Abstract

In recent years, climate change and urbanization have become major concerns for developing countries, and this will continue to exacerbate in the future. It has triggered abundant challenges, among which urban flooding is becoming one of the most important. In this study, the impact of extreme rainfall on urban drainage systems is analyzed through a case study of Rohtak City in Haryana, India. For the study, a MIKE+ one-dimensional hydrodynamic and rainfall-runoff model was adopted. The monsoon rainfall data, from June to September 2022, was retrieved from India-WRIS and incorporated into MIKE+ as a time series for the simulation of rainfall-runoff. The main objectives of the study were to assess urban flood vulnerability zones and to identify individual hotspot nodes of existing drainage networks. Flooding from extreme rainfall and future rainfall increased due to climate change by 10% (Rainfall; R1), 20% (Rainfall; R2) and 50% (Rainfall; R3) because of the monsoon rainfall. Along with extreme event analysis, predictive analysis was also made. The hydraulic parameter for water level in nodes and pipes was used to determine the hydraulic capacity of the drainage system. The simulation results indicated that the city's drainage system became hydraulically inefficient in dealing with the extreme rainfall event in 2022 that caused urban flooding. For the studied drainage system, 52 overflooding nodes, 57 pressurized links, and 07 critical catchments were found to be vulnerable, which is 9.13% of the total catchment area. Validation of the extreme rainfall event simulated in MIKE+ was done by obtaining a flood extent map using Google Earth Engine with the help of SENTINEL-1 SAR imagery data. The accuracy of the MIKE+ model is analyzed using two parameters, i.e., percentage flooded areas and pixel percentage flooded. The MIKE+ model performed significantly well in determining percentage flooded areas with an accuracy of 79.66%. When using predictive analysis, the MIKE+ model provides a great insight into R3 time series rainfall showing 22.38% of the total sub-catchment area to be flooded when a rainfall intensity of R3 occurs. Remedies to this drainage failure could be either redesigning the drainage system or designing sustainable detention ponds.

1 Introduction

Urban flooding in India has serious and diverse consequences, affecting public health, infrastructure, the economy, and the environment. India's growing urbanization, along with concerns such as poor drainage systems, unplanned buildings, and climate change, increases the frequency and severity of urban floods (NDMA 2020). Infrastructure such as roads, bridges, and drainage systems are routinely destroyed, interrupting transportation and costing significant repair expenditures. Economic activity halts, reducing productivity, and increasing insurance and recovery costs (NDMA 2010). Contaminated floodwaters contaminate soil and water resources, destroying ecosystems and increasing the vulnerability of low-lying and informal communities. Residents are frequently displaced, livelihoods are lost, and schooling is disrupted, adding to socioeconomic difficulties, which are exacerbated by climate change through erratic rainfall and increasing sea levels (Kourtis and Tsihrintzis 2021).

Since 2022, urban flooding in India has become more severe in cities like Mumbai, Chennai, Hyderabad, Bengaluru, and Kolkata which are particularly vulnerable, having seen massive floods in recent years that have affected millions. Mumbai has regular monsoon floods, with catastrophic incidents documented in 2005, 2017, and 2020 (Tripathy et al. 2024). Flooding is exacerbated by factors such as obsolete drainage, inadequate waste management, and encroaching on natural water channels. For example, the 2005 floods killed over 1,000 people and caused significant economic damage, with much of the city's low-lying neighbourhoods under water for days amid heavy rain. The Chennai floods in 2015 and 2021 emphasized the city's vulnerability, causing major damage from encroached lakes and congested drainage systems (Kartheeshwari and Elango 2022). The 2015 Chennai floods were exceptionally severe, caused by record-breaking rainfall and insufficient water management. This tragedy evacuated people, shuttered the airport, and resulted in damages surpassing ₹20,000 crores (about USD 2.7 billion) (Seenirajan et al. 2017). Bengaluru, with its many lakes, has experienced floods because of urbanization and water body encroachment. The floods of 2022 interrupted IT centers, harmed residential areas, and evacuated inhabitants, requiring greater attempts to demolish unlawful buildings on stormwater drains (Kalore et al. 2024). Kolkata also has severe monsoon floods, exacerbated by the city's low elevation and failing drainage infrastructure. The 2021 monsoon season saw communities inundated for days, and drainage systems overwhelmed by heavy rains (Mukherjee et al. 2019).

The problem of urban flooding is studied by many researchers, and traditional flood modeling techniques, such as 1D and 2D hydraulic models, have been developed to include urban-specific features (Bisht et al. 2016). Modeling software such as MIKE FLOOD and HEC-RAS 2D are commonly used to simulate water flow and estimate urban flood risk (Kourtis et al. 2017). These models now incorporate comprehensive urban topography and land use data to enhance accuracy in densely populated regions. Machine learning models are increasingly being used to predict urban floods, as more data from remote sensing, rainfall records, and GIS become available (Dang et al. 2024). These models may learn from past flood patterns and anticipate future flood occurrences, offering a more flexible approach than static physical models. Random Forest and Neural Network techniques are useful for identifying flood-prone locations, particularly when data is sparse or inadequate (Dey et al. 2024).

There are also software packages available for flood modeling, such as EPA SWMM software, which is popularly used for designing stormwater sewer networks (Rossman 2015). It is the most fundamental software used for stormwater drainage system modeling. MIKE+ is another urban water modeling software that is now in use for modeling collection systems, as well as water distribution networks for the analysis, management, and operation of stormwater drainage and wastewater systems in separate and combined systems (Pandey and Singh 2021a; 2021b). Mehta et al. (2022) showed a resilience-based evaluation of an urban drainage system (UDS) in Gurugram City under the impact of functional and structural failure modes. Resilience-based analysis is carried out using SWMM. The structural resilience method found 22 susceptible nodes in the examined UDS, whereas the functional resilience approach demonstrated that urbanization has a greater impact on UDS than climate change.

An urban drainage network model consists of basically three parts: the precipitation input, the hydrological surface and runoff processes, and the hydrodynamics of the pipe network (Rangari et al. 2020). The purpose of designing stormwater drainage systems is to have a properly efficient network that can alter the problems that are catered by frequent flooding due to insufficient loading of sewerage at the time of heavy rainfall. The objective of the study is to identify the hotspots of the critical elements of the urban drainage networks of Rohtak City, India using MIKE+. Rohtak is a key town in the National Capital Region (NCR) and economic hub for Haryana state and has experienced floods on a regular basis in recent times. The major reasons behind urban flooding in Rohtak city are due to a sudden increment in population, lack of proper infrastructure, and rapid urban sprawl in recent years. The city's drainage systems are also outdated and inadequate for heavy rainfall. Rohtak's flat terrain with poor natural drainage is also one of the major reasons behind water stagnation during heavy rains (Pandey and Singh 2024; Yadav 2013). A sudden rainfall on June 1, 2022, led to urban flooding in Rohtak city which impacted daily life and created a major threat of water contamination. Rohtak city has faced urban flooding almost every year in the past decade. A hotspot is described as the overflowing nodes that led to the local flooding in those areas. Such collective overflowing nodes are the primary reason behind urban flooding. These overflowing nodes collectively form a region defined as hotspot zones (Kansal et al. 2019). These hotspot zones are validated using the Google Earth Engine platform with SENTINEL-1 SAR data to derive a flood extent map that defines the urban flooding zones (Tazmul Islam and Meng 2022). Also, a predictive analysis is completed to determine the impact of future climate on the existing drainage system. Predictive rainfall is based on extreme rainfall events. R1, R2, and R3 are the three predictive rainfalls used for the model simulation (Mehta et al. 2022). Each predictive rainfall's impact on existing drainage is analyzed and accordingly, their disaster level is defined.

2 Study area and data used

2.1 Study area

The city of Rohtak is selected as the study area and has a geographical urban area of 1745 km², and a population of around 446,164 (according to the 2011 census report). The study area is well illustrated in Figure 1. The latitude of Rohtak is 28.895515, and the longitude is 76.606613. GPS coordinates are 28°53'43.8540''N and 76°36'23.8068''E. The district area is in the Yamuna sub-basin of the Ganga basin. The district territory is primarily drained by the artificial drain No. 8, which flows from north to south. The city experiences 592 mm of rain on average per year. Physically, the area is a flat land mass. The strategy primarily entails using MIKE+ to model and simulate the area to create a drainage system that works well. Due to the skewness of rainfall patterns brought on by climate change, flooding in the city is an issue, and as a result, the city needs a planned and efficient drainage network. RL values (elevation) are referenced in the creation of nodes (manholes) connecting them via a pipe network, taking gravity flow into consideration.

Figure 1 Study area map of Rohtak city.

2.2 Data used

The positions of manholes and outlets for runoff are determined using the AutoCAD file of the sewer storm network of the projected area (Rohtak city), which displays various drains and sewer treatment facilities. To determine the elevation of various nodes constructed in MIKE+, SRTM DEM with a 90 m spatial resolution was used. One-dimensional hydrodynamic and rainfall-runoff modeling has been done in MIKE+ for which extreme event rainfall data from June 2022 to September 2022 was taken from the India-WRIS site. An ArcGIS tool is used for the extraction of elevation values of the ground level of the manholes. Results of modeling impervious areas have been ranging from 30% to 70%, based on house density in that region. The Time Area method was used to determine the net effective rainfall-runoff using standard parameters provided by MIKE+ software, such as initial losses and reduction factors that occur due to many reasons such as interception loss, etc. In this model, a constant 0.9 mm initial loss from all catchments, along with a reduction factor of 0.85 is used, and the Time of Concentration was assigned as per the catchment size (CPHEEO recommendation).

2.3 Dimensional data

Node dimensions are classified into two groups: Manholes with a diameter of 1.5 meters and Outlets with a diameter of 2 meters. Pipes and links used were of the following dimensions according to their uses:

1. Trunk line with a diameter of 1.2 m
2. Main line with a diameter of 0.9 m
3. Sewer line with a diameter of 0.6 m
4. Ordinary line with a diameter of 0.3 m

2.4 Rainfall data

Rainfall data from the monsoon period (June to September) from the past 5 years was used for the model analysis and it was observed that extreme rainfall occurred in 2022.

Figure 2 shows the actual precipitation from the 2022 monsoon period (June to September) in Rohtak city.

Figure 2 Daily precipitation data of Rohtak in the 2022 monsoon season.
SOURCE: https://indiadataportal.com/p/climate-data/r/wris-daily_rainfall_district-dt-dl-imd

3 Model description

The MIKE+ and MIKE1D engines use the Implicit Finite Difference Method to solve Saint-Venant equations. This method discretizes the equations over space and time, creating a set of algebraic equations that can be solved iteratively. In MIKE+, the momentum equations are solved using the Implicit Finite Difference Method, which is supported by the Newton-Raphson technique to manage nonlinearities (DHI 2022).

3.1 Saint-Venant equations

The complete one-dimensional theoretical approach results in a pair of equations for gradually varied unsteady flow in open channels (including part-full pipes), which are commonly referred to as "the Saint-Venant equations" after A.J.C. Barré de Saint-Venant, who first published them in the middle of the nineteenth century. A clear derivation of these equations is available in Chow (1959).

There are two Saint-Venant equations: a dynamic equation, which when presented in terms of flow, is termed a momentum equation; and a continuity equation. Both equations are used for finite difference discretization. The dynamic equation can be written as follows:

(1)

Where:

y	=	flow depth (m),
v	=	velocity (m/s),
x	=	distance (m),
t	=	time (s),
So	=	bed slope (−),
St	=	friction slope (−), and
g	=	gravity acceleration (m/s2)

The above equation is commonly presented in terms of flow rate rather than velocity and is given together with the continuity equation:

(2)

(3)

Where:

Q	=	flow rate (m3/s),
A	=	area of flow cross-section (m2), and
B	=	water surface width (m).

The following conditions must be met for the Saint-Venant equations to be valid:

1. The pressure distribution is hydrostatic.
2. The sewer bed slope is so small that the flow depth measured vertically is nearly equal to normal for the bed.
3. At a channel cross-section, the velocity distribution is uniform.
4. The channel is prismatic.
5. Friction losses estimated by steady flow equations are valid in unsteady flow.
6. Lateral flow is negligible.

3.2 Simplifications of the full equations

For Equation 2, the bed slope, friction slope, water depth variation, and flow rate variation with distance and time, are all included in the dynamic equation. There may be chances to simplify the equations because certain terms are more important than others. The greatest simplification is to take Equation 2 and assume that most of the terms can be ignored, reducing it to:

(4)

This is the equivalent of ignoring all but part 1 of Equation 1 and implies that the relationship between flow rate and depth is the same as it would be in steady uniform flow. Combining Equations 3 and 4 gives:

(5)

The wave, called a kinematic wave, does not attenuate but translates at the wave speed c (Butler et al. 2018).

3.3 Pressurized flow

The fundamental presumption used to derive the Saint-Venant equations does not hold true when pipelines or canals are flowing water full because the flow switches from free surface flow to pressurized flow. To include pressurized flow in closed reaches, it may be possible to generalize the equations for free surface flow. The Preissmann slot, a hypothetical slot at the top of the reach, is used to accomplish this (Zhou and Li 2020).

3.4 Finite difference discretization

To solve Saint-Venant equations using the finite difference approach, the domain (the channel's length) is divided into grids, and the partial derivatives are approximated by finite differences at each grid point (Lai and Khan 2018).

Continuity Equation Discretization

Using finite difference approximations for the time (t) and spatial (x) derivatives, we can rewrite the continuity equation for each grid point, as follows:

(6)

Where:

and	=	the cross-sectional area and discharge at point i and time step n,
	=	lateral inflow or source term at grid point i and time step n. It is a term that accounts for any external addition or removal of mass (or volume) at a specific location i, and time step n,
Δt	=	time step size, and
Δx	=	spatial step size

This forward difference in time and backward difference in space technique is referred to as an explicit scheme.

Momentum equation discretization

For the momentum equation, we apply similar finite difference approximations:

(7)

Where:

	=	friction slope, which can be calculated using Manning’s equation:
	=	water depth at a specific grid point i and time step n,
S0	=	bed slope, which derives the flow downstream, and
g	=	a constant, i.e., acceleration due to gravity.

(8)

Where:

n

=

Manning's roughness coefficient.

Explicit and implicit schemes

Depending on the choice of time and spatial differences, two primary methods are often used:

Explicit scheme

In the explicit scheme, the unknowns at the next time step (Aⁿ⁺¹ and Qⁿ⁺¹) are evaluated based on known values from the current time step (Aⁿ and Qⁿ).

This scheme is simple but subject to stability constraints. Specifically, it must satisfy the Courant-Friedrichs-Lewy (CFL) condition:

(9)

which limits the time step size.

Implicit scheme

The implicit scheme involves solving equations for both Aⁿ⁺¹ and Qⁿ⁺¹ simultaneously. This requires solving a system of equations at each time step but is unconditionally stable, allowing for larger time steps. The Preissmann scheme is a popular implicit approach that combines weighted averages of spatial and temporal derivatives to get smoother and more stable solutions even with greater time increments.

Solving the discretized equations

1. Set Initial Conditions: Define the initial water depth and velocity (or discharge) across the channel.
2. Apply Boundary Conditions: Specify inflows, outflows, or fixed water levels at the boundaries.
3. Iterate Over Time Steps: Use the finite difference equations to update values of A and Q at each grid point for each time step until the simulation time is complete.

3.5 Hydrodynamic module

The 1D engine solves the problem of water flowing through a network of nodes and structures. The problem is fully specified by boundary conditions at the network boundaries and initial conditions (Kourtis et al. 2017).

A typical representation of water flowing through a sewer system and its critical points is well represented in Figure 3 (Schmitt et al. 2004). When the water level coincides with the ground level it becomes a critical point for the overflowing of the nodes which shows structural failure of that point. The ideal condition for the sewer flow should be when the water level is below the pipe top level, which maintains the gravity flow of the pipelines.

Figure 3 Mechanism of water flowing through pipes and nodes.

3.6 Time area surface rainfall-runoff model

In this model, the initial loss, the size of the contributing area, and a continuous hydrological loss all influence the amount of runoff. The time of concentration and the time-area determine the form of the runoff hydrograph (T-A) curve (DHI 2017). These parameters define catchment reaction speed and catchment shape. The amount of time needed for the flow to travel from any point over the catchment to the outlet is known as the travel time of flow. This time depends on the length of the trip as well as any other factor influencing the flow velocity, like vegetation cover, land use, slope, and rainfall intensity. Due to variability among these factors in space and time, estimating travel time with precision is difficult, if not impossible (Sabzevari 2017).

The Time/Area curve accounts for the shape of the catchment layout and determines the choice of available T/A curves used in computations. Three predefined types of T/A curves are available: Rectangular, Divergent, and Convergent catchment as shown in Figure 4.

Figure 4 Time-area curve used in MIKE+.

4 Methodology

4.1 Model construction

The area under consideration in this study is quite vast. To precisely study the projected area, the catchment delineation tool MIKE+ is used to delineate the area into smaller parts. MIKE+ uses the Thiessen Polygon method for catchment delineation which provides each node a separate sub-catchment, with the help of which, the data of each node can be analyzed (Luan et al. 2018). The land imperviousness percentage in this study is projected based on a house density evaluation, which lies between the range of 30 to 70 percent (Alley and Veenhuis 1983).

The basic construction of this project methodology starts from the selection of an area using the Google Earth Pro engine, then using MIKE+ software to create nodes and catchments, as discussed in Figure 5. In this study, the total number of nodes (diameter – 1.5 m) employed was 537, of which 533 are manholes and 4 are outlets (diameter – 2.0 m). The connection of nodes was done using their respective invert levels (bottom level) which were in the range of 0.8 and 5.2 m from the ground. The manholes are circular with a diameter of 1.5 m. Gravity is the dominant flow type in the sewer network, which is why the nodes are connected from high elevations to lower elevations. However, occasionally, significant runoff can result in pressurized flow, which can be avoided by making the appropriate assumptions. Furthermore, a total of 56 major connection lines were used for the analysis, of which 1 was a trunk line, 2 were main lines, and 7 were sewer lines. After the creation of nodes and links, the catchment connection was done, and initial and boundary conditions were defined for the project simulation.

A major part of the initial condition consists of the imperviousness area and determining the person equivalent of each sub-catchment. The person equivalent for this project is calculated according to the CPHEEO formula, i.e., PE = round (catchment area*model impervious area/100/0.014*5). Assuming 1500 square feet per house area (0.014 hectares) and 5 people per house. The person equivalent calculated using the above formula was found to be 1,170,029.

For rainfall-runoff modeling, rainfall time-series data was used for the analysis along with a rectangular time-area curve (DHI 2020). After deducting the initial loss, the runoff volume was computed by continuously reducing the precipitation on the impervious and semi-pervious surfaces. The temporal flow fluctuation on the surface is calculated as the amount of time that precipitation falls between when it hits the surface, and when it reaches the main drainage pipe (Thorndahl and Schaarup-Jensen 2007). For the below parameters, MIKE+ uses the CPHEEO recommendations: Initial Loss = 0.9, Reduction Factor = 0.85, while Time of Concentration was calculated for each sub-catchment as per Time-Area Curve by MIKE+ based on their shape and size.

Figure 5 Methodology.

4.2 Flood extent mapping using Google Earth Engine (GEE) with SENTINEL-1 SAR imagery data

The Google Earth Engine (GEE) is a cloud-based geospatial analysis platform that enables users to visualize and analyze satellite images. The GEE platform is used for the visualization of SENTINEL-1 SAR imagery data to derive the flood extent mapping of Rohtak city by analyzing before-flood conditions and after-flood conditions.

The Sentinel-1 mission provides data from a dual-polarization C-band Synthetic Aperture Radar (SAR) instrument at 5.405GHz (C-band). This collection includes the S1 Ground Range Detected (GRD) scenes, processed using the SENTINEL-1 Toolbox to generate a calibrated, ortho-corrected product.

The GEE code editor is used to extract SENTINEL-1 SAR data along with the area of interest, i.e., Rohtak region. In the pre-processing phase, the flood image is analyzed from March 2022 to April 2022 when rainfall is minimal, while the after-flood image is analyzed for the period of extreme rainfall, i.e., from June 2022 to September 2022. After this, a smoothening filter is applied to both images to remove any speckle effect. The speckle effect is due to the interference between pixels which act as waterlogging pixels. The difference between the flood image and before-flood image after applying the smoothening filter provides a flood extent map, as shown in Figure 7. The methodology for this process is shown in Figure 6.

Figure 6 Methodology for flood extent mapping using GEE and SENTINEL-1 SAR imagery data.

Figure 7 Different stages of flood extent mapping.

4.3 Predictive rainfall data mapping

Climate change, caused by human activity, is changing Earth's climate patterns at an unprecedented rate. Among its many repercussions, changes in rainfall patterns stand out as especially significant. To map the future failure of urban drainage, three predictive rainfalls are used for the analysis of model simulation, namely Rainfall1, Rainfall2, and Rainfall3. Rainfall1 (R1) is a 10% increment in actual extreme rainfall data that occurs in 2022, while Rainfall2 (R2) and Rainfall3 (R3) are 20% and 50% increments in actual extreme rainfall data, respectively, as shown in Figure 8.

Figure 8 Predictive rainfall R1, R2, and R3 mapping.

5 Results and discussion

5.1 Model simulation of actual extreme rainfall data

MIKE+ gives detailed findings for water depth and water level at nodes, as well as runoff volume for each catchment. Simulation using MIKE+ produces findings in the form of profile charting for any linkage in the study region, as well as an animation compilation of runoff volume from each sub-catchment. The outputs will be classified as profile plot discussion and animation analysis of nodes, linkages, and catchments.

Profile plot evaluation and analysis

Profile plotting in MIKE+ is essentially a hydrodynamic modeling function that displays cross-sectional data for every flow channel in the study region using the Set Flag tool. Cross-sectional data includes various parameters such as link water level, node water level, and many more, with plotting options for maximum, minimum, or average value animation. With the help of animation of links and nodes, water level profile plotting, and overflowing nodes/links can be found, as shown in Figure 9.

Figure 9 Profile plot of major trunk line.
NOTE: The green line shows the ground level, light blue and dark blue lines show nodes and link water level, respectively, while the bottom line shows the invert level.

The trunk line profile diagram shows node and connection failures at different places. Due to the collapse of these nodes and links, local urban flooding will ensue. So, when using profile plotting, any flow channel may be studied and discussed for structural behavior during the simulation period.

Simulation analysis through animation

MIKE+ simulates the model and generates animation files (res1d) that are executable within MIKE+. These animation files allow for a basic examination total runoff of a watershed. Figure 10 depicts the total runoff from each sub-catchment in the model. Critical sub-catchments are shown in red, indicating an overflooded zone. Green and yellow indicate semi-flooded zones. Simulation of these animation files yields results for nodes and pipes. The animation of pipes and canals depict the volume of water flowing through pipes, which defines whether the flow is gravity or pressure. Pressured flow in pipes will be deemed a hydraulic failure. The animation of the nodes displays the water depth of each node, which is used to detect if the nodes are overflowing, as shown in Figure 11.

Figure 10 Total runoff animation in each sub-catchment and water level depth in each node and link.

Figure 11 Illustration of overflowing nodes and failed hydraulic links with critical
sub-attachments.

Failure nodes and areas

Node failures occur when the volume of water in nodes surpasses the critical threshold, causing local flooding surrounding the nodes. Link failures are due to a full flowing condition of links, which leads to link congestion and increment in volume of water above the critical level in nodes. Sub-catchment failure is due to the local flooding condition arising from link and nodes failure. These three features are the main contributors to urban flooding. These key features are measured to estimate the degree of urban flooding. Table 1 shows the numbers from June to September 2022, where the features failed during the MIKE+ model simulation, and a graphical representation of these features is shown in Figure 11

Table 1 Number of overflowing nodes, failed links, and critical sub-catchments.

Elements	Numbers
Nodes	52
Links	57
Sub-catchments	07

5.2 Validation

To quantitatively demonstrate the consistency of MIKE+ urban flood mapping output and to show how well MIKE+ performed, the Google Earth Engine (GEE) platform along with SENTINEL-1 SAR imagery data is used to derive a flood extent map for the extreme rainfall period, i.e., from June to September 2022, as shown in Figure 12.

Figure 12 Comparison of MIKE+ and GEE with SENTINEL-1 SAR data flood mapping.

QGIS software is used for the analysis of flood maps derived from MIKE+ and GEE platforms. Flooded area percentage and pixels percentage flooded are the two parameters used for the comparison of MIKE+ and GEE flooding images, as shown in Table 2 below, and the flooded area-based analysis is given.

Table 2 Accuracy assessment between MIKE+ and GEE model based on percentage flooded area.

Urban Flood Map	Total sub-catchment area (km2)	Flooded sub-catchment area (km2)	% Flooded area	Accuracy
MIKE+	449.05	40.99	9.13%	79.66%
GEE using SENTINEL-1 SAR	449.05	51.45	11.46%

MIKE+ performs significantly well in percentage area flooded, with an accuracy of 79.66%, while in pixel percentage flooded, accuracy is reduced to 55.4%. Furthermore, the results of this study are comparable with other flood mapping studies.

5.3 Model simulation of predictive rainfall data

For the prediction of flooding area increment in the future, a predictive rainfall model is prepared based on the extreme rainfall event in 2022, which is calculated based on the incremental theory of increasing extreme rainfall to 10% (R1), 20% (R2), and 50% (R3). These predictive rainfall time series are further used as base rainfall time series for the model simulation to derive the urban flooding map when the intensities of rainfalls occur in the near future, as shown in Figure 13.

Figure 13 Change in flooding areas for different rainfall time series, i.e., R1, R2, and R3.

This data is further analyzed using ARC GIS to determine the exact number of flooding elements along with the comparative relationship between different rainfall time series and change in flooded areas with respect to total sub-catchment area.

Figure 13 shows the MIKE+ simulation of the different predictive rainfall time series, i.e., R1, R2, and R3, while the quantitative representation is shown in the tables below. In Table 3, the number of failures of nodes, links, and sub-catchments are shown for the predictive rainfall time series of R1, R2, and R3. It shows the comparative change in the number of nodes, links, and sub-catchments failures when the rainfall time series changes between R1, R2, and R3.

Table 3 Total number of overflowing nodes, hydraulic failed links, and critical sub-catchments for different predictive rainfalls.

Elements	Rainfall1 (R1)	Rainfall2 (R2)	Rainfall3 (R3)
Nodes	58	58	87
Links	72	82	113
Sub-catchments	08	10	16

Table 4 shows the percentage increment in the failure rate of nodes, links, and sub-catchments for different rainfall time series (R1, R2, and R3) when compared with the base extreme rainfall from 2022 in the MIKE+ model simulation. As Table 4 shows, the percentage rate of failure for nodes, links, and sub-catchments increases drastically for the R3 predictive rainfall time series, which makes it a most critical event.

Table 4 Percentage change in failure of critical elements for different predictive rainfalls.

Elements	% increment in R1	% increment in R2	% increment in R3
Nodes	11.54%	11.54%	67.3%
Links	26.31%	43.85%	98.24%
Sub-catchments	14.28%	42.85%	128.57%

NOTE: Colour key: light green = low-moderate conditions, dark green = moderate conditions, yellow = low-severe conditions, red = severe conditions, and dark red = extreme severe conditions.

Table 5 highlights the critical sub-catchment areas simulated under different rainfall time series with their percentage coverage of the total sub-catchment area.

Table 5 Critical sub-catchment areas (km²) for different rainfalls and their percentage of total catchment area.

Total sub-catchment area (km2)	Extreme rainfall (km2) (% total area)	R1 (km2) (% total area)	R2 (km2) (% total area)	R3 (km2) (% total area)
449.05	40.99	51.41	74.89	100.52
	(9.13%)	(11.45%)	(16.67%)	(22.38%)

Hence, the results show that the flooding increment for R1 and R2 is in the moderate zone and can be tackled with proper arrangements, but for rainfall time series R3, the increment in flooding zones is drastic and should be handled with utmost importance, otherwise it can lead to an urban flooding disaster. The R3 time series shows that 22.38% of the total sub-catchment areas will flood when an R3 rainfall intensity occurs.

6 Conclusion

Urban flooding occurs when rainwater exceeds the capacity of urban drainage systems (UDS) and inundates streets, buildings, and other infrastructure. Urban areas have obsolete or insufficient drainage systems built to handle the amount of rainwater produced by urban development. As a result, after severe rainfall, drainage systems may become overloaded, allowing water to pool on streets and flood low-lying regions. The current study focused on the identification of critical areas where functional failure of the urban drainage system could happen. An extreme rainfall event in 2022 is used for the analysis of the hydrological model which is prepared using MIKE+ and validated using the GEE platform with the help of SENTINEL-1 SAR imagery data.

Extreme rainfall data from 2022 shows 52 node failures, 57 link failures, and 7 sub-catchment failures, resulting in 9.13% of the total catchment flooding. When using SENTINEL-1 satellite imagery, the flooded area is 11.46%. Validation shows the percentage of flooded areas have a great accuracy of 79.66%, which makes the MIKE+ model a dependable model. Modeling using data from the extreme rainfall event of 2022 illustrates the areas of Rohtak city to focus on to avoid any future flooding issues.

Furthermore, predictive analysis is done for the most sensitive parameter – precipitation. In predictive analysis, rainfall mapping is done using extreme event rainfall data. R1, R2, and R3 are the three predictive rainfall categories, with 10%, 20%, and 50% increments to extreme events, respectively. R1 rainfall data shows 58 node failures, 72 link failures, and 8 sub-catchment failures. R2 rainfall data shows 58 node failures, 82 link failures, and 10 sub-catchment failures. R3 rainfall data shows 87 node failures, 113 link failures, and 16 sub-catchment failures. Also, it shows the importance of parameters like land imperviousness percentage, reduction factor, and the initial losses over the urban drainage network modeling. The cause of functional failure of UDS is predominantly high precipitation in the study area. It was observed that the result obtained from predictive rainfall R3 shows 22.38% of the total catchment area under flooding causing drastic failure of UDS.

As R3 rainfall shows about 23% of total sub-catchment areas under water, flooding is a huge concern and to avoid disaster, it should be properly managed, and the drainage capacity should be modified accordingly. The advantage of predictive analysis is to identify vulnerable spots in the catchment areas as extreme rainfall is predicted to increase in coming years due to climatic variability.

This research work can be further carried out using more sensitive parameters such as Urbanization along with Precipitation. Also, Low Impact Development (LID) techniques can be used to control urban flooding conditions. LID techniques are sustainable practices that aim to manage stormwater close to its source in urban areas, reducing flooding and improving water quality. Some of the LID techniques that can be used are Permeable Pavements, Rain Gardens, Green Roofs, Detention and Retention Basins, etc. These LID techniques provide benefit by reducing the strain on stormwater infrastructure and provide resilience against climate change-induced extreme rainfall events.

Acknowledgements

The authors are extremely thankful to DHI (India) Water & Environment Pvt Ltd, New Delhi and the office of the Sub Divisional Officer, Public Health Engineering Department, Circle Rohtak, Haryana for providing MIKE+ software and data, respectively.

References

1. Alley, W.M., and J.E. Veenhuis. 1983. “Effective Impervious Area in Urban Runoff Modeling.” Hydraulic Engineering 109, 2. https://doi.org/10.1061/(ASCE)0733-9429(1983)109:2(313)
2. Bisht, D.S., C. Chatterjee, S. Kalakoti, P Upadhyay, M. Sahoo, and A. Panda. 2016. "Modeling urban floods and drainage using SWMM and MIKE URBAN: A case study." Natural Hazards, 84 (2): 749–776. https://doi.org/10.1007/s11069-016-2455-1
3. Butler, D., C.J. Digman, C. Makropoulos, and J.W. Davies. 2018. Urban Drainage. CRC Press, Boca Ratan, USA.
4. Chow, V.T. 1959. Open Channel Hydraulics. McGraw-Hill, New York, USA.
5. Dang, T.Q., B.H. Tran, Q.N. Le, T.D. Dang, A.H. Tanim, Q.B. Pham, V.H. Bui, S.T. Mai, et al. 2024. "Application of machine learning-based surrogate models for urban flood depth modeling in Ho Chi Minh City, Vietnam." Applied Soft Computing 150, 111031. https://doi.org/10.1016/j.asoc.2023.111031
6. Dey, H., W. Shao, H. Moradkhani, and B.D. Keim. 2024. "Urban flood susceptibility mapping using frequency ratio and multiple decision tree-based machine learning models." Natural Hazards 120, 10365–10393. https://doi.org/10.1007/s11069-024-06609-x
7. DHI. 2017. MIKE 1D Reference Manual, Simulation. 1–334. DHI Water & Environment, Hørsholm, Denmark.
8. DHI. 2020. MIKE URBAN + Collection System User Guide. DHI Water & Environment, Hørsholm, Denmark.
9. DHI. 2022. MIKE 1D, DHI Simulation Engine for 1D river and urban modelling. Reference Manual, DHI Water & Environment, Hørsholm, Denmark.
10. Kalore, S., V. Yashas, A. Bagrecha, J. Nypunya, and G.L. Sivakumar Babu. 2024. "Climate responsive design for road surface drainage systems: A case study for city of Bengaluru." Urban Water Journal 21 (3): 295–307. https://doi.org/10.1080/1573062X.2023.2287049
11. Kansal, M.L., O. Mehta, and A. Tyagi. 2019. "Hotspot Identification for Urban Flooding in a Satellite Town of National Capital Region of India." World Environmental and Water Resources Congress 2019: Emerging and Innovative Technologies and International Perspectives.
12. Kartheeshwari, M.R., and L. Elango. 2022. "2021 Chennai Floods – An Overview." Journal of the Geological Society of India 98 (6): 863–864. https://doi.org/10.1007/s12594-022-2079-x
13. Kourtis. I.M., V. Bellos, and V.A. Tsihrintzis. 2017. "Comparison of 1D-1D and 1D-2D urban flood models." 15th International Conference on Environmental Science and Technology, Rhodes, Greece, 1–6.
14. Kourtis, I.M., and V.A. Tsihrintzis. 2021. "Adaptation of urban drainage networks to climate change: A review." Science of the Total Environment 771, 145431. https://doi.org/10.1016/j.scitotenv.2021.145431
15. Lai, W., and A.A. Khan. 2018. "Numerical solution of the Saint-Venant equations by an efficient hybrid finite-volume/finite-difference method." Journal of Hydrodynamics 30, 189–202. https://doi.org/10.1007/s42241-018-0020-y
16. Luan, Q., K. Zhang, J. Liu, D. Wang, and J. Ma. 2018. "The application of Mike Urban model in drainage and waterlogging in Lincheng county, China." Proceedings of the International Association of Hydrological Sciences 379, 381–386. https://doi.org/10.5194/piahs-379-381-2018
17. Mehta, O., M.L. Kansal, and D.S. Bisht. 2022. "Evaluation of an urban drainage system using functional and structural resilience approach." Urban Water Journal 20 (10): 1794–1812. https://doi.org/10.1080/1573062X.2022.2044495
18. Mukherjee, A.B., and S. Bardhan. 2019. "Studies on Flood Vulnerability Assessment of Kolkata under a Changing Climate: A Quantitative Inferential Approach an Assessment of Direct and Indirect Emission Reduction Potential of Natural Wetland Systems of Kolkata, India." International Journal of Applied Engineering Research 14 (8): 1923–1930. https://dx.doi.org/10.37622/IJAER/14.8.2019.1923-1930
19. National Disaster Management Authority (NDMA). 2010. National Disaster Management Guidelines: Management of Urban Flooding. Government of India (GofI).
20. National Disaster Management Authority (NDMA). 2020. National Disaster Management Authority Annual Report 2020, Government of India (GofI). https://www.ndma.gov.in/Resources/AnnualReports
21. Pandey, D., and K.K Singh. 2021a. "Identification of Critical Urban Drainage Flooding Points Using MIKE URBAN+: A Case Study of Rohtak, Haryana, India." SDCEE-2021: 1st International Online Conference on Sustainable Development in Civil and Electrical Engineering, 11.
22. Pandey, D., and K.K Singh. 2021b. Urban Drainage Network Modelling Using Mike Urban+. NIT Kurukshetra.
23. Pandey, D., and K.K. Singh. 2024. "Comparison of Urban Flood Susceptibility Maps of MIKE+ and AHP with GIS Integration: A Case Study of Rohtak City, India." ASCE 15 (1): 37–48.
24. Rangari, V.A., V. Sridhar, N.V. Umamahesh, and A.K. Patel. 2020. "Rainfall Runoff Modelling of Urban Area Using HEC-HMS: A Case Study of Hyderabad City." In: AlKhaddar, R., Singh, R., Dutta, S., Kumari, M. (eds) Advances in Water Resources Engineering and Management. Lecture Notes in Civil Engineering , vol. 39. Springer, Singapore. https://doi.org/10.1007/978-981-13-8181-2_9
25. Rossman, L.A. 2015. Storm Water Management Model (SWMM) User's Manual Version 5.1. U.S. Environmental Protection Agency (EPA), 1–353, Ohio, USA.
26. Sabzevari, T. 2017. "Runoff prediction in ungauged catchments using the gamma dimensionless time-area method." Arabian Journal of Geosciences 10 (6): 0–10. https://doi.org/10.1007/s12517-017-2852-0
27. Schmitt, T.G., M. Thomas, and N. Ettrich. 2004. "Analysis and modeling of flooding in urban drainage systems." Journal of Hydrology 299 (3–4): 300–311. https://doi.org/10.1016/j.jhydrol.2004.08.012
28. Seenirajan, M., M. Natarajan, R. Thangaraj, and M. Bagyaraj. 2017. "Study and Analysis of Chennai Flood 2015 using GIS and Multicriteria Technique." Journal of Geographic Information System 09 (02): 126–140. https://doi.org/10.4236/jgis.2017.92009
29. Tazmul Islam, M., and Q. Meng. 2022. "An exploratory study of Sentinel-1 SAR for rapid urban flood mapping on Google Earth Engine." International Journal of Applied Earth Observation and Geoinformation 113, 103002. https://doi.org/10.1016/j.jag.2022.103002
30. Thorndahl, S., and K. Schaarup-Jensen. 2007. "Comparative analysis of uncertainties in urban surface runoff modelling." The International Conference on Sustainable Techniques and Strategies in Urban Water Management, Lyon, France, 1631–1638. http://vbn.aau.dk/files/44971436/Comparative_Analysis_of_Uncertainties_in_Urban_Surface_Runoff_Modelling_Abstract.pdf
31. Tripathy, S.S., S. Chaudhuri, R. Murtugudde, V. Mhatre, D. Parmar, M. Pinto, P.E. Zope, et al. 2024. "Analysis of Mumbai floods in recent years with crowdsourced data." Urban Climate 53 (101815): 1–34. https://doi.org/10.1016/j.uclim.2024.101815
32. Yadav, V. 2013. "Vulnerability of a District : A Case of Rohtak , Haryana." The SPA Journal of Planing and Architecture 17 (2): 93–101.
33. Zhou, J., and Y. Li. 2020. "Modeling of the free-surface-pressurized flow of a hydropower system with a flat ceiling tail tunnel." Water (Switzerland) 12 (3): 699. https://doi.org/10.3390/w12030699