Progress in Disaster Science 14 (2022) 100235

Contents lists available at ScienceDirect

Progress in Disaster Science

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

Applying Google earth engine for flood mapping and monitoring in the
downstream provinces of Mekong river
Bui Phan Quoc Nghia a, 1, Indrajit Pal a, 1, Nuwong Chollacoop b, 2, Anirban Mukhopadhyay a, 1, *
a
Disaster Preparedness, Mitigation and Management, Asian Institute of Technology, Thailand
b
National Energy Technology Center (ENTEC), National Science and Technology Development Agency (NSTDA), Thailand

A R T I C L E I N F O A B S T R A C T

Keywords: Flood is a hazard, but it is also essential for the local communities in the Mekong Delta, as it also provides
Mekong flood nutrients for agriculture and resists saltwater intrusion. In recent years, changes in flood patterns have been
Flood shifting observed in the Mekong Delta with erratic fluctuations in the flooded area, especially in the downstream
Flooding
provinces. It causes a lot of difficulties in flood monitoring and management in this area. Therefore, a robust
GEE (Google Earth Engine)
SAR (synthetic aperture radar)
assessment of flood shifting is essential in the Mekong Delta. This work addresses the development of a logical
model based on Google Earth Engine (GEE) using Sentinel-1 synthetic aperture radar (SAR) data to assess floods
in the downstream Mekong River basin. In the present study, the observed data of Tan Chau and Chau Doc
hydrological stations have been used to test and prove the reliability of this application. The inundation scenario
generated from this study showed a gradual shift in flooding patterns of the downstream provinces of the Mekong
Delta. Overall, the flooded area is decreasing; however, the same is increasing in Bac Lieu province. This research
will help to construct a professional support system to monitor the flood inundation in the Mekong Delta. They
can give local authorities a big picture of inundation to decide on flood disaster preparedness, mitigation, and
adaptation.

1. Background benefits as well as damages and losses. Extreme floods can cause human
loss, and destruction of infrastructure and agriculture. However, sea­
The Mekong Delta of Vietnam (MDV) is the third-largest delta in the sonal floods have played an essential role in providing aquatic resources,
world, located in the downstream area of the Mekong River before the freshwater, and nutrients [30,44,45]. The regular flooding season in the
river drains into the East Sea of Vietnam [23]. According to General study area is between June and the end of November [21], with high-
Statistics Office (GSO), the total area of MDV is approximately 04 intensity flooding in September and October. The MDV is inundated
million ha, with a total population of more than 17 million [13]. This almost every year, ranging from 1.4 million hectares during small floods
area leads the rice and aquaculture production of the nation, and the to 1.9 million hectares during big floods [44]. Based on the analysis of
economic growth rate of the MDV is higher (7.8%) than the national water levels from 1990 to 2019, Le [21] found that the trend of the big
average (6.8%) [13]. Historically, the MDV had a favourable climate flood is decreasing, and small flood is increasing. However, flood is
condition and a low rate of natural disasters compared to other regions impacted by tides and surges in the coastal provinces more than up­
of Vietnam; however, this area has been faced many severe hazards such stream river discharge [41].
as water scarcity, drought, saltwater intrusion, and coastal erosion in Flood hazard pattern has been observed to be shifted in the MDV
recent years [23,30,42,43]. with the erratic fluctuations of the flooded area in the downstream
Based on the country report of the ASEAN Coordinating Centre for provinces and the increasing flooding trend in urban areas because of
Humanitarian Assistance on disaster management (AHA Centre), the the development of irrigation systems, sea level rise, high tides, and
MDV is a hot spot for flood and inundation in Vietnam [2]. Flood brings subsidence (T. A. [21,41,45]). For example, based on observed data, the

* Corresponding author.
E-mail addresses: [email protected] (B.P.Q. Nghia), [email protected] (I. Pal), [email protected] (N. Chollacoop), [email protected]
(A. Mukhopadhyay).
1
Disaster Preparedness, Mitigation and Management, Asian Institute of Technology, Thailand, Pathumthani,12,120
2
The National Energy Technology Center, National Energy Technology Center, Thailand, Pathumthani,12,120

https://doi.org/10.1016/j.pdisas.2022.100235
Received 23 March 2022; Received in revised form 16 May 2022; Accepted 20 May 2022
Available online 24 May 2022
2590-0617/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

water level at Can Tho station – a downstream station in MDV – was concerned authorities mitigate flood hazards, water security, and con­
2.15 m in 2011, higher than that of 2000 (1.79 m). Meanwhile, the water struction planning.
level of Tan Chau – an upstream station in MDV – was 4.27 m in 2011,
lower than 4.90 m in 2000 [9]. The shift of flooding areas may impact 2. Study area, data, and methodology
policymakers’ decisions in many aspects such as water management,
land use, urban management, etc. Therefore, monitoring flooding areas 2.1. Study area and data
and the shift of flood patterns is vital for understanding the concurrent
situation of flood and helps decision-makers give appropriate solutions The study area was the MDV’s eastern coastal provinces: Tien Giang,
promptly. Ben Tre, Tra Vinh, Soc Trang, Bac Lieu, and Ca Mau. The flooding area in
Synthetic aperture radar (SAR), a type of space-based sensor, has these provinces is usually less than in other upstream provinces. The
proven its ability to monitor and map flood extent [3,17]. SAR is locations and boundaries of the study area, MDV, An Giang province,
beneficial compared to other sensors since its radar transmission in the and Tan Chau and Chau Doc stations are shown in Fig. 1.
microwave spectrum is unaffected by cloud, heavy rain, and visibility The Mekong River is the longest (4800 km) in Southeast Asia, with a
[39,40]. Flood hazard mappings derived from SAR data could be vital drainage basin of approximately 7.95 × 105 km2 area (T. V. H. [22]).
evidence for efficient disaster risk management enabling social relief The Mekong River Delta (MRD) encompasses a 4.95 × 104 km2 area, of
groups and key stakeholders to gather spatially accurate intelligence on which 74% falls in Vietnam. The MRD in Vietnam could be character­
flood events in a prompt and cost-effective approach [49]. Floods are ized by a vast floodplain with numerous rivers and channels, creating a
emerging as a grave issue in our changing climate. Flood mapping is complex network. The MRD experiences dual tidal activity, one from the
crucial for both risk assessment and forecast improvement. Since SAR South China Sea and the other from the Gulf of Thailand. Climatic
data is adequate for flood risk assessment, many flood mapping methods conditions follow a typical monsoonal pattern with two distinct seasons.
have been devised in recent years [35,46]. conducted a study using The rainy season starts in May and ends in October, and the dry season
Sentinel-1-SAR-based flood assessment by incorporating daily and starts in November and ends in April. Numerous canals across the delta
monthly rainfall data obtained from the IMD (“Indian Meteorological in Vietnam were constructed for various purposes, and dykes, and weirs
Department”) and TRMM (“Tropical Rainfall Measuring Mission”) for were made to fight against floods and associated hazards. These con­
delineating the regions swamped by floodwaters across Kerala, India, structions disrupt the seasonal variability of the freshwater flow of the
during the monsoon season of 2018. Landuyt et al. [20] reviewed and Mekong River [34].
assessed different flood mapping approaches, including enhanced Three different types of data have been used in this study (Table 1).
thresholds, contour modelling, and change detection, while evaluating Firstly, the Sentinel-1 SAR Vertically transmitted Vertically receives
the strategies using intermediate resolution SAR data from various (VV) polarization data. VV is used instead of vertically transmitted
flooding events in the UK and Ireland. Long et al. [24] used SAR data and Horizontally received (VH) because it is more efficient in flood mapping
a novel flood estimation approach termed CDAT (“Change Detection and [1,5,46]. Secondly, the administrative boundaries of the whole study
Thresholding”) to outline the flood extent of Namibia’s Chobe flood­ area and each province were downloaded from the Human Data Ex­
plain. Pulvirenti et al. [51] utilized SAR data to map submerged zones change and updated in 2020. Finally, the Mekong River Commission
using fuzzy logic. It is planned to integrate this tool into the Italian data have been collected for the highest water level data at Tan Chau
Public Security’s operation flood management structure and to test it on and Chau Doc hydrometeorological stations during the flood season
a recent flooding event in Albania. from June to December from 2015 to 2020. Tan Chau and Chau Doc
Google Earth Engine (GEE) is introduced as the world’s most hydrometeorological stations are two upstream stations of the Tien and
advanced cloud-based geospatial processing platform that can overcome Hau rivers. They are located in An Giang province. The amount of water
processing problems encountered by traditional satellite image pro­ discharged through these two rivers covers 80 to 85% of the amount of
cessing methods [12,40]. GEE provides a cloud computing platform to water in the MDV (T. A. [21]).
store and analyse huge data sets up to petabytes in size for ultimate
decision making and analysis [18]. Google has incorporated the Landsat 2.2. Methodology
data sets into its cloud computing platform since the series became
freely available in 2008 [28]. On a regional and global scale, researchers The methodology is divided into two figures; Fig. 2 describes the
have utilized the GEE to investigate the dynamics of land regions such as methods adopted for flood area calculation using GEE, and Fig. 3 depicts
forest/ vegetation cover [14,16], surface waterways [8,33,38], and the statistical operations performed to analyse the results.
human habitation [8]. It has been employed to assess current agrarian Fig. 3 depicts the general approach of this work, which depicts the
diversification [29,31,50] and for other purposes. The GEE maintains a logical model built up to assess the reliability of the flooded area given
huge geospatial data repository, including Sentinel-1 GRD (“Ground by GEE and Sentinel-1 SAR data, and then detects changes in flooding
Range Detected”) data that is constantly updated [6]. Due to the in­ patterns across the study area and each province. It can be divided into
tricacy of SAR preprocessing, analysis-ready SAR datasets available on three main processes as follows: The first process is to check the flooded
GEE represent a huge step forward for application-based remote sensing area’s reliability generated by the GEE database of Sentinel-1 SAR im­
through SAR. GEE’s dataset comprises over 40 years of archival earth ageries. Two factors further prove the reliability. First, from June to
observation imageries regularly updated in quantity and quality, October of each year, a correlation was performed between the flooded
including SAR data [12]. area of An Giang province from GEE and observed data of the highest
This study aims to assess the flood shifting pattern in the MDV by water level in Tan Chau and Chau Doc stations. This study uses the
utilising GEE (Google Earth Engine) and Sentinel-1 SAR (synthetic coefficient of determination (R2) to determine the correlation between
aperture radar) data products to visualise flooded areas in the MDV. The flooding areas from GEE and water level. In addition, P-value shows the
flood maps were cross-checked with available datasets and observation statistically significant difference between flooding areas from GEE and
stations. Then the flooded areas were correlated with water levels and water level. If the results are good with R2 ≥ 0.7 and p-value <0.05, the
different provinces. The correlation could be served as a baseline report reliability is high [10,36]. If not, the code of flood calculation needs to
for much-talked flood shifting in the downstream Mekong River. be revised. The reason for choosing 02 stations Tan Chau and Chau Doc
Moreover, the coherence of the flooded area and the water level of the to evaluate the correlation is 2.1. These stations are 02 upstream stations
upstream stations are examined annually. Finally, the shift in flooding of Tien River and Hau River, through which the amount of water
area per province has been detected, along with the quantification of the discharge covers from 80 to 85% of the amount of water in the MDV.
relationships established. This study will help the policymakers, and Secondly, this study comprises the coherence of the flooding area of

2
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

Fig. 1. Map of the study area (06 East coastal provinces of the MDV).

1.0 is strong, from 0.4 to less than 0.7 is medium, and from 0.0 to less
Table 1
than 0.4 is weak [36]. Similarly, P-values have been applied to show the
Data specification.
statistically significant difference between the MDV and the study region
No. Data Description Sources and between each province’s studied area. The threshold is 0.05 [10]. P-
1 Sentinel-1 - Polarization: VV polarization 1 Google Earth Engine: value less than or equal to 0.05 proves no significant flood shifting; on
SAR - Bands: 1 https://developers. the other hand, greater than 0.05 means significant flood shifting.
- Wavelength: 5.5 google.com Because the study area is within MDV, and each Eastern coastal province
- Spatial resolution: 20 m × 20 - Acquisition date: 29
m October 2021
is within the study area, if the flooding area in the MDV increases, the
- Swath: 250 flooding area inside the study area and each province will increase and
- Temporal resolution: 12 days vice-versa (Fig. 4).
2 Shapefile of Administrative boundary of The Humanitarian Data The last part is to conclude flood pattern shifting in the study area.
the study MDV, An Giang province, study Exchange
Local reports about flood shifting and hazards are reviewed to under­
area area https://data.humdata.
org/ stand the reliability and relevance of the results of flood shifting from
- Acquisition date: 29 GEE and Sentinel-1 SAR images.
October 2021 The most important part of the methodology is decoding Sentinel-1
3 Highest - Station: Tan Chau (10o48’ Mekong River SAR data to identify the flooded areas. This section involves deploying
water level 105o13’), Chau Doc (10o42’ Commission for
105o06’) in An Giang province sustainable
the JavaScript programming language directly implemented on the GEE
- Temporal resolution: Highest development interface (https://developers.google.com/earth-engine). In this study,
water level from June to http://ffw.mrcmekong. the code is developed based on the UN online course about flood map­
October each year from 2017 to org/ ping using Sentinel-1 SAR data in GEE (https://t.ly/0FdC). It includes
2020 in Tan Chau and Chau Doc - Acquisition date: 29
command declaration tasks to import image data to the platform, image
station October 2021
processing, image analysis, result visualisation, and result exporting. In
this study, pixel values of pre-flood and post-flood scenario images have
the MDV and yearly local reports about floods and drought. In addition, been compared. This difference helps to separate pixels that belong to
the coherence of distribution of flood area and MDV terrain is also the permanent water bodies such as rivers and wetlands and pixels that
considered. The process entails detecting changes in flood patterns in the represent waterlogged areas during the flooding season.
study area and each province. In terms of flooded areas, the R2 value is Flood mapping and inundated area calculation methodology are as
used to estimate the correlation between MDV and the total study area follows:
and between the total study area and each province. R2 value from 0.7 to The Sentinel-1 data package and study area boundary have been

3
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

Fig. 2. Processing of Sentinel-1 SAR data using GEE platform.

Fig. 3. Schematic representation of the statistical operations performed in this study.

imported to GEE in the first step. The boundaries are designated as the November of each year because flood usually starts in June and ends in
shapefiles of the MDV on a province level. They were directly imported November every year (T. A. [21]). By setting periods, not single dates,
into the Assets of GEE. In the next step, Time frame and sensor param­ the selection can cover enough tiles of the selected area. Sensor pa­
eters were specified for this study. The base period selected for flooding rameters specified are polarization “VV” and a “descending” pass di­
area comparison is 1st to 31st May of each year (right before the start of rection. The difference threshold was “1.25” because this value is
the flood season). The pre and post-flood period are 1 June and 30st usually used for GEE flood detection [19,37]. The flood extent was

4
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

Fig. 4. All cases of correlation of flood area between MDV, study area and each province.

determined when the pixels representing the flooded area were classi­ from GEE and visualised in QGIS. It is a cross-platform desktop
fied correctly. The primary comparison is the difference between the geographic information system (GIS) tool to browse, edit, and analyse
before and after imageries. After applying the predefined threshold, the geographical data without payment.
flood extent mask is created, and the flood result is refined. Each year,
the flood extent area is calculated. Finally, the flooded area was 3. Results
approximated by summing all the pixels representing the flooded re­
gions and translating the flood extent to hectares. Finally, the inundated 3.1. The coherence of flooding area from GEE and observed data
area was visualised by importing the geographic information system
application results. The result of the flood extent area has been acquired The coherence of the flooding area from GEE and observed data is

5 5
R² = 0.791
4 4
Water level (m)

Water level (m)

R² = 0.9006
3 3

2 2

1 1

0 0
0 50000 100000 0 50000 100000
Flooding area (ha) Flooding area (ha)
2017 2018
5 5

4 4
Water level (m)

Water level (m)

R² = 0.9292
3 3 R² = 0.9394
2 2

1 1

0 0
0 50000 100000 0 50000 100000
Flooding area (ha) Flooding area (ha)

2019 2020
Fig. 5. Correlation between flooding area in An Giang province and water level in Tan Chau station in June to October from 2017 to 2020.

5
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

proved by the correlation and the p-value between the flooding area in Table 2
An Giang province and the water level in Tan Chau and Chau Doc sta­ The P-value of flooding area in An Giang province and water level in Tan Chau
tions of An Giang province from June to October 2017 to 2020. Fig. 5 and Chau Doc stations from June to October from 2017 to 2020.
shows the correlation between Tan Chau station’s water level and the Flooding area in An Giang province
flooding area in the Giang province. R2 ranges from 0.791 to 0.9394 2017 2018 2019 2020
from 2017 to 2020. R2 in 2018 is the lowest, with 0.791. In other years, Tan Chau
0.013715 0.043423 0.008168 0.006443
the R2 value is higher than 0.90. Water station
Fig. 6 describes the coherence of the water level in Chau Doc station level Chau Doc
0.036041 0.016001 0.006453 0.004858
station
and the flooding area in An Giang province. The R2 value is between
0.8144 and 0.9497. The year 2016 is the lowest with 0.8144, and 2020 is
the highest with 0.9497. These graphs also show the linkage between
water level and flooding area. When the water level increases, the
10,00,000
flooding area will increase and vice versa. Table 2 describes the p-value
9,00,000
of the flooded area in An Giang province and the water level in Tan Chau 8,00,000
and Chau Doc stations from June to October from 2017 to 2020. The 7,00,000
highest one is 0.043423, which is still below the threshold of 0.05.

Area (ha)
6,00,000
Another indicator of the code’s dependability is the coherence be­ 5,00,000
tween flooding areas in MDV and weather changes from 2015 to 2020. 4,00,000
The flooding area of the whole MDV from 2015 to 2020 is demonstrated 3,00,000
in Fig. 7 and visualised in Fig. 8. In Fig. 7, it can be seen that the periods 2,00,000
1,00,000
from 2015 to 2016 and 2019 to 2020 have low flooded areas, while the
0
period from 2017 to 2018 represents high flood incidents. Because there 2015 2016 2017 2018 2019 2020
were severe droughts in the dry season during 2015 /2016 and 2019 Flood area 6,27,15 7,52,46 9,06,21 7,97,90 6,53,80 6,29,37
/2020, which was reported in the Global Facility for Disaster Reduction Year
and Recovery (GFDRR) and Ministry of Agriculture and Rural Devel­
opment (MARD) [11,25]. The negative trend in flooded areas is also Fig. 7. Flooded area of the MDV over the years.
coherent with recent research about the long-term decrease of water
discharge to the MDV because of climate change, riverbed mining and than the downstream and coastal areas. Only one part of the upstream
upstream dams [15,30,47]. Fig. 8 also shows the flood area distribution, area that is not flooded during flood season is the Bay Nui mountainous
and it is close to reality. The upstream area is covered with floods more area of An Giang province, the highest elevated area in the MDV, with a

4 4

R² = 0.8144
3 3
Water level (m)

Water level (m)

R² = 0.8901
2 2

1 1

0 0
0 50000 100000 0 50000 100000
Flooding area (ha) Flooding area (ha)

2017 2018
4 4

R² = 0.9394
3 3
Water level (m)

Water level (m)

R² = 0.9497

2 2

1 1

0 0
0 50000 100000 0 50000 100000
Flooding area (ha) Flooding area (ha)

2019 2020
Fig. 6. Correlation between flooding area in An Giang province and water level in Chau Doc station in June to October from 2017 to 2020.

6
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

2015 2016

2017 2018

2019 2020
Fig. 8. Spatial representation of flooding area of the MDV.

peak of 705 m [4]. area of the MDV [13]; however, the total flooding area is only from 15%
The correlation and the p-value between the flooded area in An to 20% of the total flooding area of Mekong Delta. It is understandable
Giang province and the water level in the chosen stations; and the because the study area is downstream of the MDV. A correlation be­
consistency between the flooded area in MDV and local reports about tween the flooded area of the study area and the MDV from 2015 to 2020
disasters and topology show that the code in this study is reliable. These is shown in Fig. 9b; the p-value is 0.008498, which is less than 0.05;
results also support the findings of Vanama et al. [48], Moothedan et al. therefore, the flooded area of the study area and the MDV is coherent.
[27], and Moharrami et al. [26] that the resultant flood area and maps The correlation between the flooded area of MDV and the study area
from GEE have high accuracies, and it proves that GEE is an efficient is high, but it is not strong between each province and the whole study
way for rapid flood mapping. area. Fig. 10 describes the correlation of the flooding area between the
study area and each province in it. It can be recognized that the R2 value
of Soc Trang province is 0.9634, which is high; the R2 value of Ca Mau is
3.2. The shift of flooding area of the eastern coastal provinces from 2015 0.4758, which is medium; and the R2 values of Tien Giang, Ben Tre, Tra
to 2020 Vinh, and Bac Lieu are 0.1338, 0.0302, 0.1331, and 0.148 respectively
which are in low range. Table 3 describes the p-value of the flooding
Fig. 9a shows the flood area of the study area and its trend from 2015 area between the study area and each province from 2015 to 2020. The
to 2020. It can be observed that the trend decreases slightly, the same as p-value in Soc Trang is 0.000508, which is the only one below 0.05; the
the trend of the MDV. According to GSO, the total area of 06 coastal others are above. They prove that the correlation between the flooding
provinces is 1846.6 thousand ha accounting for about 45.2% of the total

7
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

200,000 2000
180,000 1800
160,000 R² = 0.8533
1600
140,000

Study area (km2)

1400
120,000
1200
ha
100,000
80,000 1000
60,000 800
40,000 600
20,000 400
0
2015 2016 2017 2018 2019 2020 200
Flood area 104,226 131,901 184,199 122,023 112,667 101,864 0
6000 7000 8000 9000 10000
year
MDV (km2)
(a) (b)
Fig. 9. Flooding area of the study area (a), and correlation of the flooding area between the MDV and study area from 2015 to 2020 (b).

400 20
Tien Giang (km2)

350 R² = 0.1338

Ben Tre (km2)

15
R² = 0.0302
300
10
250
5
200

150 0
1000 1200 1400 1600 1800 2000 1000 1200 1400 1600 1800 2000
Study area (km2) Study area (km2)

Tien Giang Ben Tre

155 950
135 850 R² = 0.9634
Soc Trang (km2)
Tra Vinh (km2)

115 750
R² = 0.1331
95 650
75 550
55 450
35 350
1000 1200 1400 1600 1800 2000 1000 1200 1400 1600 1800 2000
Study area (km2) Study area (km2)

Tra Vinh Soc Trang

450 250
400 200 R² = 0.4758
Bac Lieu (km2)

Ca Mau (km2)

350
R² = 0.148 150
300
100
250
200 50
150 0
1000 1200 1400 1600 1800 2000 1000 1200 1400 1600 1800 2000
Study area (km2) Study area (km2)

Bac Lieu Ca Mau

Fig. 10. The correlation of the flooding area between the study area and each province from 2015 to 2020.

8
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

Table 3
P-value of the flooding area between the study area and each province from 2015 to 2020.
Flooded areas in each province

Tien Giang Ben Tre Tra Vinh Soc Trang Bac Lieu Ca Mau

Flooding area in study area 0.475762 0.741978 0.476956 0.000508 0.451452 0.129423

area, the study area and Tien Giang, Ben Tre, Tra Vinh, and Bac Lieu is in Bac Lieu province in the rainy season. It damaged 18,625.734 ha of
low. Therefore, the shift in the flooded area of these 04 provinces is high. agricultural products, including rice, vegetables, and fruit trees [32].
In Fig. 11, it can be observed that the decreasing trend is the same as Massive flooding has been mentioned in the report are with high pre­
the trend of the MDV and study area in five provinces, Tien Giang, Ben cipitation, tidal peaks, and impacts of storms [32].
Tre, Tra Vinh, Soc Trang, and Ca Mau. These negative trends can be the According to the results above, it can be concluded that flood shifting
same as the negative trend of water discharge to the MDV: upstream is happening in the Eastern coastal area of the MDV. This conclusion is in
dam building, riverbed mining, and climate change impacts mentioned some way consistent with the study of [30], Dieu and Thao [7], and Triet
in part 3.1. However, the decrease is not the same within these five et al. [45]. This study argued that the flood regime is changing in the
provinces. While Soc Trang, Ca Mau, and Tra Vinh have a slightly MDV due to urbanisation, climate change, sand mining, and irrigation
decreasing trend, the trend in Tien Giang and Ben Tre decreases infrastructure building.
dramatically. An increasing trend was observed in Bac Lieu province,
which is different from other provinces in the study area and MDV. 4. Conclusion
Especially in 2020, the flooding area in Bac Lieu province is highest. This
result is very consistent with the report of the People’s Committee of Bac This study attempts to determine the dynamics in flood patterns of
Lieu Province (PCBL) in 2020. The report mentioned a huge flood event six East coastal provinces downstream of the MDV. Using GEE and its

50,000 2,000
40,000 1,500
30,000
1,000
ha

ha

20,000
10,000 500

0 0
2015 2016 2017 2018 2019 2020 2015 2016 2017 2018 2019 2020
Year Year

Tien Giang Ben Tre

15,000 100,000
80,000
10,000
60,000
ha
ha

40,000
5,000
20,000
0 0
2015 2016 2017 2018 2019 2020 2015 2016 2017 2018 2019 2020
Year Year

Tra Vinh Soc Trang

50,000 25,000
40,000 20,000
30,000 15,000
ha

ha

20,000 10,000
10,000 5,000
0 0
2015 2016 2017 2018 2019 2020 2015 2016 2017 2018 2019 2020
Year Year

Bac Lieu Ca Mau

Fig. 11. The trend of flooding area of the 06 East coastal provinces from 2015 to 2020.

9
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

Sentinel-1 SAR data database, the model has developed a logic-based [4] Canh DN, Thi NTA. Developing Khmer community based rural tourism in Tinh Bien
district, an Giang province. Can Tho Univ J Sci 2018;54(6):148–57. https://doi.
approach to map and monitor flooding downstream of the MDV and
org/10.22144/ctu.jvn.2018.107.
recognize the correlation between regions regarding flooding. It also [5] Carreño Conde F, De Mata Muñoz M. Flood monitoring based on the study of
compares the results from coding to the observed data and available Sentinel-1 SAR images: the Ebro River case study. Water 2019;11(12):2454.
local government reports, checking the results’ reliability. [6] DeVries B, Huang C, Armston J, Huang W, Jones JW, Lang MW. Rapid and robust
monitoring of flood events using Sentinel-1 and Landsat data on the Google earth
The study proves that GEE is an admirable platform to monitor and engine. Remote Sens Environ 2020;240(October 2018):111664. https://doi.org/
map the flooding across the study area; therefore, it can also monitor the 10.1016/j.rse.2020.111664.
flood in other areas. It is beneficial not only for scientists but also for the [7] Dieu PQ, Thao PTT. Urbanizing Mekong Delta in Vietnam: The challenges of urban
expansion adapting to floods. In: The 5th International Conference of the
local governments, decision-makers, or other persons interested in doing International Forum on Urbanism (IFoU)–2011 the National University of
research following flood progression. The result shows that the flood Singapore, Department of Architecture–Global Visions: Risk and Opportunities for
pattern of the whole six east coastal provinces is consistent with the the Urban Planet; 2011.
[8] Donchyts G, Baart F, Winsemius H, Gorelick N, Kwadijk J, Van De Giesen N. Earth’s
flood pattern of the MDV. However, the dynamics of the flooded area surface water change over the past 30 years. Nature Climate Change 2016;6(9):
within six East coastal provinces is high. There is potentially a shift in 810–3.
flooding areas within provinces in the study area and the MDV. How­ [9] Duc Tran D, Van Halsema G, Hellegers PJGJ, Phi Hoang L, Quang Tran T,
Kummu M, et al. Assessing impacts of dike construction on the flood dynamics of
ever, to conclude the shift in flooding areas in MDV, more profound the Mekong Delta. Hydrol Earth Syst Sci 2018;22(3):1875–96.
research is needed. [10] Fisher RA. Statistical methods for research workers. In breakthroughs in statistics.
There are many ways to develop more profound research by fields, Springer; 1992. p. 66–70.
[11] GFDRR. Toward integrated disaster risk Management in Vietnam. World Bank
areas, and methods based on this research. Disaster management, water
2017. https://doi.org/10.1596/28871.
management, and urban planning are promising fields that can use the [12] Gorelick N, Hancher M, Dixon M, Ilyushchenko S, Thau D, Moore R. Google earth
result of this research. Each province can rely on this to adapt suitable engine: planetary-scale geospatial analysis for everyone. Remote Sens Environ
socio-economic development plans for its own. Some methods, such as 2017;202:18–27.
[13] GSO. Statistical Yearbook of Vietnam 2020. Statistical Publishing House; 2021.
combining with observed data or other satellite data, can apply by using [14] Hansen MC, Potapov PV, Moore R, Hancher M, Turubanova SA, Tyukavina A, et al.
the direction of this thesis to compare and give better results for flood High-resolution Global Maps. 2021.
mapping and monitoring. [15] Hoang LP, Lauri H, Kummu M, Koponen J, Van Vliet MTH, Supit I, et al. Mekong
River flow and hydrological extremes under climate change. Hydrol Earth Syst Sci
This study has some shortcomings in the result. Firstly, the result is 2016;20(7):3027–41.
based on the outcome of Sentinel-1 SAR images analysis; therefore, there [16] Johansen K, Phinn S, Taylor M. Mapping woody vegetation clearing in Queensland,
might be some bias compared to the result from observed data. How­ Australia from Landsat imagery using the Google earth engine. Remote Sensing
Appl Soc Environ 2015;1:36–49.
ever, it is still good to quickly track the flood in case of insufficient [17] Khan SI, Hong Y, Wang J, Yilmaz KK, Gourley JJ, Adler RF, et al. Satellite remote
observed data. Secondly, Sentinel-1 SAR data is only available from 3 sensing and hydrologic modeling for flood inundation mapping in Lake victoria
October 2014 to the present; thus, this method cannot be used to find the basin: implications for hydrologic prediction in ungauged basins. IEEE Trans
Geosci Remote Sens 2011;49(1 PART 1):85–95. https://doi.org/10.1109/
flood progression before 2014. TGRS.2010.2057513.
[18] Kumar L, Mutanga O. Google earth engine applications since inception: usage,
CRediT authorship contribution statement trends, and potential. Remote Sens (Basel) 2018;10(10):1–15. https://doi.org/
10.3390/rs10101509.
[19] Lal P, Prakash A, Kumar A, Srivastava PK, Saikia P, Pandey AC, et al. Evaluating
Bui Phan Quoc Nghia: Conceptualization, Methodology, Formal the 2018 extreme flood hazard events in Kerala, India. Remote Sensing Lett 2020;
analysis, Writing – original draft. Indrajit Pal: Conceptualization, 11(5):436–45.
[20] Landuyt L, Van Wesemael A, Schumann GJP, Hostache R, Verhoest NEC, Van
Methodology, Writing – review & editing. Nuwong Chollacoop: Coillie FMB. Flood mapping based on synthetic aperture radar: an assessment of
Conceptualization, Methodology. Anirban Mukhopadhyay: Method­ established approaches. IEEE Trans Geosci Remote Sens 2019;57(2):722–39.
ology, Writing – review & editing. https://doi.org/10.1109/TGRS.2018.2860054.
[21] Le TA. Phân tích diễn biến lũ lụt và khô hạn ở Đồng bằng sông Cửu Long trong 20
năm gần đây. Vietnam J Sci Technol Eng 2020;62(11):22–7.
[22] Le TVH, Nguyen HN, Wolanski E, Tran TC, Haruyama S. The combined impact on
Declaration of Competing Interest the flooding in Vietnam’s Mekong River delta of local man-made structures, sea
level rise, and dams upstream in the river catchment. Estuar Coast Shelf Sci 2007;
71(1–2):110–6.
The authors declare that they have no known competing financial
[23] Loc HH, Van Binh D, Park E, Shrestha S, Dung TD, Son VH, et al. Intensifying saline
interests or personal relationships that could have appeared to influence water intrusion and drought in the Mekong Delta: from physical evidence to policy
the work reported in this paper. outlooks. Sci Total Environ 2021;757:143919.
[24] Long S, Fatoyinbo TE, Policelli F. Flood extent mapping for Namibia using change
detection and thresholding with SAR. Environ Res Lett 2014;9(3):1–9. https://doi.
Acknowledgements org/10.1088/1748-9326/9/3/035002.
[25] MARD. Summary of the steering and management of responding drought, water
shortage, saltwater intrusion, ensuring water sources for agricultural production.
The first author of this study, Mr. Bui Phan Quoc Nghia, has received In: people in the Mekong Delta in the dry season 2019–2020. MARD (Ministry of
financial support from the Education and Research Collaboration be­ Agriculture and Rural Development); 2020.
tween the National Science and Technology Development Agency [26] Moharrami M, Javanbakht M, Attarchi S. Automatic flood detection using sentinel-
1 images on the google earth engine. Environ Monit Assess 2021;193(5):1–17.
(NSTDA) and the Asian Institute of Technology (AIT). The authors
[27] Moothedan AJ, Dhote PR, Thakur PK, Garg V, Aggarwal SP, Mohapatra M.
would like to thank the Mekong River Commission for sustainable Automatic flood mapping using Sentinel-1 GRD SAR images and Google earth
development, ESA, Google Earth Engine, and The Humanitarian Data engine: A case study of Darbhangah, Bihar. In: The Proceedings of National
Exchange for the data used in this study. Seminar on ‘Recent Advances in Geospatial Technology and Applications; 2020.
[28] Mutanga O, Kumar L. Google earth engine applications. Remote Sens (Basel) 2019;
11(5):11–4. https://doi.org/10.3390/rs11050591.
References [29] Padarian J, Minasny B, McBratney AB. Using Google’s cloud-based platform for
digital soil mapping. Comput Geosci 2015;83:80–8. https://doi.org/10.1016/j.
cageo.2015.06.023.
[1] Agnihotri AK, Ohri A, Gaur S, Das N, Mishra S. Flood inundation mapping and
[30] Park E, Ho HL, Tran DD, Yang X, Alcantara E, Merino E, et al. Dramatic decrease of
monitoring using SAR data and its impact on Ramganga River in ganga basin.
flood frequency in the Mekong Delta due to riverbed mining and dyke construction.
Environ Monit Assess 2019;191(12):1–16.
Sci Total Environ 2020;723:138066.
[2] AHA Centre. Country report Vietnam. Natural disaster risk assessment and area
[31] Patela NN, Angiuli E, Gamba P, Gaughan A, Lisini G, Stevens FR, et al.
business continuity plan formulation for industrial agglomerated areas in the ASEAN
Multitemporal settlement and population mapping from landsatusing google earth
region. Jpn Int Cooperat Agency 2015:10–2.
engine. Int J Appl Earth Observ Geoinforma 2015;35(PB):199–208. https://doi.
[3] Amitrano D, Di Martino G, Iodice A, Riccio D, Ruello G. Unsupervised rapid flood
org/10.1016/j.jag.2014.09.005.
mapping using Sentinel-1 GRD SAR images. IEEE Trans Geosci Remote Sens 2018;
56(6):3290–9.

10
B.P.Q. Nghia et al. Progress in Disaster Science 14 (2022) 100235

[32] PCBL. Report on the Socio-economic Situation in November and the First. 2020 [11 the Vietnamese Mekong Delta. Int J Disaster Risk Reduct 2021;57:102183. https://
months of 2020]. doi.org/10.1016/J.IJDRR.2021.102183.
[33] Pekel J-F, Cottam A, Gorelick N, Belward AS. High-resolution mapping of global [43] Trang PTH, Tuan TV. Flooding in the Mekong Delta: causes and solutions. J Sci Ho
surface water and its long-term changes. Nature 2016;540(7633):418–22. Chi Minh City Univ Educ 2016;3(158):158–69.
[34] Pokhrel Y, Shin S, Lin Z, Yamazaki D, Qi J. Potential disruption of flood dynamics [44] Trang Pham Thi Huyen, Van Tuan T. Flooding in the Mekong Delta: causes and
in the lower Mekong River basin due to upstream flow regulation. Sci Rep 2018;8 solutions. J Sci Ho Chi Minh City Univ Educ 2016;3(81):158–69.
(1):1–13. [45] Triet NVK, Dung NV, Fujii H, Kummu M, Merz B, Apel H. Has dyke development in
[35] Pramanick N, Acharyya R, Mukherjee S, Mukherjee S, Pal I, Mitra D, et al. SAR the Vietnamese Mekong Delta shifted flood hazard downstream? Hydrol Earth Syst
based flood risk analysis: a case study Kerala flood 2018. Adv Space Res 2021. Sci 2017;21(8):3991–4010.
https://doi.org/10.1016/j.asr.2021.07.003. [46] Twele A, Cao W, Plank S, Martinis S. Sentinel-1-based flood mapping: a fully
[36] Ratner B. The correlation coefficient: its values range between+ 1/− 1, or do they? automated processing chain. Int J Remote Sens 2016;37(13):2990–3004. https://
J Target Measure Anal Market 2009;17(2):139–42. doi.org/10.1080/01431161.2016.1192304.
[37] Singh G, Pandey A. Mapping Punjab flood using multi-temporal open-access [47] Van Binh D, Kantoush S, Sumi T. Changes to long-term discharge and sediment
synthetic aperture radar data in Google earth engine. In: Hydrological Extremes. loads in the Vietnamese Mekong Delta caused by upstream dams. Geomorphology
Springer; 2021. p. 75–85. 2020;353:107011.
[38] Tang Z, Li Y, Gu Y, Jiang W, Xue Y, Hu Q, et al. Assessing Nebraska playa wetland [48] Vanama VSK, Mandal D, Rao YS. GEE4FLOOD: rapid mapping of FLOOD areas
inundation status during 1985–2015 using Landsat data and Google earth engine. using temporal Sentinel-1 SAR images with Google earth engine cloud platform.
Environ Monit Assess 2016;188(12):1–14. J Appl Remote Sens 2020;14(3):34505.
[39] Tay CWJ, Yun S-H, Chin ST, Bhardwaj A, Jung J, Hill EM. Rapid flood and damage [49] Voigt S, Kemper T, Riedlinger T, Kiefl R, Scholte K, Mehl H. Satellite image analysis
mapping using synthetic aperture radar in response to typhoon Hagibis, Japan. for disaster and crisis-management support. IEEE Trans Geosci Remote Sens 2007;
Scientific Data 2020;7(1):1–9. 45(6):1520–8. https://doi.org/10.1109/TGRS.2007.895830.
[40] Tiwari V, Kumar V, Matin MA, Thapa A, Ellenburg WL, Gupta N, et al. Flood [50] Xiong J, Thenkabail PS, Gumma MK, Teluguntla P, Poehnelt J, Congalton RG, et al.
inundation mapping-Kerala 2018; harnessing the power of SAR, automatic Automated cropland mapping of continental Africa using Google earth engine
threshold detection method and Google earth engine. Plos One 2020;15(8): cloud computing. ISPRS J Photogr Remote Sens 2017;126:225–44. https://doi.org/
e0237324. 10.1016/j.isprsjprs.2017.01.019.
[41] Toan TQ. Climate change and sea level rise in the Mekong Delta: Flood, tidal [51] Pulvirenti L, Pierdicca N, Chini M, Guerriero L. An algorithm for operational flood
inundation, salinity intrusion, and irrigation adaptation methods. In: Coastal mapping from Synthetic Aperture Radar (SAR) data using fuzzy logic. Natural
disasters and climate change in Vietnam. Elsevier; 2014. p. 199–218. Hazards and Earth System Sciences 2011;11(2):529–40.
[42] Tran DD, Dang MM, Du Duong B, Sea W, Vo TT. Livelihood vulnerability and
adaptability of coastal communities to extreme drought and salinity intrusion in

11