Regional Studies in Marine Science 66 (2023) 103118

Contents lists available at ScienceDirect

Regional Studies in Marine Science

journal homepage: www.elsevier.com/locate/rsma

Shoreline change rate detection applying the DSAS technique on low

and medium resolution data: Case study along Ash Shu’aybah-Al
Mujayrimah coastal Area of the Eastern Red Sea, Saudi Arabia
∗
Omar A. Alharbi a , , Samia S. Hasan b , Amr S. Fahil c,d , Ammar Mannaa e ,
Nelson Rangel-Buitrago f , Abdullah F. Alqurashi a
a
Geography Department, College of Social Sciences, Umm Al- Qura University, Makkah 21955, Saudi Arabia
b
Desert Research Center, El Matareya, Cairo Governorate 11753, Egypt
c
Geology Department, Faculty of Science, Tanta University, Tanta 31527, Egypt
d
Department of Earth and Ocean Sciences, University of North Carolina Wilmington, 601 South College Road, Wilmington, NC 28403-5944, USA
e
Marine Geology Department, Faculty of Marine Science, King Abdulaziz University, P.O. Box 80200, Jeddah 21589, Saudi Arabia
f
Programas de Física y Biologia, Facultad de Ciencias Basicas, Universidad del Atlantico, Barranquilla, Atlantico, Colombia

article info a b s t r a c t

Article history: The Ash Shu’aybah and Al Mujayrimah coast area in Saudi Arabia is threatened by constantly changing
Received 12 January 2023 shoreline dynamics. This work explains how to appropriately quantify shoreline change rates using the
Received in revised form 15 July 2023 Landsat and Sentinel-2 datasets, predict the coastline after twenty years (2022–2042) and compare the
Accepted 21 July 2023
results from both satellite data. The technique used depends on the integration of remote sensing data
Available online 25 July 2023
and Geographic Information Systems (GIS) with a Digital Shoreline Analysis System tool (DSAS). For
Keywords: calculating shoreline change rates and predicting future shoreline positions in 2042, eight Landsat
Shoreline dynamics satellite images over 36 years from 1986 to 2022 were analysed along with four Sentinel-2 satellite
Sentinel-2 images from 2015 to 2022. Based on the LRR models’ data, the shoreline is exposed to erosion with
Landsat average values of −3.25 and −2.12 m/yr for the Sentinel-2 and the Landsat datasets, respectively.
Shoreline changes Segment 2 in the central part of the coastline is the area most affected by the dynamics of the
Shoreline predication shoreline. To assess the accuracy of the obtained results, the root mean square error (RMSE), the
Statistical approach
mean annual error (MAE) and coefficient of determination (R2 ) were used. The results indicated that
both the Sentinel-2 and the Landsat datasets can reliably calculate the shoreline change rate of the
study area based on the similar patterns of their data, although the Sentinel-2 had a somewhat better
ability in calculating the change of the shoreline.
© 2023 Elsevier B.V. All rights reserved.

1. Introduction more than 80% of the world’s coastal areas are exposed to ero-
sion. Consequently, detecting the geometry and the position of
Coastal shorelines are presently undergoing intensified nat- the shoreline consistently has become an absolute necessity for
ural and anthropogenic activities (e.g., waves, tides, winds, pe- understanding coastal line dynamics and continued management
riodic storms, phenomena associated with climate change, and of coastal areas. The position of the shoreline provides an index
the geomorphic processes of erosion and accretion and resource for coastal erosion or accumulation (Ruiz-Beltran et al., 2019;
exploitation). Shorelines undergo three processes over time: ero- Abou Samra and Ali, 2021; Mishra et al., 2022; Nath et al., 2022).
sion, deposition, and stability (Shin and Kim, 2015; Ratnayake Typically, the shoreline is the physical boundary between
et al., 2018a,b; Palamakumbure et al., 2020). Erosion is a constant land and water (Boak and Turner, 2005). The shoreline position
issue that affects coastal populations and the economic progress changes constantly, so precisely detecting the movement of the
for many countries; according to (Appeaning Addo et al., 2008) shoreline is considered a particularly important investigation.
The spatiotemporal variation of shoreline can be delineated by
many approaches such as the traditional field survey and the
∗ Corresponding author.
integration method between remote sensing and Geographic
E-mail addresses: oaoharbi@uqu.edu.sa (O.A. Alharbi),
Information Systems (GIS) (Dolan et al., 1980; Masria et al., 2015;
samia.samir@science.tanta.edu.eg (S.S. Hasan), amr_shaban@science.tanta.edu.eg
(A.S. Fahil), amannaa@kau.edu.sa (A. Mannaa),
Salghuna and Bharathvaj, 2015; Nandi et al., 2016; Nassar et al.,
nelsonrangel@mail.uniatlantico.edu.co (N. Rangel-Buitrago), 2019; N. Apostolopoulos and G. Nikolakopoulos, 2020; Dey et al.,
afalqurashi@uqu.edu.sa (A.F. Alqurashi). 2021; Parthasarathy and Deka, 2021; Aladwani, 2022; Ankrah

https://doi.org/10.1016/j.rsma.2023.103118
2352-4855/© 2023 Elsevier B.V. All rights reserved.
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Fig. 1. Location of the area under investigation.

et al., 2022; Rahman et al., 2022). The traditional field survey services (https://www.stats.gov.sa/). This research was carried
approach has many disadvantages including extensive labour, out to explain shoreline dynamics and predict the coastline after
the difficulty – or perhaps impossibility – of obtaining data over twenty years (2022–2042) for the purpose of providing a guide to
longer periods of time, and the cost required to complete the sur- better management practices and future economic development
vey. These disadvantages are magnified for large research areas, along the Ash Shu’aybah and Al Mujayrimah coast in South Jed-
such as the current study area. In contrast, the remote sensing dah, Saudi Arabia. The DSAS 5.0 and ArcGIS10.8 software were
technique provides universal continuous coverage and multispec- used to process and analyse the Landsat and Sentinel-2 imagery.
tral satellite data, indicating that it is a much more powerful
tool in monitoring and detecting shoreline changes. Also, remote 2. Materials and methods
sensing is linked to some applications that are more powerful for
analysing shoreline change on ArcGIS and support researchers 2.1. Study area
with remarkable findings for shoreline change (Liu et al., 2012;
Almonacid-Caballer et al., 2016; Chen et al., 2019; Gunasinghe
The investigated study area is located along the Ash Shu’aybah
et al., 2021; Apostolopoulos and Nikolakopoulos, 2022; Siyal et al.,
and Al Mujayrimah coastline and stretches 73 km, between lati-
2022; Weerasingha and Ratnayake, 2022). Additionally, the avail-
tudes 20◦ 16′ 9N to 20◦ 43′ 40N and longitudes 39◦ 29′ 23E to
ability of satellite images increases the use of remote sensing and
40◦ 5′ 46E (Fig. 1). The area under investigation includes a variety
GIS techniques for monitoring and assessing shoreline positions.
of characteristics and landmarks, such as sabkhas, lagoons, tidal
The Digital Shoreline Analysis System (DSAS) is a GIS tool that
flats, coral reefs and sand dunes. The quaternary sedimentary rock
evaluates shoreline position and predicts shoreline patterns by
deposits connect extensively throughout the area so they may be
calculating the rate of change statistics for a series of shoreline
positions over time. The data enables researchers to observe considered united beneath their textural overlays. The numerous
and assess the nature of shoreline dynamics and any changing sand dunes and sheets present in this study zone represent the
patterns (Thieler et al., 2009; Bera and Maiti, 2019; Gunasinghe many aeolian deposits that can be found west of the basement
et al., 2021; Weerasingha and Ratnayake, 2022). hill, stretching out to the far coast of the Red Sea. In local
The Ash Shu’aybah and Al Mujayrimah coastline it is located wadis, alluvial deposits are starkly apparent as terraces in both
about 118 km to the south of Jeddah city in the Kingdom of the water and the banks that surround these rivers. There are
Saudi Arabia. This area is distinguished by the presence of many mainly fining-upward channel-lag gravel cycles in these deposits,
development projects, such as the water desalination plant and which are covered with thin layers of sands, silts and clays.
electric company. The water desalination plant of Ash Shu’aybah Sabkha deposits are flat intertidal deposits that are formed during
has gone through many stages and expansions on the shoreline. high tides as the sea rises. The lagoonal faces below them are
Stage one involved the plant being operated in 1989 with an divided from the intertidal-supratidal deposits above by saline
estimated cost of $1.06 billion. Stage two involved the plant being - rich sabkha deposits, which are particularly prevalent in this
operated in 2002, and stage three included the plant operating in region (Hakami and Seif, 2019). Finally, reefal limestone deposits
2009. The final stage, which is stage four, began when the plant create huge, cavernous systems that may house a range of marine
was operated in 2020. Also, the concerned area is considered an flora and fauna close to the coast of this research region. These
attractive place for tourism and national heritage, and the govern- systems form into lengthy, discontinuous belts of porous corals
ment now is providing the village with many basic development and molluscs over the Red Sea floor (Aqeel, 2016).
2
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Fig. 2. Physiographic layers obtained from remote sensing multi–spectral and elevation data. (a) Surface elevation and drainage network, (b) Surface lineaments for
the area under investigation.

As shown in Fig. 2a, the DEM of the interested area was In this study, lineaments were extracted using Google Earth Pro
extracted using data from the Shuttle Radar Topography Mission by visual interpretation and on-screen digitization of accentuated
(SRTM) with a resolution of 30 m (SRTM 1 arc-second for global topography imaged in an eight-direction summed hill shade using
coverage) (https://earthexplorer.usgs.gov/). The surface elevation SRTM/DEM and very high-resolution imagery. Using ArcGIS, the
of the study area ranges from −22 m by the shoreline to 964 m in density kernel analysis was used to calculate the total length
the northeast. The SRTM/DEM data was used in the ArcGIS 10.8 of lineaments per unit area for the entire area. Remote sensing
hydrology tool to create a drainage map of the concerned area data revealed lineament tendencies in the northwest and north-
(Fig. 2a). Drainage systems adhere to water flow and network east that correspond to mapped Precambrian faults, while the
connectivity patterns primarily driven by geological structures, predominant trend of lineaments is toward the northwest (Fig. 2).
surface processes, topographical characteristics and rainfall be-
haviour. The usual stable condition of the terrain surface on which
2.2. Multi–spectral data
they evolved is frequently reflected in these patterns. Since the
advent of remote sensing methods, previously unseen terrain
features may now be observed by satellites with a range of optical Landsat data was acquired from the United States Geological
characteristics and image resolutions. In this work, three different Survey (USGS) and was utilized to measure shoreline change in
types of satellite images were used (Landsat 8 OLI, Sentinel-2 the study area (https://earthexplorer.usgs.gov/). This study uses
and STRM), revealing distinct drainage patterns and anomalies in Landsat TM, ETM multi–temporal satellite data and OLI (Oper-
the area of the study and its surroundings. The Ash Shu’aybah ational Land Imager) to cover the Ash Shu’aybah and Al Mu-
and Al Mujayrimah region drains into the Red Sea in a south- jayrimah coast for the years 1986, 1992, 1997, 2002, 2007, 2012,
western direction after being recharged at the height of 964 m. 2017 and 2022. The selected satellite images were orthorectified
3
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

using polynomial georectification methods to detect shoreline 2.5. Shoreline uncertainty

changes along the Ash Shu’aybah and Al Mujayrimah coast. In
Sentinel-2, the image covers a wide area at high resolution. There It is essential to estimate shoreline errors and uncertainties ac-
are 13 spectral bands sampled by this optical instrument: four curately (Kankara et al., 2015), but there are several factors have
at 10 m, six at 20 m, and three at 60 m. Sentinel-2 data was an impact on the shoreline shift estimation’s accuracy. Hapke
downloaded from the Copernicus Open Access Hub. The multi– et al. (2011) evaluated four uncertainty terms. As mentioned
spectral imaging mission of Sentinel-2 covered the shoreline of before, the satellite images are subjected to geometric correction
the Ash Shu’aybah and Al Mujayrimah region between 2015, so georeferencing uncertainty (Ug) of the Landsat and Sentinel-2
2017, 2019 and 2022. imagery are assigned 0. In the case of imagery, the pixel un-
certainty (Up) depends on the image’s pixel resolution. Landsat
2.3. Image processing satellite images have a 30 m resolution, whereas the Sentinel-
2 images have 10 m resolution. ‘‘Digitising uncertainty’’ (Ud)
The principal image processing methods applied in this ap- describes the degree of uncertainty associated with digitising
proach were gap filling, geometric rectifying and radiometric shorelines. The digitising uncertainty (Ud) value is assigned 0
correction. All images were checked against the striping (Bustil- as the extracted shorelines are verified using Google Earth Pro.
los, 2012) modelling was used, and Python algorithms in ArcGIS To assess uncertainty due to tidal variation (Ut), the tide range
10.8 filled any gaps. ENVI 5.5.1 software implements geometric is obtained from nearest station (https://www.gasgi.gov.sa/). The
adjustments to eliminate oblique scaling and lens deformities. All tide range within the study area is 0.24 to 1.02 m. Islam and
satellite images captured in Level 1 are rectified to the projected Crawford (2022) reported a 1 m tidal variance, which causes a
coordinate system (WGS 1984 - UTM - Zone 37 N). To recover 7 m shoreline shift. Each shoreline’s total uncertainty value was
surface reflectance data, the dark subtraction atmospheric cor- calculated according to Eq. (1). The total uncertainty of Landsat
rection procedure is employed for all datasets (Chavez, 1988). data ranges from 30 to 30.8 with an average value of 30.4. The
This approach is preferred for satellite imagery with large water total uncertainty of Sentinel-2 dataset varies from 10.14 to 12.2
surfaces to correct atmospheric inaccuracies, which are caused with an average value of 11.17.
by water vapour and other airborne particles in the atmosphere. √
Ut = Ug 2 + Up2 + Ud2 + Ut 2 (1)
The radiometric calibration tool in ENVI v.5.5.1 is applied to
radiometrically calibrate the original satellite data to the top of
atmosphere reflectance. Next, the dark subtraction tool in ENVI 2.6. Shoreline analysis and forecasting
is used to process the calibrated data, resulting in an accurate
assessment of the Earth’s surface excluding errors induced by The DSAS approach is an extension in ArcGIS, established
weather and light (Maanan et al., 2014; Vermote et al., 2016; by the US Geological Survey that calculates rates of shoreline
Ratnayake et al., 2018a,b). The georeferenced imagery is then change based on measured differences between shoreline loca-
transferred to ArcGIS 10.8, where it is cropped to border the tions relative to a specific time period (Thieler et al., 2009; Nandi
research region and vectorised shoreline. et al., 2016; Weerasingha and Ratnayake, 2022) DSAS allows
the following statistical measures: (a) Shoreline Change Envelope
2.4. Shoreline digitising and extraction (SCE): considers all available shoreline positions and measures
the highest distance among all the available shorelines imported
Coastline location is constantly changing due to natural and in the geodatabase. (b) Net Shoreline Movement (NSM): the dis-
anthropogenic causes, as mentioned above. The coastline repre- tance between the oldest and youngest shorelines. (c) End Point
sents the high-water line as observed by GPS devices in kinematic Rate (EPR): the distance travelled by the shoreline divided by
mode (Moore, 2000). Conversely, since the land-water boundary the elapsed time between the oldest and the youngest shoreline
is covered by water-saturated zones, automatic coastline deter- positions. (d) Linear Regression Rate (LRR): the statistical rate of
mination from moderate and high satellite imagery is a com- change by fitting the least squares regression to all shorelines of
plex process (Maiti and Bhattacharya, 2009). Various approaches a given cross-section. All of these parameters can be plotted in
have been created to determine the shore-water interface such coastal change statistics to show the spatial evolution of change.
as the Normalized Difference Water Index (NDWI), the Modi- Based on 193 transect lines intersecting eight extracted shore-
fied Normalized Difference Water Index (MNDWI), the Normal- lines from Landsat and four extracted shorelines from Sentinel-2,
ized Difference Vegetation Index (NDVI) and the Tasseled Can we calculated shoreline change rate results and statistics. The
Transformation (McFEETERS, 1996; Rasuly et al., 2010; Bera and interval was 400 m to cover the length of the coast (73 km).
Maiti, 2019; Chen et al., 2019). However, the disadvantage of In this investigation, the DSAS 5.0 beta forecast technique was
these semi-automatic approaches is that they are created using used. This technique is based on the Kalman filter, and combines
high expensive high-resolution multispectral images, particularly the current shoreline positions with predicted future coastline
in large areas (Apostolopoulos et al., 2023). The shoreline extrac- positions. This tool was used to estimate the future shoreline
tions in this study depend on the fact that water absorbs most (N. Apostolopoulos and G. Nikolakopoulos, 2020; Apostolopoulos
of the radiation in the near infrared (NIR) and mid infrared (MIR) et al., 2022). To more precisely anticipate the position of the
regions of the spectrum. So, at these wavelengths, the reflectivity approaching shoreline, the Kalman filter uses of a set of LRR
of water is almost zero. At the same time, the reflectivity of values that were statistically calculated (Long and Plant, 2012;
different land covers in both regions is higher than that of water. Chrisben Sam and Gurugnanam, 2022). A coastline prediction has
Consequently, the coastline can be extracted from the single been produced for 2042.
- band image because water reflectivity is almost zero in the
infrared reflection band. Then, a binary image can be acquired 2.7. Accuracy assessment
simply by thresholding the histogram of band 5 of the infrared
bands from a TM or ETM image, band 10 or 11 from an OLI image The Python scikit-learn package was used in this study to
and band 8 from Sentinel-2 imagery (Alesheikh et al., 2007). The correlate the findings from the DSAS approach (Armstrong and
extracted shorelines are compared to images from Google Earth Collopy, 1992; Pedregosa et al., 2011; Wan et al., 2018; Apos-
Pro to verify and confirm. tolopoulos and Nikolakopoulos, 2022). The root mean square
4
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Fig. 3. Multi–temporal shoreline sites extracted from the Landsat data over 36 years (1986–2022).

error (RMSE) and the mean annual error (MAE) were used to between LRRs’ Landsat and Sentinel-2 outputs. It is recommended
evaluate the results of Landsat and Sentinel-2 dataset. The calcu- to use MAE to detect the variations of errors in the results. The
lations of these metrics are referred by Chai and Draxler (2014). findings of both satellite dataset are useful when the values of
RMSE is a standard metric is used to quantify the differences RMSE and MAE close to 0.
5
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Fig. 4. Multi–temporal shoreline sites extracted from the Sentinel-2 data over seven years (2015–2022).

3. Results Fig. 5 shows the shoreline change rate based on the LRR
for both the Sentinel-2 and the Landsat datasets. According to
DSAS was originally used to compute shoreline movement and Sentinel-2 data, LRR values range from −9.3 to 2.58 m/yr, with
changes. As shown in Fig. 3, the shoreline positions were gen- an average value of −3.25 m/yr (Table 1). About 75.41% (54.9
erated from the Landsat dataset from eight historical shoreline km) of the total length of the coastal line is displaying signs of
positions for the years: 1986, 1992, 1997, 2002, 2007, 2012, 2017 erosion. Segment 2 in the central zone was exposed to a very
and 2022. Fig. 4 shows the shoreline positions extracted from the high level of erosion (Table 2), while the lowest rate of erosion
Sentinel-2 dataset from 2015, 2017, 2019 and 2022. was randomly distributed in Segment 1 and segment 3, varying
6
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Fig. 5. Shoreline change rates by the LRR model, (a) Sentinel-2 dataset and (b) Landsat dataset.

Table 1 Table 2
Statistical summary of shoreline changes rates by the EPR and LRR models. Shoreline classification based on EPR and LRR (Natesan et al., 2015).
LRR EPR Category Rate of shoreline (m/yr) Shoreline classification
Sentinel-2 Landsat Sentinel-2 Landsat 1 >−2 Very high erosion
Average rate (m/yr) −3.25 −2.12 −3.68 −1.98 2 >−1 and <−2 High erosion
Percent of erosional 75.41% 89.07% 80.22% 87.43% 3 >0 and <−1 Moderate erosion
Maximum erosion (m/yr) −9.3 −10.09 −12.02 −14.59 4 0 Stable
Percent of accretional 24.59% 10.93% 19.78% 12.57% 5 >0 and <+1 Moderate accretion
Maximum accretion (m/yr) 2.58 2.97 2.53 6.22 6 >+1 and <+2 High accretion
7 >+2 Very high accretion
SCE (m) NSM (m)
Sentinel-2 Landsat Sentinel-2 Landsat
Minimum 0.2 39.68 −80.09 −523.68
Maximum 298.3 830.1 17.3 223.13 with some variation in the rate. Fig. 7a illustrates the relationship
Average 28.8 154.6 −12.7 −63.3 between the values of the LRRs for the Sentinel-2 and Landsat
datasets. Most of the points are plotted around the 1:1 line with
coefficient of determination (R2 ) value equals 0.52, indicating a
moderate relationship between the obtained results from Land-
from −2 to 0 m/yr (Fig. 5a). On the other hand, accretion areas
sat and Sentinel-2 imagery. For accurate differentiation between
could be characterised by their short length and were found in the Sentinel-2 and Landsat results, a statistical comparison was
Segment 1 and 3, interspersed with erosion areas (Fig. 6a). These conducted using the LRR residuals’ value. The LRR residuals are
segments are characterized by high to very high accretion rate the difference between the LRR of Sentinel-2 and the LRR of
representing 24.59% (18.1 km) of the shoreline. Landsat. The calibration residuals histogram (Fig. 7b) displays
Based on the Landsat dataset, the LRR varies from −10.09 to the frequency of LRRs, for specified intervals of the normalized
2.97 m/yr, with an average value of −2.12 m/yr (Table 1). Erosion calibration residual values. About 48.8% of the LRR residuals are 0
areas were found in Segment 2 in the central portion of the to ±1 m/yr, proving that the results from both satellite datasets
shoreline, with 89.07% (64.9 km) of the total shoreline. Segment were similar. Considering the input uncertainty value, the total
2 is exposed to high and very high erosional rates (Fig. 5b, shoreline has LRR residuals ±4.18 m/yr, which affirms that the
Table 2). Segments 1 and 3 in the northern and southern areas, results of either Landsat or Sentinel-2 can be relied upon for the
respectively, show the lowest values of erosion interspersed with concerning shoreline.
some accumulation transects (Fig. 5b). The depositional transects Table 3 describes the specific value for the two criteria of
represent 8.1 km of the total length of the shoreline. As men- the LRR values of Landsat and Sentinel-2 results. Generally, the
tioned before, the accretional areas occupied small areas of the calculation showed that the RMSE and MAE values are close to
northern and southern regions of the coast. each other. As shown in Table 3, the LRR of Sentinel-2 has a
Additionally, Fig. 6 presents the calculated LRR values accom- RMSE value of 3.41713 m/yr and a MAE value of 1.52872 m/yr.
panied by the confidence band for each segment’s LRR values. The Landsat LRR has 3.92567 m/yr RMSE value and 1.93825 m/yr
The positive LRR values in Fig. 6 illustrate accumulation (above MAE value, indicating that Sentinel-2 data has a relatively better
the zero-change line), and the negative values indicate erosion generalization capability in this research.
(under the zero-change line). The results of both the Sentinel- Fig. 8 shows the evolution of the shoreline change rate with
2 and Landsat data showed that almost the whole coastal area regard to EPRR. Positive values reflect seaward migration, while
is subjected to erosion. The LRR change rate lines of Sentinel-2 negative values indicate landward recession. EPR values range
and Landsat reflect almost the same process for each segment, from −12.02 to 2.53 m/yr, with a mean of −3.68 m/yr and −14.59
7
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Fig. 6. LRR values for each segment with the confidence band (LCI90%) of Landsat and Sentinel-2 data. (a) Segment 1 (b) Segment 2 (c) Segment 3.

Fig. 7. (a) Correlation between LRR (m/yr) of the Landsat dataset and Sentinel-2 dataset in conjugation with coefficient of determination (R2 ) and (b) Histogram
representing the frequency range of LRR’s residuals.

Table 3 central sections of the shore (Segment 2). Fig. 10 shows the
RMSE (m/yr) and MAE values (m/yr) of the LRRs’ Sentinel-2 and Landsat distances between the oldest and the newest shoreline based on
values.
the Sentinel-2 and Landsat data. According to Sentinel-2, NSM
RMSE MAE
ranges from −80.09 to 17.3 m, with an average of −12.7 (Table 1).
LRR of Sentinel-2 data 3.41713 1.52872 The NSM Landsat results vary from −523.68 to 223.13 m, with a
LRR of Landsat data 3.92567 1.93825
mean of −63.3 m (Table 1). Segment 2 in the central portion of
the coast had the greatest values of NSM for both the Sentinel-2
data and Landsat data, as shown in Fig. 10 a and b, respectively.
to 6.22 m/yr with an average value of −1.98 for Sentinel-2 and The shoreline position was predicted for 2042 using the
Landsat, respectively (Table 1). The shoreline with the highest Kalman filter model in the beta forecast. Fig. 11a displays the
level of erosion was in Segment 2, according to both satellites shoreline forecast for the year 2042 based on the Sentinel-2 data
from 2015–2022. Fig. 11b shows the shoreline position for the
(Table 2). Further, Segments 1 and 3 were occupied with high
year 2042 according to 36 years of Landsat’s data (1986–2022).
to moderate erosion transects scattered with accretion transects.
By examining the predicted shoreline, it was found that most of
According to the Shoreline Change Envelope (SCE), the results
the coastline is exposed to erosional processes based on the his-
of the Sentinel-2 data showed that the shoreline movement dis- torical shoreline dynamics. Segments 1 and 3 experienced erosion
tances varied from 0.2 to 298.3 m, with an average value of and accretion processes. Segment 2 of the shoreline was highly
28.8 m (Table 1). The Landsat data displayed distances ranging exposed to erosion, which will affect any development projects
from 39.68 to 830.1 m, with an average value of 154.6 m (Ta- in this region. Therefore, governmental bodies and policymakers
ble 1). Fig. 9 displays that the greatest movement of the shoreline must pay more attention to this region, and develop plans to
from both the Sentinel-2 and Landsat results was found in the protect and manage the shoreline in regards to erosion.
8
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Fig. 8. Shoreline change rates by EPR model with the shoreline segments plotted, (a) Sentinel-2 dataset and (b) Landsat dataset.

Fig. 9. Shoreline change rates by the SCE model, (a) Sentinel-2 dataset and (b) Landsat dataset.

4. Discussion maximum erosion and accretion rates reaching –4.30 m/yr (Al-
harbi, 2020). There are numerous factors contributing to changes
In the present study, we investigate the Ash Shu’aybah and in the shoreline shape of the Red Sea, such as shallow tidal
flats, distributary channels of wadies, sabkhas, sharms, khors, and
Al Mujayrimah coastline to trace the dynamic behaviour of the
lagoons. The Strait of Bab al-Mandab connects the Red Sea to
shoreline induced by erosion and accretion processes over time.
the Indian Ocean, which replenishes waters through a unique
Analysis of remote sensing data revealed that there is signifi-
process of water exchange (Al-Zubieri et al., 2020). Furthermore,
cant erosion along the shoreline with an average rate of −2.685 shoreline changes were linked to a variety of causes, including
m/y. Also, the shoreline position will continue to shift land- strong wave action, silt drifts along the coast caused by long
ward, according to the prediction model. The eastern coast of shore currents, and severe storms (Amalan et al., 2018; Gunas-
the Red Sea, particularly the Ash Shuqayq, Turfah, and Jizan inghe et al., 2022). Regarding monsoons, the Red Sea is subjected
areas, have undergone similar changes and have had erosion rates to two main wind systems. During the winter, northwesterly
of −0.59, −1.80, and −3.53 m/yr, respectively (Alharbi et al., winds associated with storms in the Mediterranean Sea affect
2017). The Rabigh Coast of Saudi Arabia, has been subjected to the northern part of coast (up to 20 m/s). The winter monsoon
9
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Fig. 10. Shoreline change rates by NSM model, (a) Sentinel-2 dataset and (b) Landsat dataset.

Fig. 11. Spatial distributions of the shoreline in 2042 based on (a) Sentinel-2 and (b) Landsat data.

event in the Arabian Sea is associated with southeasterly winds 5. Conclusions

that are present mainly in the southern part of the Red Sea
coast (Langodan et al., 2017). Because of these factors, coastal This research aimed to determine the shoreline change rate
zones need more in-depth monitoring and management. Many and predict the location of the Ash Shu’aybah and Al Mujayrimah
researchers have attempted to create strategies regarding coastal coastline in the Eastern Red Sea, Saudi Arabia, after twenty years
(2022–2042). Satellite images derived from Landsat and Sentinel-
area management e.g. Christie and White (1997), Cetin (2016),
2 data were used, and the results of both datasets were compared.
Ouellette and Getinet (2016), Balasuriya (2018) and Ratnayake
As a result, the DSAS v5.0 tool, combined with ArcGIS10.8 and
et al. (2018a,b). Ratnayake and Perera (2022) proposed a frame-
satellite images, was implemented to accomplish these goals. The
work for integrated coastal zone management in order to improve shoreline change rate was determined using the results of the
the quality of decisions being made in coastal zones. Coastal en- SCE, NSM, EPR and LRR models. The results of the Sentinel-2
vironment components are categorized into four administrative and Landsat data agreed that the shoreline (75.41% and 89.07%,
levels: national, district, divisional, and village. This policy seeks respectively) is subjected primarily to erosion with an average
to conserve nature and manage the coastal areas sustainably. value −3.25 m/yr and −2.12 m/yr, respectively Segment 2 in
10
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

the central area specifically is exposed to very high erosion. Apostolopoulos, D.N., et al., 2023. Monitoring the Kotychi Lagoon in Western
Various statistical approaches are used to validate the findings of Peloponnese, Greece, using remote sensing techniques and environmental
assessment. J. Mar. Sci. Eng. 11 (2), 411, Available at: https://doi.org/10.3390/
Landsat and Sentinel-2 data. The shoreline patterns and statistics
jmse11020411.
of both the satellites data reveal a correspondence, indicating Appeaning Addo, K., Walkden, M., Mills, J.P., 2008. Detection, measurement and
that both Sentinel-2 and Landsat datasets are suitable and trust- prediction of shoreline recession in Accra, Ghana. ISPRS J. Photogramm. Re-
worthy resources for predicting coastlines. Finally, the results of mote Sens. 63 (5), 543–558, Available at: https://doi.org/10.1016/j.isprsjprs.
2008.04.001.
this approach emphasised the capacity of multi-spectral satellite
Aqeel, A., 2016. Investigation of expansive soils in Obhor Sabkha, Jeddah-Saudi
imagery, geospatial technology techniques and the DSAS tool in Arabia. Arab. J. Geosci. 9 (4), 314, Available at: https://doi.org/10.1007/
providing critical data on coastal morphodynamics and coastal s12517-016-2341-x.
zone administration. Armstrong, J.S., Collopy, F., 1992. Error measures for generalizing about forecast-
ing methods: Empirical comparisons. Int. J. Forecast. 8 (1), 69–80, Available
at: https://doi.org/10.1016/0169-2070(92)90008-W.
CRediT authorship contribution statement Balasuriya, A., 2018. Chapter 25 - coastal area management: Biodiversity and
ecological sustainability in Sri Lankan perspective. In: Sivaperuman, C., et
al. (Eds.), Biodiversity and Climate Change Adaptation in Tropical IslandS.
Omar A. Alharbi: Proposal of the idea, Conceptualization, Data
Academic Press, pp. 701–724, Available at: https://doi.org/10.1016/B978-0-
analysis, Writing – original draft, Writing – review & editing. 12-813064-3.00025-9.
Samia S. Hasan: Proposal of the idea, Conceptualization, Data Bera, R., Maiti, R., 2019. Quantitative analysis of erosion and accretion (1975–
analysis, Writing – original draft, Writing – review & editing. 2017) using DSAS — A study on Indian sundarbans. Reg. Stud. Mar. Sci. 28,
100583, Available at: https://doi.org/10.1016/j.rsma.2019.100583.
Amr S. Fahil: Proposal of the idea, Conceptualization, Data anal-
Boak, E.H., Turner, I.L., 2005. Shoreline definition and detection: A review. J.
ysis, Writing – original draft, Writing – review & editing. Am- Coast. Res. 2005 (214), 688–703, Available at: https://doi.org/10.2112/03-
mar Mannaa: Reviewing the article & editing. Nelson Rangel- 0071.1.
Buitrago: Reviewing the article & editing. Abdullah F. Alqurashi: Bustillos, L.V., 2012. Gap fill for landsat 7, images - a correction of SLC-off.
Cetin, M., 2016. Sustainability of urban coastal area management: A case study
Reviewing the article & editing.
on Cide. J. Sustain. For. 35 (7), 527–541, Available at: https://doi.org/10.1080/
10549811.2016.1228072.
Declaration of competing interest Chai, T., Draxler, R.R., 2014. Root mean square error (RMSE) or mean absolute
error (MAE)? – Arguments against avoiding RMSE in the literature. Geosci.
Model Dev. 7 (3), 1247–1250, Available at: https://doi.org/10.5194/gmd-7-
The authors declares that there are no potential conflicts of 1247-2014.
interest. Chavez, S., 1988. An improved dark-object subtraction technique for atmospheric
scattering correction of multispectral data. Remote Sens. Environ. 24 (3),
459–479, Available at: http://dx.doi.org/10.1016/0034-4257(88)90019-3.
Data availability Chen, C., et al., 2019. Coastline information extraction based on the tasseled
cap transformation of Landsat-8, OLI images. Estuar. Coast. Shelf Sci. 217,
Data will be made available on request. 281–291, Available at: https://doi.org/10.1016/j.ecss.2018.10.021.
Chrisben Sam, S., Gurugnanam, B., 2022. Coastal transgression and regression
from 1980 to 2020 and shoreline forecasting for 2030 and 2040, using DSAS
References along the southern coastal tip of Peninsular India. Geodesy Geodyn. 13 (6),
585–594, Available at: https://doi.org/10.1016/j.geog.2022.04.004.
Abou Samra, R.M., Ali, R.R., 2021. Applying DSAS tool to detect coastal changes Christie, P., White, A.T., 1997. Trends in development of coastal area man-
along Nile Delta, Egypt. Egypt. J. Remote Sens. Space Sci. 24 (3), 463–470, agement in tropical countries: From central to community orientation.
Available at: https://doi.org/10.1016/j.ejrs.2020.11.002. Coast. Manage. 25 (2), 155–181, Available at: https://doi.org/10.1080/
Al-Zubieri, A., et al., 2020. Shoreline evolution between Al Lith and Ras Mahāsin 08920759709362316.
on the Red Sea Coast, Saudi Arabia Using GIS and DSAS Techniques. J. Indian Dey, M., S, S.P., Jena, B.K., 2021. A shoreline change detection (2012–2021) and
Soc. Remote Sens. 48, 1_16, Available at: https://doi.org/10.1007/s12524- forecasting using digital shoreline analysis system (DSAS) tool: A case study
020-01169-6. of Dahej Coast, Gulf of Khambhat, Gujarat, India. Indones. J. Geogr. 53 (2),
Aladwani, N.S., 2022. Shoreline change rate dynamics analysis and prediction of Available at: https://doi.org/10.22146/ijg.56297.
future positions using satellite imagery for the southern coast of Kuwait: Dolan, R., et al., 1980. The reliability of shoreline change measurements from
A case study. Oceanologia 64 (3), 417–432, Available at: https://doi.org/10. aerial photographs. Shore Beach 48 (4), 22–29.
1016/j.oceano.2022.02.002. Gunasinghe, G.P., et al., 2021. Influence of manmade effects on geomorphology,
Alesheikh, A.A., Ghorbanali, A., Nouri, N., 2007. Coastline change detection using bathymetry and coastal dynamics in a monsoon-affected river outlet in
remote sensing. Int. J. Environ. Sci. Technol. 4 (1), 61–66, Available at: Southwest coast of Sri Lanka. Environ. Earth Sci. 80 (7), 238, Available at:
https://doi.org/10.1007/BF03325962. https://doi.org/10.1007/s12665-021-09555-0.
Gunasinghe, G.P., et al., 2022. Monsoon-driven geomorphological changes along
Alharbi, O., 2020. Shoreline change analysis along the Rabigh coast of Saudi
the West Coast of Sri Lanka: A combined approach utilizing CoastSat and
Arabia, using multi-temporal satellite imagery. J. King Abdulaziz Univ. 30,
Google Earth Engine. Ocean Sci. J. 57 (3), 475–492, Available at: https:
33–57, Available at: https://doi.org/10.4197/Mar.30-2.3.
//doi.org/10.1007/s12601-022-00081-z.
Alharbi, O.A., et al., 2017. Temporal shoreline change and infrastructure influ-
Hakami, B.A., Seif, E.-S.S.A., 2019. Geotechnical aspects and associated problems
ences along the southern Red Sea coast of Saudi Arabia. Arab. J. Geosci. 10
of Al-Shuaiba Lagoon soil, Red Sea coast, Saudi Arabia. Environ. Earth Sci. 78
(16), 360, Available at: https://doi.org/10.1007/s1517-017-3109-7.
(5), 158, Available at: https://doi.org/10.1007/s12665-019-8136-0.
Almonacid-Caballer, J., et al., 2016. Evaluation of annual mean shoreline position Hapke, C., et al., 2011. National Assessment of Shoreline Change: Historical
deduced from Landsat imagery as a mid-term coastal evolution indicator. Shoreline Change Along the New England and Mid-Atlantic Coasts. U.S.
Mar. Geol. 372, 79–88, Available at: https://doi.org/10.1016/j.margeo.2015. Geological Survey, [Preprint]. Available at: https://digitalcommons.usf.edu/
12.015. msc_facpub/2498.
Amalan, K., et al., 2018. Influence of nearshore sediment dynamics on the Islam, M.S., Crawford, T.W., 2022. Assessment of spatio-temporal empirical
distribution of heavy mineral placer deposits in Sri Lanka. Environ. Earth forecasting performance of future shoreline positions. Remote Sens. 14 (24),
Sci. 77 (21), 737, Available at: https://doi.org/10.1007/s12665-018-7914-4. 6364, Available at: https://doi.org/10.3390/rs14246364.
Ankrah, J., Monteiro, A., Madureira, H., 2022. Bibliometric analysis of data sources Kankara, R.S., et al., 2015. Estimation of long and short term shoreline changes
and tools for shoreline change analysis and detection. Sustainability 14 (9), along andhra Pradesh Coast using remote sensing and GIS techniques.
4895, Available at: https://doi.org/10.3390/su14094895. Procedia Eng. 116, 855–862, Available at: https://doi.org/10.1016/j.proeng.
Apostolopoulos, D.N., Avramidis, P., Nikolakopoulos, K.G., 2022. Estimating 2015.08.374.
quantitative morphometric parameters and spatiotemporal evolution of the Langodan, S., et al., 2017. The climatology of the Red Sea – part 1,: the wind.
Prokopos Lagoon using remote sensing techniques. J. Mar. Sci. Eng. 10 (7), Int. J. Climatol. 37 (13), 4509–4517, Available at: https://doi.org/10.1002/joc.
931, Available at: https://doi.org/10.3390/jmse10070931. 5103.
Apostolopoulos, D.N., Nikolakopoulos, K.G., 2022. Spot vs landsat satellite images Liu, Y.Y., et al., 2012. Trend-preserving blending of passive and active microwave
for the evolution of the north Peloponnese coastline, Greece. Reg. Stud. Mar. soil moisture retrievals. Remote Sens. Environ. 123, 280–297, Available at:
Sci. 56, 102691, Available at: https://doi.org/10.1016/j.rsma.2022.102691. https://doi.org/10.1016/j.rse.2012.03.014.

11
O.A. Alharbi, S.S. Hasan, A.S. Fahil et al. Regional Studies in Marine Science 66 (2023) 103118

Long, J.W., Plant, N.G., 2012. Extended Kalman filter framework for forecasting Pedregosa, F., et al., 2011. Scikit-learn: Machine learning in Python. In: Machine
shoreline evolution. Geophys. Res. Lett. 39 (13), Available at: https://doi.org/ Learning in Python. [Preprint].
10.1029/2012GL052180. Rahman, M.K., Crawford, T.W., Islam, M.S., 2022. Shoreline change analysis along
Maanan, Mehdi, et al., 2014. A long-term record of land use change impacts on rivers and deltas: A systematic review and bibliometric analysis of the
sediments in Oualidia lagoon, Morocco. Int. J. Sediment Res. 29 (1), 1–10, shoreline study literature from 2000 to 2021. Geosciences 12 (11), 410,
Available at: http://dx.doi.org/10.1016/S1001-6279(14)60017-2. Available at: https://doi.org/10.3390/geosciences12110410.
Maiti, S., Bhattacharya, A.K., 2009. Shoreline change analysis and its application Rasuly, A., Naghdifar, R., Rasoli, M., 2010. Monitoring of caspian Sea Coastline
to prediction: A remote sensing and statistics based approach. Mar. Geol. changes using object-oriented techniques. Procedia Environ. Sci. 2, 416–426,
257 (1), 11–23, Available at: https://doi.org/10.1016/j.margeo.2008.10.006. Available at: https://doi.org/10.1016/j.proenv.2010.10.046.
Masria, A., et al., 2015. Detection of shoreline and land cover changes around Ratnayake, A.S., Perera, U.L.H.P., 2022. Coastal zone management in Sri Lanka: A
rosetta promontory, Egypt, based on remote sensing analysis. Land 4 (1), lesson after recent naval accidents. Mar. Pollut. Bull. 182, 113994, Available
216–230, Available at: https://doi.org/10.3390/l{and}4010216. at: https://doi.org/10.1016/j.marpolbul.2022.113994.
McFEETERS, S.K., 1996. The use of the normalized difference water index (NDWI) Ratnayake, N.P., Ratnayake, A.S., Azoor, R.M., et al., 2018a. Erosion processes
in the delineation of open water features. Int. J. Remote Sens. 17 (7), driven by monsoon events after a beach nourishment and breakwater
1425–1432, Available at: https://doi.org/10.1080/01431169608948714. construction at Uswetakeiyawa beach, Sri Lanka. SN Appl. Sci. 1 (1), 52,
Mishra, M., et al., 2022. Assessment of impacts to the sequence of the tropical Available at: https://doi.org/10.1007/s42452-018-0050-7.
cyclone Nisarga and monsoon events in shoreline changes and vegetation Ratnayake, N.P., Ratnayake, A.S., Keegle, V., et al., 2018b. An analysis of beach
damage in the coastal zone of Maharashtra, India. Mar. Pollut. Bull. 174, profile changes subsequent to the Colombo Harbor Expansion Project, Sri
113262, Available at: https://doi.org/10.1016/j.marpolbul.2021.113262. Lanka. Environ. Earth Sci. 77 (1), 24, Available at: https://doi.org/10.1007/
Moore, L.J., 2000. Shoreline mapping techniques. J. Coast. Res. 16 (1), 111–124. s12665-018-7234-8.
N. Apostolopoulos, D., G. Nikolakopoulos, K., 2020. Assessment and quantification Ruiz-Beltran, A.P., et al., 2019. Short-term shoreline trend detection patterns
of the accuracy of low- and high-resolution remote sensing data for shoreline using SPOT-5, image fusion in the Northwest of Yucatan, Mexico. Estuar.
monitoring. ISPRS Int. J. Geo-Inf. 9 (6), 391, Available at: https://doi.org/10. Coasts 42 (7), 1761–1773, Available at: https://doi.org/10.1007/s12237-019-
3390/ijgi9060391. 00573-7.
Nandi, S., et al., 2016. Shoreline shifting and its prediction using remote sensing Salghuna, N.N., Bharathvaj, S.A., 2015. Shoreline change analysis for northern
and GIS techniques: a case study of Sagar Island, West Bengal (India). J. part of the coromandel coast. Aquat. Proc. 4, 317–324, Available at: https:
Coast. Conserv. 20 (1), 61–80, Available at: https://doi.org/10.1007/s11852- //doi.org/10.1016/j.aqpro.2015.02.043.
015-0418-4. Shin, B., Kim, K., 2015. Estimation of shoreline change using high resolution
Nassar, K., et al., 2019. Shoreline change detection using DSAS technique: Case of images. Procedia Eng. 116, 994–1001, Available at: https://doi.org/10.1016/j.
North Sinai coast, Egypt. Mar. Georesour. Geotechnol. 37 (1), 81–95, Available proeng.2015.08.391.
at: https://doi.org/10.1080/1064119X.2018.1448912. Siyal, A.A., et al., 2022. Shoreline change assessment of indus delta using GIS-
Natesan, U., et al., 2015. Monitoring longterm shoreline changes along Tamil DSAS and satellite data. Reg. Stud. Mar. Sci. 53, 102405, Available at: https:
Nadu, India using geospatial techniques. Aquat. Proc. 4, 325–332, Available //doi.org/10.1016/j.rsma.2022.102405.
at: https://doi.org/10.1016/j.aqpro.2015.02.044. Thieler, E.R., et al., 2009. The Digital Shoreline Analysis System (DSAS) Version
Nath, A., et al., 2022. Geospatial analysis of short term shoreline change 4.0 - an ArcGIS Extension for Calculating Shoreline Change. In: USGS
behavior between Subarnarekha and rasulpur estuary, east coast of India Numbered Series 2008–1278, U.S. Geological Survey, Reston, VA, Available at:
using intelligent techniques (DSAS). GeoJournal [Preprint] Available at: https: https://doi.org/10.3133/ofr20081278, The Digital Shoreline Analysis System
//doi.org/10.1007/s10708-022-10683-8. (DSAS) Version 4.0 - An ArcGIS extension for calculating shoreline change.
Ouellette, W., Getinet, W., 2016. Remote sensing for marine spatial planning and Vermote, E., et al., 2016. Preliminary analysis of the performance of the landsat
Integrated Coastal Areas management: Achievements, challenges, opportu- 8,/OLI, land surface reflectance product. Remote Sens. Environ. 185, 46–56,
nities and future prospects. Remote Sens. Appl.: Soc. Environ. 4, 138–157, Available at: https://doi.org/10.1016/j.rse.2016.04.008.
Available at: https://doi.org/10.1016/j.rsase.2016.07.003. Wan, R., et al., 2018. Modeling wetland aboveground biomass in the Poyang Lake
Palamakumbure, L., et al., 2020. Sea-level inundation and risk assessment along National Nature Reserve using machine learning algorithms and Landsat-8,
the south and southwest coasts of Sri Lanka. Geoenviron. Disasters 7 (1), 17, imagery. J. Appl. Remote Sens. 12 (4), 046029, Available at: https://doi.org/
Available at: https://doi.org/10.1186/s40677-020-00154-y. 10.1117/1.JRS.12.046029.
Parthasarathy, K.S.S., Deka, C., 2021. Remote sensing and GIS application in Weerasingha, W.A.D.B., Ratnayake, A.S., 2022. Coastal landform changes on the
assessment of coastal vulnerability and shoreline changes: a review. ISH east coast of Sri Lanka using remote sensing and geographic information
J. Hydraul. Eng. 27 (sup1), 588–600, Available at: https://doi.org/10.1080/ system (GIS) techniques. Remote Sens. Appl.: Soc. Environ. 26, 100763,
09715010.2019.1603086. Available at: https://doi.org/10.1016/j.rsase.2022.100763.

12