water

Article
Impact of Urbanization on Pluvial Flooding: Insights from a
Fast Growing Megacity, Dhaka
Md Shadman Sakib , Siam Alam, Shampa * , Sonia Binte Murshed, Ripan Kirtunia , M. Shahjahan Mondal
and Ahmed Ishtiaque Amin Chowdhury

Institute of Water and Flood Management (IWFM), Bangladesh University of Engineering and
Technology (BUET); Dhaka 1000, Bangladesh; shadman@[Link] (M.S.S.); siam94015@[Link] (S.A.);
sonia@[Link] (S.B.M.); ripankirtuniajust@[Link] (R.K.);
mshahjahanmondal@[Link] (M.S.M.); ishtiaquechowdhury@[Link] (A.I.A.C.)
* Correspondence: shampa_iwfm@[Link]

Abstract: The 400-year history of Dhaka says that the city once had several well-known natural
canals (khals) that drained stormwater and graywater. In addition to city’s combined sewer system,
these water bodies offered an essential natural drainage system that allowed to manage the monsoon
rainfall effectively. However, over the past three decades, due to rapid urbanization, these khals have
significantly depleted to the point where they are no longer capable of draining the city’s monsoon
runoff. Using past, present, and future Land Use and Land Cover (LULC) and urban drainage
modeling, this study identified the effects of such LULC change on pluvial flooding of the northern
part of the city. Analysis shows that the rapid and extensive changes in LULC over the past decades
have resulted in significant shrinkage of these khals, consequently leading to escalated rates of urban
flooding in this region. The western part of Turag thana, low-lying areas close to the Baunia Khal
depression, and the upstream region of Abdullahpur Khal are highly vulnerable to future urban
floods. The projected LULC change indicates an increase of 8.47%, 8.11%, and 4.05% in the total
inundation area by 2042 for rainfall events with return periods of 50 years, 25 years, and 2.33 years,
respectively. The findings also indicate that 11% more area is likely to experience long-duration
flooding due to LULC change.
Citation: Sakib, M.S.; Alam, S.;
Shampa; Murshed, S.B.; Kirtunia, R.; Keywords: LULC; pluvial flooding; SWMM; MOLUCSE; Dhaka city
Mondal, M.S.; Chowdhury, A.I.A.
Impact of Urbanization on Pluvial
Flooding: Insights from a Fast
Growing Megacity, Dhaka. Water 1. Introduction
2023, 15, 3834. [Link]
Pluvial floods in megacities are a serious issue worldwide, particularly for rapidly
10.3390/w15213834
growing cities [1–4]. Rapid urban development in these megacities has led to the unchecked
Academic Editor: Francesco De Paola conversion of surface water bodies, greeneries, and buffer zones to paved urban land,
thereby increasing its susceptibility to flooding [5]. It also alters the hydrological balance of
Received: 16 August 2023
the city’s intricate canal-stream network which forms the backbone of the natural drainage
Revised: 10 October 2023
system [6]. However, this is not the only prime factor that contributes to the waterlogging
Accepted: 11 October 2023
Published: 2 November 2023
problem, rather variables including heavy seasonal precipitation, rapid shift in built-up
land use patterns, inadequate drainage facilities, and ill waste management practice all
play critical roles [7–11]. Over the past four decades, South Asian countries have shown
a characteristic trend of unplanned urban expansion, which leads to water insecurities,
Copyright: © 2023 by the authors. environmental degradation, public health, and sanitation issues [12]. By 2050, it is expected
Licensee MDPI, Basel, Switzerland. that approximately 70% population of the world will have migrated to cities, with Asian
This article is an open access article and African continents accounting for 93% of the migration [13]. This demographic shift
distributed under the terms and adds to the already depleted environmental resource (green space 5%), particularly in
conditions of the Creative Commons densely populated South Asian capitals such as Dhaka, Kathmandu, and Karachi, and is
Attribution (CC BY) license (https://
expected to exaggerate the problem [14]. This paper focuses on such a problem in one of
[Link]/licenses/by/
the South Asian capitals-Dhaka in Bangladesh, home to over 22 million as of 2022 [15].
4.0/).

Water 2023, 15, 3834. [Link] [Link]

Water 2023, 15, 3834 2 of 26

The 400-year history of Dhaka shows that it was a low-lying floodplain of multiple
river systems [16].
It is surrounded by the Turag-Buriganga river system to the west and the Balu-
Shitalakhya rivers to the east [17]. Records show that, in the latter half of the 19th century,
Dhaka was subjected to severe fluvial flood events at regular intervals in 1954, 1955, 1962,
1966, 1974, 1987, 1988, and 1998. Among these 1988 flood was the most devastating. It
was triggered by excessive fluvial inflow from the surrounding rivers [18]. The flood
peak lasted for three weeks, inundating 85% of the city and setting record-high water
levels [19]. In response, the government initiated the ‘Dhaka City Integrated Flood Pro-
tection Project (DCIFPP)’ in 1991 to protect existing and future development in low-lying
built-up areas [20]. The project oversaw the construction of 30.2 km of dyke, 9.25 km of
flood wall, 12 numbers of floodproofing sluices, and the installation of 3 pumping stations
at Dholai Khal, Kalyanpur, and Goranchatbari. Before the construction of the embankment
following the flood in 1991, the city experienced several instances of pluvial flooding caused
by the Turag-Buriganga river system. Following the implementation of DCIFPP in 1991,
the city was safeguarded against flood occurrences. Therefore, currently, it is not prone to
pluvial floods.
However, rainwater that once flowed into the river was now trapped within em-
banked areas, resulting in localized flooding. This was due to poor maintenance of the
embankment’s flow control structures, blocked sluice gates from solid waste dumping, and
rampant urbanization which collectively disrupted Dhaka’s core drainage network [21].
As a consequence, in 1998, Dhaka suffered its worst pluvial flood resulting in very high
rainfall. Most of eastern Dhaka was inundated, along with 20% of the western part, which
was newly embanked under the flood action plan [19]. Over time, the majority of the sluice
gates became inoperable. Despite flood embankment in the west, the flooding problem had
not been solved; rather, the inundation dynamics had shifted from fluvial to pluvial floods.
Pluvial flood dynamics are linked to urbanization, which alters the natural land
surface through impervious artificial structures such as asphalt, concrete, brick, stone, and
buildings. This reduces the amount of rainfall-runoff that can infiltrate into the substrata,
resulting in flooding. Especially in the case of rapidly growing megacities like Dhaka, it is
pivotal to understand its Land Use and Land Cover (LULC) dynamics. Denser inner-city
areas with high runoff potential have nearly tripled in the last four decades [22]. A survey
by [23] shows the primary arterial canals in Dhaka to be infringed up to 40% of their
original demarcation. In addition, the widespread illegal encroachment of waterways and
ponds (small waterbodies) has reduced the amount of open space available for water to
flow and be stored [24]. The problem is aggravated by poor waste management practices
and rapid population growth which have put pressure on the already inadequate drainage
infrastructure [11]. These factors collectively reduce the city’s ability to manage the sudden
influx of runoff following moderate to heavy rainfall, leading to severe pluvial flooding.
As previously stated, the three pump stations, Kallyanpur, Goranchatbari, and Dholai
Khal, act as flood control structures to protect Dhaka. However, the retention ponds at
Kallyanpur and Dholai Khal have been significantly reduced in recent decades due to
aggressive encroachment, landfilling, and residential development [25].
This advertently reduces the runoff accumulation lead time, resulting in severe flood-
ing under medium to heavy rainfall. The only exception is the Goranchatbari Pump Station
(GPS), which still has a functioning retention pond of 676 acres [26]. It is responsible
for draining out stormwater from the north-western part of greater Dhaka (Figure 1),
i.e., Mirpur, Pallabi, Cantonment, Uttara, Diabari, and Tongi. Nevertheless, GPS is ex-
periencing immense pressure due to aggressive urbanization, which has compelled it to
give up a considerable portion of its retention pond for urban settlements and residential
expansion [27]. Therefore, it is critical to understand and predict the system’s response to
this disruptive LULC trajectory and understand its potential impact on escalating urban
flood vulnerability.
 Water 2023, 15, x FOR PEER REVIEW 3 of 27

expansion [27]. Therefore, it is critical to understand and predict the system’s response to
Water 2023, 15, 3834 3 of 26
this disruptive LULC trajectory and understand its potential impact on escalating urban
flood vulnerability.

Figure 1.
1. Map
MapofofGoranchatbari
Goranchatbaricatchment forfor
catchment drainage study.
drainage study.

The flood dangers

The dangers in in Dhaka
Dhaka were were evaluated
evaluated by by[28]
[28]using
usingthetheutilization
utilizationofofgeographic
geo-
graphic information
information systemssystems
(GIS) and(GIS) and remote
remote sensing sensing (RS) imagery.
(RS) imagery. This assessment
This assessment in- the
involved
volved the
analysis analysis
of flood of flood and
frequency frequency and flood
flood depth depth [Link]
[Link] on the flood onextent
the flood
maps extent
spanning
maps
the spanning
years 1988 to the2009,
yearsa 1988 to 2009,
frequency mapa frequency
was generatedmap was generatedthe
to illustrate to illustrate the by
areas affected
areas affected
floods. by floods.
The analysis The analysis
revealed that arevealed
minimum thatofa 23%
minimumof theoftotal
23%area
of the total
fell areathe
inside fell high-
inside the high-flooded zone. On the other hand, the flood depth
flooded zone. On the other hand, the flood depth map indicated that a significant portion, map indicated that a
significant portion, specifically 45%, of the surveyed region was characterized
specifically 45%, of the surveyed region was characterized by a substantial level of flooding. by a sub-
stantial
That levelproposes
study of flooding. That study
to create proposes to create
a comprehensive a comprehensive
rainfall-runoff modelrainfall-runoff
that incorporates
model that incorporates hourly or daily time stamps
hourly or daily time stamps in order to accurately capture the dynamics in order to accurately capture the and
of floods
dynamics of floods and their seasonal variations. In their study, Ref.
their seasonal variations. In their study, Ref. [29] utilized an integrated hydrologic and [29] utilized an inte-
grated hydrologic and hydrodynamic model to investigate the phenomenon of urban
hydrodynamic model to investigate the phenomenon of urban flooding in Dhaka City.
flooding in Dhaka City. However, it is important to note that their research primarily con-
However, it is important to note that their research primarily concentrated on the specific
centrated on the specific area of Segunbagicha kahl. [25] conducted an assessment of the
area of Segunbagicha kahl. [25] conducted an assessment of the efficacy of several flood
efficacy of several flood control measures implemented in the vicinity of Dhaka city. Their
control measures implemented in the vicinity of Dhaka city. Their objective was to identify
objective was to identify the factors contributing to both external and internal flooding of
the factors contributing to both external and internal flooding of the protected areas in
the protected areas in Dhaka during significant flood events that occurred in the recent
Dhaka during
past. Their significant
primary emphasis flood
was events
on fluvialthatflooding
occurred [Link] In
thetheir
recent past.
study, [Link] primary
[30] con-
emphasis
ducted an assessment of long-term urban surface water changes utilizing multi-year sat- of
was on fluvial flooding only. In their study, Ref. [30] conducted an assessment
long-term urban
ellite imagery. surface
Their water
findings changes
revealed theutilizing
occurrence multi-year satellitein
and transitions imagery. Their findings
surface water are
revealed
significantly influenced by the nature of the city’s expansion, seasonal variations, and ge-by the
the occurrence and transitions in surface water are significantly influenced
nature
ographic oflocations.
the city’sHowever,
expansion, theseasonal
study did variations,
not take intoandaccount
geographic locations.
the dynamics of However,
flooding the
study did not
and future landtake
[Link]
Ref. account
[31] focusthe ondynamics of flooding
the identification andprone
of areas future to land
wateruse. Ref. [31]
logging
focus on the identification of areas prone to water logging risks in Dhaka city and the
assessment of their flood susceptibility. The flood susceptibility map was generated in
order to assess the potential for waterlogging. This study also lacks the future projection of
land use.
Under these backdrops, this research investigates the pattern of LULC change in the
Goranchatbari sub-catchment and then identifies its potential impact on pluvial flooding.
Water 2023, 15, 3834 4 of 26

Specifically, we attempt to determine the pattern of change in LULC over the last five
decades; forecast the future LULC and finally examine how the pluvial flooding dynamics
are changing due to altered LULC using historical precipitation data and one-dimensional
(1D) urban drainage modeling.

Study Area
Dhaka, the capital of Bangladesh, is located between 23◦ 400 N to 23◦ 540 N latitude
and 90◦ 020 E to 90◦ 310 E longitude. It is the most densely populated megacity in the
world with an area of approximately 258.78 km2 . Its ground elevation varies between
1.5 and 15 m above Mean Sea Level (MSL), with an average elevation of 6 m [32]. The
city is divided into 41 administrative sub-divisions called ‘Thana’, which are managed by
two independent organizations, i.e., Dhaka South and Dhaka North City Corporations
(DSCC and DNCC). The greater Dhaka is divided into 10 major artificial catchments for
stormwater and drainage management [33]. Among these Goranchatbari catchment stands
to be the largest under DNCC covering an estimated area of 67.52 km2 . It stands out from
other drainage catchments in Dhaka city due to its environmental diversity. The catchment
still retains most of its natural canal pathways, despite aggressive urban development. In
addition, the presence of a comparatively larger amount of vegetation or greeneries and
waterbodies has been one of the prime reasons behind selecting it as the study area [34,34].
However, just like other parts of Dhaka, the waterbodies of this region are also facing
severe anthropogenic encroachments over the last decade [35].
The catchment (Figure 1) is drained by a 66 m3 /s capacity pumphouse located at
Goranchatbari. Its 676-acre ponding area stores stormwater runoff from Cantonment,
Turag, Kafrul, Mirpur, Shah Ali, Pallabi, and Uttara which is pumped out into the Turag
River. The study area consists of 11 major natural canals (Table 1) having a total length of
37.06 km. Although originally designed to serve as stormwater channels, these are subjected
to additional loads to carry industrial and domestic wastewater. Consequently, these canals
suffer from severe solid waste accumulation, and blockage and fail to function at their
design capacity [36]. The detailed status of several khals is discussed in Supplementary S1.

Table 1. Average width, length, and drainage area of major natural canals within the Goranchatbari
catchment.

SL Khal Name Average Width (m) Length (km) Contributing Thana

Cantonment, Turag, Kafrul,
1 Baunia Khal 22.50 9.61
Mirpur
2 Diabari Khal 81.28 6.89 Turag
3 Digun Khal 34.11 4.33 Turag
4 Eastern Housing Khal 14.65 4.27 Shah Ali, Mirpur, Pallabi
5 Mirpur Cantonment Khal 16.94 3.30 Pallabi, Turag
6 Abdullahpur Khal 74.84 2.64 Uttara, Turag
7 Uttara Lake Khal 58.46 2.34 Uttara
8 Shangbadik Colony Khal 6.32 1.27 Pallabi
9 Paris Road Khal 6.61 1.14 Pallabi
10 Arambag Khal 12.80 0.87 Pallabi, Shah Ali
11 Bayeshteki Khal 10.96 0.41 Pallabi, Kafrul

The present area of the Goranchatbari detention pond measures 636.5 acres. It serves as
a reservoir for waste and stormwater originating from the localities of Mirpur, Cantonment,
Airport, Pallabi, Shah Ali, and Uttara thana. In the dry season, the amassed wastewater
is discharged via gravitational force into the Turag River by means of a drainage sluice
positioned near the pump house. However, during the rainy season, it becomes necessary
to employ pumping mechanisms to transfer the excess stormwater into the Turag River, as
the river’s water level is significantly higher. The GPS comprises three pump houses, each
containing three vertical turbine pumps with an individual pumping capacity of 7.33 m3 /s.
 capacity of 7.33 m3/s. However, at the time of the study, one of the pumps in pump hous
1 was undergoing maintenance, resulting in a reduction of the current pumping capaci
to 58.64 m3/s. In the monsoon season, the pumps are operated to maintain a water level
3.5 m in the retention pond. This is achieved by running one or two pumps in a period
manner over the course of several days following a heavy precipitation event.
Water 2023, 15, 3834 Figure 2a shows pump house 1′s intake channel, while Figure 2b illustrates 5 of 26 the o
site solid waste disposal site situated adjacent to the intake channel within the premise
the pump station. Various forms of household waste, such as plastic bags, used furnitur
However, at theandtimedead tree study,
of the branches,
onealong
of thewith inorganic
pumps in pumpmaterials,
house-1enter
wastheundergoing
pumping channel an
maintenance, resulting in a reduction of the current pumping capacity to 58.64 mprovides
necessitate manual screening to prevent pump clogging. Figure 2c 3 /s. In thea view of th
monsoon season, the pumps are operated to maintain a water level of 3.5 m in the retention reductio
636.5-acre retention basin of the pump house, which has undergone a 50-acre
due to the
pond. This is achieved development
by running one orof two
a newly
pumpsconstructed residential
in a periodic manner area visible
over in the far right
the course
the image. This expansion took
of several days following a heavy precipitation [Link] after 2016 to meet the residential demands of the ci
by land reclamation from 0 the retention basin.
Figure 2a shows pump house 1 s intake channel, while Figure 2b illustrates the on-site
The average bed elevation of the retention basin ranges from −0.5 to −1.0 m PWD
solid waste disposal site situated adjacent to the intake channel within the premise of the
Additionally, an artificially excavated canal with a bed level of −1.5 m runs perpendicul
pump station. Various forms of household waste, such as plastic bags, used furniture,
to the pump station’s intake openings, facilitating the transfer of retained water from th
and dead tree branches, along with inorganic materials, enter the pumping channel and
reservoir to the pump intake channels. Figure 2d exhibits an image near the inlet of Digu
necessitate manual screening to prevent pump clogging. Figure 2c provides a view of the
Khal, which connects the Cantonment and Eastern Housing Area. Notably, a substanti
636.5-acre retention
amount basin of the pump
of floating wastehouse, which
and debris canhasbe undergone a 50-acre
observed lining reductionof the kha
the periphery
due to the development
During the time of the study, the water in these areas was stagnant andright
of a newly constructed residential area visible in the far of bioha
exhibited
the image. This ardous
expansion took place after 2016 to meet the residential demands of the city
quality. The detailed status of the khals in the study area is presented in Suppl
by land reclamation
mentaryfrom the retention
Section S1. basin.

Figure 2. Field Observation—GPS

Figure 2. Field Observation—GPS and Digun Khal: and Digunhouse-1
(a) pump Khal: (a) pump
intake house-1
channel atintake channel
GPS (b) solid at GPS (
solid waste disposal site adjacent to the pump house-2 at GPS; (c) GPS retention pond and RAJU
waste disposal site adjacent to the pump house-2 at GPS; (c) GPS retention pond and RAJUK Uttara
Uttara Apartment Project residential area in the upper right quadrant (d) Solid waste and floatin
Apartment Project residential
debris area inatthe
accumulation theupper rightofquadrant
periphery (d) Solidbasin.
the GPS retention waste and floating debris
accumulation at the periphery of the GPS retention basin.

The average bed elevation of the retention basin ranges from −0.5 to −1.0 m PWD.
Additionally, an artificially excavated canal with a bed level of −1.5 m runs perpendicular
to the pump station’s intake openings, facilitating the transfer of retained water from
the reservoir to the pump intake channels. Figure 2d exhibits an image near the inlet
of Digun Khal, which connects the Cantonment and Eastern Housing Area. Notably, a
substantial amount of floating waste and debris can be observed lining the periphery of
the khal. During the time of the study, the water in these areas was stagnant and exhibited
biohazardous quality. The detailed status of the khals in the study area is presented in
Supplementary Section S1.

2. Materials and Methods

The research is divided into two parts: detecting changes in LULC and assessing
the impact of changed LULC on pluvial flooding using 1D modeling. Satellite image
analysis revealed the evolution of LULC over the last five decades. The spatial variable (i.e.,
Waterbodies, Green areas or builtup areas) for LULC was identified while analyzing the
LULC, and future land cover was predicted using these variables. From the last 66 years
of measured precipitation, 2.33-, 25-, and 50-year return period rains were calculated, and
 2. Materials and Methods
The research is divided into two parts: detecting changes in LULC and assessing the
impact of changed LULC on pluvial flooding using 1D modeling. Satellite image analysis
revealed the evolution of LULC over the last five decades. The spatial variable (i.e., Wa-
Water 2023, 15, 3834 terbodies, Green areas or builtup areas) for LULC was identified while analyzing 6 of 26 the
LULC, and future land cover was predicted using these variables. From the last 66 years
of measured precipitation, 2.33-, 25-, and 50-year return period rains were calculated, and
thesereturn
these returnperiod
period rains
rains were
were used
used to calculate
to calculate the impact
the impact of pluvial
of pluvial floodingflooding onuse
on land land use
change. Statistically, the 2.33-year return period flood is considered as an average [Link].
change. Statistically, the 2.33-year return period flood is considered as an average
Therefore,
Therefore, thethe 2.33-year
2.33-year storm
storm event
event was was
chosenchosen
as theas the reference
reference storm to storm to establish
establish a control a con-
trol case.
case. It alsoItaligns
also aligns with
with the theperiod
return returnofperiod of the
the mean mean
annual annual
flood, which flood, which
is also is also 2.33
2.33 years.
Figure 3 shows the overall methodology of the study.
years. Figure 3 shows the overall methodology of the study.

[Link]
Figure Theoverall
overall methodology
methodology of the
of the study.
study.

2.1. Assessment of LULC

2.1. Assessment of LULC
The analysis of Landsat satellite images from the years 1973, 1983, 1993, 2003, 2013, and
The analysis of Landsat satellite images from the years 1973, 1983, 1993, 2003, 2013,
2022 is conducted to detect changes in LULC. The RGB composite images were subjected
and 2022 is conducted to detect changes in LULC. The RGB composite images were sub-
to classification using a clustering algorithm that relied on their spectral features. Each
jected to classification using awith
year, the iteration commenced clustering
a total ofalgorithm that relied
thirty courses on their spectral
and concluded features.
with three
Each year,
classes the iteration
representing commenced
waterbodies, greenwith a total
areas, andof thirty courses
built-up and each
areas. After concluded with three
processing
classes representing waterbodies, green areas, and built-up areas. After
iteration, the clusters that were formed underwent labeling based on the class to which each processing
iteration,
they might the clusters that
be assigned. This were formed
labeling underwent
procedure labeling using
was conducted basedimage
on the class to which
processing
they mightavailable
capabilities be assigned. ThisThe
in ArcGIS. labeling procedure
methodology was assessment
of LULC conductedisusingshown image processing
in Figure 4.
Accuracy assessment
capabilities availablewas performed
in ArcGIS. Theusing the Kappaofstatistic
methodology by [37]. As the
LULC assessment reference
is shown in Figure
dataset, Google
4. Accuracy Earth Pro was
assessment images were used
performed to determine
using the Kappathe accuracy
statistic of theAs
by [37]. classified
the reference
result.
dataset, Google Earth Pro images were used to determine the accuracy of thewith
Randomly 60 points were generated in several LULC types and were compared classified
the reference dataset. The Kappa coefficient was 0.85 which indicates a
result. Randomly 60 points were generated in several LULC types and were compared good agreement
according to [37]. This assessment was done for the land use of the year 2022. This is due to
with the reference dataset. The Kappa coefficient was 0.85 which indicates a good agree-
the fact that the future LULC was validated using the LULC of 2022. The confusion matrix
ment according to [37]. This assessment was done for the land use of the year 2022. This
is shown in Table 2.
is due to the fact that the future LULC was validated using the LULC of 2022. The confu-
sion matrix is shown in Table 2.
 Water 2023, 15, x FOR PEER REVIEW 7 of
Water 2023, 15, 3834 7 of 26

Figure 4. Overall Figure 4. Overall

Workflow of theWorkflow of the Landuse Analysis.
Landuse Analysis.

Table 2. Accuracy assessment of LULC.

Table 2. Accuracy assessment of LULC.
Class Value Waterbodies Green Area Built-Up Area User Accuracy
Class Value Waterbodies Green Area Built-Up Area User Accuracy
Waterbodies 15.00 2.00 0.00 0.88
Green AreaWaterbodies 0.00 15.00 15.00 2.00 1.00 0.00 0.88
0.94
Built-up AreaGreen Area 0.00 0.00 2.00 15.00 15.00 1.00 0.94
0.88
Built-up Area 0.00 2.00 15.00 0.88
Producer Accuracy 1.00 0.79 0.94 0.90
Producer Accuracy 1.00 0.79 0.94 0.90
Kappa 0.85
Kappa 0.85
2.2. LULC Prediction
2.2. LULC Prediction This study applied a hybrid Cellular Automata (CA)—Artificial Neural Netwo
This study(ANN)
applied a hybrid
based Cellular Automata
Multi-Objective (CA)—Artificial
Land Use Change Simulation Neural Network
Environment (MOLUCS
(ANN) based Multi-Objective Land Use Change Simulation Environment (MOLUCSE) mode
model [38–42] in QGIS to predict the future LULC change. Similar mathematical
model [38–42] ine.g., Markov
QGIS Chain Model
to predict (MCM)
the future [43,44],
LULC Cellular
change. Automata
Similar (CA) [45,46],
mathematical Artificial Ne
mod-
els e.g., Markov Chain Model (MCM) [43,44], Cellular Automata (CA) [45,46], Artificial have bee
ral Network (ANN) [38,47], and advanced remote sensing techniques [22,48]
Neural Network applied
(ANN) to [38,47],
predict theandLULC changes
advanced in other
remote studies.
sensing Landsat satellite
techniques [22,48]imagery
have for 197
been applied to predict the LULC changes in other studies. Landsat satellite imagery forsupervise
1980, 1993, 2003, 2013, and 2022 is analyzed for LULC change using the ArcGIS
classification
1973, 1980, 1993, 2003, 2013,[49]. andThe elements
2022 of the LULC
is analyzed are clustered
for LULC change basedusingonthetheir spectral chara
ArcGIS
teristics. Following
supervised classification [49]. The that, pixels were
elements of thegrouped
LULC based on their based
are clustered spectralonvalues,
their and noi
was filtered
spectral characteristics. out using
Following that,image
pixelsprocessing
were groupedtools based
in ArcGIS. The spectral
on their classification
values,was verifie
using the field observations around the retention pond
and noise was filtered out using image processing tools in ArcGIS. The classification was of Goranchatbari pumphous
low-lying areas of Baunia khal, and highly built-up areas of Diabari, Uttara, Pallab
verified using the field observations around the retention pond of Goranchatbari pump-
Mirpur, and Vashantech. Subsequently, the images were reclassified into three categorie
house, low-lying areas of Baunia khal, and highly built-up areas of Diabari, Uttara, Pallabi,
i.e., waterbodies and canals, vegetation, and settlements plus bare lands [50]. These da
Mirpur, and Vashantech. Subsequently, the images were reclassified into three categories,
were fed into MOLUCSE along with a digital elevation model, population density, an
i.e., waterbodies and canals, vegetation, and settlements plus bare lands [50]. These data
distances from roads and waterbodies raster [46,51] to predict the future LULC pattern f
were fed into MOLUCSE along with a digital elevation model, population density, and
the Goranchatbari catchment. These spatial input variable for LULC prediction over th
distances from roads and waterbodies raster [46,51] to predict the future LULC pattern for
Goranchatbari catchment area is shown in Figure 5. Whereas, Figure 6 shows the wor
the Goranchatbari
flowcatchment.
on which the These spatialofinput
prediction future variable
land usefor LULC
was prediction
prepared over the algorith
using CA-ANN
Goranchatbari catchment area is shown in Figure 5. Whereas, Figure 6 shows the workflow
in QGIS. Number of hidden layers used in this step was 10. The rest of the parameters lik
on which the prediction
neighborhood,of future
number land of use was prepared
samples, using CA-ANN
etc. were prepared through aalgorithm in metho
trial-and-error
QGIS. Number to offind
hidden layers used in this step was 10. The rest of the parameters
the best possible case in training the model. The year 2042 was chosen to assess th like
neighborhood, number
drainage of samples,immediately
condition etc. were prepared throughthe
after achieving a trial-and-error method2041”. Th
national goal “Vision
to find the best possible case in training the model. The year 2042 was chosen to assess
the drainage condition immediately after achieving the national goal “Vision 2041”. The
Bangladesh Government has a plan to achieve High Income Country status by 2041 which
 Water 2023, 15, x FOR PEER REVIEW 8 of 27
Water 2023,
Water 15,15,
2023, 3834
x FOR PEER REVIEW 8 of 278 of 26

Bangladesh Government has a plan to achieve High Income Country status by 2041 which
Bangladesh
is
is why
why the Government
the prediction
prediction has afor
period
period plan
forthetoland
the achieve
land
use High
use was
was Income
set set at Country
2042
at 2042 status bythe
to incorporate 2041
to incorporate which
the possible
possible
is why
negativethe prediction
sides of period
wealth and for the land
urbanization use was set
[52,53].
negative sides of wealth and urbanization [52,53]. at 2042 to incorporate the possible
negative sides of wealth and urbanization [52,53].

Figure 5.
Figure 5. Spatial
Spatialvariable
variablefor
forLULC
LULCprediction using
prediction MOLUSCE
using plugin
MOLUSCE in QGIS.
plugin in QGIS.
Figure 5. Spatial variable for LULC prediction using MOLUSCE plugin in QGIS.

Figure 6. CA ANN Model Algorithm Work Flow.

Figure 6. CA
Figure CAANN
ANNModel
ModelAlgorithm
AlgorithmWork
WorkFlow.
Flow.

Kappa Coefficient (κ) was the determinant factor in assessing the accuracy of the
prediction [54,55]. The model parameters are set to have a high κ value. According
to [56,57] a model is considered to be of high accuracy for prediction if its kappa value
is over 0.8, while a value in the range of 0.61 to 0.8 is seen as being acceptable for use in
making future predictions. The observed and predicted LULC using MOLUSE is shown in
 Water 2023, 15, x FOR PEER REVIEW 9 of 27

Kappa Coefficient (κ) was the determinant factor in assessing the accuracy of the pre-
Water 2023, 15, 3834 diction [54,55]. The model parameters are set to have a high κ value. According 9 of 26 to [56,57]
a model is considered to be of high accuracy for prediction if its kappa value is over 0.8
while a value in the range of 0.61 to 0.8 is seen as being acceptable for use in making future
predictions. The observed and predicted LULC using MOLUSE is shown in Figure 7. for
Figure 7. for which
which the
the validation kappawas
validation kappa wascalculated
calculatedtotobebe0.625.
0.625. The
The prediction
prediction waswasfound to be
found to be representative
representativewith
withan
anoverall
overall accuracy
accuracy of of 86.46%.
86.46%.

Figure 7. LULCFigure 7. LULC

classification classification
validation validation
between between
observed observed
Landsat Landsat
data (2022) and data (2022) and MOLUSCE-
MOLUSCE-based
based
simulated case (2022).simulated case (2022).

2.3. Urban
2.3. Urban Drainage Drainage
Modeling andModeling and its Validation
its Validation
Governing Equations
Governing Equations
In Personal
In this study, the this study, the Personal
Computer Computer
Storm Water Storm Modeling
Management Water Management
(PCSWMM)Modeling
(PCSWMM) software (Version 7.4.3240) is used to model the urban drainage
software (Version 7.4.3240) is used to model the urban drainage system of the Goranchatbari system of the
catchment. The hydrodynamic component of the PCSWMM model is based on a one-model is
Goranchatbari catchment. The hydrodynamic component of the PCSWMM
dimensional SaintbasedVenant
on a one-dimensional Saint Venant
momentum equation momentum
[58] for equation
simulating [58] for
gradually simulating
varied and gradu-
unsteady flow conditions. Manning’s equation [59] was applied to assess the hydraulic to assess
ally varied and unsteady flow conditions. Manning’s equation [59] was applied
characteristicsthe of hydraulic characteristics
the conduit flow, withofManning’s
the conduit “n”
flow,value
with Manning’s
serving as“n”thevalue serving as the
primary
primary parameter for calibration. To model the infiltration of rainfall, the Green-Ampt
parameter for calibration. To model the infiltration of rainfall, the Green-Ampt model
model is employed, while for flood routing, a dynamic routing algorithm proposed by
is employed, while for flood routing, a dynamic routing algorithm proposed by [60] is
[60] is utilized. These are part of dynamic wave analysis in EXTRAN [61]. The next para-
utilized. These are part of dynamic wave analysis in EXTRAN [61]. The next paragraphs
graphs demonstrate the governing equations used in the model.
demonstrate the governing equations used in the model.
=𝑟+𝑖−𝑒−𝑓−𝑄 (1)
∂d
= r+i−e− f −Q (1)
Equation (1) is the
∂t hydrologic governing equation. It is derived from the Conserva-
tion of Mass Law [62]. Here, the change of depth per unit of time in sub-catchments is
Equation expressed
(1) is the hydrologic
in the formgoverning equation.
of r, i, e, f, and It is
q. r is the derived
rate from the
of upstream Conservation
run-on, i is the intensity of
of Mass Law [62]. Here,
rainfall, the
e is thechange
rate ofofevaporation,
depth per unit of time
f is the rate in
of sub-catchments
infiltration, and is q isexpressed
the rate of runoff
in the form of r, i, e, f, and q. r is the rate of upstream run-on, i is the
All of them can be expressed as flow rates per unit area (m/s). In the 1D model intensity of rainfall, e iseach sub-
the rate of evaporation, f is the rate of infiltration, and q is the rate of runoff.
catchment is treated as a nonlinear reservoir. It is assumed that the capacity of this reser- All of them
can be expressed
voiras flow
is the rates perwhen
maximum unit the
areatopographic
(m/s). In the 1D model
depression Surface runoff 𝑄, occurs
each sub-catchment
is higher.
is treated as awhen
nonlinear reservoir. It is assumed that the capacity
the depth of water in the sub-catchment exceeds the maximum of this reservoir is the storage
depression
maximum when the topographic depression is higher. Surface runoff Q, occurs when the
depth of water in the sub-catchment exceeds the maximum depression storage. The depth
of water over the sub-catchment is updated in real time by numerically solving a water
balance equation over the sub-catchment.

∂A ∂Q
+ =0 (2)
∂t ∂x
Water 2023, 15, 3834 10 of 26

Equation (2) is the continuity equation where x is referred to as distance, t is time, A is

the flowing cross-sectional area, and Q is the flow rate. It is also the hydraulic governing
equation of PCSWMM.
 2
Q
∂Q ∂ A ∂H
+ + gA + gAS f + gAhl = 0 (3)
∂t ∂x ∂x
Equation (3) is the 1D Saint Venant momentum equation. Here, H is the hydraulic
head of water in the conduit which is the summation of invert elevation and water depth,
and Sf is the friction slope. Manning’s equation is used to determine the flow inside the
conduits where Manning’s coefficient represents the roughness parameter, as shown in
Equation (4). Here, Q is the flow in the conduit, the cross-sectional area of the conduit is
marked as A, the hydraulic radius is denoted as R, and S is the slope of the ends of the
conduits. The roughness coefficient, n is used as a calibration parameter.

1.49 2 1
Q= AR 3 S 2 (4)
n
hl is the local energy loss per unit length of conduit. In the case of solving a network
of conduits, an additional continuity relationship is needed for the junction nodes that
connect two or more conduits together. To solve this a continuous water surface is assumed
to exist between the water elevation at the node and in the conduits that enter and leave
the node. The change in hydraulic head H at the node with respect to time was calculated
using Equation (5).
∂H ∑Q
= (5)
∂t An + ∑ Ac
Here where An is the surface area of the node, ∑ Ac is the surface area of the conduits
connected to the node, and ∑ Q is the net flow into the node contributed by all conduits
connected to the node as well as any externally imposed inflows. Dynamic Wave routing
was utilized because it couples the solution for both water levels at nodes and flow in
conduits. For calculation of infiltration Green-Ampt Infiltration Method has been implicated
in this study [63]. This method for modeling infiltration assumes that a sharp wetting front
exists in the soil column, separating soil with some initial moisture content below from
saturated soil above. Equation (6) shows the basic equation of the Green-Ampt method.
Here, f is the rate of infiltration which is expressed in terms of K, ∅, θ i , ψ, and F. K is the
hydraulic conductivity, ∅ is the porosity, θ i is the initial water content, ψ is the wetting
front soil suction head, and F is the cumulative infiltration.

(∅ − θi )ψ
 
f = K 1+ (6)
F

To model pumps that connect two nodes together, the flow is calculated based on the
heads at their end nodes.
Model Schematization
The topography of the urban drainage model is based on the Copernicus Digital
Elevation Model (COP-30 DEM). It provides elevation data with a spatial resolution of 30
m. The data source for precipitation, LULC analysis, and projection is listed in Table 3. A
professional survey team equipped with echo sounder and survey equipment was assigned
to measure canal depth and cross-section. This survey conducted in this study consisted of
a total of 126 cross-sections, which were distributed among 11 major canals listed in Table 1
and whose location is shown in Figure 1. The distribution of surveyed points within each
canal was determined based on canal length, alignment, and variations in width. This was
guided by engineering judgement and was further substantiated by multiple field visits.
The spacing between surveyed points ranged from 40 m to 900 m on an average of 300 m.
But it is crucial to note that our survey strategy was designed to ensure comprehensive
Water 2023, 15, 3834 11 of 26

coverage of all significant features of the canals, including abrupt constrictions and width
variations. PCSWMM hydraulic model does not facilitate interpolation between channel
cross-sections, therefore, a uniform open channel cross-section was considered between the
surveyed points.

Table 3. Input data used and their sources for urban drainage modeling.

Data Type Data Source Function

Daily Rainfall data Bangladesh Meteorological
Frequency analysis
(1965–2020) Department (BMD)
Bangladesh Meteorological
3 Hourly Rainfall (2020) Model validation
Department (BMD)
Model input as
Sentinel-2 Image NASA Earth Data
percent impervious
Landsat 4, 5, 8, 9 Image USGS Earth Explorer LULC analysis and prediction
Digital elevation Open Topography
Model DEM input
model (DEM) (Copernicus 30)
Population and Housing
Population Density LULC Prediction
Census 2011
Road Network Google Earth Pro LULC Prediction
Canal Cross-Section
Bathymetry Survey Model input
(April 2022)

This drainage network is shown in Figure 8. which illustrates sub-catchment boundaries,

the layout of the natural canal (surveyed cross-sections) as well the primary artificial conduits
within the study area. The model consists of 76 sub-catchments, which represent different
areas of land that contribute to the overall drainage system. It has a total of 374 junctions,
which represent the points in the system where the flow converges and diverges. It also
includes one outfall at the Goranchatbari pumphouse (Figure 8), opposite the Turag River. A
pump element is used to simulate the effect of the pump station on the drainage system. It
simulates the effect of pumping water from lower areas within the embanked Goranchatbari
catchment to the Turag River to maintain water levels below 3.5 m.
Land use parameters, such as imperviousness and storage capacity, are calculated
from the LULC analysis. These parameters are used to estimate the amount of runoff
generated by different land uses and to model the behavior of storage facilities, such as
ponds and tanks. Calibration parameters, such as Manning’s roughness coefficient (n)
and storage coefficients, are used to adjust the model to match the observed behavior of
the drainage system. The n value was taken as 0.04 for the canal bed and 0.06 for the
banks. This indicates that the canal bed had a smoother surface than the banks, which
would result in less resistance to flow in the channel. The higher coefficient for the banks
suggests that the banks were rougher, which would increase the resistance to flow. This
was considered to take into account the effect of significant pollution and waste disposal in
the canal system [64]. The pollutants can build up on the channel bed and banks, making
them more resistant to flow. A list of input parameters for the PCSWMM model is given in
Table 4. In addition to these parameters, the study also reported flow routing and continuity
errors below 1%. These errors were within acceptable limits and suggest that the model
was able to accurately simulate the flow of water through the drainage system [65].
Water 2023, 15, 3834 12 of 26
Water 2023, 15, x FOR PEER REVIEW 12 of 27

Figure
[Link]
PCSWMM model: Goranchatbari
model: drainage
Goranchatbari network
drainage (natural
network canalscanals
(natural and artificial conduits)
and artificial conduits)
and sub-catchment boundary.
and sub-catchment boundary.

Land use parameters, such as imperviousness and storage capacity, are calculated
Table 4. PCSWMM model parameters and methods.
from the LULC analysis. These parameters are used to estimate the amount of runoff gen-
erated by different land uses and to model the behavior of storage facilities,
Parameter Valuessuch as ponds
and tanks. Calibration parameters, such as Manning’s roughness coefficient (n) and stor-
Junction baseline flow 0.002 m3 /s
age coefficients, are used to adjust the model to match the observed behavior of the drain-
age system. The Sub-catchment
n value was slope
taken as 0.04 for the canal bed and 0.060.002%
for the banks. This
indicates that the canal bedmethod
Infiltration had a smoother surface than the banks,Green-Ampt
which would result in
less resistance to flow in the channel. The higher coefficient for the banks suggests that the
Suction head 3.5 mm
banks were rougher, which would increase the resistance to flow. This was considered to
take into accountSoil
theconductivity 0.5 in
effect of significant pollution and waste disposal mm/h
the canal system
[64]. The pollutants can build
Routing methodup on the channel bed and banks,Dynamic
makingWavethem more re-
sistant to flow. A list of input parameters for the PCSWMM model is given in Table 4. In
Force equation Hazen-Williams
addition to these parameters, the study also reported flow routing and continuity errors
below 1%. TheseSurcharge
errors weremethod
within acceptable limits and suggest thatExtran
the model was able
to accurately simulate the flow of water through the drainage system [65].
2.4. Boundary Condition
Table 4. PCSWMM model parameters and methods.
The rainfall frequency analysis was done to calculate the total precipitation for 2.33-,
25-, and 50-years Parameter
return period to use as the boundary conditionsValues for three different
cases. The yearly maximum precipitation
Junction baseline flow between 1965–2020 was
0.002 fed into the Gumbel
m3/s
distribution [66] to generate rainfalls
Sub-catchment slope of 2.33-, 25-, and 50-years
0.002% return period. The
of the
24-h rainfall hydrographs (Figure 9) were generated based on
Infiltration method the [67] reports for Dhaka’s
Green-Ampt
stormwater improvement
Suction headproject. The precipitation attained3.5 itsmm
maximum intensity of
149.28 mm for the 50-year return period, while for the 25-year return period, this value was
recorded as 118.04 mm. Similarly, for the 2.33-year return period, the precipitation peak
was found to be 59.66 mm.
 The yearly maximum precipitation between 1965–2020 was fed into the Gumbel distribu-
tion [66] to generate rainfalls of 2.33-, 25-, and 50-years of the return period. The 24-h rain-
fall hydrographs (Figure 9) were generated based on the [67] reports for Dhaka’s storm-
water improvement project. The precipitation attained its maximum intensity of 149.28
Water 2023, 15, 3834 mm for the 50-year return period, while for the 25-year return period, this value was 13 rec-
of 26
orded as 118.04 mm. Similarly, for the 2.33-year return period, the precipitation peak was
found to be 59.66 mm.

Figure9.9. Rainfall
Figure Rainfall hyetograph
hyetograph for
for 2.33,
2.33, 25,
25, and
and 50
50 Years
Years of
of Return
Return Period.
Period.

3.3. Results
Results and
and Discussions
Discussions
3.1. Validation Results
3.1. Validation Results
The
The GPS gaugewas
GPS gauge wasthe the only
only source
source of observed
of observed water water
level level
data todata to validate
validate the
the drain-
drainage model. The validation period covered 25 days, from
age model. The validation period covered 25 days, from 4 May to 28 May 2020. This 4 May to 28 May 2020. This
timeframe
timeframe allowed
allowed us us to
to analyze
analyze aa series
series of of events: initially, there
events: initially, there was
was sporadic
sporadic rainfall
rainfall
from 6 May 2020, to 18 May 2020, followed by two significant rainfall
from 6 May 2020, to 18 May 2020, followed by two significant rainfall events on 20 May events on 20 May and
24 May 2020. We chose this validation window to accurately represent
and 24 May 2020. We chose this validation window to accurately represent the water level the water level at
the GCGC
at the pump
pump house
house throughout
throughout thisthis
entire period.
entire It’sIt’s
period. important
important to to
notenotethat our
that ourchoice
choiceof
this timeframe was also influenced by the availability of reliable water
of this timeframe was also influenced by the availability of reliable water level records at level records at GPS.
[Link] has a total of 8 fully functional pumps, each capable of draining out water at a flow
rate ofGPS 3 /s, giving the pump station a total capacity of 66 m3 /s. However, at any given
8 mhas a total of 8 fully functional pumps, each capable of draining out water at a
time,
flow only one
rate of 8m pump is keptthe
3/s, giving operational.
pump station According to the official
a total capacity of 66 statement,
m3/s. However, the pumpat any is
turned on when the water level reaches 3.5 m. To reflect this in
given time, only one pump is kept operational. According to the official statement, thethe PCSWMM model, the
pump
pumpelement
is turned wasonset
whento have a capacity
the water levelof 8 m3 /s,
reaches 3.5with
m. Toa conditional
reflect this operational
in the PCSWMM clause
that would
model, the turn
pump onelement
the pump as set
was soonto as
havethe awater levelofrose
capacity 8 mabove 3.5 am.
3/s, with However, opera-
conditional during
the
tional clause that would turn on the pump as soon as the water level rose above 3.5 not
month of May 2020, the pump operation log revealed that the pump operator did m.
always
However,turnduring
on andthe offmonth
the pump at the
of May set threshold
2020, the pumpofoperation
3.5 m. Instead, the pump
log revealed thatwas
the turned
pump
on when the
operator didwater level reached
not always turn onaand rangeoffbetween
the pump 3.25atand
the 3.75 m. Additionally,
set threshold of 3.5 m. there were
Instead,
times when more than one pump was operational for a short
the pump was turned on when the water level reached a range between 3.25 and 3.75 m. period of time in order to
reduce the water level rapidly during heavy precipitation events. This variability in pump
operations cannot be simulated by the pump element incorporated in the PCSWMM model,
and as a result, a constant draw out of 8 m3 /s was assumed for the entire model validation
period. Even with these practical limitations, the model responds to the sudden influx of
rainwater during high precipitation events. In addition, for the simulation period from
4–28 May 2020, Figure 10a shows the correlation coefficient of the model simulated and
observed data to be 0.84. Figure 10b shows the sub-catchment runoff coefficient with high
accuracy having an R2 value of 0.99. This analysis was done by manually calculating the
runoff coefficients of different sub-catchments using land use data and Table 5. The per-
centage of different land type was the driving factor in calculating this and later comparing
it with the model simulated ones [68,69].
WaterWater 15, 3834
2023,2023, 15, x FOR PEER REVIEW 15 of1427of 26

(a) (b)

Figure10.
Figure 10. Model
ModelValidation: (a) Correlation
Validation: between
(a) Correlation observed
between and simulated
observed water level water
and simulated (24-h in-
level
terval)
(24-h at GPS;at(b)
interval) Correlation
GPS; betweenbetween
(b) Correlation observedobserved
and simulated runoff coefficient
and simulated runoff over the Go-over
coefficient
ranchatbari catchment.
the Goranchatbari catchment.
Table 6. Statistical goodness-of-fit test scores of numerical model results.
Table 5. Runoff Coefficient Chart by Ohio Department of Transportation Hydraulics Manual.
Statistical Parameter WL Validation Runoff Coeff. Validation
PBIAS (%) Runoff Coefficients for Rational Method
3.494 1.573
NSE 0.748Flat Rolling 0.988 Hilly
RSR
Pavement & Roofs 0.4910.90 0.90 0.110 0.90
Earth Shoulders 0.50 0.50 0.50
3.2. LULC Change
Drivers & Walks 0.75 0.80 0.85
FigureGravel Pavement
11 depicts 0.85
the Goranchatbari catchment’s classified 0.85 0.85
LULC maps (1973–2022),
City Business Areas 0.80 0.85
which are divided into three groups: waterbodies and canals, vegetation, and settlements, 0.85
Apartment Dwelling Areas 0.50 0.60 0.70
and bare lands. Figure 11a–c depicts the spatial patterns of LULC changes in 1973, 1980,
Light Residential: 1 to 3 units/acre 0.35 0.40 0.45
and 1993. Natural vegetation, as well as canal and water bodies, were the dominant land
Normal Residential: 3 to 6 units/acre 0.50 0.55 0.60
use types
Dense during this
Residential: period,
6 to while the trajectory
15 units/acre of urban expansion
0.70 0.75 (settlement and0.80bare
land) shows a Lawns
southward progression toward0.17 Pallabi and Kafrul.
0.22 Figure 11c,d depicts
0.35
how, between 1993 and
Grass Shoulders 2003, the built-up or settlement
0.25 area replaced
0.25 the majority of the
0.25
water bodies,
Side canals, and low-lying depressions,
Slopes, Earth 0.60 as well as the0.60
natural vegetation.0.60This
resulted inSide
the Slopes,
depletion
Turfof wetland areas, which0.30 serve as ecological
0.30 buffers and0.30 runoff
Median
sinks during Areas,
heavy Turf
precipitation 0.25
events. The north-eastern 0.30 region experienced
settlement 0.30
Cultivated
significant Land, Clay
expansion, & Loamat the expense0.50
primarily 0.55The analysis of0.60
of vegetative land. LULC
Cultivated
changes in theLand, Sand & Gravel
Goranchatbari catchment from 0.25 0.3011d,e) shows that
2003 to 2013 (Figure 0.35 the
Industrial Areas, Light 0.50 0.70 0.80
settlement area increased by 23.5 km2 or more than 2.3 km2/year on average. Prior to this
Industrial Areas, Heavy 0.60 0.80 0.90
rapid increase in settlement, urban growth averaged 0.19 km2/year from 1973 to 1993. In
Parks & Cemeteries 0.10 0.15 0.25
fact, betweenPlaygrounds
2003 and 2013, the area of vegetation0.20 decreased by 0.25
60 percent, from 25.7
0.30km
2

(19.24 percent) to 10.6 km

Woodland & Forests
2. Similarly, the area of waterbodies and canals fell to 9.8 km2
0.10 0.15 0.20
(17.79 Meadows
percent) in& 2013 from
Pasture an already low 17.7
Land km2 (32.12 percent)
0.25 0.30 in 2003. 0.35
Unimproved Areas 0.10 0.20 0.30
Rolling = ground slope between 2% to 10%
Hilly = ground slope greater than 10%

Table 6 shows the values of Percent bias (PBIAS), Nash-Sutcliffe efficiency (NSE), and
RMSE-observations standard deviation ratio (RSR), and all of these scores fell into the
category of satisfactory range. Therefore, the results indicate that the model is reliable and
can be used to study the extreme scenario response and urban flooding.
Water 2023, 15, 3834 15 of 26

Table 6. Statistical goodness-of-fit test scores of numerical model results.

Statistical Parameter WL Validation Runoff Coeff. Validation

PBIAS (%) 3.494 1.573
NSE 0.748 0.988
RSR 0.491 0.110

3.2. LULC Change

Figure 11 depicts the Goranchatbari catchment’s classified LULC maps (1973–2022),
which are divided into three groups: waterbodies and canals, vegetation, and settlements,
and bare lands. Figure 11a–c depicts the spatial patterns of LULC changes in 1973, 1980,
and 1993. Natural vegetation, as well as canal and water bodies, were the dominant land
use types during this period, while the trajectory of urban expansion (settlement and bare
land) shows a southward progression toward Pallabi and Kafrul. Figure 11c,d depicts
how, between 1993 and 2003, the built-up or settlement area replaced the majority of the
water bodies, canals, and low-lying depressions, as well as the natural vegetation. This
resulted in the depletion of wetland areas, which serve as ecological buffers and runoff
sinks during heavy precipitation events. The north-eastern settlement region experienced
significant expansion, primarily at the expense of vegetative land. The analysis of LULC
changes in the Goranchatbari catchment from 2003 to 2013 (Figure 11d,e) shows that the
settlement area increased by 23.5 km2 or more than 2.3 km2 /year on average. Prior to this
rapid increase in settlement, urban growth averaged 0.19 km2 /year from 1973 to 1993. In
fact, between 2003 and 2013, the area of vegetation decreased by 60 percent, from 25.7 km2
(19.24 percent) to 10.6 km2 . Similarly, the area of waterbodies and canals fell to 9.8 km2
(17.79 percent) in 2013 from an already low 17.7 km2 (32.12 percent) in 2003.
According to the 2003 LULC analysis (Figure 11d), the areas around Uttara Khal in the
northeast and the Eastern Housing Khal in the lower southwest experienced a significant
increase in settlement growth between 1993 and 2003. Despite this rapid settlement growth,
Figure 11d shows that in 2003, a significant proportion of the drainage basin (78.77 percent)
was still a central waterbody, depressed regions, and vegetation. The primary drainage
routes of the northern and south-central regions are found to be Diabari Khal and Buania
Khal, respectively. However, Figure 11e shows that the Baunia Khal’s extent decreased
significantly between 2003 and 2013, with a significant portion of it becoming an urban set-
tlement by 2022. According to [34], Dhaka experienced a 40.17 percent decline in wetlands
and waterbody between 1978 and 2009. Within a comparable time range of 1980–2013, a
reduction rate of 22.3 percent is found in this study. This suggests that the level of intrusion
into aquatic wetlands and waterbodies within the Goranchatbari catchment is lower than
the overall patterns observed throughout Dhaka. In fact, Figure 11d,e the GPS retention
pond is the only natural water body that has been preserved between 2003 and 2022. How-
ever, Landsat images (Figure 11f) show that between 2013 and 2022, approximately 50 acres
of land on the northern side of the retention basin were converted to settlement. Over
a 50-year period (Figure 12), the water body and vegetation decreased to 15.58 percent
and 6.88 percent, respectively, from 43.19 percent and 44.83 percent. Settlements and bare
lands, on the other hand, increased by 65 percent in five decades to 77.54 percent from
11.98 percent. Notably, Ref. [22] reported a 42 percent increase in Dhaka’s built-up area
between 1978 and 2018. It should be noted that the reported rate of increase is lower than
that observed in the Goranchatbari catchment area. This strongly suggests that our study
area is experiencing an even faster rate of urbanization, surpassing the already rapid rate
of urban development within Dhaka city itself.
 Water2023,
Water 2023,15,
15,3834
x FOR PEER REVIEW 16 of 2716 of 26

Water 2023, 15, x FOR PEER REVIEW 17 of 27

decreased to 15.58 percent and 6.88 percent, respectively, from 43.19 percent and 44.83
percent. Settlements and bare lands, on the other hand, increased by 65 percent in five
decades to 77.54 percent from 11.98 percent. Notably, Ref. [22] reported a 42 percent in-
crease in Dhaka’s built-up area between 1978 and 2018. It should be noted that the re-
ported rate of increase is lower than that observed in the Goranchatbari catchment area.
Figure
This
Figure 11.
11. LULC
strongly
LULC change
change for for
suggests (a)our
that
(a) 1973;
1973; (b) 1980;
study
(b) area
1980; (c)experiencing
is
(c) 1993; 1993; (d) (e)
(d) 2003; 2003;
an(e) 2013;
even
2013; (f) (f) 2022—classified
faster into 3 into
rate of urbaniza-
2022—classified
tion, surpassing
3 subgroups—canal
subgroups—canal andthe
and already rapid
waterbodies,
waterbodies, rate ofsettlement,
urban
vegetations,
vegetations, development
settlement,
and bare within
andlands.
bare Dhaka city itself.
lands.

90.00 According to the 2003 LULC analysis (Figure 11d), the areas around Uttara Khal in
Canalsthe
andnortheast
Waterbodies
and the Vegetations Settlements
Eastern Housing and lower
Khal in the Barelands
southwest experienced a signifi-
80.00
cant increase in settlement growth between 1993 and 2003. Despite this rapid settlement
70.00 growth, Figure 11d shows that in 2003, a significant proportion of the drainage basin
60.00 (78.77 percent) was still a central waterbody, depressed regions, and vegetation. The pri-
Percentage (%)

mary drainage routes of the northern and south-central regions are found to be Diabari
50.00
Khal and Buania Khal, respectively. However, Figure 11e shows that the Baunia Khal’s
40.00 extent decreased significantly between 2003 and 2013, with a significant portion of it be-
30.00 coming an urban settlement by 2022. According to [34], Dhaka experienced a 40.17 percent
decline in wetlands and waterbody between 1978 and 2009. Within a comparable time
20.00
range of 1980–2013, a reduction rate of 22.3 percent is found in this study. This suggests
10.00 that the level of intrusion into aquatic wetlands and waterbodies within the Goranchatbari
0.00 catchment is lower than the overall patterns observed throughout Dhaka. In fact, Figure
1973 11d,e the 1980
GPS retention pond1993is the only natural
2003 water body2013
that has been preserved
2022 be-
tween 2003 and 2022. However, Landsat images (Figure 11f) show that between 2013 and
Figure 12. LULC change percentage for 3 subgroups—canal and waterbodies, vegetations, settle-
Figure
2022, 12. LULC change
approximately percentage
50 acres foron
of land 3 subgroups—canal
the northern side and waterbodies,
of the vegetations,
retention basin settlement,
were con-
ment, and bare lands.
and bare
verted to lands.
settlement. Over a 50-year period (Figure 12), the water body and vegetation
Illegal encroachment on waterways and ponds has clearly reduced the amount of
Illegal encroachment on waterways and ponds has clearly reduced the amount of
open space available for water to flow and be stored. Furthermore, the rate of urban de-
open space available for water to flow and be stored. Furthermore, the rate of urban
velopment, population influx, and drainage technology adaptation determine how the
development,
Goranchatbari population influx,
catchment’s and
future drainage
pluvial floodtechnology adaptation
conditions will determine
be. Figure how
13 depicts thethe
predicted LULC in 2042 to help understand these implications. According to the graph,
the settlement area will increase from 77.54 percent in 2022 to 89.49 percent in 2042. This
upward trend implies that a greater proportion of land will be used for residential and
commercial purposes, potentially depleting natural waterbodies and green spaces. The
Water 2023, 15, 3834 17 of 26

Goranchatbari catchment’s future pluvial flood conditions will be. Figure 13 depicts the
predicted LULC in 2042 to help understand these implications. According to the graph,
the settlement area will increase from 77.54 percent in 2022 to 89.49 percent in 2042. This
upward trend implies that a greater proportion of land will be used for residential and
commercial purposes, potentially depleting natural waterbodies and green spaces. The map
also highlights the significant depletion of low-lying terrain suitable for urban settlements,
particularly in the vicinity of Baunia Khal. Furthermore, the map indicates that Diabari
Khal and Abdullahpur Khal in the northwest region are expected to lose connectivity
with Uttara Khal, potentially causing drainage congestion in the northern region. Mirpur
Housing Khal and Shangbadik Colony Khal, which collect rainwater runoff from Pallabi
and Kafrul, would also be disconnected from the shrinking Baunia Khal. This could
exacerbate flooding and other water-related problems in the southeast. The proposed
LULC also implies that the Goranchatbari retention pond will not be directly linked to the
southwest region’s Rupnagar Khal and Bayesteki Khal. Furthermore, the western portion
of the Goranchatbari retention pond, adjacent to the embankment, is expected to lose
21.35 percent of its retention area by 2042. Given that the retention pond serves as a
Water 2023, 15, x FOR PEER REVIEWreservoir for water before it is discharged into the Turag River, such encroachment18 of has
27 the

potential to exacerbate the urban flood vulnerability.

Figure 13.
Figure [Link]
PredictedLULC
LULCclassification forfor
classification 2042 using
2042 MOLUSCE.
using MOLUSCE.

3.3. GoranchatbariFlooding
3.3. Goranchatbari Floodingfrom
from Heavy
Heavy Rain:
Rain: Present
Present andand Future
Future
The hydraulic
The hydraulicmodel
modelis isapplied
appliedto to rainfall
rainfall events
events withwith return
return periods
periods of 2.33-,
of 2.33-, 25-, 25-,
and 50-years in both current and future scenarios to investigate the potential
and 50-years in both current and future scenarios to investigate the potential impact of impact of
rapid urbanization on urban flooding. The first three scenarios (Figure 14a–c)
rapid urbanization on urban flooding. The first three scenarios (Figure 14a–c) are used to are used to
assess the capacity of the existing drainage system in the Goranchatbari
assess the capacity of the existing drainage system in the Goranchatbari area to handle area to handle
extreme rainfallevents.
extreme rainfall events. These
These scenarios
scenarios are denoted
are denoted as EHas EHRP,
2.33 2.33EHRP, EH and
25 RP, 25 RP,
EHand50 EH
50
RP,RP, based
based on existing
on the the existing drainage
drainage network
network and historical
and historical [Link]. Thethree
The second second
sce-three
nario runs (Figure 14d–f) are designed to assess the future flooding condition in the Go-
ranchatbari area as a result of rapid urbanization. The drainage network is updated using
the CA-ANN-based MOLUSCE plugin in QGIS based on the projected LULC maps for
2042. These scenarios are denoted as PH 2.33 RP, PH 25 RP, and PH 50 RP, based on the
projected drainage network and historical rainfall.
Water 2023, 15, 3834 18 of 26

scenario runs (Figure 14d–f) are designed to assess the future flooding condition in the
Goranchatbari area as a result of rapid urbanization. The drainage network is updated
using the CA-ANN-based MOLUSCE plugin in QGIS based on the projected LULC maps
Water 2023, 15, x FOR PEER REVIEW 19 of 27
for 2042. These scenarios are denoted as PH 2.33 RP, PH 25 RP, and PH 50 RP, based on the
projected drainage network and historical rainfall.

Figure14.14.
Figure Maximum
Maximum floodflood inundation
inundation depth depth under
under (a) (a) EH_2.33
EH_2.33 RP, (b)
RP, (b) EH_25 RP,EH_25 RP, RP,
(c) EH_50 (c) EH_50
(d)
RP,PH_2.33 RP, (e) PH_25
(d) PH_2.33 RP, (e)RP, (f) PH_50
PH_25 RP PH_50
RP, (f) within the
RP Goranchatbari Catchment Maps
within the Goranchatbari Scenario Maps
Catchment
conditions.
Scenario conditions.

According
Figure 14todepicts
the EHtheflood inundation
flood inundationanalysis
area,(Figure 14a–c), the west
which represents side of
the depth andthespatial
Diabari
extent khal, which
of flood is located
waters. When under Turagdepth
the water thana,isisless
prone to 0.1
than repeated
m, thewaterlogging in
area is waterlogged,
several
but thereturn
water period
only floods.
reachesThe areas
ankle near the
height. Baunia
The urban khal that
area is carry stormwater
considered runoff if the
vulnerable
from
waterMirpur,
depth Pallabi, and Kafrul
ranges from 0.1 m are vulnerable
to 0.6 to highreaching
m, with water water levels
up tounder
knee2.33, 25, The
height. and urban
50-year rainfall events. The EH 2.33 RP inundation map in Figure 14a shows an inundation
area is classified as severely waterlogged when the water depth ranges from 0.6 m to 1.0 m
depth of 1.5–2.0 m around Baunia Khal. However, as the connectivity is disrupted in the
projected LULC scenario: PH 2.33 RP (Figure 14d), the inundation depth and spatial ex-
tent increase in the upstream region. Similar characteristics are observed in the western
Water 2023, 15, 3834 19 of 26

and the water level is around waist height. If the water depth is greater than 1.0 m, the area
is classified as low-lying lands, khals, natural drainage canals, and retention ponds within
the sub-catchment. The classification was done based on the response of field surveys
carried out to assess the extent of waterlogging severity within the study area. Residents
conveyed the depth of flooding by using relatable terms like “ankle-deep”, “waist-deep”,
“knee-deep”, and so on. Consequently, the classification in this paper was designed to align
with these field observations for consistency.
According to the EH flood inundation analysis (Figure 14a–c), the west side of the
Diabari khal, which is located under Turag thana, is prone to repeated waterlogging in
several return period floods. The areas near the Baunia khal that carry stormwater runoff
from Mirpur, Pallabi, and Kafrul are vulnerable to high water levels under 2.33, 25, and
50-year rainfall events. The EH 2.33 RP inundation map in Figure 14a shows an inundation
depth of 1.5–2.0 m around Baunia Khal. However, as the connectivity is disrupted in the
projected LULC scenario: PH 2.33 RP (Figure 14d), the inundation depth and spatial extent
increase in the upstream region. Similar characteristics are observed in the western part of
the Diabari khal, where the extended flood inundation area increases by 8.61 percent under
PH 2.33 RP.
The inundated area around Abdullahpur Khal, which serves as the primary drainage
channel for Uttara and Turag thanas, sees a significant increase in water depth in the
low-lying floodplain zone under EH 25 RP (Figure 14b). The inundation map shows that
due to the LULC change in 2042, a 0.23 km2 area near the intersection of Abdullahpur and
Buania khals shifts from a water level below 2 m (Figure 14b) to a water level above 2 m
(Figure 14e). Because of increased urbanization and encroachment in the area, this change
in water depth is attributed to a decrease in connectivity and disruption in the natural
drainage system. Similarly, the downstream portion of the Baunia khal exhibits a shift
in water depth and spatial inundation extent toward greater depth. These findings are
consistent with recent research by [31], which shows a high vulnerability to waterlogging
around the Baunia Khal depression.
The EH 50 RP (Figure 14c) and PH 50 RP (Figure 14f) runs show the inundation extent
to be more severe around the west portion of Turag thana, Diabari Khal. This area is
naturally low-lying and depressed, and it is bounded on the west side by the Turag River
embankment, making it more vulnerable to flooding. The Diabari Khal’s lack of drainage
connectivity exacerbates the problem, as heavy rainfall causes runoff to accumulate faster,
resulting in severe waterlogged conditions. The upper portion of the Abdullahpur Khal is
seen to have increased water depth due to channel width shrinkage, which is primarily
attributed to the LULC change and increase in urban settlements in PH 50 RP. As cities
grow and encroach on the natural environment, natural drainage systems are disrupted,
resulting in increased flood inundation.
Figure 15 depicts a comparison of waterlogged areas with a 1 m inundation depth under
current (EH) and projected (PH) LULC conditions for 2.33, 25, and 50 years of return period
runs. The depth of inundation is divided into five categories, each with a 0.2 m increment from
0 to 1 m. The results show that the inundation area between 0.4–0.8 m of water depth remains
nearly the same in both current and projected LULC conditions, but the PH 50 year has a
31% increase in inundation area over 0.8 m compared to EH 50-year. In the 25-year return
period simulation (Figure 15b), PH 25 year increases by 9.52 percent and 19.05 percent in the
0.6–0.8 m and 0.8–1.0 m ranges, respectively. The PH 25 year average increase for inundation
areas below 0.6 m water depth is 5.18 percent. Figure 15c shows a significant increase in
inundation levels between 0.6 and 0.8 m, indicating a 21.34 percent increase over baseline
conditions. According to the analysis, the total inundation area below the 1.0 m water depth
threshold increases by 8.47 percent, 8.11 percent, and 4.05 percent in PH cases for 50-, 25-, and
2.33-year return period rainfall events, respectively, compared to the current conditions in
2022. This means that more land is at risk of being inundated during heavy rainfall events,
potentially resulting in increased flood damage and asset loss.
 areas, increasing to 8 to 12% in altered land use conditions. Due to a
ditions, an additional 15% of areas would like to experience 1 h to
Figure 17 indicates that Baunia Khal, Parise Road Khal, Shangbadik C
Water 2023, 15, 3834
Cantonment Khal, Abdullahpur Khal, and Digun Khal will20face of 26
devas
that will likely last 2–5 days.

Figure
Figure 15. Inundated
15. Inundated area underarea
(a) 50under
years (b)(a) 50 years
25 years (c) 2.33 (b)
years25 years
return (c)precipitation
period 2.33 years retu
within the Goranchatbari catchment.
within the Goranchatbari catchment.
Another special projected scenario (SPH) has also been simulated to assess the impact
of extreme events in 2042 if minimum amount of connectivity among the dead spots
of the canals are maintained (1 m width). It has been found that, the total inundation
extent does not quite vary with the PH scenario (Figure 16). However, the depth of the
inundation varies to some extent. For, 50 years return period, the area of the zone in which
the inundation layer is above 0.8 m will decrease by a small margin of 2.32%. The rest of
the analysis for this return period remained unchanged. Scenarios like, SPH 2.33 year, and
SPH 25 year faces negligible amount of change comparing to PH 2.33 year, and PH 25 year
return period scenarios respectively. This proves the fact that, ensuring a minimum amount
of connectivity will not bring much positives on a larger scale. So, to ensure a sustainable
solution, the canals need to be freed from every sort of encroachments.
Water2023,
Water 2023, 15,
15, 3834
x FOR PEER REVIEW 22 of 27 21 of 26

Maximum
[Link]
Figure flood
flood inundation
inundation depthdepth
underunder (a) SPH_2.33
(a) SPH_2.33 RP, (b)RP, (b) SPH_25
SPH_25 RP, (c) RP, (c) SPH_50
SPH_50
RP,(d)(d)
RP, PH_2.33
PH_2.33 RP,PH_25
RP, (e) (e) PH_25 RP, (f)RP
RP, (f) PH_50 PH_50
withinRP
the within the Goranchatbari
Goranchatbari Catchment Maps Catchment
Scenario Maps
Conditions.
Scenario Conditions.

Table 7 and Figure 17 show the duration of flooding caused by different return period
rain patterns in existing (EH) and altered (PH) land use. When these cases are compared,
the no-flood area is reduced by 3% to 10% as a result of the altered land use. In the
case of EH cases, the high duration (2 to 5 days) was observed in nearly 4 to 8% of the
catchment areas, increasing to 8 to 12% in altered land use conditions. Due to altered land
use conditions, an additional 15% of areas would like to experience 1 h to 5 days of flooding.
Figure 17 indicates that Baunia Khal, Parise Road Khal, Shangbadik Colony Khal, Mirpur
Cantonment Khal, Abdullahpur Khal, and Digun Khal will face devastating floods in 2042
that will likely last 2–5 days.
Water 2023, 15, 3834 22 of 26
Water 2023, 15, x FOR PEER REVIEW 23 of 27

Table 7. Duration of the flooding.

Table 7. Duration of the flooding.
Percentage of Inundation
Percentage of Inundation
Duration of Duration
EH of Flooding EH PH PH
Flooding
EH_2.33 EH_25 EH_50 PH_2.33
EH_2.33 EH_25 PH_25 PH_2.33
EH_50 PH_50
PH_25 PH_50
No Flooding 55 50
No Flooding 43 55 48 50 43 40 48 40
40 40
<30 min 29 32
<30 min 39 29 31 32 39 38 31 38
38 38
1–6 h 1 1 1–6 h 0 1 2 1 0 1 2 11 1
6–24 h 5 3 6–24 h 4 5 5 3 4 3 5 33 3
1–2 Days 6 51–2 Days 5 6 6 5 5 7 6 76 6
2–5 Days 4 8 8 8 11 12
2–5 Days 4 8 8 8 11 12

Sub-catchment-wise
[Link]-catchment-wise
Figure FloodFlood Duration
Duration Map
Map (a) (a) EH_2.33
EH_2.33 RP, (b)RP,
RP, (b) EH_25 EH_25 RP, (c)
(c) EH_50 RP,EH_50
(d) PH_2.33 RP, (e) PH_25 RP, (f) PH_50 RP within the Goranchatbari Catchment Maps Scenario Maps
RP, (d) PH_2.33 RP, (e) PH_25 RP, (f) PH_50 RP within the Goranchatbari Catchment
conditions.
Scenario conditions.

4. Conclusions
The Goranchatbari catchment is one the last remaining interconnected natural drainage
systems of Dhaka city. The rapid pace of urban development has put significant demo-
Water 2023, 15, 3834 23 of 26

graphic pressure on an already fragile system. Field observations revealed that unchecked
urban development and poor waste management practices were crippling the natural
drainage system of the catchment. Furthermore, the construction of residential infrastruc-
ture over already encroached khals and low-lying areas, as well as dumping of unchecked
solid waste and industrial effluent, are the primary factors reducing the conveyance ca-
pacity of the khals. According to LULC analysis, settlement/built-up area has increased
by more than 2.3 km2 /year on average between 2003 and 2013. This rate of urban growth
during 2003–2013 is 12 times of the growth between 1973 and 1993. The western portion of
Turag thana is highly susceptible to urban flooding due to its naturally depressed topog-
raphy and absence of a well-integrated drainage system with the primary khal network.
Low-lying areas near the Baunia Khal and the upstream region of the Abdullahpur Khal
will be highly vulnerable to urban flooding in the future. For 50 years, 25 years, and
2.33 years of rainfall events, the current trajectory of LULC change results in an increase
of 8.47%, 8.11%, and 4.05% in the total inundated areas (<1.0 m) in 2042, respectively, in
comparison to the current condition. Flood duration is also expected to increase as a result
of LULC change. Long-term flooding will result in 11% more areas.
The findings of the study can assist policymakers and city planners in understanding
the relationship between urban flooding and unchecked urban growth. Stakeholders can
take proactive steps to mitigate future flooding risks and enhance long-term sustainability
and resilience of the city by assessing the potential risks of future waterlogging.

Supplementary Materials: The following supporting information can be downloaded at: https:
//[Link]/article/10.3390/w15213834/s1. Section S1: Status of Natural Canals or Khals in
the Study Area.
Author Contributions: Conceptualization, S., M.S.M. and M.S.S.; methodology, S., M.S.S. and M.S.M.;
software, S.A. and M.S.S.; validation, S.A. and M.S.S.; formal analysis, S.A., M.S.S., S., S.B.M. and R.K.;
investigation, S.A., M.S.S., S., M.S.M., S.B.M., R.K. and A.I.A.C.; resources, M.S.M.; data curation,
S.A., M.S.S. and R.K.; writing—original draft preparation, M.S.S., S.A. and S.; writing—review and
editing, M.S.S., S.A., S. and M.S.M.; visualization, S.A., M.S.S., S. and M.S.M.; supervision, M.S.M.;
project administration, M.S.M.; funding acquisition, M.S.M. All authors have read and agreed to the
published version of the manuscript.
Funding: This research was funded by the U.S. Department of State under the project “Assessing
Health Impacts of Urban Flooding under Changing Climate: A Case Study in Dhaka City”.
Data Availability Statement: Data is available on request.
Acknowledgments: The authors acknowledged the contributions of different key stakeholders who
provided valuable information during the field survey.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Xiao, S.; Zou, L.; Xia, J.; Dong, Y.; Yang, Z.; Yao, T. Assessment of the urban waterlogging resilience and identification of its
driving factors: A case study of Wuhan City, China. Sci. Total Environ. 2023, 866, 161321. [CrossRef] [PubMed]
2. Akther, H.; Ahmad, M.M. Livelihood in the pluvial flood prone slum communities in Dhaka, Bangladesh. Prog. Disaster Sci. 2022,
14, 100227. [CrossRef]
3. Seleem, O.; Heistermann, M.; Bronstert, A. Efficient Hazard Assessment for Pluvial Floods in Urban Environments: A Bench-
marking Case Study for the City of Berlin, Germany. Water 2021, 13, 2476. [CrossRef]
4. Mediero, L.; Soriano, E.; Oria, P.; Bagli, S.; Castellarin, A.; Garrote, L.; Mazzoli, P.; Mysiak, J.; Pasetti, S.; Persiano, S.; et al. Pluvial
flooding: High-resolution stochastic hazard mapping in urban areas by using fast-processing DEM-based algorithms. J. Hydrol.
2022, 608, 127649. [CrossRef]
5. Yahya, S.M.; Shams, S.; Mahmud, K. Climate Change Impacts on Flood Vulnerability for Dhaka City. In Proceedings of the
International Conference on Enviromental Aspects of Bangladesh, Kitakyushu, Japan, 4 September 2010.
6. WB. Flood Risk Management in Dhaka: A Case for Eco-Engineering Approaches and Institutional Reform; World Bank: Dhaka,
Bangladesh, 2018.
7. Akter, A.; Mohit, S.A.; Chowdhury, M.A.H. Predicting urban storm water-logging for Chittagong city in Bangladesh. Int. J.
Sustain. Built Environ. 2017, 6, 238–249. [CrossRef]
Water 2023, 15, 3834 24 of 26

8. Quan, R.-S.; Liu, M.; Lu, M.; Zhang, L.-J.; Wang, J.-J.; Xu, S.-Y. Waterlogging risk assessment based on land use/cover change: A
case study in Pudong New Area, Shanghai. Environ. Earth Sci. 2010, 61, 1113–1121. [CrossRef]
9. Roy, S.; Bose, A.; Singha, N.; Basak, D.; Chowdhury, I.R. Urban waterlogging risk as an undervalued environmental challenge:
An Integrated MCDA-GIS based modeling approach. Environ. Chall. 2021, 4, 100194. [CrossRef]
10. Shao, Z.; Ding, L.; Li, D.; Altan, O.; Huq, M.E.; Li, C. Exploring the Relationship between Urbanization and Ecological Environment
Using Remote Sensing Images and Statistical Data: A Case Study in the Yangtze River Delta, China. Sustainability 2020, 12, 5620.
[CrossRef]
11. Hossain, M.A.; Mahiuddin, S.; Ahmad, A.U.; Mamun, A.H.M.M. Causes and Effects of Water Logging in Dhaka City. In Climate,
Environment and Disaster in Developing Countries; Springer Nature: Singapore, 2022.
12. WHO. Unplanned Urbanization a Challenge for Public Health. Available online: [Link]
detail/05-04-2010-unplanned-urbanization-a-challenge-for-public-health (accessed on 5 April 2010).
13. Ojo, S.S.; Tpl, D.B.; Pojwan, M.A. Urbanization and Urban Growth: Challenges and Prospects for National Development.
J. Humanit. Soc. Policy 2017, 3, 65–71.
14. Adhikari, B.; Pokharel, S.; Mishra, S.R. Shrinking Urban Greenspace and the Rise in Non-communicable Diseases in South Asia:
An Urgent Need for an Advocacy. Front. Sustain. Cities 2019, 1, 2019. [CrossRef]
15. World Pop. WorldPop; University of Southampton: Southampton, UK, 2023.
16. Dewan, A.; Corner, R. Dhaka Megacity: Geospatial Perspectives on Urbanisation, Environment and Health; Springer Science & Business
Media: Berlin, Germany, 2014.
17. BWDB. Feasibility and Mathematical Model Study of Approaching and Investigating Strategy for Rehabilitating the Buriganga-turag-
Shitalakhya River System and Augmentation of Dry Season Flow in the Buriganga River; Bangladesh Water Development Board: Dhaka,
Bangladesh, 2004.
18. Faisal, I.M.; Kabir, M.R.; Nishat, A. Nonstructura Flood Mitigation Measures for Dhaka City. Urban Water 1999, 1, 145–153.
[CrossRef]
19. WB. Building Safer Cities: The Future of Disaster Risk; World Bank: Washington, DC, USA, 2003.
20. ADB. Project Completion Report: Dhaka Integrated Flood Protection Project; Asian Development Bank: Mandaluyong,
Philippines, 2002.
21. IWFM. Investigation of Hydrological Aspect of Flood—2004 with Special Emphasis on Dhaka City; IWFM: Dhaka, Bangladesh, 2005.
22. Moniruzzaman, M.; Thakur, P.K.; Kumar, P.; Ashraful Alam, M.; Garg, V.; Rousta, I.; Olafsson, H. Decadal Urban Land Use/Land
Cover Changes and Its Impact on Surface Runoff Potential for the Dhaka City and Surroundings Using Remote Sensing. Remote
Sens. 2020, 13, 83. [CrossRef]
23. Ishtiaque, A.; Mahmud, M.S.; Rafi, M.H. Encroachment of Canals of Dhaka City Bangladesh: An Investigative Approach. GeoScape
2014, 8, 48–64. [CrossRef]
24. Hossain, S.; Islam, A.S.; Ali, M.; Billah, M.; Rabbani, F. Impcat of Climate Change of Brahmaputra River Basin on Urban Drainage
of Goranchatbari, Dhaka. Int. J. Sci. Eng. Res. 2015, 7, 765–779.
25. Bala, S.K.; Islam, A.S.; Chowdhury, J.U.; Rahman, M.R.; Haque, M.A.; Khan, M.S.A.; Salehin, M. Performance of Flood Control
Works around Dhaka City during Major Floods in Bangladesh. In Proceedings of the 2nd International Conference on Water &
Flood Management (ICWFM-2009), Dhaka, Bangladesh, 15–17 March 2009.
26. Alam, H. Why Wasa’s Reservoirs Dying. The Daily Star, 5 September 2016.
27. Das, B.; Islam, I. Analyzing the Proposals of FAP 8B Project of Dhaka and Present Context of Retention Pond Areas and Canals . J.
Bangladesh Inst. Plan. 2010, 3, 89–102.
28. Dewan, A.M. Floods in a Megacity: Geospatial Techniques in Assessing Hazards, Risk and Vulnerability; Springer Geography; Springer:
Dordrecht, The Netherlands, 2012. [CrossRef]
29. Mark, O.; Apirumanekul, C.; Kamal, M.M.; Praydal, G. Modelling of Urban Flooding in Dhaka City. Urban. Drain. Model. 2001,
333–343. [CrossRef]
30. Faridatul, M.I.; Wu, B.; Zhu, X. Assessing Long-Term Urban Surface Water Changes Using Multi-Year Satellite Images: A Tale of
Two Cities, Dhaka and Hong Kong. J. Environ. Manage. 2019, 243, 287–298. [CrossRef]
31. Alam, R.; Quayyum, Z.; Moulds, S.; Radia, M.A.; Sara, H.H.; Hasan, M.T.; Butler, A. Dhaka City Water Logging Hazards: Area
Identification and Vulnerability Assessment through GIS-Remote Sensing Techniques. Environ. Monit. Assess. 2023, 195, 543.
[CrossRef]
32. Sultana, M.S.; Tarekul Islam, G.M.; Islam, Z. Pre- and Post-Urban Wetland Area in Dhaka City, Bangladesh: A Remote Sensing
and GIS Analysis. J. Water Resour. Prot. 2009, 1, 414–421. [CrossRef]
33. DWASA. Updating/Preparation of the Stormwater Drainage Master Plan for Dhaka City: Master Plan Report; Dhaka Water Supply and
Sewerage Authority: Dhaka, Bangladesh, 2016; p. 262.
34. Habiba, U.; Haider, F.; Ishtiaque, A.; Mahmud, M.S.; Masrur, A. Remote Sensing & GIS Based Spatio-Temporal Change Analysis
of Wetland in Dhaka City, Bangladesh. J. Water Resour. Prot. 2011, 3, 781–787.
35. Subrina, S.; Chowdhury, F.K. Urban Dynamics: An undervalued issue for water logging disaster risk management in case of
Dhaka city, Bangladesh. Procedia Eng. 2017, 212, 801–808. [CrossRef]
36. Mowla, Q.A.; Islam, M.S. Natural Drainage System and Water Logging in Dhaka: Measures to address the Problems. J. Bangladesh
Inst. Plan. 2013, 6, 23–33.
Water 2023, 15, 3834 25 of 26

37. Monserud, R.A.; Leemans, R. Comparing global vegetation maps with the Kappa statistic. Ecol. Model. 1992, 62, 275–293.
[CrossRef]
38. Kafy, A.A.; Faisal, A.-A.; Shuvo, R.M.; Naim, M.N.H.; Sikdar, M.S.; Chowdhury, R.R.; Islam, M.A.; Sarker, M.H.S.; Khan, M.H.H.;
Kona, M.A. Remote sensing approach to simulate the land use/land cover and seasonal land surface temperature change using
machine learning algorithms in a fastest-growing megacity of Bangladesh. Remote Sens. Appl. Soc. Environ. 2021, 21, 100463.
[CrossRef]
39. Kafy, A.A.; Faisal, A.A.; Al Rakib, A.; Roy, S.; Ferdousi, J.; Raikwar, V.; Kona, M.A.; Fatin, S. Predicting changes in land use/land
cover and seasonal land surface temperature using multi-temporal landsat images in the northwest region of Bangladesh. Heliyon
2021, 7, e07623. [CrossRef]
40. Talukdar, S.; Eibek, K.U.; Akhter, S.; Ziaul, S.; Towfiqul Islam, A.R.M.; Mallick, J. Modeling fragmentation probability of land-use
and land-cover using the bagging, random forest and random subspace in the Teesta River Basin, Bangladesh. Ecol. Indic. 2021,
126, 107612. [CrossRef]
41. Biratu, A.A.; Bedadi, B.; Gebrehiwot, S.G.; Melesse, A.M.; Nebi, T.H.; Abera, W.; Tamene, L.; Egeru, A. Ecosystem Service
Valuation along Landscape Transformation in Central Ethiopia. Land 2022, 11, 500. [CrossRef]
42. Alshari, E.A.; Gawali, B.W.; Kaluri, R. Modeling Land Use Change in Sana’a City of Yemen with MOLUSCE. J. Sens. 2022,
2022, 7419031. [CrossRef]
43. Dey, N.N.; Al Rakib, A.; Kafy, A.A.; Raikwar, V. Geospatial modelling of changes in land use/land cover dynamics using
Multi-layer Perceptron Markov chain model in Rajshahi City, Bangladesh. Environ. Chall. 2021, 4, 100148. [CrossRef]
44. Rimal, B.; Zhang, L.; Keshtkar, H.; Wang, N.; Lin, Y. Monitoring and modeling of spatiotemporal urban expansion and land-
use/land-cover change using integrated Markov chain cellular automata model. SPRS Int. J. Geo-Inf. 2017, 6, 288. [CrossRef]
45. Gebresellase, S.H.; Wu, Z.; Xu, H.; Muhammad, W.I. Scenario-Based LULC Dynamics Projection Using the CA–Markov Model on
Upper Awash Basin (UAB), Ethiopia. Sustain. Environ. Resour. Manag. Under Glob. Change 2023, 15, 1683. [CrossRef]
46. Norizah, K.; Jamhuri, J.; Balqis, M.; Hasmadi, I.M.; Akmar, A.A.N. Land Use and Land Cover Change Prediction Using ANN-CA
Model. In Tropical Forest Ecosystem Services in Improving Livelihoods For Local Communities; Springer: Singapore, 2022; pp. 107–125.
47. Shatnawi, N.; Abu Qdais, H. Mapping urban land surface temperature using remote sensing techniques and artificial neural
network modelling. Int. J. Remote Sens. 2019, 40, 3968–3983. [CrossRef]
48. Rahman, M.H. A Study on Determining Land Use/Land Cover Changes in Dhaka over the Last 20 Years and Observing the
Impact of Population Growth on Land Use/Land Cover Using Remote Sensing. Malays. J. Civ. Eng. 2022, 34, 1–9. [CrossRef]
49. Ismail, M.A.; Ludin, A.N.M.; Hosni, N. Comparative Assessment of the Unsupervised Land Use Classification by Using
Proprietary GIS and Open Source Software. In Proceedings of the 10th IGRSM International Conference and Exhibition on
Geospatial & Remote Sensing, Kuala Lumpur, Malaysia, 20–21 October 2020.
50. Butt, A.; Shabbir, R.; Ahmad, S.S.; Aziz, N. Land use change mapping and analysis using Remote Sensing and GIS: A case study
of Simly watershed, Islamabad, Pakistan. Egypt. J. Remote Sens. Space Sci. 2015, 18, 251–259. [CrossRef]
51. Kamaraj, M.; Rangarajan, S. Predicting the future land use and land cover changes for Bhavani basin, Tamil Nadu, India, using
QGIS MOLUSCE plugin. Environ. Sci. Pollut. Res. Int. 2022, 29, 86337–86348. [CrossRef] [PubMed]
52. Kabir, K.H.; Hossain, M.R.; Shams, S.M.N.; Rahman, M.S.; Islam, M.R. Post-COVID-19 Strategies for Power and Energy Sector of
Bangladesh. SSRN Electron. J. 2022, 50, 101176. [CrossRef]
53. Taha, H.; Kamruzzaman. Priority Areas for Bangladesh: Roadmap to 2041 as Developed Country. Arch. Community Med. Public
Health 2020, 6, 277–280. [CrossRef]
54. Congalton, R.G. A Review of Assessing the Accuracy of Classifications of Remotely Sensed Data. Remote Sens. Environ. 1991, 37,
35–46. [CrossRef]
55. Sankarrao, L.; Ghose, D.K.; Rathinsamy, M. Predicting land-use change: Intercomparison of different hybrid machine learning
models. Environ. Model. Softw. 2021, 145, 105207. [CrossRef]
56. Halder, B.; Bandyopadhyay, J.; Banik, P. Monitoring the effect of urban development on urban heat island based on remote
sensing and geo-spatial approach in Kolkata and adjacent areas, India. Sustain. Cities Soc. 2021, 74, 103186. [CrossRef]
57. Alam, S.; Hasan, F.; Debnath, M.; Rahman, A. Morphology and Land Use Change Analysis of Lower Padma River Floodplain of
Bangladesh. Environ. Monit. Assess. 2023, 195, 886. [CrossRef] [PubMed]
58. Milne-Thomson, L.M. Theoretical Hydrodynamics; Dover Publications: Mineola, NY, USA, 1968; p. 672.
59. Arcement, G.J.; Schneider, V.R. Guide for Selecting Manning’s Roughness Coefficients for Natural Channels and Flood Plains; U.S.
Department of Transportation, Federal Highway Administration: Washington, DC, USA, 1989.
60. Roesner, L.; Shubinski, R.; Aldrich, J. Storm Water Management Model User’s Manual, Version 3. Addendum 1: Extran (Extended
Transport); EPA/600/2-84/109B; United States Environmental Protection Agency: Washington, DC, USA, 2002.
61. Rangari, V.A.; Prashanth, S.S.; Umamahesh, N.V.; Patel, A.K. Simulation of Urban Drainage System Using a Storm Water
Management Model (SWMM). Asian J. Eng. Appl. Technol. 2018, 7, 7–10. [CrossRef]
62. Trenberth, K. Storm Tracks in the Southern Hemisphere. J. Atmos. Sci. 1991, 48, 2159–2178. [CrossRef]
63. Rawls Walter, J.; Brakensiek Donald, L.; Miller, N. Green-ampt Infiltration Parameters from Soils Data. J. Hydraul. Eng. 1983, 109,
62–70. [CrossRef]
64. Chow, V.T. Open-Channel Hydraulics; McGraw-Hill: New York, NY, USA, 1959.
Water 2023, 15, 3834 26 of 26

65. Rossman, L.A. Storm Water Management Model Quality Assurance Report: Dynamic Wave Flow Routing; Water Supply and Water
Resources Division, National Risk Management Research Laboratory: Cincinnati, OH, USA, 2006.
66. Kotz, S.; Nadarajah, S. Extreme Value Distributions; World Scientific: Singapore, 2000; p. 196.
67. JICA. Study on Storm Water Drainage System Improvement Project in Dhaka City; Department of Public Health and Engineering:
Dhaka, Bangladesh, 1987.
68. Alam, S.; Rahman, A.; Yunus, A. Designing Stormwater Drainage Network for Urban Flood Mitigation Using SWMM: A Case
Study on Dhaka City of Bangladesh. Am. J. Water Resour. 2023, 11, 65–78. [CrossRef]
69. Maliha, M.; Khan, S.M. Design of Stormwater Drainage Network with Limited Data Using GeoSWMM: Case Study for Bogra
City. Ph.D. Thesis, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh, 2017.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.