Assessment of Flood Susceptibility Mapping in the Krishna River Basin using Geospatial Techniques

Floods are a natural part of the water cycle (Vojtek and Vojtekova, 2016) mainly caused by snowmelt, sudden cloudbursts, and heavy rainfall in semi-arid regions or wind-driven waves in coastal and estuarine areas (Tehrany et al., 2013; Merkuryeva et al., 2015) significantly impacting human life and infrastructure (Das, 2018; Debnath et al., 2023; Winsemiu et al., 2016). The flood hazards are heightened by climatic changes, land use modifications, and human activities such as urban sprawl, stream diversion, deforestation, and excessive groundwater extraction (Kourgialas and Karatzas, 2011; Khodaei et al., 2025; Abijith et al., 2025; Kvocka et al., 2016). However, modern tools can be used to effectively monitor river overflow floods (Jonkman, 2005; Das, 2018).

Floods are a complex event that has been widely studied by researchers to better understand their causes and develop effective methods for control and prevention (Miguez, and Magalhaes, 2010; Wang, et al., 2022). In addition to natural as well as man-made factors continuous rainfall is also a major factor in flooding (Kim and Kim, 2014). Furthermore, human changes in land use in upstream areas of flood zones increase the risk and impact of flooding (Gigović et al., 2017). Recently, extreme flooding events have occured due to the climate change phenomenon (Rojas et al., 2012; Halgamuge and Nirmalathas, 2017). Fluctuations in climate patterns, such as increased precipitation, rising sea levels, and altered weather systems, have the potential to intensify flooding, thereby increasing both the regularity and rigorousness of such events. Apart from these reasons, human factors are also involved in the creation of flood situation such as, poor management, improper planning, mismanaged reservoir discharges, limited drainage capacity, monsoons, and inadequate flood mitigation measures. According to Saha and Agrawal (2020), applying specific, effective mitigation strategies can help reduce the intensity of a disaster, even though it is not possible to fully eliminate the hazards associated with flood events. Therefore, understanding the impact of climate change is crucial for accurately evaluating flood threats and formulating effective mitigation measures (Wang, et al., 2022; Das, 2018). Floods are a significant threat to the economy and the livelihoods of individuals, underscoring the need for the proper flood risk management to reduce the damage caused by these events.

According to Lim and Lee (2017), Flood conditions are influenced by geographic features such as the slope inclination of the land, the river course, the geology of riverbed, distance from the main channel, terrain curvature as well as the elevation and geological composition of the area. The rapid urbanization and industrialization along riverbanks, driven by population growth, disrupt the natural river flow and lead to the narrowing of river channels. In recent years, rivers worldwide have been experiencing more frequent and severe flooding events. Flooding frequently occurs across Asian countries, including China and Nepal, affecting the major rivers such as the Yangtze and Yellow in China, the Kosi and Pokhara region in Nepal (2008, 2012), along with several Indian rivers like the Ganga (2005) and Krishna (Dimri et al., 2016). In Maharashtra, where the Krishna River basin is originated, major rivers respond rapidly to intense monsoon rainfall, often resulting in severe flooding in low-lying and densely populated areas of Krishna (2019, 2021), Mutha (2024), and Godavari (2006, 2008, 2024). These flood events lead to the degradation of both natural and human resources.

Recent research highlights the significance of flood hazard mapping and demarcation of flood-susceptibility using deep learning and machine learning approach (Efraimidou and Spiliotis, 2024; Al-Ruzouq et al., 2024; Razavi-Termeh et al., 2025; Arora et al., 2025; Pourzangbar et al., 2025; Tepetidis et al., 2025; Abijith et al., 2025; Wale et al., 2025). Ekmekcioglu et al., (2022) emphasized the role of geospatial technologies in flood assessment. Similarly, Gupta and Dixit (2022) applied GIS-based models to evaluate flood risks, enhancing flood management strategies. Malekian and Azarnivand (2016) also used GIS and RS to assess the flood susceptible areas, demonstrating the effectiveness of spatial data in flood risk evaluation. Many studies on flood hazards have been done at different levels such as Tique et al., (2025) utilized remote sensing and machine learning to develop 30 m flood susceptibility maps for Nigeria, revealing that approximately 11 million people reside in high-risk flood zones. The Copernicus DEM with D8/FD8 methods gave the most accurate results of Nigeria.  Dou et al. (2017) studied regional flood risks, while Muis et al. (2015) examined floods globally. Su et al. (2011) focused on China, studying floods and snow disasters across the country and how they affect people and places. Das et al. (2017, 2018) emphasized that RS and GIS are crucial tools for disaster risk management with spatial mapping using multi-criteria decision models. Diyali Chattaraj et al. (2021) study flood and waterlogging risks at a local level. They use an AHP and GIS model to improve hazard assessment and management. Many studies have been conducted around the world, including India, to understand flood causes and damage (Tehrany et al., 2013; Suykens et al., 2016; Risi et al., 2017; Marvi, 2020; Atif et al., 2021; Yu et al., 2022). Furthermore, numerous researchers have integrated GIS technique with AHP and fuzzy AHP methods to generate flood risk maps (Hammami et al., 2019; Meshram et al., 2019; Ekmekcioglu et al., 2021; Zhran et al., 2024). Islam (2024) has attempted to evaluate the flood susceptibility using bivariate statistical method. However, most of these studies focus on improving specific flood control methods for particular assets (Zhang and Alipour, 2021).

Flood risk assessment using geospatial techniques involves analyzing spatial data to identify flood-prone areas and assessing the potential impacts of flooding. The Krishna River, the third-longest river in India, originates in the Western Ghats near Mahabaleshwar, Maharashtra, at an elevation of 1300 meters. It flows eastward through Wai and eventually drains into the Bay of Bengal. Flooding in the Krishna River is exacerbated by inadequate river management, with Maharashtra and Karnataka filling reservoir to capacity and subsequently releasing large volume of water downstream leads to flooding in the lower reaches and backwater areas. Geospatial techniques enable detailed mapping and analysis of terrain hydrology, land use to support effective flood management and mitigation strategies. The study assesses the factors causing floods and mapped flood-prone regions in the Krishna River Basin (KRB) applying the geospatial methods. This will help equip the residents of the region with the ability to forecast potential flood risks and promptly execute localized interventions to manage flood situations as they arise.

density, Flow accumulation, Rainfall, LULC, Geology, Soil type, SPI, TWI, and Curvature. This technique offers a rigorous, efficient tactic for detecting flood-prone areas, thereby supporting effective water resource management and mitigation policies within the basin. The analytical procedure comprised six sequential steps: criteria selection, ranking of criteria, pairwise comparison, weight calculation, score assignment, and execution of a weighted overlay analysis. These methodological steps were pivotal in delineating areas susceptible to flooding, thereby enhancing the understanding of flood risk dynamics and informing strategic planning in the KRB.

3.1 Analytical Hierarchy Process (AHP) model

The major factors influencing flood susceptibility include elevation, slope gradient, river proximity, Rainfall, flow accumulation, LULC, geological structure, TWI, and surface curvature. To demarcate the flood prone areas in KRB multiple parameters have been taken into account on the basis of their relative significance in the decision-making method. The structured categorization, optimization of efficiency, and generation of precise outcomes are fundamental attributes of the AHP (Das, 2019; Gaikwad, 2019). The AHP is extensively applied for managing water resources. The analysis rigorously incorporated a comprehensive suite of thirteen physiographic parameters. Each criterion has gave a rank based on its significance, aiding in the organization and prioritization of spatial decisions (Saaty, 1977; Das, 2019; Navale and Bhagat, 2021). The most important criteria were assigned higher weights whereas the least important were given lower weights. The Pairwise Comparison Matrix (PWCM) in AHP assesses factors influencing flood susceptibility using Saaty’s scale (1977), with values ranging from 9 to 1/9 to indicate relative significance. These comparisons help calculate priority weights for decision-making in flood prone mapping. In the present work, criteria were ranked from 1 to 9, with expert input determining the significance of each factor. AHP is an innovative and flexible method in statistics, widely used in research for analyzing multiple factors and predicting outcomes accurately to solve complex problems. The AHP method has been applied to assign weights and ranks to different parameters in the flood prone index model, as shown below:

​\(FH = \sum_{i=1}^{n} W_i * R_i\)​  (1)

Where, FH is the flood hazard index, Wi is the weight of each factor, and Ri is the rank of each factor’s value. The final map in this study is categorized into five classes: very low, low, moderate, high, and very high, using the FH model to assess the probability of flood occurrence (Das, 2018).

AHP calculates parameter weights using the Pairwise Comparison Matrix (PCM) based on the relative importance and influence of each factor on the basis of extensive literature review and the expert opinion. These weights establish the importance of each parameter in a hierarchical structure. After calculating the weights in PCM, the matrix is normalized by dividing respectively judgment value by the total of its column. The quotient values for each row are summed and then divided by the number of parameters in the PCM (Table 1). The resultant value for each parameter represents its weight, extending from 0 to 1, with the sum of all weights constantly equal to 1.

A PCM is established, with diagonal elements set to 1 (Eq. 2). The relative prominence of the criteria has determined based on predefined criteria, assigning ranks as follows: 1 for equally important, 3 for moderately more important, 5 for strongly more important and 7 for very strongly more important. Intermediate values, such as 2, 4, and 6, are used to represent varying degrees of importance (Dejen and Soni 2021). Each row shows how important one parameter is compared to others. For example, the first row compares elevation to eight other parameters. In AHP, parameters deemed more important receive higher numerical weights, while less important ones are assigned reciprocal values. For example, if basin elevation is more significant than slope, it may be assigned 5, with slope receiving 1/5. This reciprocal system ensures consistent and proportional weighting across all parameters for flood-susceptibility analysis (Saaty, 1977; Das, 2018).

The quantitative and qualitative approach is a major strength of AHP (Forman, 1993; Doke et al., 2021).

​\(A = (a_{ij})_{n \times n} = \begin{bmatrix} a_{11} & a_{12} & \dots & a_{1n} \\ a_{21} & a_{22} & \dots & a_{2n} \\ a_{n1} & a_{n2} & \dots & a_{nn} \end{bmatrix}, \quad a_{ii} = 1, \quad a_{ij} = \frac{1}{a_{ji}}, \; a_{ij} \ne 0\)​  (2)

Where,

A: The matrix A represents the collection of all pairwise comparisons between the parameters n × n, indicating a square matrix with n rows and n columns.

​\(aij = \)​  Element at the ​\(i\)​ -th row and ​\(j\)​ -th column

​\(a_{ii}\)​ =1: Diagonal elements of the matrix are always 1, as each parameter is equally important to itself

​\(a_{ij} = \frac{1}{a_{ij}}\)​ Elements are reciprocals of each other, ensuring consistency in the comparison (if parameter iii is more important than ​\(j\)​  , then ​\(j\)​  is less important than ​\(i\)​ ). ​\(a_{ij} \neq 0\)​ : No element in the matrix is zero, as comparisons are always non-zero values.

Afterward, the weighted arithmetic mean methodology is employed to compute the weights in the PCM. In this process, the values in the matrix are standardized to obtain the standardized values in the standard pairwise comparison matrix (Table 2). This normalization step confirms that all values are scaled in a consistent method, facilitating the calculation of relative importance or weights for each parameter. Using Equation (3), the pairwise comparison matrix (Table 1) was normalized by dividing each parameter by its respective column sum, ensuring all column totals equal to 1. The average of each row produced the relative weight of each criterion (Table 2). (Equation 3 and 4).

​\(a_{ij} = \frac{a_{ij}}{\sum_{i=1}^{n} a_{ij}}\)​  (3)

​\(w_i = \frac{1}{n} \sum_{j=1}^{n} a'_{ij}\)​  (4)

​\(A_w=λ max⁡W\)​  (5)

The equation ​\(A_w=λ max⁡W\)​ is used to find the importance of criteria in decision-making. w is the weight factor and A is a comparison matrix, and. By averaging rows and solving the equation, we determine the relative weights of each criterion (Equation 5).                                                                                                              

3.1.1 Consistency ratio (CR)

Consistency in judgments is assessed using the CR calculated as (Biswas et al., 2021),

CR ​\(=\frac{CI}{RI}\)​  (6)

Where the Consistency Index,

CI ​\(=\frac{\lambda_{\text{max}} - n}{n - 1}\)​  (7)

λ max ​as the largest eigenvalue and n as the matrix order (Pawluszek and Borkowski, 2017).

Equation (7) was then applied to assess judgment consistency. The maximum eigenvalue ( ​\(λ_{max}\)​ =13.78)) was derived from the matrix–weight product, yielding (CI=0.065) and (CR=0.041). As CR < 0.10, the pairwise comparisons are considered consistent and reliable for flood susceptibility analysis in the KRB. Eigenvalues, particularly ​\(λ_{max}\)​ , are derived from the sum of total and normalized weights, with Saaty’s RI providing consistency benchmarks, such as 1.56 for 13 criteria. Table 3 represents the AHP method has been applied to a total of 13 parameters and all their subtypes considered for the research, and those parameters have been given a relative ranking according to their importance.

3.2 Preparations of Thematic Layers

To delineate the flood hazard zones in the KRB, high-resolution satellite imagery and ancillary data were procured from a range of digital repositories and governmental institutions. The detailed methodological framework is illustrated in Figure 2.

3.2.1 Elevation

According Botzen et al (2012) elevation is key factor in mapping flood risk vulnerability. High elevated land areas have less risk of flooding, while lowland plains are more vulnerable. The areas at higher elevations exhibits lower flood susceptibility (Tehrany et al., 2013; Das, 2018). The relief in Krishna basin ranges from 0 to 1865m. About 10% of the basin lies around 250m elevation, while nearly 20% of the area falls between 250 and 450m (Figure 3a).

3.2.2 Slope

Mapping flood-susceptible areas, slope is a crucial topographic factor because it controls surface water flow. In the present study area, the slope values range from 0° to 67°. Surface slope acts as an amplifying factor by increasing water-flow velocity and reducing infiltration, which collectively heightens flood risk (Elkhrachy, 2015). As the slope gradient increases, runoff becomes faster and infiltration decreases; however, where the slope suddenly flattens, the reduced flow velocity causes water to accumulate, leading to stagnation and potential flooding (Das, 2019; Doke et al., 2021).

3.2.3 Distance from the river

Proximity to river channels is considered a significant geomorphic factor while preparing accurate scientific maps of flood prone area. The areas closer to the river generally have lower elevation and gentler slopes. Additionally, a river channel represents the lowest topographic point within a given area. Consequently, regions located farther from the river exhibit reduced susceptibility to flooding. In the present study the buffer tool of ArcGIS software is used to calculate the distance from the river (Figure 3c). Multiple buffers were generated at intervals of 0.5km, which were then merged into a single vector layer and subsequently converted into a raster format for further analysis. (Fernandez and Lutz, 2010; Das, 2018). Flooding raises groundwater levels, especially in flat areas, causing prolonged waterlogging and slower drainage. However, the areas near rivers, ponds, and lakes are prone to flooding due to elevation changes, with flash floods also occurring based on meteorological and topographical factors (Pham et al., 2020).