLR, respectively. The
result of a similar analysis conducted at the 21 tide gauge stations and the corresponding
sea level change information is summarized in Table 4 below. The relationship between
shoreline displacement and sea-level change based on the correlation of mean annual
shoreline displacement and sea-level data is also provided in this table. Sea-level data
were obtained from tide gauges and satellite altimetry to estimate the relative and absolute
sea-level trend, respectively. However, the tide-gauge data had missing values for some
timestamps, and some stations had stopped providing records. In addition, some stations
were flagged for errors in the PSMSL database. In contrast, absolute sea-level data were
not affected by these limitations. So, we used absolute sea-level change for the correlation
analysis because it was not affected by missing data and had more recent records than
those for relative sea level, which are usually affected by tide-gauge instability and vertical
land movement.
 Remote Sens.Sens.
Remote 2021,2021,
13, x13,FOR
3587PEER REVIEW 10 of 2210 of 23

Remote Sens. 2021, 13, x FOR PEER REVIEW 10 of 22

(a)

(b)
(b)
Figure 4.4.Relative
Figure
Figure Relative sea-level
4. Relative
sea-level trends
sea-level trendsatat
trends at(a)
(a) the
(a)the Peninsular
Peninsular
the Malaysia
Malaysia
Peninsular coastline
coastline
Malaysia and
and (b)
coastline (b)
Sabah
and (b)Sabah
and and
andSarawak.
Sarawak.
Sabah The
The values
Sarawak. values
valuesesti-
estimated
The esti-
mated
mated for
for thethe
for the stations
stations ininred
in red
stations redrectangles
rectangles were were
rectangles wereuncertain
uncertain due to due
uncertain the
due totothe
theunreliability
unreliability ofofthe
thetide-gauge
of the tide-gauge
unreliability tide-gaugerecord.
record. record.
 (b)
Remote Sens.
Figure 2021, 13, 3587
4. Relative 11 of
sea-level trends at (a) the Peninsular Malaysia coastline and (b) Sabah and Sarawak. The values 22
esti-
mated for the stations in red rectangles were uncertain due to the unreliability of the tide-gauge record.

Remote Sens. 2021, 13, x FOR PEER REVIEW 11 of 22

(a)

(b)
Figure 5. Absolute
Figure sea-level
5. Absolute trends
sea-level trendsatat(a)
(a)the
thePeninsular Malaysiacoastline
Peninsular Malaysia coastline and
and (b) (b) Sabah
Sabah and and Sarawak.
Sarawak.

Shorelines were extracted from the multi-sensor satellites at all 21 locations with tide
gauges using the shoreline-detection method described above. The application of the au-
tomated shoreline-detection method was first validated at a test site prior to its use at
other tide gauge sites. Teluk Nipah was selected as the test site because a previous study
[52] had been conducted at this location using the conventional approach. The change
 PSMSL database. In contrast, absolute sea-level data were not affected by these limita-
tions. So, we used absolute sea-level change for the correlation analysis because it was not
Remote Sens. 2021, 13, 3587 affected by missing data and had more recent records than those for relative sea12level,of 22
which are usually affected by tide-gauge instability and vertical land movement.

(a)

(b)
Figure 6.
Figure (a) Map
6. (a) Map of
of shoreline
shoreline change
change rate
rate (EPR).
(EPR). (b)
(b) Plot
Plot of
of shoreline
shoreline change
change rate
rate for
for EPR,
EPR, LRR,
LRR, and
and WLR.
WLR.
Remote Sens. 2021, 13,
Sens. 2021, 13, 3587
x FOR PEER REVIEW 13 of 22

Figure 7. (a)
Figure 7. (a) Map
Map of
of shoreline
shoreline change (SCE). (b)
change (SCE). (b) Plot
Plot of
of shoreline
shoreline distance
distance change
change for
for SCE
SCE and
and NSM.
NSM.
Remote Sens. 2021, 13, 3587 14 of 22

Table 3. Shoreline change statistics at Teluk Nipah Beach indicating accretion/erosion at each transect.

Transect ID TCD (m) Length (m) SCE (m) NSM (m) EPR (m/y) LRR (m/y) WLR (m/y)
1 100 139.69 72.36 4.24 0.14 1.58 1.66
2 200 86.88 79.38 −6.2 −0.2 1.19 1.3
3 300 97.37 61.87 −1.92 −0.06 0.96 1.09
4 400 111.26 35.49 −25.07 −0.83 −0.13 −0.13
5 500 174.30 91.89 −43.36 −1.43 −1.37 −1.36
6 600 121.59 89.55 −43.39 −1.44 0.48 1.02
7 700 55.76 50.23 −22.95 −0.76 0.09 0.22
8 800 40.66 39.65 −26.95 −0.89 −0.15 −0.12
9 900 58.57 50.37 −25.32 −0.84 −0.31 −0.28
10 1000 84.49 46.37 −32.36 −1.07 −0.02 0.03
11 1100 118.66 39.21 −14.15 −0.47 0.22 0.27
12 1200 172.65 40.05 −21.61 −0.71 −0.02 0.06
13 1300 263.21 77.31 −49.29 −1.63 0.05 0.18
14 1400 308.90 54.18 6.88 0.23 −0.14 −0.26
Mean 59.14 −21.53 −0.71 −0.23 −0.30

Table 4. Summary of trends in relative sea level, absolute sea level, and shoreline displacement at all tide-gauge sites in Malaysia.

Correlation of Mean
Relative Sea Absolute Shoreline
Annual Shoreline
Stations Level Sea Level SCE (m) NSM (m) EPR (m/y) LRR (m/y) Change
and Absolute
(mm/y) (mm/y) Pattern
Sea-Level Change
Pulau
3.98 ± 2.28 2.78 ± 1.30 45.11 −12.28 −0.21 −0.41 Erosion 0.23
Langkawi
Pulau
4.44 ± 2.48 3.47 ± 1.36 34.37 −2.26 −0.42 −0.24 Erosion 0.43
Pinang
Lumut 3.76 ± 2.04 3.89 ± 1.42 63.53 5.37 0.03 0.1 Accretion 0.13
Port Kelang 3.85 ± 2.47 3.81 ± 1.53 60.75 −8.17 −0.75 −0.2 Erosion 0.32
Tanjung
3.08 ± 2.07 3.91 ± 1.33 41.78 −4.49 −0.18 −0.23 Erosion 0.12
Keling
Kukup 5.91 ± 1.85 2.63 ± 1.22 52.05 2.30 0.94 0.09 Accretion 0.05
Johor Bahru 4.52 ± 1.99 2.95 ± 1.15 29.44 −20.64 −0.15 −0.24 Erosion 0.36
Getting 3.22 ± 1.14 3.58 ± 0.92 19.6 0.06 −0.29 −0.33 Erosion 0.26
Cendering 3.63 ± 1.55 3.86 ± 0.89 56.35 2.06 0.17 0.31 Accretion 0.12
Tanjung
3.96 ± 1.34 3.76 ± 0.86 25.54 −17.81 0.03 −0.36 Erosion 0.16
Gelang
Pulau
3.36 ± 1.60 3.72 ± 0.85 53.88 −19.08 −0.17 −0.44 Erosion 0.06
Tioman
Tanjung
2.47 ± 1.62 3.38 ± 0.97 31.9 −3.75 −0.67 −0.32 Erosion 0.54
Sedili
Sandakan 3.81 ± 2.46 4.03 ± 1.75 50.91 0.61 0.05 0.2 Accretion 0.08
Bintulu 2.86 ± 1.76 3.58 ± 1.23 26.2 −12.37 −0.57 −0.19 Erosion 0.34
Kota
4.31 ± 2.00 4.04 ± 1.63 45.83 −7.75 −0.12 −0.37 Erosion 0.23
Kinabalu
Lahad Datu 2.97 ± 3.03 4.46 ± 2.13 53.59 15.20 0.38 0.15 Accretion 0.02
Tawau 3.83 ± 2.82 4.09 ± 2.17 53.8 −2.28 −0.46 −0.36 Erosion 0.16
Kudat 2.81 ± 2.78 4.20 ± 1.6 48.92 −12.08 0.2 −0.41 Erosion 0.10
Labuan 2 3.24 ± 2.57 3.79 ± 1.44 30.61 −19.84 −0.38 −0.42 Erosion 0.32
Sejingkat −3.99 ± 5.80 4.11 ± 0.98 32.42 −17.90 −0.74 −0.41 Erosion 0.17
Miri 10.51 ± 2.39 3.99 ± 1.34 24.7 −7.15 −0.24 −0.40 Erosion 0.42
 Tawau 3.83 ± 2.82 4.09 ± 2.17 53.8 −2.28 −0.46 −0.36 Erosion 0.16
Kudat 2.81 ± 2.78 4.20 ± 1.6 48.92 −12.08 0.2 −0.41 Erosion 0.10
Labuan 2 3.24 ± 2.57 3.79 ± 1.44 30.61 −19.84 −0.38 −0.42 Erosion 0.32
Sejingkat −3.99 ± 5.80 4.11 ± 0.98 32.42 −17.90 −0.74 −0.41 Erosion 0.17
Miri
Remote Sens. 2021,±13,2.39
10.51 3587 3.99 ± 1.34 24.7 −7.15 −0.24 −0.40 Erosion 150.42
of 22

To evaluate the automatic shoreline-extraction techniques employed in this study,

we compared the results with shorelines extracted from selected satellite imagery ac-
To evaluate the automatic shoreline-extraction techniques employed in this study, we
quired from the
compared USGS andwith
results Sci-hub usingextracted
shorelines manual and
fromsemi-automatic
selected satellite techniques. The semi-
imagery acquired
automatic extraction was conducted in ArcGIS Pro, while the manual on-screen
from USGS and Sci-hub using manual and semi-automatic techniques. The semi-automatic digitizing
of shorelines
extraction waswas done ininArcGIS
conducted Desktop.
ArcGIS Pro, Figure
while the 8 shows
manual thedigitizing
on-screen shorelineof extracted
shorelines from
Landsat
was done7 with the three
in ArcGIS techniques.
Desktop. Figure 8 The
showsmanually extracted
the shoreline shoreline
extracted was taken
from Landsat 7 withas the
the three
actual techniques.
shoreline for theThe manuallyof
validation extracted shorelineand
the automatic wassemi-automatic
taken as the actual shorelineshore-
extracted
for The
lines. the validation of the automatic
mean deviations and semi-automatic
of the transect points fromextracted shorelines.
the manually The mean
extracted shorelines
deviations
were computed. of the transect points from the manually extracted shorelines were computed.

Figure 8. (a)8.Band
Figure 4, 5,4,75,composite
(a) Band ofofLandsat
7 composite Landsat77 and extractedshorelines.
and extracted shorelines.(b)(b) Manual,
Manual, semi-automatic,
semi-automatic, and automatic
and automatic
shorelines extracted from the image. (c) Base map imagery of the study area (base map by Google Satellite).
shorelines extracted from the image. (c) Base map imagery of the study area (base map by Google Satellite).

Furthermore,
Furthermore,shoreline-change assessmentofofthethe
shoreline-change assessment manual
manual andand semi-automatic
semi-automatic ex- ex-
tracted
tracted shorelineswas
shorelines wasperformed
performed atat low-frequency
low-frequency sampling
samplingat an
at interval of approxi-
an interval of approxi-
mately
mately a decade.The
a decade. Theresulting
resulting change
change statistics
statisticsfor
forthe shoreline
the areare
shoreline presented in Table
presented 5.
in Table 5.
Figure 9 below shows the elevation profile and developments along the test sites.
Figure 9 below shows the elevation profile and developments along the test sites.
Table 5. Shoreline-change statistics for low-frequency sampling indicating accretion/erosion at each transect.
Table 5. Shoreline-change statistics for low-frequency sampling indicating accretion/erosion at each transect.
Transect ID TCD Length (m) SCE (m) NSM (m) EPR (m/y) LRR (m/y) WLR (m/y)
Transect ID
1 TCD 100 Length (m)
139.6885251 SCE (m)
13.63 NSM
−2.2(m) EPR (m/y)
−0.07 LRR
−0.16(m/y) −WLR
0.16 (m/y)
1 2
100 200
139.6885251
86.88467384
13.63
25.75
−2.2
2.96
−0.07
0.1
−0.16
−0.02 −0.02−0.16
2 3
200 300
86.88467384
97.36654348
25.75
24.42
2.96
11.26
0.1
0.37
−0.02
0.38 0.38
−0.02
3 4
300 400
97.36654348
111.264345
24.42
23.42
11.26
−17.34
0.37
−0.58
0.38
−0.63 −0.63
0.38
5 500 174.2997707 76.64 −48.69 −1.62 −2.26 −2.26
6 600 121.5914144 40.36 −31.05 −1.03 −1.36 −1.36
Remote Sens. 2021, 13, 3587 16 of 22

Remote Sens. 2021, 13, x FOR PEER REVIEW 16 of 22

Table 5. Cont.

4 ID
Transect 400
TCD 111.264345
Length (m) 23.42
SCE (m) −17.34
NSM (m) −0.58(m/y)
EPR −0.63
LRR (m/y) −0.63
WLR (m/y)
75 500
700 174.2997707
55.76084843 76.64
25.35 −48.69
−22.11 −1.62
−0.73 −2.26
−0.85 −2.26
−0.85
6 600 121.5914144 40.36 −31.05 −1.03 −1.36 −1.36
8 800 40.65624084 24.91 −21.74 −0.72 −0.9 −0.9
7 700 55.76084843 25.35 −22.11 −0.73 −0.85 −0.85
9 900 58.57434371 32.65 −24.35 −0.81 −1.06 −1.06
8 800 40.65624084 24.91 −21.74 −0.72 −0.9 −0.9
10 1000 84.49062081 29.02 −18.32 −0.61 −0.84 −0.84
9 900 58.57434371 32.65 −24.35 −0.81 −1.06 −1.06
11
10 1100
1000 118.6596668
84.49062081 20.12
29.02 −9.05
−18.32 −0.3
−0.61 −0.51
−0.84 −0.51
−0.84
12
11 1200
1100 172.6533679
118.6596668 32.53
20.12 −16.94
−9.05 −0.56
−0.3 −0.7
−0.51 −0.7
−0.51
12
13 1200
1300 172.6533679
263.2099826 32.53
43.55 −16.94
−34.83 −0.56
−1.16 −0.7
−1.38 −0.7
− 1.38
13
14 1300
1400 263.2099826
308.8970508 43.55
2.85 −34.83
−2.85 −1.16
−0.41 −1.38 −1.38
14 1400 308.8970508 2.85 −2.85 −0.41

Figure 9.
Figure 9. Elevation
Elevationprofile
profileand
anddevelopment
development along
along thethe coast
coast of Teluk
of Teluk Nipah
Nipah Beach.
Beach. The The profile
profile was drawn
was drawn overJuly
over the the2020
July
2020 shoreline, along the north direction.
shoreline, along the north direction.

5. Discussion
5.1. High-Frequency Sampling and Automatic Extraction
Extraction with
with GEE
GEE
Regular
Regular andandperiodic
periodicmonitoring
monitoringofof shoreline
shorelinechanges
changesis vital for coastal
is vital management
for coastal manage-
and
mentfuture planning.
and future WhileWhile
planning. the contemporaneous
the contemporaneous observation of the
observation ofEarth by multiple
the Earth by mul-
satellite missions facilitates regular and long-term monitoring of shoreline
tiple satellite missions facilitates regular and long-term monitoring of shoreline changes, changes, big-
data management has been a challenge in leveraging this opportunity.
big-data management has been a challenge in leveraging this opportunity. Consequently, Consequently,
earlier
earlier studies
studies employed
employed low-frequency
low-frequency data data sampling
sampling for for shoreline
shoreline analysis.
analysis. A A major
major
limitation with low-frequency sampling is that it does not resolve the shoreline
limitation with low-frequency sampling is that it does not resolve the shoreline dynamics dynamics
well
well enough
enough to to quantify
quantify the impact of
the impact of sea
sea level,
level, which
which isis usually
usually sampled
sampled at at hourly,
hourly, daily,
daily,
or monthly frequencies. In addition, there are several uncertainties due
or monthly frequencies. In addition, there are several uncertainties due to short-term var- to short-term
variability
iability andand
tidaltidal effects
effects on instantaneous
on instantaneous shorelines.
shorelines. So, a method
So, a method to automatically
to automatically extract
extract shorelines from multi-mission optical satellites at high-frequency
shorelines from multi-mission optical satellites at high-frequency sampling samplingwithwith
im-
improved
proved data data management
management and
and a arobust
robustextraction
extractionalgorithm
algorithmwas wasapplied
appliedin in this
this study.
study.
Between January 1993 and October 2019, a total of 845 images were available for shoreline-
change analysis at the test site. Consequently, 845 satellite images (266 from Landsat 5,
Remote Sens. 2021, 13, 3587 17 of 22

Between January 1993 and October 2019, a total of 845 images were available for shoreline-
change analysis at the test site. Consequently, 845 satellite images (266 from Landsat 5,
252 from Landsat 7, 135 from Landsat 8, and 192 from Sentinel-2) returned after temporal
and spatial filtering were processed with CoastSat, a GEE-enabled toolkit. GEE supports
the acquisition of only the required bands of satellites and cropping to the region before
downloading, thus significantly reducing the data size. A total of 428 of the 845 satellite
images that met the quality requirement of cloud cover of less than 20% were just 20.83 MB
(4.78 MB for 128 Landsat 5 images, 4.53 MB for 140 Landsat 7, 2.42 MB for 73 Landsat 8,
and 9.1 MB for 87 Sentinel-2) after band selection and cropping to AOI. For performance
comparison, a low-frequency shoreline-sampling method using a conventional approach
was also investigated. Four satellites images at an interval of approximately a decade were
acquired from USGS (Landsat) and Sci-hub (Sentinel-2). Landsat 5 images acquired on
13 March 1990 were 116.21 MB, Landsat 7 images acquired on 22 November 1999 were
245.23 MB, Landsat 8 images acquired on 31 May 2013 were 825.47 MB, while Sentinel-2
images acquired on 8 April 2020 were 471.36 MB. The total size of the four satellite images
was 1.7 GB, and they were required to be downloaded in their entirety before further
processing such as clipping and pan sharpening. One image from Landsat 8 acquired from
USGS was 41 times larger than the total size of 428 multi-sensor satellite images retrieved
from GEE. Thus, GEE addressed the big-data storage and management issue of SDS change
analysis. This facilitated the effective combination of multiple satellite missions at high
sampling frequency, which made it possible to offer continuous data throughout the year.
By ensuring that shorelines were extracted for almost every month of the study period, the
impact of sea-level rise on shoreline displacement could be estimated.
When compared to the manually digitized shoreline, the automatically extracted
shoreline appeared to be in closer proximity than the semi-automatic extracted shorelines
as shown in Figure 8. The mean deviation of the transect points of the automatic and the
semi-automatic extracted shorelines from the actual shoreline were 0.104 m and 0.119 m,
respectively. The shoreline extracted manually through digitizing of the land–sea boundary
from a false-colour composite of satellite scenes was accurate, but time-consuming and
labour-intensive for high-frequency shoreline sampling. Semi-automatic extraction of
shorelines using tasseled cap transformation and NDVI also takes time depending on
the number of images and user experience. In addition, there is a high propensity for
false classification due to clouds and breaking waves. In contrast, automatic shoreline
extraction enables faster and easier extraction of multi-shorelines while eliminating errors
due to human intervention. It took approximately 10 min to extract 428 shorelines from
the multi-mission satellites after training with the robust ML algorithm. The extracted
shorelines based on the automated shoreline detection method were as accurate as the
manually and carefully digitized shorelines.
In an earlier study, Foo, Teh and Babatunde [52] reported the decadal shoreline change
rates of Teluk Nipah in the 1990s, 2000s, and 2010s using EPR and LRR. The shoreline
variation for the 1990s was relatively stable, with maximum erosion and accretion rates
of −1.64 m/y (−0.74 m/y) and 2.48 m/y (2.95 m/y), respectively, based on EPR (LRR).
Contrary to the 1990s, Teluk Nipah experienced severe erosion in the 2010s. The study
estimated the maximum erosion rate for the 2010s as −6.83 m/y and −5.68 m/y based
on EPR and LRR, respectively, but acknowledged that it is most likely that this indication
of critical erosion was an overestimation of the actual result. The result was obtained
from the analysis of four instantaneous satellite-derived shorelines extracted using a semi-
automatic approach with no tidal correction. The study suggested that the unusually high
erosion rate was perhaps due to the tidal effect on the instantaneous shorelines acquired at
different times. The use of high-frequency sampling and tidal correction of a time series of
shoreline displacement in the present study effectively reduced uncertainties in the satellite-
derived shoreline analysis due to short-term variability and tidal effects, respectively. It
was noteworthy that there was no difference in some change analyses between high- and
low-frequency sampling. The NSM and EPR estimates were approximately the same for
Remote Sens. 2021, 13, 3587 18 of 22

high- and low-frequency analyses, both indicating that 85.71% and 14.29% of the shoreline
experienced erosion and accretion, respectively. This was not unexpected, because these
change statistics only considered two endpoints (oldest and earliest date) and ignored
other sampled data in between. Other change statistics; however, showed a significant
difference between the low- and high-frequency sampling analyses. The linear regression
rate (LRR) statistic for the low-frequency sampling analysis indicated that 92.31% of the
shoreline experienced erosion at an average rate of −0.79 ± 1.32 m/y, while the high-
frequency sampling analysis indicated 50% shoreline erosion at a rate of −0.17 ± 0.07 m/y.
The maximum erosion rates of −2.26 m/y and −1.37 m/y were estimated based on LRR
statistics of the low- and high-frequency sampling, respectively. With this acceptable lower-
intensity erosion rate, this study established the potential of high-frequency-sampled, tidal-
corrected, satellite-derived shorelines in reducing uncertainties in instantaneous shorelines.

5.2. SLR and Shoreline Dynamics

Sea-level change is usually perceived to be responsible for long-term shoreline dis-
placement, but our study showed that both erosion and accretion occurred along the
shoreline, whereas there was a consistent rise in sea level at all stations along the Malaysian
coastline. Results of shoreline displacement (Table 4) demonstrated that the larger part
of the Malaysian coastline (about 76%) was eroding, but some coastal areas also showed
an accretion pattern. An uptrend was observed in the absolute sea-level change at all
stations. The highest trend of 4.46 ± 2.13 mm/y was estimated at the Sejinkat station in
Sabah and Sarawak (Figure 5b), while Kukup in West Peninsular Malaysia (Figure 5a) had
the lowest trend estimate of 2.63 ± 1.22 mm/y. The shoreline displacement pattern at these
two tide-gauge stations was accretion (for Kukup) and erosional (for Sejinkat) at an LRR of
0.09 m/y and −0.41 m/y, respectively. LRR statistics showed that Pulau Tioman Beach,
with an absolute sea-level trend of 3.72 ± 0.85 mm/y, was the most erosional, at a rate
of −0.44 m/y. Cendering Beach, with an absolute sea-level trend of 3.86 ± 0.89 mm/y,
was the most accretional, at a rate of 0.31 m/y. The correlation of yearly mean absolute
sea-level change and shoreline displacement showed a positive relationship at all stations.
Tanjung Sedili, Pulua Pinang, and Miri showed a positive correlation of 0.54, 0.43, and 0.42,
respectively. Positive correlation values greater than 0.2 were also observed at the majority
of coasts with erosion patterns. These correlation estimates were small but significant in
the context of SLR–shoreline dynamics, considering that shoreline erosion is also connected
to other energetic processes such as waves and winds. However, positive but low correla-
tion was also found at the coasts with an accretion pattern—Lumut, Kukup, Cendering,
Sandaku, and Lahad Datu. It is possible that the input of other factors, such as sediment
supply, overrode the contribution of SLR in those areas. This implies that SLR contributed
to shoreline erosion, but was not the major driver of shoreline displacement for this study
area. Investigation of other processes and potential factors might be required to holistically
explain the causes of shoreline dynamics experienced along the coastline.

5.3. Other Processes Responsible for Shifting Shorelines

The impact of the high density of human population and activities, such as the
development of residential areas, industries, tourist destinations, and commercial projects,
as well as construction of harbours and jetties, was observed in some coastal areas. As one
of Malaysia’s most popular recreational beaches [52], Teluk Nipah, for example, witnesses
a high inflow of tourists annually, and is thus a region of high anthropogenic activity. The
growing human activities have influenced commercial developments in the area. Figure 9
shows the development and profile of Teluk Nipah Beach. Restaurants, hotels and resorts,
homestays, and hawker stalls were distributed along the coastline, but were concentrated
at the mid-section, thereby exerting enormous pressure on this region. Inspection of the
landscape revealed erosion in this region of intense activities and development. The erosion
regions were concentrated in areas showing restaurants, homestays, resorts, hawkers stalls,
and commercial shops along the coastline. The beach profile revealed a downward slope
Remote Sens. 2021, 13, 3587 19 of 22

from the eroding mid-section to the upper north section of the beach. We believe that this
arrangement enabled the eroded sediments from the mid-sections to be transported to the
lower elevation upper section. The accretion observed at the end section was perhaps due
to this longshore transportation of eroded sediments from the higher elevation mid-sections
of the coast. Foo, Teh and Babatunde [52] had previously reported this contribution of
sediment supply over the test site. The study conducted an analysis of the grain-size
distribution of the soil sample using a sieve analysis, and discovered that the sediment
in the northern region was finer than that in the southern region. It was later discovered
during site studies that the southern beach’s top sand had been covered by a thin layer of
finer, brighter sand. The fine sand was reported to have been transported from the beach
in the northern region by sediment movement along the shore. Although shoreline erosion
is connected to energetic (e.g., sea-level rise, marine regime, and winds) and mass balances
(e.g., the inputs/outputs of sediments), this study largely considered the impact of SLR on
shoreline dynamics. In future studies, adopting a more holistic approach by considering
other pertinent parameters could provide further insights into shoreline dynamics.

5.4. Future Impact of Rising Sea Level

From this study, the rate of SLR along the Malaysian coastline estimated from the av-
erage sea-level trend at all stations was 3.72 mm/y for the relative sea level and 3.68 mm/y
for the absolute sea level. Based on these rates, absolute sea-level rise of about 0.11 m and
0.29 m is expected by 2050 and 2100, respectively. The relative sea level is predicted to rise
by 0.12 m and 0.30 m over the same period. This is threatening, considering that 76% of the
coastline already shows erosion patterns. For a high global temperature rise scenario due
to increased greenhouse gas emissions (RCP 8.5), some studies based on semi-empirical
models [65] and the Intergovernmental Panel on Climate Change (IPCC) [13] predicted
a global rise of more than 1 m by the end of the 21st century. This estimate would cause
a devastating submersion of most of the Malaysian coastline. Pulau Pinang, Port Kelang,
Johor Bahru, Tanjung Sedili, Bintulu, Lanbuan 2, and Miri, where a moderate positive
correlation (greater than 0.3) was observed between the sea level and shoreline change,
might be most affected by this future projection of the sea level.

6. Conclusions
Satellites provide long-term, continuous, and quasi-global data that are vital for assess-
ing the morpho-dynamics of shorelines; however, there are challenges in data management,
processing, and analysis. By employing a Google Earth Engine-enabled Python toolkit and
multi-mission optical satellites, we demonstrated an approach to automatically extracting
instantaneous shorelines at high-frequency sampling towards quantifying the response
of the coastlines to rising sea levels. These techniques were employed to assess 30 years
of shoreline displacement at a test site and 21 coastal areas in Malaysia with tide gauges.
Compared to semi-automatic and manual approaches, the approach was faster, less labori-
ous, and required minimal storage space, and automatically extracted multiple shorelines
from Landsat 5, 7, and 8 and Sentinel-2 images in a short time with high accuracy. By
ensuring that shorelines were extracted for almost every month of the study period, the
impact of sea-level rise on shoreline displacement could be estimated. Rising absolute
sea level was discovered at all stations, but the shoreline displacement pattern showed
both erosion and accretion trends, with the former more prevalent. A significant positive
correlation existed between SLR and shoreline dynamics at all eroding coasts. A positive
but low correlation was also found at the coasts with an accretion pattern. This confirmed
that a rising sea level is contributing to shoreline erosion in the study area, but is not the
only driver of shoreline displacement. Further investigation revealed that a combination
of high population density, anthropogenic activities, and long shore-sediment transport
are contributing to shoreline displacement over the study area. Malaysia’s coastline might
experience a considerable coastal retreat in the face of future estimates of SLR.
Remote Sens. 2021, 13, 3587 20 of 22

Author Contributions: Conceptualization, A.-L.B. and N.A.; methodology, N.A.; software, N.A.;
validation, A.-L.B., N.A., M.M. and T.H.M.; formal analysis, A.-L.B. and N.A.; investigation, A.-L.B.
and N.A.; resources, A.-L.B., N.A. and M.M.; data curation, N.A.; writing—original draft preparation,
N.A.; writing—review and editing, A.-L.B., N.A., M.M. and T.H.M.; visualization, A.-L.B., N.A.,
M.M. and T.H.M.; supervision, A.-L.B.; project administration, A.-L.B., M.M. and T.H.M.; funding
acquisition, A.-L.B. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Data Availability Statement: Publicly available datasets were analyzed in this study. The relative
sea level data can be found here: [https://www.psmsl.org/data/] (accessed on 1 May 2021). Ab-
solute sea level data is available here: [https://marine.copernicus.eu/] (accessed on 1 May 2021).
Optical satellites data used for the study can be obtained here: [https://earthexplorer.usgs.gov/]
(accessed on 1 May 2021).
Acknowledgments: We would like to acknowledge the Permanent Service of Mean Sea Level
(PSMSL) and Copernicus Marine Environment Service (CMEMS) for making sea levels openly
available for public use; the U.S. Geological Survey (USGS) for making satellite imagery freely
available; Archiving, Validation, and Interpretation of Satellite Oceanographic data (AVISO) for
developing and making the ocean-tide model available; and Google for making Google Earth Engine,
the cloud computing platform, free for researchers.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Vitousek, P.M.; Mooney, H.A.; Lubchenco, J.; Melillo, J.M. Human domination of earth’s ecosystems. Science 1997, 277, 494.
[CrossRef]
2. Luijendijk, A.; Hagenaars, G.; Ranasinghe, R.; Baart, F.; Donchyts, G.; Aarninkhof, S. The state of the world’s beaches. Sci. Rep.
2018, 8, 6641. [CrossRef] [PubMed]
3. Cabezas-Rabadan, C.; Pardo-Pascual, J.E.; Palomar-Vazquez, J.; Fernandez-Sarria, A. Characterizing beach changes using high-
frequency Sentinel-2 derived shorelines on the Valencian coast (Spanish Mediterranean). Sci. Total Environ. 2019, 691, 216–231.
[CrossRef] [PubMed]
4. Vos, K.; Splinter, K.D.; Harley, M.D.; Simmons, J.A.; Turner, I.L. CoastSat: A google earth engine-enabled python toolkit to extract
shorelines from publicly available satellite imagery. Environ. Model. Softw. 2019, 122, 104528. [CrossRef]
5. Mills, A.K.; Ruggiero, P.; Bolte, J.P.; Serafin, K.A.; Lipiec, E. Quantifying uncertainty in exposure to coastal hazards associated
with both climate change and adaptation strategies: A U.S. pacific northwest alternative coastal futures analysis. Water 2021,
13, 545. [CrossRef]
6. Anfuso, G.; Postacchini, M.; Di Luccio, D.; Benassai, G. Coastal sensitivity/vulnerability characterization and adaptation
strategies: A review. J. Mar. Sci. Eng. 2021, 9, 72. [CrossRef]
7. Le Cozannet, G.; Garcin, M.; Yates, M.; Idier, D.; Meyssignac, B. Approaches to evaluate the recent impacts of sea-level rise on
shoreline changes. Earth-Sci. Rev. 2014, 138, 47–60. [CrossRef]
8. Anzidei, M.; Scicchitano, G.; Scardino, G.; Bignami, C.; Tolomei, C.; Vecchio, A.; Serpelloni, E.; De Santis, V.; Monaco, C.; Milella,
M.; et al. Relative sea-level rise scenario for 2100 along the coast of south eastern sicily (Italy) by InSAR data, satellite images and
high-resolution topography. Remote Sens. 2021, 13, 1108. [CrossRef]
9. Di Stefano, A.; De Pietro, R.; Monaco, C.; Zanini, A. Anthropogenic influence on coastal evolution: A case history from the
Catania Gulf shoreline (eastern Sicily, Italy). Ocean. Coast. Manag. 2013, 80, 133–148. [CrossRef]
10. Vousdoukas, M.I.; Mentaschi, L.; Voukouvalas, E.; Verlaan, M.; Jevrejeva, S.; Jackson, L.P.; Feyen, L. Global probabilistic projections
of extreme sea levels show intensification of coastal flood hazard. Nat. Commun. 2018, 9, 2360. [CrossRef]
11. Mengel, M.; Levermann, A.; Frieler, K.; Robinson, A.; Marzeion, B.; Winkelmann, R. Future sea level rise constrained by
observations and long-term commitment. Proc. Natl. Acad. Sci. USA 2016, 113, 2597–2602. [CrossRef]
12. IPCC. Climate Change 2014: Synthesis Report; Contribution of Working Groups I, II and III to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change; IPCC: Geneva, Switzerland, 2014; 151p.
13. Oppenheimer, M.; Glavovic, B.; Hinkel, J.; van de Wal, R.; Magnan, A.K.; Abd-Elgawad, A.; Cai, R.; Cifuentes-Jara, M.; Deconto,
R.M.; Ghosh, T.; et al. Sea level rise and implications for low-lying islands, coasts and communities. In IPCC Special Report on the
Ocean and Cryosphere in a Changing Climate; Pörtner, H.-O., Roberts, D.C., Masson-Delmotte, V., Zhai, P., Tignor, M., Poloczanska,
E., Mintenbeck, K., Alegría, A., Nicolai, M., Okem, A., et al., Eds.; IPCC: Geneva, Switzerland, 2019; In Press.
14. Rahmstorf, S.J.S. A semi-empirical approach to projecting future sea-level rise. Science 2007, 315, 368–370. [CrossRef]
15. Horton, R.; Herweijer, C.; Rosenzweig, C.; Liu, J.; Gornitz, V.; Ruane, A.C. Sea level rise projections for current generation CGCMs
based on the semi-empirical method. Geophys. Res. Lett. 2008, 35, 2. [CrossRef]
16. Nicholls, R.J.; Cazenave, A. Sea-level rise and its impact on coastal zones. Science 2010, 328, 1517–1520. [CrossRef]
17. Mitrovica, J.X.; Gomez, N.; Clark, P.U.J.S. The sea-level fingerprint of West Antarctic collapse. Science 2009, 323, 753. [CrossRef]
Remote Sens. 2021, 13, 3587 21 of 22

18. Peltier, W. Global glacial isostatic adjustment and modern instrumental records of relative sea level history. In International
Geophysics; Elsevier: Amsterdam, The Netherlands, 2001; Volume 75, pp. 65–95.
19. Ezer, T.; Corlett, W.B. Analysis of relative sea level variations and trends in the Chesapeake Bay: Is there evidence for acceleration
in sea level rise? In Proceedings of the OCEANS 2012 MTS/IEEE Harnessing the Power of the Ocean, Virginia, VA, USA,
14–19 October 2012.
20. Werner, A.D.; Simmons, C.T.J.G. Impact of sea-level rise on sea water intrusion in coastal aquifers. Groundwater 2009, 47, 197–204.
[CrossRef]
21. Church, J.A.; White, N.J. Sea-level rise from the late 19th to the early 21st century. Surv. Geophys. 2011, 32, 585–602. [CrossRef]
22. Cazenave, A.; Llovel, W. Contemporary sea level rise. Annu. Rev. Mar. Sci. 2010, 2, 145–173. [CrossRef]
23. Ranjbar, A.; Cherubini, C.; Saber, A. Investigation of transient sea level rise impacts on water quality of unconfined shallow
coastal aquifers. Int. J. Environ. Sci. Technol. 2020, 17, 2607–2622. [CrossRef]
24. Church, J.; Clark, P.; Cazenave, A.; Gregory, J.; Jevrejeva, S.; Levermann, A.; Merrifield, M.; Milne, G.; Nerem, R.; Nunn, P. Climate
change 2013: The physical science basis. In Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental
Panel on Climate Change; IPCC: Geneva, Switzerland, 2013; p. 1137.
25. Gutierrez, B.T.; Plant, N.G.; Thieler, E.R. A Bayesian network to predict coastal vulnerability to sea level rise. J. Geophys. Res.
Earth Surf. 2011, 116. [CrossRef]
26. Adebisi, N.; Balogun, A.-L.; Min, T.H.; Tella, A. Advances in estimating Sea Level Rise: A review of tide gauge, satellite altimetry
and spatial data science approaches. Ocean. Coast. Manag. 2021, 208, 105632. [CrossRef]
27. Webb, A.P.; Kench, P.S. The dynamic response of reef islands to sea-level rise: Evidence from multi-decadal analysis of island
change in the Central Pacific. Glob. Planet. Chang. 2010, 72, 234–246. [CrossRef]
28. Toure, S.; Diop, O.; Kpalma, K.; Maiga, A. Shoreline detection using optical remote sensing: A review. ISPRS Int. J. Geo-Inf. 2019,
8, 75. [CrossRef]
29. Boak, E.H.; Turner, I.L. Shoreline definition and detection: A review. J. Coast. Res. 2005, 21, 688–703. [CrossRef]
30. Ribas, F.; Simarro, G.; Arriaga, J.; Luque, P. Automatic shoreline detection from video images by combining information from
different methods. Remote Sens. 2020, 12, 3717. [CrossRef]
31. Jóia Santos, C.; Andriolo, U.; Ferreira, J.C. Shoreline response to a sandy nourishment in a wave-dominated coast using video
monitoring. Water 2020, 12, 1632. [CrossRef]
32. Valentini, N.; Saponieri, A.; Molfetta, M.G.; Damiani, L. New algorithms for shoreline monitoring from coastal video systems.
Earth Sci. Inform. 2017, 10, 495–506. [CrossRef]
33. Vélez-Nicolás, M.; García-López, S.; Barbero, L.; Ruiz-Ortiz, V.; Sánchez-Bellón, Á. Applications of unmanned aerial systems
(UASs) in hydrology: A review. Remote Sens. 2021, 13, 1359. [CrossRef]
34. Vos, K.; Harley, M.D.; Splinter, K.D.; Simmons, J.A.; Turner, I.L. Sub-annual to multi-decadal shoreline variability from publicly
available satellite imagery. Coast. Eng. 2019, 150, 160–174. [CrossRef]
35. Li, J.; Roy, D.P. A global analysis of sentinel-2A, sentinel-2B and landsat-8 data revisit intervals and implications for terrestrial
monitoring. Remote Sens. 2017, 9, 902. [CrossRef]
36. Ozturk, D.; Sesli, F.A. Shoreline change analysis of the Kizilirmak Lagoon Series. Ocean. Coast. Manag. 2015, 118, 290–308.
[CrossRef]
37. Sekovski, I.; Stecchi, F.; Mancini, F.; Del Rio, L. Image classification methods applied to shoreline extraction on very high-resolution
multispectral imagery. Int. J. Remote Sens. 2014, 35, 3556–3578. [CrossRef]
38. Aedla, R.; Dwarakish, G.S.; Reddy, D.V. Automatic Shoreline Detection and Change Detection Analysis of Netravati-
GurpurRivermouth Using Histogram Equalization and Adaptive Thresholding Techniques. Aquat. Procedia 2015, 4, 563–570.
[CrossRef]
39. Ghoneim, E.; Mashaly, J.; Gamble, D.; Halls, J.; AbuBakr, M. Nile Delta exhibited a spatial reversal in the rates of shoreline retreat
on the Rosetta promontory comparing pre- and post-beach protection. Geomorphology 2015, 228, 1–14. [CrossRef]
40. McFeeters, S.K. The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. Int. J.
Remote Sens. 1996, 17, 1425–1432. [CrossRef]
41. Xu, H. Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery.
Int. J. Remote Sens. 2006, 27, 3025–3033. [CrossRef]
42. Feyisa, G.L.; Meilby, H.; Fensholt, R.; Proud, S.R. Automated water extraction index: A new technique for surface water mapping
using Landsat imagery. Remote Sens. Environ. 2014, 140, 23–35. [CrossRef]
43. Liu, H.; Wang, L.; Sherman, D.J.; Wu, Q.; Su, H. Algorithmic foundation and software tools for extracting shoreline features from
remote sensing imagery and LiDAR data. J. Geogr. Inf. Syst. 2011, 3, 99. [CrossRef]
44. García-Rubio, G.; Huntley, D.; Russell, P. Evaluating shoreline identification using optical satellite images. Mar. Geol. 2015,
359, 96–105. [CrossRef]
45. Elnabwy, M.T.; Elbeltagi, E.; El Banna, M.M.; Elshikh, M.M.Y.; Motawa, I.; Kaloop, M.R. An approach based on landsat images
for shoreline monitoring to support integrated coastal management—A case study, ezbet elborg, nile delta, egypt. ISPRS Int. J.
Geo-Inf. 2020, 9, 199. [CrossRef]
46. Foody, G. Pattern recognition and classification of remotely sensed images by artificial neural networks. In Ecological Informatics;
Springer: Berlin/Heidelberg, Germany, 2006; pp. 459–477.
Remote Sens. 2021, 13, 3587 22 of 22

47. Tsekouras, G.E.; Trygonis, V.; Maniatopoulos, A.; Rigos, A.; Chatzipavlis, A.; Tsimikas, J.; Mitianoudis, N.; Velegrakis, A.F. A
Hermite neural network incorporating artificial bee colony optimization to model shoreline realignment at a reef-fronted beach.
Neurocomputing 2018, 280, 32–45. [CrossRef]
48. Kerh, T.; Lu, H.; Saunders, R. Construct. Architect. Eng. Shoreline change estimation from survey image coordinates and neural
network approximation. Int. J. Civil Struct. Construct. Architect. Eng. 2014, 8, 372–377.
49. Nazeer, M.; Waqas, M.; Shahzad, M.I.; Zia, I.; Wu, W. Coastline vulnerability assessment through landsat and cubesats in a coastal
mega city. Remote Sens. 2020, 12, 749. [CrossRef]
50. Balogun, A.-L.; Adebisi, N. Sea level prediction using ARIMA, SVR and LSTM neural network: Assessing the impact of ensemble
Ocean-Atmospheric processes on models’ accuracy. Geomat. Nat. Hazards Risk 2021, 12, 653–674. [CrossRef]
51. Adebisi, N.; Balogun, A.-L. A deep-learning model for national scale modelling and mapping of Sea level rise in Malaysia: The
past, present, and future. Geocarto Int. 2021, 1–23. [CrossRef]
52. Foo, W.Y.; Teh, H.M.; Babatunde, A.-L. Morphodynamics of the teluk nipah shorelines. J. Eng. 2020, 4, 29–40.
53. Ehsan, S.; Ara Begum, R.; Ghani Md Nor, N.; Nizam Abdul Maulud, K. Current and potential impacts of sea level rise in the
coastal areas of Malaysia. IOP Conf. Ser. Earth Environ. Sci. 2019, 228, 012023. [CrossRef]
54. Gorelick, N.; Hancher, M.; Dixon, M.; Ilyushchenko, S.; Thau, D.; Moore, R. Google earth engine: Planetary-scale geospatial
analysis for everyone. Remote Sens. Environ. 2017, 202, 18–27. [CrossRef]
55. Tamiminia, H.; Salehi, B.; Mahdianpari, M.; Quackenbush, L.; Adeli, S.; Brisco, B. Google earth engine for geo-big data applications:
A meta-analysis and systematic review. ISPRS J. Photogramm. Remote Sens. 2020, 164, 152–170. [CrossRef]
56. Hamid, A.I.A.; Din, A.H.M.; Hwang, C.; Khalid, N.F.; Tugi, A.; Omar, K.M. Contemporary sea level rise rates around Malaysia:
Altimeter data optimization for assessing coastal impact. J. Asian Earth Sci. 2018, 166, 247–259. [CrossRef]
57. USGS. Landsat collection 1 level 1 product definition. In United States Geological Survey, 1st ed.; 2017. Available online:
https://prd-wret.s3.us-west-2.amazonaws.com/assets/palladium/production/atoms/files/LSDS-1656_%20Landsat_Collec
tion1_L1_Product_Definition-v2.pdf (accessed on 1 May 2021).
58. Chander, G.; Markham, B.L.; Helder, D.L. Summary of current radiometric calibration coefficients for Landsat MSS, TM, ETM+,
and EO-1 ALI sensors. Remote Sens. Environ. 2009, 113, 893–903. [CrossRef]
59. Din, A.H.M.; Hamid, A.I.A.; Yazid, N.M.; Tugi, A.; Khalid, N.F.; Omar, K.M.; Ahmad, A.J.J.T. Malaysian sea water level pattern
derived from 19 years tidal data. J. Teknol. 2017, 79, 5.
60. Holland, P.W.; Welsch, R.E. Robust regression using iteratively reweighted least-squares. Commun. Stat.-Theory Methods 1977,
6, 813–827. [CrossRef]
61. Wicaksono, A.; Wicaksono, P.; Khakhim, N.; Mohammad, N.; Farda, M.A.M.J.G.I. Tidal correction effects analysis on shoreline
mapping in Jepara Regency. Geosp. Inf. 2018, 2, 2. [CrossRef]
62. Carrère, L.; Lyard, F.; Cancet, M.; Guillot, A.; Picot, N. FES 2014, a new tidal model—Validation results and perspectives for
improvements. In Proceedings of the ESA living planet symposium, Prague, Czech Republic, 9–13 May 2016; pp. 9–13.
63. Himmelstoss, E.A.; Henderson, R.E.; Kratzmann, M.G.; Farris, A.S. Digital Shoreline Analysis System (DSAS) Version 5.0 User Guide;
US Geological Survey: Reston, VA, USA, 2018.
64. Berman, J.J. Understanding your data. In Data Simplification; Berman, J.J., Ed.; Morgan Kaufmann: Boston, MA, USA, 2016;
pp. 135–187.
65. Church, J.A.; Clark, P.U.; Cazenave, A.; Gregory, J.M.; Jevrejeva, S.; Levermann, A.; Merrifield, M.A.; Milne, G.A.; Nerem, R.S.;
Nunn, P.D. Sea Level Change; PM Cambridge University Press: Cambridge, UK, 2013.