Analysis of Coastline Evolution using Landsat and Sentinel 2 Images
from 2001 to 2020 in Callao Bay, Peru
Amanda Muñoz
a
, Luis Mendoza
b
, Emanuel Guzmán
c
and Carmela Ramos
d
Facultad y Escuela de Ingeniería Civil, Universidad Peruana de Ciencias Aplicadas,
Av. Prolongación Primavera 2390, Santiago de Surco, Lima, Peru
Keywords: Coastal Erosion, Coastal Accretion, Satellite Images, Callao Bay.
Abstract: The study presented an analysis of the shoreline evolution in Callao Bay Lima, Perú; which is one of the
most important bays in Perú due to economics and touristic activities. Study areas include La Punta, Callao
and Ventanilla districts with an approximate 32 km of length. The study area was divided into six sectors, and
the analysis was focused mainly from the 2001 to 2020 period, identifying areas affected by coastal erosion
or accretion throughout. Satellite data images were obtained from Landsat (5, 7 and 8) and Sentinel 2, they
were processed to correctly identify the shoreline. Shoreline variations were analyzed using the DSAS (Digital
Shoreline Analysis System) utility, applying a statistical method called "Linear Regression Ratio". Shoreline
variations showed different rates of changes along different sectors of the study area. In general terms, the
accretion or erosion trend in Callao Bay was a low accretion with average rates from 3.77 m / year to 4.20 m
/ year, except in the sector which is closed to the Rímac river with change rate of around 11.85 m/year.
1 INTRODUCTION
The constant growth of the population, the flow of
economic activities and mismanagement of water use,
have directly affected the deterioration of the water
resources (Sánchez, 2019). This is directly related to
the shoreline, which is of vital importance for the
improvement of social, economic, and recreational
opportunities; in other words, it is fundamental for the
development of the economic and natural
environment (Yasir et al., 2020). However, the
hazards affecting the coastal zone have increased over
the years, resulting from rapid changes in various
physical and geological variables that have been
influenced by dynamic coastal processes. Likewise,
the coastal zone and the water quality has undergone
changes due to the influences of anthropogenic
activities such as piers and beach protection structures
(Sheik Mujabar & Chandrasekar, 2013). This
problem is mentioned in a study by Soto (2018),
where he highlights how the northern and central
coast of Perú has been affected by sediments
generated by marine currents and the erosion of river
a
https://orcid.org/0000-0002-3512-587X
b
https://orcid.org/0000-0001-7890-9597
c
https://orcid.org/0000-0001-8381-4509
d
https://orcid.org/0000-0002-4269-2944
basins such as the Rímac and Chillón (Soto, 2018). In
addition, the metropolitan area of Lima has been
affected by constant changes in the shoreline of the
Callao Bay (Guzman et al., 2020); specifically in the
area near the Callao Port Terminal with a rate of
3m/year between the years 1984 to 2016 (Luijendijk
et al., 2018). On the other hand, El Niño, an extreme
factor, has a great influence on the displacement of
the shoreline in Callao Bay (Guzman et al., 2020).
Therefore, the study of the constant change of the
coastline is important so that, based on this, a water
resource management strategy can be developed and
the negative effects, such as chemical and dynamic
imbalance of the coast, loss of coastal biodiversity,
and decrease in gross national product (GDP), can be
avoided (Rangel-Buitrago et al., 2015). Satellite
imagery has been used, to monitor changes along the
coastal zone, because it provides repeatable and
consistent statistics of variations. In addition, the
combination of this methodology with Geographic
Information System (GIS) for monitoring the
evolution of coastlines on a temporal scale, presents
Muñoz, A., Mendoza, L., Guzmán, E. and Ramos, C.
Analysis of Coastline Evolution using Landsat and Sentinel 2 Images from 2001 to 2020 in Callao Bay, Peru.
DOI: 10.5220/0011037500003185
In Proceedings of the 8th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2022), pages 115-122
ISBN: 978-989-758-571-5; ISSN: 2184-500X
Copyright
c
2022 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
115
a digital structure that facilitates the detection of
vulnerable areas (Yasir et al., 2020).
The present study includes the use of Landsat (5,7
and 8) and Sentinel 2 satellite images, to visually
identify land and water coverage, also manually
obtaining the annual shorelines variations from 2001
to 2020. An analysis of the accretion or erosion
tendency of the shoreline will be present through the
application of Digital Shoreline Analysis System
(DSAS) which is a tool that allows statistical methods
such as Linear Regression Ratio (LRR) to be applied
to a set of annual coastlines to determine their
movement and trends.
2 METHODOLOGY
2.1 Study Area
Callao Bay is located on the central coast of Perú and
belongs to the department of Lima and the Callao
Constitutional Province (Figure 1); it also covers La
Punta, Callao and Ventanilla districts. Callao Bay is
one of the most important bays in Perú due to
economics activities that are developed (Callao Port,
fishing industry, touristic activities such fishing and
water sport). One of the special characteristics of the
study area is the presence of the Chillon and Rimac
Rivers mouths and Callao port installations.
Figure 1: a) Peru, b) Lima and c) Districts covered by the
study area.
2.2 Data Sources
Images from Landsat and Sentinel 2 satellites were
used for the development of the study since they
facilitate the remote sensing of the characteristics in
the area to be studied. Both satellite image sources are
freely available and were obtained from the US
Geological Survey’s (USGS) Earth explorer website
(http://earthexplorer.usgs.gov)
To analyze shoreline variations in study area, one
satellite image per year was selected, taking into
consideration that an annual coastline was required,
and the area shows a constant low tide height with an
average of 54 cm per month that was maintained even
during the 2017 El Niño phenomenon
(DIHIDRONAV, 2020)
The area shows a lot of cloud cover, which made
its identification difficult. Therefore, the resolution of
the band sets belonging to each satellite were
important characteristics to take into consideration
(Table 1).
Table 1: Annual satellite image and resolution.
SATELLITE YEARS RESOLUTION
Landsat 7
2001, 2002,
2003, 2004,
2012, 2013,
2014 and 2015
Bands 1 to 7
(30 m/pixel)
Bands 8
(
15 m/
p
ixel
)
Landsat 5
2005, 2006,
2007, 2008,
2009, 2010
and 2011
Bands 1 to 7
(30 m/pixel)
Landsat 8 2020
Bands 1 to 7 and 9
(30 m/pixel)
Bands 8
(15 m/pixel)
Bands 10 and 11
(100 m/pixel)
Sentinel 2
2016, 2017,
2018 and 2019
Bands 1, 9 and 10
(60 m/pixel)
Bands 2 to 4 and 8
(10 m/pixel)
Bands 5 to 7, 8, 11
and 12 (20 m/pixel)
Therefore Table 2 shows the satellite images, the day
they were taken, along with the name and bands that
were used.
2.3 Image Pre-Processing and
Coastline Detection
For this research work, pre-processing steps were
applied, using the "Semi-Automatic Classification"
plugin of the QGIS program, which simplified the
manual extraction of the shoreline.
GISTAM 2022 - 8th International Conference on Geographical Information Systems Theory, Applications and Management
116
Table 2: Characteristics of the satellite images used.
YEAR
CHARACTERISTICS OF THE SATELLITE IMAGES
USED (2001-2020)
SAT. NAME DAY
START
TIME
BANDS
2001 L7
LT05_L1TP_007068_200
11122_20161209_01_T1
11/22 14:50:15
4, 5, 6
and 7
2002 L7
LE07_L1TP_008068_200
21226_20170127_01_T1
12/26 15:05:15
1, 2, 3, 4,
5,6 and 7
2003 L7
LE07_L1TP_008068_200
30401_20170125_01_T1
04/01 15:05:38
4,5,6,7,
and 8
2004 L7
LE07_L1TP_007068_200
40123_20170123_01_T1
01/23 14:59:45
3,4,5,6,7
and 8
2005 L5
LT05_L1TP_008068_200
50430
_
20161126
_
01
_
T1
04/30 15:04:07 5 and 7
2006 L5
LT05_L1TP_008068_200
60503
_
20161122
_
01
_
T1
04/03 15:08:52 5
2007 L5
LT05_L1TP_007068_2007
0328_20161116_01_T1_d
03/28 15:05:50
1, 4, 5
and 7
2008 L5
LT05_L1TP_008068_200
80321_20161101_01_T1
03/21 15:06:48
2, 3, 4, 5
and 7
2009 L5
LT05_L1TP_008068_200
90425
_
20161026
_
01
_
T1
04/25 15:04:18 5 and 7
2010 L5
LT05_L1TP_008068_201
00514
_
20161015
_
01
_
T1
05/14 15:07:42 4, 5 and 7
2011 L5
LT05_L1TP_007068_201
10307
_
20200823
_
02
_
T1
03/7 15:00:40 5, 6 and 7
2012 L7
LE07_L1TP_008068_201
20104_20200909_02_T1
01/04 15:10:56
1, 2, 3,
4,5,6,7
and 8
2013 L7
LE07_L1TP_007068_201
30405_20200907_02_T1
04/05 15:06:47
1, 6, 7
and 8
2014 L7
LE07_L1TP_008068_201
40125_20200906_02_T1
01/25 15:13:41
2, 3, 4, 5,
6 and 7
2015 L7
LE07_L1TP_008068_201
50317_20200905_02_T1
03/17 15:15:54
5, 6, 7
and 8
2016 S2
L1C_T18LTM_A007887_
20161225T152814
12/25 15:28:14
2, 3,4, 8,
11 and 12
2017 S2
L1C_T18LTM_A008416_
20170131T152508
01/31 15:25:08 2 and 3
2018 S2
L1C_T18LTM_A004584_
20180121T151656
01/21 15:16:56
2, 3, 4
and 8
2019 S2
L1C_T18LTM_A018855_
20190131T152414
01/31 15:24:14
2, 3, 4
and 8
2020 L8
LC08_L1TP_007068_202
01126_20210316_02_T1
11/26 15:11:02
1, 2, 6
and 7
2.3.1 Image Pre-processing using QGIS
With the QGIS version 3 plugin, the preprocessing
was performed to achieve the highest sharpness of the
study area comprising the satellite images for the
separation of water and land covers from the annual
images.
A clip was created for all the bands and the
respective atmospheric corrections using the
DarkObject Subtraction 1 (DOS1) method (Luca
Congedo, 2016), to achieve a complete visualization
without noise (visual obstacles in the raster).
Subsequently, the combination of bands was
performed with different resolutions such as
10m/pixel for Sentinel 2 ,15 m/pixel for Landsat 7
and Landsat 8, and 30m/pixel for Landsat 5. This to
achieve a sharper raster that allowed the identification
of the contrast between land and water.
On the other hand, for the years 2004, 2012, 2013
and 2015 only the Landsat 7 satellite image was
available, which since 2002 had a banding error due
to sensor failure. However, a reliability analysis using
a percentage fraction of the parts that are not visible,
due to the band error, with the total length of the zone.
Led to the conclusion that this failure would not
significantly affect the visual identification of the
coastline. Figure 2 shows the corrected satellite
images for the years between 2001 and 2010 and
Figure 3 shows the corrected satellite images for the
years between 2011 and 2020.
Figure 2: Processed satellite images for years 2001 to 2010.
Figure 3: Processed satellite images for years 2011 to 2020.
2.3.2 Shoreline Detection
The detection of the shoreline, belonging to the
Callao Bay, was performed manually, and validated
by importing the polyline into Google Earth where it
was georeferenced by date and UTM coordinates.
The difference in the range of colors, presented by the
satellite images (raster), was considered and in this
way the limit between the white pixels (land covers)
and black pixels (sea covers) was determined. In
addition, the creation of the polyline for the coastline
representation was performed with the multiline tool
and in the most precise way to avoid taking non-
Analysis of Coastline Evolution using Landsat and Sentinel 2 Images from 2001 to 2020 in Callao Bay, Peru
117
coastal structures. To perform the analysis, study area
was divided into 6 sectors (Figure 4).
The creation of these sectors was made
considering the extension of the shoreline; as well as
the presence of structures such as the Perú Port
Terminal and natural elements such as the mouths of
the Rímac and Chillón rivers (Table 3).
2.4 DSAS Applications
After collecting the coastlines, belonging to the
period 2001 to 2020, with QGIS these were exported
in “shape” format to ArcGIS 10.5 and the “Merge”
tool was applied to create a single image that would
comply all the polylines (20 shorelines). Then, the
“Buffer” tool was applied to the compilation of lines
to obtain an equidistant margin at 150 meters.
Afterwards, geographic database file was created,
where two parameters were added for analysis; with
names of Coastal Lines and Baseline and were
georeferenced in the UTM WGS84 18S zone.
The “Digital Shoreline Analysis System” (DSAS)
was used for the analysis of the evolution in the
shoreline. Then, the transects were designed with a
spacing of 20 meters and a maximum distance from
the baseline of 500 meters, considering the sinuosity
of the terrain.
Table 3: Mainly characteristics of sectors analysis in study
area.
Secto
r
Length (km) Observations
1 6.95 Presence of the Perú Port Terminal
2 7.18 Rímac mouth is located
3 8.44 Chillón mouth is located
4 3.49 Start zone of Ventanilla’s wetlands
5 2.72 End zone of Ventanilla’s wetlands
6 3.32 Presence of littoral cliffs
Finally, once the transects were obtained, the data
of the intersections between the shorelines and the
transects were extracted.
The specified distances indicate how many meters
there are between the shorelines of the period and the
baseline that was previously created equidistant from
the union of the shorelines.
In Figure 5 a map with the transects generated for
the shorelines by year and belonging to the period
2001-2020. In this regard, it is important to mention
that, although for greater precision the intersection of
the transects should be avoided, due to the concave or
sinuous shape of the bay, the intersection of the
transects could not be totally reduced. However, the
few areas where these intersections exist, the close up
view 2 of Figure 5, would not be generating a great
impact on the analysis for the proposed study.
Figure 4: Shorelines from 2001 to 2020 for each sector, a) Sector 1, b) Sector 2, c) Sector 3, d) Sector 4, e) Sector 5, f) Sector
6 and g) Sector Overview.
GISTAM 2022 - 8th International Conference on Geographical Information Systems Theory, Applications and Management
118
2.5 Statistical Analysis of Shoreline
Change Rates
After obtaining the transects, a statistical analysis was
performed using LRR method; since one of the
advantages of this method is that it considers all the
shorelines, thus giving a more detailed analysis (Yasir
et al., 2020). For this reason, the DSAS “Calculation
of ratios” tool was used, where the rates of change of
the shoreline were calculated throughout the 20-year
period with the LRR method.
Figure 5: Transects created from the shorelines from 2001
to 2020. In approach 2 it was observed that intersections of
the transects that were created due to the sinuousness of the
land were created; however, its presence did not
significantly affect the calculations performed. Then rates
obtained by the LRR method represented with color gamut.
In Figure 6 the different trends for each sector are
shown. Likewise, the transects (132 to transect 317)
that belonged to the “Terminal Portuario del Perú”
were identified, since these transects would not enter
the statistical analysis.
To perform the analysis of the variation of the
shoreline, the distance data between the shorelines
and the baseline extracted from each transect,
manually taking 100% of the rates. A subtraction was
performed between the total distances of the
consecutive years, to then average the data and obtain
trends for each sector created.
Figure 6: LRR method in the sectorization of the Bay of
Callao and detail of sectors.
3 RESULTS
Table 4 shows a summary of the average rates of
sedimentation, erosion, maximum and minimum,
taking 95% of the significant rates. It was identified
that some sectors did not show negative erosion data;
however, rate trends have been declining over the
years.
Table 4: Table of average annual rates by sector applying
the DSAS tool (2001-2020).
From 2001 to 2020
Sector
1
Sector
2
Sector
3
Sector
4
Sector
5
Sector
6
Accretion
(m/year)
1.53 11.89 2.00 2.47 5.83 0.95
Erosion
(
m/
y
ear
)
-1.06 -0.24 -0.44 DP DP -0.54
Maximum
(
m/
y
ear
)
3.93 22.15 6.19 3.93 7.56 7.18
Minimum
(
m/
y
ear
)
-3.85 -0.24 -1.57 0.23 3.62 -2.14
Average
(
m/
y
ear
)
0.13 11.85 1.96 2.47 5.83 0.38
Note: DP means "Don't present"
In the same way, a graph was made (Figure 8) of the
average historical displacement of the shoreline from
2001 to 2020, which were scaled to achieve a global
visualization of the changes per year that have
suffered shoreline. Results are described as follows:
It was observed that sector 1 remains stable,
having negative and positive rates of lower value.
It was also identified that most of the structures,
including springs and the Perú Port Terminal, are
in the sector, with a high anthropogenic influence.
Analysis of Coastline Evolution using Landsat and Sentinel 2 Images from 2001 to 2020 in Callao Bay, Peru
119
In sectors 2, 4 and 5, an abrupt change was
identified that shows high sedimentation between
2016 and 2017, in which the El Niño phenomenon
occurred, generating great changes in the natural
factors that are present on the coast.
Sector 3 has shown erosion and sedimentation
rates that have increased over the years, with a
global tendency to sediment slightly.
Finally, sector 6, where it is noteworthy to
mention, has a large presence of cliffs, showed
higher moderate erosion rates, maintaining this
trend throughout the 20-year period.
In Figure 7 the displacement of the shoreline in m /
year is presented, where the positive value represents
sedimentation and the negative value represents
erosion, according to the transect number. The
highest peak occurs in the sector 2 (transect 349),
with a value of 22.15 m / year which is in the northern
area of the outlet of the Rímac River and the
maximum erosion occurs in sector 1 (transect 322),
with a value of -3.85 m / year located in the southern
part of the mouth of the Rímac River.
Figure 7: Transects created from shorelines from 2001 to 2020.
Figure 8: a) Historical analysis by LRR of sector 1, b) Historical analysis by LRR of sector 2, c) Historical analysis by LRR of
sector 3, d) Historical analysis by LRR of sector 4, e) Historical analysis by LRR of sector 5, f) Historical analysis by LRR of
sector 6.
GISTAM 2022 - 8th International Conference on Geographical Information Systems Theory, Applications and Management
120
4 DISCUSSIONS
After a comparison with the background of the study
presented, it was observed that the tendencies to
sediment or erode in some places along the coast of
Callao Bay have been maintained despite the
presence of extreme factors such as the El Niño
phenomenon. As mentioned by Teves and San
Román (2012) study, the area between the mouths of
the Rímac and Chillón rivers (Sector 2 and Sector 3)
has a large presence of sediments that are transported
towards the northern part of the mouths due to marine
currents (Teves & San Román, 2012). Similarly, the
creation of the trend map of the present study showed
this accretion between the mouths and the beaches
north to the area of the beginning of the cliffs, with a
notable increase for the year 2017; however, for
subsequent years, sedimentation rates were
decreasing minimally.
On the other hand, erosion is a very recurrent
process in the surroundings of “Mirador Playa
Pachacutec” in sector 6 (Figure 4), this largely due to
the steep slopes that occur in the place (Teves & San
Román, 2012). The trend map (Figure 8) created
shows high erosion rates that are decreasing until
showing slight sedimentation rates in certain parts of
the sector 6.
In addition, the R
2
factor obtained from the
historical analysis (Figure 8), shows a trend to present
accretion as it gets closer to 1 and the data present a
constant increase, otherwise, while the annual data
shows more negative points of erosion, the factor is
almost 0.
Likewise, the results obtained from this study
have the same sedimentation and coastal erosion
trends as Luijendijk's Aquamonitor interface. In this
study, the author analyzed the trends of coastlines
along the world's continents, showing very general
rates of coastal dynamics that could be taken as a base
for the present study. (Luijendijk et al., 2018). The
comparisons at the total average level are:
This study shows a sedimentation rate of
0.13m/yr, 11.85m/yr, 1.96m/yr, 2.47m/yr,
5.83m/yr and 0.38m/yr with a difference with the
Aquamonitor of 0.26m/yr, 4.50m/yr, 0.37m/yr, -
3.31m/yr, -3.09m/yr and -1.00m/yr for sectors 1,
2,3,4,5 and 6 respectively.
On the one hand, it is worth mentioning that the
Aquamonitor presents data from 1984 to 2016 and
is a global level study. On the other hand, our
study presents data from 2001 to 2020. The
differences in rates in sector 2 are mainly due to
the occurrence of the 2017 El Niño phenomenon
because there was a sediment peak of 60 meters.
5 CONCLUSIONS
From the analysis of the sectorization and in general
lines of the Callao Bay, it was observed that the trend
of the coastline is to present a slight sedimentation
with average rates between 3.77 to 4.2 m/year. In
addition, sector 2; showed a moderate and constant
trend throughout the period that, for the year 2017
suffered an accumulation of sediment that reached up
to 60 meters offshore. This is due to the presence of
the extreme phenomenon called El Niño, which
generated an accumulation of sediment on the
beaches located north of the mouths of the Rímac and
Chillón rivers.
The temporal analysis shows that sector 1, which
includes the Perú Port Terminal, has remained
constant when compared to the other sectors. The
rates for the 20-year period have had a minimal but
progressive increase. Likewise, sectors 2 and 5 have
high average sedimentation rates, which have been
decreasing in lower values for the last 3 years, since
the El Niño phenomenon.
Sector 6 presents high and constant erosion
values, due mostly to the presence of cliffs and high
slopes in the area; however, the average rate of the
sector is to present a slight sedimentation because a
large percentage of the rates are positive (Figure 7).
As mentioned above, a Linear Regression Ratio
statistical analysis was performed, which was
obtained manually and with the application of the
DSAS extension. It was observed that both methods
show similar trends in the long term. On the one hand,
the manual calculation allows to see the annual
evolution of changes in average rates, while the
DSAS shows an average rate for the 20-year period.
In other words, the extension applies statistical
methods and takes into consideration the shoreline
variations in each confidence interval.
Although the study manages to present erosion or
accretion trends for the study area, there is an error
range of 5 to 30 meters due to the different resolutions
of the satellites used. This affects manual shoreline
detection and subsequent statistical analysis.
Therefore, the use of higher resolution satellite
images or digital elevation models will allow an
automatic extraction of the shoreline and,
consequently, will improve the extracted data with a
smaller error interval than the one presented by the
study.
Analysis of Coastline Evolution using Landsat and Sentinel 2 Images from 2001 to 2020 in Callao Bay, Peru
121
REFERENCES
Congedo, L. Semi-Automatic Classification Plugin
Documentation (Release 6.0. 1.1). 2016.
DIHIDRONAV. (2020). Tabla de mareas. Retrieved from
https://www.dhn.mil.pe/portal/tabla-mareas#.
Guzman, E., Ramos, C., & Dastgheib, A. (2020). Influence
of the el nino phenomenon on shoreline evolution.
Case study: Callao bay, Peru. Journal of Marine
Science and Engineering, 8(2). https://doi.org/
10.3390/jmse8020090
Himmelstoss, E. A., Henderson, R. E., Kratzmann, M. G.,
& Farris, A. S. (2018). Digital Shoreline Analysis
System (DSAS) Version 5.0 User Guide. Open-File
Report 2018-1179, 126.
Luijendijk, A., Hagenaars, G., Ranasinghe, R., Baart, F.,
Donchyts, G., & Aarninkhof, S. (2018). The State of
the World’s Beaches. Scientific Reports, 8(1), 1–11.
https://doi.org/10.1038/s41598-018-24630-6
Rangel-Buitrago, N. G., Anfuso, G., & Williams, A. T.
(2015). Coastal erosion along the Caribbean coast of
Colombia: Magnitudes, causes and management.
Ocean and Coastal Management, 114, 129–144.
https://doi.org/10.1016/j.ocecoaman.2015.06.024
Sánchez, L. (2019). Evaluación De La Calidad Del Agua
De Mar En La Playa Cantolao – Sector Espigón Del
Abtao En La Bahía Del Callao. 135.
Sheik Mujabar, P., & Chandrasekar, N. (2013). Coastal
erosion hazard and vulnerability assessment for
southern coastal Tamil Nadu of India by using remote
sensing and GIS. Natural Hazards, 69(3), 1295–1314.
https://doi.org/10.1007/s11069-011-9962-x
Soto, R. G. (2018). Dunas y procesos costeros en una isla
tropical caribeña amenazada por erosión, actividades
humanas y aumento del nivel del mar. Caribbean
Studies, 46(2), 57–77. https://doi.org/10.1353/
crb.2018.0023
Teves, N., & San Román, C. (2012). Geología Marina y
Ambiental del litoral limeño entre Ancón y Pucusana.
In XIV Congreso Peruano de Geología.
Yasir, M., Sheng, H., Fan, H., Nazir, S., Niang, A. J.,
Salauddin, M., & Khan, S. (2020). Automatic Coastline
Extraction and Changes Analysis Using Remote
Sensing and GIS Technology. IEEE Access, 8, 180156–
180170. https://doi.org/10.1109/ACCESS.2020.30278
81
GISTAM 2022 - 8th International Conference on Geographical Information Systems Theory, Applications and Management
122