Flood hazard analysis in Mumbai using geospatial and multi-criteria decision-making techniques

Floods as a threat have attracted more attention than others for many years now as they influence cities and countryside equally. Flood annually affects millions of people, highlighting them as a frequent but lethal phenomenon (Halgamuge & Nirmalathas 2017). What is even more disquieting is the remarkable upsurge in the frequency of these events over the past three decades, positioning floods at an unparalleled level of risk (Mishra et al. 2022). Some of these flood occurrences have unleashed catastrophic and unforeseen devastation, as witnessed in the form of glacial lake outburst floods and flash floods, prominently illustrated by the tragic events across the world, South Asia Floods (2017) (Palash et al. 2020), Central Europe Flooding (2013) (Grams et al. 2014), Japan's Typhoon Hagibis (2019), Jakarta Indonesia (2020), Bangladesh Monsoon Floods (2020) (Aishi et al. 2023), India Monsoon Floods (2013, 2015, 2018) (Kumar & Yadav 2020; Mishra 2021), etc. The escalating frequency and severity of these flood-related disasters demand our unwavering attention and concerted efforts to understand and address their underlying causes, impacts, and mitigation measures. The urgency of this matter cannot be overstated, as floods continue to jeopardize the safety and well-being of communities around the world (Sivakumar 2011).

Although it may be an insurmountable challenge to eliminate the threat of flood hazards, the adoption of effective mitigation strategies can significantly diminish both the severity of the disaster and its protracted consequences. There are a lot of causes that lead to flooding. Among them is climate change, which seems nowadays to have an increasing influence on the risk of flooding (Tsatsaris et al. 2021; Jamal et al. 2024). Besides weather changes, human beings have also played a significant role by causing alterations such as changes in land uses, urban expansion, removal of forests, and road development in the flood-prone areas (Kumar et al. 2023). It is observed that these human-induced interventions break the natural pathways for draining and this results in increased runoff which then leads to flooding thus causing destruction (Amoateng et al. 2018). In other words, the interactivity of people and nature results in increasing danger of flooding. Changes in nature's landscape, excessive urbanization, and pollution of rivers have made cities more vulnerable to floods (Mortoja & Yigitcanlar 2020). Global warming, besides these changes in climatic conditions, has turned what used to be only occasional flooding into a permanent threat of flooding to cities around the globe (Whitfield 2012). As a result, it is important to recognize this complex linkage and introduce comprehensive flood mitigation and management measures that will ensure that lives, properties, as well as the environment are not subjected to destruction by floods (Umrigar et al. 2023).

The ever-expanding urbanization and proliferating artificial surfaces have reconfigured how rainwater runs-off (Ahmed et al. 2019). This metamorphosis implies significant repercussions as regards inundation occurrence during flooding. These incidents extend even to the most developed city zones as well as rural settlements inhabited by populations which chose them for living. Such artificial surfaces which are made up of concrete, asphalt and many more do not allow sufficient absorption and retention of water (Scalenghe & Marsan 2009). They function as effective pipes through which the heavy rain trickles onto the streams resulting in devastating floods during rainstorms. Consequently, previous regions become susceptible to flooding, and the actual area affected by floods becomes much larger. The growing challenge requires effective utilization of land resources as well as improved urban planning techniques (Lehmann 2018). Proactive actions such as sustainable urban design, inclusion of green spaces, promotion of porous pavements, and preserving the original drainage channel can mitigate flooding problems (Andrés-Doménech et al. 2021). This is vital as it helps lessen the impact of floods, particularly in areas being affected by the increased risk of flooding. The significance of proper land management should be recognized as this plays a pivotal role in reducing the increasing flood impacts towards our towns.

Flood risk mapping is an essential tool that offers an advanced approach to predicting areas and extents of flood devastations, as well as providing an adequate response. However, this mapping takes place as the result of a complicated process which requires thorough investigations. This is a very complex endeavour, and it involves the convergence of geographic information systems (GIS) and multi-criteria decision-making (MCDM) approaches, such as the analytical hierarchy process (AHP). These essential spatial technologies help to integrate vast input datasets and form spatial analysis on which the resultant flood risk map is based. GIS together with the AHP increases the accuracy as well as speed of flood analysis while at the same time giving instantaneous broadcast of crucial flood-related data and warnings to people (Wang et al. 2018; Hammami et al. 2019; Kanani-Sadat et al. 2019). Spatial data infrastructures as well as web-based GIS systems are critical because they make it possible for communities to be informed and prepared for impending flood occurrences (Karymbalis et al. 2021).

However, flood risk assessment is vital and mitigation measures are required yet very few studies are found especially in South Asia. India suffered from several tragic floods requiring the performance of flood risk analysis with GIS mapping all over the country (Matheswaran et al. 2019). However, this paucity stands out, especially in terms of the application of satellite imagery in unravelling the effects of floods in land use and land cover (LULC), at a sub-district zone. The present study tries to fill a void in this research by venturing into uncharted waters. It seeks to come up with a novel and comprehensive risk assessment approach for floods integrating the powers of GIS, remote sensing, and AHP. The study would not only look at the intricacies of flood risks, but it would also give direction towards educated and smart flood management, particularly for South Asia. This research therefore aims at identifying flood-prone areas and developing vulnerability zones using MCDM methods due to challenges involved in flood risk assessment.

This study intends to spatially link AHP, GIS, and remote sensing to prepare a flood risk map for Mumbai. Criteria employed in the development of flood hazards layer could also vary depending on slope, land use, proximity to road and river, rainfall intensity, elevation, density, and topographic wetness index (TWI). The aim of this paper is to find out flooded locations in Mumbai that are associated with hazard and vulnerability considerations.

In 2005, Mumbai was hit by the worst floods, and experienced extensive damage due to monsoon rains. It was a terrible event that led to massive injuries and destruction. Flooding in Mumbai in 2005 underscores the vulnerability of the city to heavy rainfall, with proper mitigation measures needed to ensure effective flood control practices. It was not only in 2005 but also in 2017 that Mumbai experienced serious flooding. Heavily pounded monsoons flooded many streets and made parts of the city look like islands in the ocean. Floods have been a major problem in Mumbai as well in 2019. Water logging in several places occurred during July's heavy rains thus interrupting the day-to-day activities and movements among others. These regular flood episodes arise from the town's location, proximity to the Arabian ocean, and a redundant sewer system incapable of handling heavy rains and runoffs effectively. However, such events are mainly associated with an older drainage system that fails to handle excess water appropriately. This happens repeatedly, creating a huge flood problem that needs tackling thoroughly to solve Mumbai's flood problems. The data for this research were collected in a systematic way through various institutions such as governments, reports in publications, journals, and other map sources. This allowed for a more complete approach that was credible enough. Topography mapping is an important aspect when studying the land. It was, therefore, necessary to use remote sensing data from Landsat 8, ASTER Global Digital Elevation data, and thematic layers from Survey of India toposheets to carry out a detailed analysis.

The methodology for mapping flood risk and assessing vulnerability in Mumbai is as described below.

• (a) Data collection

This study used data from different sources such as government agencies, published documents, books, research magazines, and libraries.

• (b) Criteria selection

Caution was exercised in determining criteria for flood hazard layers to develop flood risk estimation. These criteria included a variety of elements, including:

• Slope: to evaluate the topographical characteristics of the area.

• Land use: to understand how different land uses impact flood risk.

• Proximity to roads and rivers: considering the influence of transportation infrastructure on flooding.

• Intensity of rainfall: to account for the impact of major rainfall occurrences.

• Elevation: used to determine the vulnerability of low-lying locations.

• Drainage density (DD): this is used to determine the effectiveness of drainage systems.

• Topographic wetness index (TWI): a measure of a terrain's moisture.

• (c) Flood risk mapping

A GIS was employed to develop flood hazard layers using the previously mentioned criteria. The layers represent space-wise indicators of flood hazards in the study area.

• (d) Analytic hierarchy process (AHP)

In applying the AHP, the research aimed to identify flood-prone areas. It interrogates issues on dangers and vulnerabilities. The decision-making process using the AHP involves analysing every criterion at the same level, giving weight to all of them.

• (e) Vulnerability assessment

Through multi-hierarchical layers of hazards and AHP methodology, a more robust assessment of vulnerability can be done covering various flood risks in Mumbai.

The AHP has become a systemic and commonly applied technique in complex decision environments, which helps to break down difficult decision-making processes into understandable parts. The AHP goes through these stages.

• Step 1: Determine the decision problem

Begin by defining the decision problem or goal that must be accomplished. This lays the groundwork for the AHP analysis.

• Step 2: Create a hierarchy

Make a hierarchical structure that depicts the primary goal, criteria, and sub-criteria.

• Step 3: Pairwise comparisons

Calculate the weights for criteria and sub-criteria based on the pairwise comparisons. These weights represent the relative importance of each element in the hierarchy. Table 1 shows Satty's scale of relative importance.

• Step 5: Consistency check

Aggregation involves combining the weights of criteria and sub-criteria to make informed decisions or rank alternatives.

• Step 7: Develop overall priority ranking

Utilize the weight of criteria to create an overall priority ranking, which aids in decision-making processes.

The amalgamation of GIS and MCDM tools has been effectively applied in the city of Mumbai. This study encompassed an intricate examination of seven critical flood susceptibility factors, comprising annual runoff, normalized difference vegetation index (NDVI), DD, elevation, topography wetness index (TWI), distance to the main channel (river and road), and LULC. These characteristics were rigorously evaluated to create a thorough flood hazard index map, which provided significant insights into the city's flood risk landscape. This is an interdisciplinary approach that epitomizes the use of GIS and MCDM in solving complex issues such as flood vulnerability and is one of the many modern research approaches.

Rainfall is a critical factor that significantly contributes to the onset of flooding. Flooding occurs when increased surface runoff, driven by heavy rainfall, overwhelms the capacity of the river channels to handle the excess water flow (Kantharia et al. 2024). In the study area, there is an uneven distribution of rainfall, with a notable increase in rainfall towards the western side of the region. Figure 3(b) depicts the mean annual rainfall (averaged from 2018 to 2022) classes in the study area, categorized as very low (1,015.08–1,063.25 mm), low (1,063.26–1,091.99 mm), medium (1,092–1,110.24 mm), high (1,110.25–1,133.0 mm), and very high (1,133.06–1,184.94 mm). However, it is important to note that the region is characterized by high slope angles and elevation. Even though the upstream areas receive slightly higher rainfall compared to the central part of the watershed, the central and downstream regions have relatively flat topography, very gentle slopes, and poor drainage conditions. These geomorphic characteristics tend to overshadow the influence of rainfall on flood hazards in these areas.

Elevation raster layers were generated using the GIS and the digital elevation model (DEM) data. Subsequently, the elevation raster layers were classified into five distinct groups using the reclassification tool. This classification was based on the elevation's impact on flood hazard, with higher elevations being less prone to flooding, and lower elevations more susceptible. The elevation categories for the research area were defined as very high (ranging from 258.1 to 494 m), high (164.1–258 m), moderate (87.01–164 m), low (29.01–87 m), and very low (−89 to 29 m), as illustrated in Figure 3(c). A DEM serves as a graphical representation of the terrain's surface. In this study area, Shuttle Radar Topography Mission (SRTM) DEM data were acquired from the USGS Earth Explorer. The study area was covered by SRTM one arc-second global DEM data tiles, and four elevation images corresponding to the study area were downloaded and seamlessly combined (mosaicked). The resulting mosaicked DEM file was then clipped to match the study area boundary, while Landsat 8 OLI/TIRS data from the USGS Earth Explorer were obtained for the Mumbai City area.

A higher value of DD indicates a relatively dense network of streams. The DD map of Mumbai, as shown in Figure 3(d), is categorized into five classes based on flood hazard guidelines: very low (0–0.8869), low (0.887–1.7739), moderate (1.774–2.6608), high (2.6609–3.5478), and very high (3.5479–4.4347). These categories are assigned weights of 1, 2, 3, 4, and 5, respectively, as illustrated in Figure 3(d).

In this equation, RED and NIR represent the spectral reflectance measurements obtained in the visible (red) and near-infrared regions, respectively.

The topographic wetness index (TWI) is defined as⁠, where represents flow accumulation and is the slope in radians. The TWI is a useful technique for determining the link between an area's geomorphology and hydrographic features. It aids in the prediction of saturated land surfaces that have the potential to create overland flow during heavy rain events. TWI layers are useful in detecting regions prone to flooding because of excessive runoff following heavy rainfall. The TWI is calculated using the ArcGIS Model Maker tool. It is crucial to remember that not every cell in the TWI layer has a valid TWI magnitude, particularly in places near or within bodies of water. To remedy this, the TWI layer is subjected to the Focal Statistics tool, which eliminates cells having undefined values. As shown in Figure 3(f), the TWI layer is manually divided into five categories: very low (−9.174 to −4.961), low (−4.96 to −2.969), moderate (−2.968 to −0.5181), high (−0.518 to 2.393), and very high (2.394 to 10.36). These classes provide a visual representation of the TWI values and their association with flood susceptibility.

The proximity to a river is a critical indicator of flood hazard because areas near rivers are more susceptible to frequent flooding compared to areas located farther away. Areas situated very near rivers are associated with a very high flood hazard, while areas within certain distances from rivers are considered to have a lower flood hazard. Figure 3(g) provides a map displaying the distances from rivers in the study area, which range from 0 to 0.1372. Based on these distance values, the classification into five distinct classes is as follows: very low, low, moderate, high, and very high. The highest weight, assigned a value of 5, is attributed to the class with a range of 0.1099–0.1372, indicating the highest flood hazard. In contrast, the class with a range of 0–0.02744, signifying the lowest flood threat, receives the lowest weight, with a value of 1. This categorization approach aids in visualizing the flood danger potential in various river proximity zones.

Roads are an important manmade component that adds to flood dangers. A road map was prepared to identify probable road construction barriers in the research region. Such impediments can hamper the flow of floodwaters and make their passage difficult. Roads, trains, bridges, and other infrastructure in a watershed can restrict floodwater discharge capacity. These constructions, which frequently cover a large amount of the ground surface, have a limited ability to hold rainwater and snowmelt. Furthermore, road and building development frequently includes the loss of flora, soil, and depressions from the natural environment. Figure 3(h) depicts a map with distances from roads inside the study area ranging from 0 to 0.1346. Based on their closeness to roadways, these distances are classified into five categories: very low, low, moderate, high, and very high. The highest weighted class, with a score of 5, corresponds to a range of 0.1078–0.1346, indicating the highest flood hazard associated with near proximity to roads. In contrast, the class with the lowest weight, with a value of 1, is designated for the class with a range of 0–0.02693, signifying the lowest flood threat in locations away from roads. This classification helps assess the flood hazard potential in relation to road infrastructure within the study area.

In the creation of the flood risk map, parameter layers are developed, and the reclassification of criteria classes is conducted to assign weights. The AHP is employed as an MCDM tool, which involves the assignment of weights and consideration of the relative importance of thematic layers at each criterion level. In the AHP process, pairwise comparisons are a vital initial step for all potential criteria, and in this case, there are eight criteria involved. As shown in Table 3, these pairwise comparisons result in the construction of a comparison matrix. This matrix is important in assessing the relative relevance of the criteria used to generate the flood risk map. Table 4 shows the normalized pairwise comparison matrix, which is produced from the AHP's pairwise comparisons. The values in this matrix show the normalized relative importance of each element in relation to the others. The diagonal values are all set to 1 because they represent the comparison of a factor with itself.

Table 3

Pairwise comparison of parameters

View Large

Table 4

Normalized pairwise comparison matrix

View Large

Priority vectors and weights are calculated for each component in the AHP using pairwise comparisons. The is used to assess the consistency of these comparisons. These are the computed values: the eigenvalue of the judgment matrix is 8.78. equals 0.78. equals 7, the equals 0.11, the is 1.41, and the equals 0.079, all of which are less than 1%. A of less than 0.1 (10%) is typically considered acceptable, suggesting that pairwise comparisons are relatively consistent. This guarantees the weights and priority vectors are valid for use in the decision-making process for creating the flood risk map. Table 5 shows the weights allocated to each of the characteristics or variables based on the AHP's priority vector. These weights show the relative relevance of each aspect in the flood risk map decision-making process.

These zones give essential information on the risk of flood-related occurrences, allowing for informed decision-making and preparing for prospective flood catastrophes. The created flood risk map assists in understanding and prioritizing regions of varied flood risk within the research area. The flood risk map shows that a large portion of the study region falls into the ‘very high’ and ‘high’ risk categories. Notably, the ‘high’ danger zones are prominently located around the main water channels, with some reaching into the research area's northeastern and eastern sections. These high-risk zones make up a significant section of the overall landscape, accounting for roughly 25.31% of the total area of the research area. Local authorities should always be monitoring these flood-prone areas, especially during the rainy seasons, to proactively prevent tragedies associated with the floods. As per Table 6, specific localities in Mumbai link with different hazard areas. Such monitoring and preparation are vital in safeguarding the lives, livelihoods, and health of those living nearby, minimizing loss in case flood-related disasters might occur.

Based on the flood risk assessment for Mumbai, key recommendations for mitigating flood impacts include establishing early warning systems for timely alerts, integrating green infrastructure in urban planning, adopting sustainable building designs, upgrading drainage systems, conducting public awareness campaigns, fostering collaboration among stakeholders, incentivizing flood-resilient measures, utilizing remote sensing technologies, and investing in climate adaptation research. These measures aim to enhance preparedness, response, and recovery, safeguarding lives, and infrastructure from future flood events. The study's findings offer a roadmap for policymakers and urban planners to enhance Mumbai's flood resilience. By prioritizing targeted infrastructure improvements, promoting green infrastructure initiatives, integrating flood risk maps into urban planning, and implementing regulations on development in high-risk areas, policymakers can proactively mitigate flood hazards and safeguard the city and its residents from potential disasters. These recommendations, when translated into concrete policies and actions, can contribute significantly to Mumbai's resilience against the increasing threat of floods, ensuring sustainable urban development in the face of climate challenges.

The present study on the comprehensive flood risk assessment of Mumbai, India, integrating GIS and the AHP, while providing valuable insights, also has limitations that should be acknowledged. Firstly, it heavily relies on the availability and quality of input data, which may vary in accuracy and resolution, potentially affecting the precision of the flood risk assessment. Limited historical flood data for validation purposes also poses a challenge to the accuracy of the flood risk model. Additionally, the temporal scope of the study might be limited to the time frame of available data, potentially missing long-term trends or changing patterns in flood risks. The spatial resolution of input data, such as DEMs and satellite imagery, might not capture localized variations in flood risk adequately, especially in complex urban areas. Furthermore, the simplification of complex factors into discrete categories, such as elevation ranges or land cover types, might overlook nuances and gradients within these parameters.

Moreover, the subjectivity involved in the AHP during pairwise comparisons could introduce bias into the weightings assigned to the criteria. Ensuring consistency in the pairwise comparisons within the AHP framework is crucial, despite the reported acceptable CR of 0.079, which is acceptable but still leaves room for improvement. The study includes a specific set of criteria (elevation, NDVI, rainfall, DD, LULC, distance from roads, TWI, and distance from rivers), which might not cover all potential factors influencing flood risk. Uncertainties in the weight assignments based on the AHP and the lack of validation against observed flood events further add to the limitations. Additionally, the static nature of the model developed in the study, not accounting for dynamic changes in flood risk over time, might limit its applicability for future scenarios.

Future research could focus on enhancing data collection and analysis methodologies. This includes the integration of long-term data trends on rainfall patterns, land use changes, and river dynamics. Additionally, exploring the integration of real-time data sources, such as weather forecasts and river water levels, could provide more dynamic and accurate flood risk assessments. Secondly, given the impacts of climate change on flood susceptibility, conducting climate scenario modelling to understand future flood risks in Mumbai is crucial. This could lead to the development of adaptive strategies and infrastructure improvements to mitigate projected flood risks under different climate scenarios. Further research could also delve into localized vulnerability studies, engaging with local communities to assess specific vulnerabilities and resilience factors. Developing detailed flood risk maps at a micro-level for vulnerable neighbourhoods or informal settlements could provide valuable insights for targeted interventions. Advanced modelling techniques, such as hydrodynamic modelling, could be implemented to simulate flood inundation scenarios with greater accuracy. Coupling GIS with hydrological and hydraulic models could lead to more precise flood risk assessments. Expanding the scope to include assessments of other natural hazards, such as landslides, cyclones, or sea-level rise, could result in a comprehensive multi-hazard risk map. Understanding the interactions and synergies between different hazards could inform holistic risk management strategies. Policy implications are significant, and future research could provide evidence-based policy recommendations for urban planning, land use zoning, and infrastructure development to enhance Mumbai's resilience to floods. Cost–benefit analyses of proposed flood mitigation measures could guide decision-makers on investment priorities.

The comprehensive flood risk assessment of Mumbai, India is presented via the integration of GIS and AHP systems. The investigation comprehensively studied the flood risk throughout the study area considering many imperative factors. The FRI generated by the AHP provides valuable information on the weighted importance of every aspect towards a complete comprehension of flooding susceptibility. The flood risk map categorizes points into different risk levels including ‘very low’, ‘low’, ‘moderate’, ‘high’, and ‘very high’. This map helps people to determine areas of high risk for floods allowing people to be prepared and implement measures for mitigation. The conclusions of this study indicate that Mumbai should take a more aggressive approach in controlling flooding, especially around major waterways and other vulnerable areas. The ‘high’ and ‘very high’ risk zones occupying large portions of the study area require additional vigilance and measures during the rainy season. The integration between GIS and AHP has played an important role in giving a comprehensive assessment of flood hazards as well as setting a standard for today's flood risk appraisals. This is important research since future climatic changes will continue increasing the amount and severity of extreme weather events. This study can also serve as a model for flooding-prone locations across the globe.

The participants have consented to the submission of the paper to the journal.

Y.P.S. contributed to conceptualization, methodology, technical investigation, validation, visualization, data collection, original draft preparation, and reviewing the manuscript; V.K. contributed to conceptualization, methodology, visualization, reviewing, and supervising the manuscript; K.V.S. contributed to data collection and reviewing the manuscript; A.D. contributed to conceptualization, validation, and data collection; D.K.T. contributed to conceptualization, validation, and data collection. All authors read and approved the final manuscript.

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

The authors have no relevant financial or non-financial interests to disclose.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.