Flood detection in the Upper Krishna Basin through integrated geospatial analysis: leveraging decision frameworks and statistical measures

Natural disasters, including volcanic eruptions, earthquakes, landslides, and floods, results from both natural hazards and human activities. Flooding is a major global disaster that causes extensive environmental and economic damage, resulting in fatalities and casualties, and is exacerbated by climate change and human activities, affecting approximately 170 million people each year. The Asian region experiences a higher frequency of these disasters, causing significant losses in human life, infrastructure, stability, and economic progress. Floods are the most prevalent and destructive natural disaster, causing hundreds of thousands of fatalities, widespread infrastructure destruction, and estimated annual economic damages in the past three decades. Human activities upstream, such as deforestation and land use changes, are increasing the size and frequency of floods while increasing populations downstream worsen the damage (Al-Omari et al. 2024). Research suggests that unfavorable weather patterns and land use changes will exacerbate these losses in the coming years (Vashist & Singh 2024). Future floods are expected to be more frequent and intense, with low population resilience, particularly in less developed countries. They are complex phenomena that attract researchers worldwide to understand and explore prevention and management mechanisms (Tadesse et al. 2024).

India is the most disaster-prone country in South Asia, primarily due to its geographical location, climatic conditions, and geological features. It faces risks from landslides, avalanches, earthquakes, cyclones, droughts, floods, and fires. India ranks second in the world for the greatest flood damage, behind Bangladesh. According to the National Flood Commission, an average of 18.6 million hectares of damage is caused by floods each year on 40 million hectares of sensitive land. The Indo-Gangetic-Brahmaputra lowlands include the Brahmaputra and Ganga River basins, which are the most vulnerable to flooding. Other vulnerable areas include the northwest with west-flowing rivers like Narmada and Tapti and the Deccan region with significant east-flowing rivers like Mahanadi, Krishna, and Cauvery (Pakhale & Nale 2023). The monsoon season in India, which lasts from June to September, accounts for three-quarters of the country's total precipitation. This leads to high river flow, causing severe flooding in major regions like West Bengal, Uttar Pradesh, Bihar, and Assam. Himalayan river silt contributes to bank erosion and overtopping, causing drainage issues. Long-lasting floods and clogged drainage arteries also coexist. Deforestation, which loosens topsoil during rainy seasons, contributes to floods by causing soil to rush into rivers instead of absorbing rain (AlAli et al. 2023).

Flooding, a catastrophic hydro-meteorological disaster, results in significant losses annually due to infrastructural damage, environmental degradation, and loss of livelihood. Floods are occurring more frequently, and the reasons for this are unusual rainfall, climate change, and fast urbanization (Lazzarin et al. 2023). Infrastructural damage impacts service delivery and water supplies across industries. Despite the enactment of flood protection laws in various countries, including India, many of these measures remain inadequately implemented. Therefore, it is crucial to predict flooding risks and implement appropriate measures to mitigate their effects. Researchers are using machine learning and multi-criteria decision-making to study flood vulnerability globally, where flood vulnerability refers to the susceptibility of an area to flooding due to factors like geography, infrastructure, and climate conditions. (Hitouri et al. 2024). Despite challenges in flood management due to timing, location, and geophysical interactions, advancements in technology and remote sensing (RS) significantly reduce the impact of natural disasters and prevent many fatalities (Van Westen 2000). RS gathers information about objects or events without physical contact, which is then used in Geographic Information System (GIS) tools for various analyses (Chaulagain et al. 2023). The hydrological aspect of flood estimation reduces death toll and financial damage, thanks to advancements in satellite observations and meteorological and hydrological modeling (Gupta 2020).

The Upper Krishna Basin (UKB), with its distinct hydrological and geological features, is an important area for flood risk assessment. Because of its intricate drainage network and high water flow from its rivers, this region is prone to flooding. The Krishna River, together with its major tributaries – Bhima, Tungabhadra, and Ghataprabha – contributes to the basin's significant flood risk. The combination of monsoon rains and geographical changes heightens the flood danger, affecting residents and infrastructure (Bhatt & Ahmed 2014). The annual monsoon season, from June to September, brings heavy rainfall, causing the rivers in the basin to swell and overflow. This results in widespread flooding, affecting agricultural lands, settlements, and infrastructure. The silt carried by these rivers contributes to bank erosion and drainage issues, further complicating flood management efforts (Pakhale et al. 2023). Additionally, deforestation in the region has loosened topsoil, which rushes into rivers during rainy seasons, contributing to flooding. Floods are frequent natural disasters that put towns, infrastructure, and human lives in danger. It is anticipated that as more people migrate to urban areas in developing nations, urban floods will occur. Flood relief shelters play a vital role in times of extreme weather, but they are susceptible to flooding, which can negatively affect the mental health of displaced people. Research highlights the necessity for psychological support services in relief camps to attend to the emotional needs of participants. Finding appropriate locations for flood shelters necessitates a multifaceted strategy that integrates statistical research and geographic methodologies for effective flood risk management (FSM) (Aju et al. 2024).

The development of GIS tools and RS data sources has enhanced the capacity to create and apply prediction models for regions vulnerable to natural disasters, allowing for the integration of many data sources for superior flood simulations (Ashtekar et al. 2019). Accordingly, the primary goal of this study is to use GIS-based AHP modeling to assess the size of the UKB's flood-prone areas and to pinpoint the most important places for the development of risk-reduction or mitigation programs. Considering hazards, vulnerability, and exposure, the suggested system coincides with past studies, proving its dependability. It is a valuable tool for better flood control and decision-making in urbanizing areas, with applicability to other places with adaptation (Guoyi et al. 2023). A hierarchical structure of criteria and sub-criteria has been developed to represent the complex interactions that influence flood risk. Expert opinion and stakeholder opinions are sought to assign weights to these criteria, reflecting their relative importance. Using the AHP framework, the study uses pairwise comparisons to quantify the relationship between criteria and sub-criteria (Yu et al. 2023).

By utilizing the frequency ratio (FR) approach to identify flood vulnerability zones in the Kulik River Basin and classifying them into five groups according to size and accuracy, the study helped reduce the risk of flooding (Sarkar & Mondal 2020). Using the weight of evidence, Shannon's entropy, and FR, the study assessed the Kopai River Basin's flood susceptibility. Results indicated that there is a need for more study and mitigation strategies because there were non-flooded upper-reach zones and significant flooding in the lower area (Sarkar et al. 2022). To identify locations with extremely high flood risk, the study focuses on using RS and GIS for mapping flood risk in the Patna district using the FR and entropy model (Saha et al. 2022). To construct highly predictive maps useful to land use planners and decision-makers, the research employs 20 indicators to identify flood-vulnerable zones in the Western Ghats-Arab Sea region using AHP and F-AHP models. Using FR classifiers and the analytical hierarchy method, the study evaluated the flood vulnerability of the Dakshin Dinajpur District. The FR model proved to be the most successful in identifying zones with different levels of sensitivity (Sarkar et al. 2023).

This study aims to address a critical research gap in flood hazard zone mapping by integrating Analytical Hierarchy Process (AHP), FR, and Statistical Index (SI) methodologies with GIS and RS techniques. This research aims to enhance the accuracy of flood hazard zone mapping in the UKB by developing an integrated framework that combines GIS, AHP, FR, and SI models. This approach not only improves the precision of flood risk assessments but also provides actionable insights for effective flood mitigation and management strategies.

The research uses GIS and RS techniques to identify and map flood-prone areas, integrating them with AHP, FR, and SI methodologies to produce a comprehensive flood hazard zonation map. RS data are used to capture real-time information about the region's hydrological and geographical conditions. The framework assigns appropriate weights to various criteria influencing flood risks, such as rainfall patterns, soil type, land use, and topography. The study provides a scalable and adaptable framework for regions facing similar hydrogeological challenges, offering valuable tools for water resource management and sustainable development. It significantly contributes to flood management by presenting a robust, multi-criteria decision-making framework that can be applied to various regions, ultimately aiding in the reduction of flood-related risks and promoting resilience against natural disasters.

The Krishna River is the second largest river basin in Peninsular India, originating at an elevation of 1,338 m near Mahabaleshwar in Maharashtra. Krishna River travels around 1,400 km from West to East through Maharashtra, Karnataka, and Andhra Pradesh before entering the Bay of Bengal. It shares borders with Andhra Pradesh, Kerala, Maharashtra, and Goa, as well as an Arabian Sea coastline. Krishna Basin covers 8% of the country's total area, with a geographic expanse of 258,948 square kilometers. (Ashtekar et al. 2019). The command area for the Upper Krishna Project is located in the Northern districts. It goes through hilly terrain with considerable rainfall in the upper parts, while it unites over a plain area with diminishing rainfall at the bottom levels. The main tributaries of the Krishna River are Ghataprabha, Panchganga, Dudhganga, Warna, and Koyna. The basin receives an average of 600–6,208 mm of rainfall per year, for a total of 1,347 mm. As of the 2011 census, the UKB has around 8,170,973 inhabitants. The Krishna River and its tributaries frequently experience flooding during the monsoon season.

The Upper Krishna initiative, led by the Government of Karnataka, covers 6,22,120 hectares in Karnataka's rainy shadow region and 17.13 lakh hectares in Maharashtra, primarily in Satara, Sangli, and Kolhapur districts. The region is shaped by the Krishna River and its tributaries, characterized by a geological composition predominantly of the Deccan Trap. Around 44.82% of the area has undergone land development treatment. The agro-climatic zones influence land use patterns, with 28% facing drought and 44% receiving rainfall. The region supports diverse crops like sugarcane, oilseeds, soybeans, rice, and vegetables. Current data shows 2.33 lakh hectares of forest, 1.04 lakh hectares of net sown land, 14.22 lakh hectares of cultivable land, and 13.03 lakh hectares of gross cropped land.

Creating supplementary thematic maps that capture elements relevant to flood hazard assessment involves a series of processing stages conducted after data collection. These might contain slope maps, drainage density maps, maps showing the distance from roads and rivers, normalized difference vegetation index (NDVI) maps, and topographic wetness index (TWI) maps. Every theme map provides insightful information about variables including surface water flow, proximity to bodies of water, steepness of the topography, vegetation cover, and soil moisture that affect a region's susceptibility to flooding.

The data phase involves data scanning, cleaning, digitization, and verification to assure accuracy, consistency, and uniformity. After the data phase, the data were pre-processed to ensure accuracy, consistency, and uniformity. This step includes data scanning, cleaning, digitalization, and verification, all by international organizational standards. The map is produced, with 70% of the data separated into a training dataset and 30% for testing (Elsadek et al. 2024). Thematic maps are integrated into a flood hazard model. This model combines various spatial datasets to create a comprehensive flood hazard map showing areas at varying levels of danger. The model is subsequently validated to establish its accuracy and reliability by splitting the data into training and testing datasets, training the model on the former, and assessing its performance on the latter (Samanta et al. 2018).

The selected 10 GIS parameters, such as rainfall, slope, elevation, land use/land cover (LULC), soil, distance from river, distance from road, NDVI, TWI, and drainage density, were chosen based on their significant impact on flood hazard assessment as demonstrated in previous studies. Rainfall is the primary factor contributing to flood hazards, with higher rainfall generally increasing the risk of flooding (Vashist & Singh 2024). Slope influences the rate of water flow, where steeper slopes result in faster runoff, thus increasing flood potential (Senan et al. 2023). Elevation affects both rainfall distribution and surface runoff, influencing flood susceptibility (Tadesse et al. 2024). Areas with high drainage density typically have higher flood potential due to increased surface runoff (Al-Omari et al. 2024). LULC affects water runoff and infiltration, as different land cover types, such as urban areas and forests, have varying capacities to absorb water (Cao et al. 2016). Soil type is crucial for flood hazard assessment, determining the infiltration rate and water-holding capacity (Bhatt & Ahmed 2014). Distance from rivers and roads is also critical, as proximity to these features can increase flood risk due to higher runoff and the potential for obstruction in water flow paths (Ashtekar et al. 2019). TWI is one of the factors that might indicate the presence of base flow, intermediate flow, and surface runoff accumulation. Flood-prone areas use the TWI value as a guidance in several higher runoff indicators (Seejata et al. 2018). Higher NDVI values indicate healthier vegetation cover, which plays a crucial role in flood regulation Low NDVI areas may experience more severe floods due to reduced natural regulation (Sarker & Adnan 2023). By integrating these parameters, a comprehensive flood hazard map for the UKB can be developed, providing valuable insights for effective flood management strategies.

This study employs three distinct models: AHP, FR, and SI, each contributing uniquely to flood hazard mapping. AHP is used to assign weights to various flood-influencing parameters based on their relative importance through pairwise comparisons and consistency checks, prioritizing factors like rainfall, slope, and land use/ and cover (Sulaiman & Mustafa 2023). FR analyzes the relationship between historical flood occurrences and the spatial distribution of each parameter, calculating the probability of flooding based on past events to provide a statistical measure of flood susceptibility (Natarajan et al. 2021). The SI evaluates the influence of each parameter by comparing the spatial occurrence of floods to the total study area, quantifying each factor's contribution to the flood hazard (Abdo 2022). Integrating the results from the AHP, FR, and SI models ensures a robust and multifaceted approach to flood hazard assessment in the UKB , enhancing the accuracy and reliability of the flood hazard map.

The receiver operating characteristic (ROC) curve is used to evaluate the model's ability to differentiate between flooded and non-flooded areas, with the area under the curve (AUC) serving as an important performance parameter. The zone is split into five groups after the data are analyzed: very high, high, moderate, low, and very low. This precisely created flood hazard map is an essential tool for flood risk assessment and mitigation efforts, providing important insights to aid decision-making and supporting initiative-taking measures to minimize vulnerability and boost resistance to flooding events.

DEM was obtained from the SRTM provides high-resolution topographic data. The rainfall data were obtained from the Indian Meteorological Department (IMD) from 1980 through 2015. The soil data were obtained from the National Bureau of Soil Survey and Land Use Planning (NBSSLUP) dataset. The LULC data were obtained from the National Remote Sensing Centre (NRSC) dataset. NDVI data were obtained from the National Oceanic and Atmospheric Administration (NOAA) dataset.

The preprocessing step comes after the data phase checks the accuracy, consistency, and uniformity of the input features data. This phase includes data scanning, cleaning, digitization, and verification. The map created for the input parameters data has been transformed into standard format. The data are divided into a training dataset (70%) and a testing dataset (30%). The datasets were cleaned by removing missing values, outliers, and inconsistent data. The datasets were transformed into a suitable format for analysis and joined based on spatial coordinates to combine information from different sources. Interpolation, spatial autocorrelation, and spatial regression have been applied to the datasets to extract relevant information.

ArcGIS was used for spatial analysis, data manipulation, and mapping. Python programming language was used for scripting, automating tasks, and statistical analysis. ERDAS Imagine and ENVI software were used for spatial analysis and data processing. The slope of the terrain has been determined using the DEM data by the ‘Slope’ tool in ArcGIS and reclassified for analysis. Drainage density has been obtained from hydrological maps and importing the drainage network data by measuring the length of the drainage network per unit area (e.g., km/km²). The drainage density has been classified into different classes based on their magnitude. Each parameter's mean, median, and standard deviation is used to understand the distribution in the study area.

Incorporating thematic maps into the flood hazard model necessitates a methodical approach. Thematic maps have been weighted based on their impact on flood hazard potential, including slope steepness, land cover characteristics, soil permeability, and drainage network density. Integration begins with GIS overlay techniques combined with intersection and union. AHP facilitated the integration process by systematically combining weighted criteria and Weighted Overlay Analysis. To assess the robustness of the model, sensitivity analyses are performed by comparing integrated flood hazard maps with historical flood data.

The FR model uses a simple geographic evaluation tool to understand the relationship between flood occurrences and factors like rainfall and slope. FR = (Number of pixels with flooding hazards in each parameter)/(Total number of pixels with flooding hazards in the study area). The SI model was implemented using R software. The SI model calculates weights by taking the natural logarithm of the ratio of flood occurrences in each class of a conditioning factor to the overall flood density in the study area. SI = ln((Number of pixels with flooding hazards in each class of each parameter)/(Total flood density in the study area)). The models were validated using ROC curves, which measure the accuracy of the models. The AUC values range from 0 to 1, with 1 denoting excellent performance. The results of the FR and SI models are presented in Table 2, which shows the frequency ratio and SI values for each theme, class, and parameter.

The AHP method assigns weights to various aspects to illustrate their interrelationships in a schematic diagram. A factor's relative importance is determined by its influence on other factors, resulting in a higher weight for that factor. The AHP grading scale, ranging from 1 to 9, specifies the intensity levels of relevance to assess the relative value of different variables. (Hirwa et al. 2023).On this scale, a value of 1 represents ‘equal importance,’ indicating that both components contribute equally to the goal being evaluated. A score of 3 signifies ‘moderate importance,’ where judgment and experience slightly prefer one component over another. Level 5 denotes ‘strong importance,’ showing that judgment and experience significantly favor one component over the other. A score of 7 indicates ‘very strong importance,’ meaning one aspect is preferred over another, with this dominance being evident in practice. The highest value, 9, represents ‘extreme importance,’ where evidence strongly supports one aspect over another, achieving the highest level of affirmation. Intermediate values of 2, 4, 6, and 8 are used to express more nuanced evaluations between the established levels. This scale is utilized to organize criteria hierarchically through a pairwise comparison matrix, ranking and prioritizing the themes accordingly (Mekonnen et al. 2023).

The research focuses on identifying hazard zones for flooding using ten parametric layers. It evaluates their interactions and assigns weights for the AHP procedure. Eight characteristics were chosen and ranked using a 5-point rating system. The methodology allows for a more detailed and context-specific ranking of each attribute, making mapping flood hazards easier. The study's approach ensures accurate and comprehensive mapping of flood hazards in the region. The study used a weighting distribution method to assign a total weight of 100–10 parameters, ensuring each element had equal importance in assessing the flood hazard zone. This resulted in a weighted hierarchy, emphasizing the relevance of each element in the overall flood danger zone evaluation while also offering a systematic framework for analysis. The method effectively highlighted the significance of each parameter in the flood hazard zone evaluation.

A comparison matrix of theme layers is generated in pairings. The AHP has been utilized to equalize the weights assigned to each layer of themes and each type of theme layer (Figures 3 and 4). The predicted CR values for each layer and the corresponding classes showed that the pairwise matrix has been consistent; all the values were within acceptable bounds. The heatmap's color scale, seen on the right, showed that darker colors, closer to the bottom of the scale, represented larger values, while brighter colors, closer to the top of the scale, reflected lower values. From light (low values, close to 0) to dark (high values, close to 1), the color scale moves

The accuracy of the FR and SI models has been confirmed by calculating success and prediction rates using ROC curves (Berhane & Tadesse 2021). The AUC measurements demonstrate the forecast's accuracy. These metrics were computed using the training (70%) and testing (30%) datasets. AUC values range from 0 to 1, with 1 denoting excellent performance.

Following a detailed description of the methodology, we provide the findings of our flood danger mapping in the UKB. This section begins with a description of the input parameters, which include rainfall, slope, elevation, LULC, soil, distance from rivers and highways, NDVI, drainage density, and TWI. The following sections will give a thorough evaluation of the flood hazard zones identified by each model.

Rainfall is a major factor in floods and river overflow. The spatial distribution and intensity of rainfall were analyzed to identify areas with high precipitation that could contribute to flooding. Historical rainfall data and seasonal variations were considered to understand patterns and extreme events. Flood runoff increases when rainfall increases, and vice versa. The Western Ghats receive the most rainfall, while the Eastern Ghats receive the least. The 23 tehsils that comprise the basin are Bellari, Chitra Durga, Gadag, Guntur, Koppal, Mandya, Mysuru, Raichur, Shivamogga, Tumakuru, and Yadgir. These tehsils collectively account for about 25% of the basin's water resources, as shown in Figure 5(b). The three tehsils of Bhima, Tungabhadra, and Kaveri are the sources of the main rivers. The Southwest monsoon peaks in June and September, but the basin experiences year-round variations in rainfall. Droughts and flash floods are due to their geography. The basin receives most of its rainfall from the SouthWest monsoon. Table 4 shows the area experiences 582–3,935 mm of annual precipitation on average. Transition zone, arid zone, and Western Ghats region. Among these, the Western Ghats zone receives the highest annual precipitation of 3,265–3,935 mm, which covers an area of about 1,215 sq km, whereas the transition zone receives moderate rainfall. 1,924–2,594 covers an area of 4,744.21 sq km, and the dry zone receives less rainfall. 582–1,253 mm covers most of the area of 28,515.49 sq km, as shown in Table 3.

Surface runoff velocity and water infiltration rate are directly governed by slope. Slope gradient was considered to understand the runoff potential, with steeper slopes likely to experience faster water flow, influencing flood dynamics. The largest sub-basin in the Krishna Basin, known for its undulating plains, hill ranges, and flat peaks, comprises 21.4% of the entire Krishna Basin, as shown in Figure 5(c). The western portion of the study area is in the Sahyadri range, also known as the Western Ghats. Table 4 shows that the study area has a very steep slope of 71.99–89.9%, covering 0.74% of the study area, and a steep slope of 53.99–71.99%, covering 2.64% of the study area. In contrast, there is a moderate slope of about 35.99–53.99% covering 6.53%, a gentle slope of 17.99–35.99% covering 29.21% of the study area, and a flat region with a slope of 0–17.99% covering an area of approximately 60.86%.

The UKB, situated within the Deccan Plateau and Western Ghats, exhibits a diverse topography featuring rolling plains and flat-topped hills. Elevation data were used to determine the topographic variations within the basin, affecting water accumulation and flow paths. Low-lying areas are generally more susceptible to water accumulation and flooding. The basin has a tropical climate, with an annual rainfall average of 859 millimeters. From Figure 5(d) and from Table 4, an elevation range of 500–750 meters encompass the majority, constituting approximately 55% of the basin's total area. This elevation becomes a crucial parameter for flood susceptibility mapping, as lower-lying areas are more prone to flooding disasters. The elevation map categorizes the region into five classes: very low elevation areas (36–389 m) covering 0.08% of the total area; low elevation areas (389–594 m) dominating Bijapur and Sangli districts with a substantial 43.45% coverage; moderate elevation areas (594–708 m) constituting around 35.64% of the study area; high elevation areas (708–887 m) occupying approximately 17.49% of the study area; and very high elevation areas (887–1,467 m) found in the Western Ghats, representing about 3.13% of the study area. This classification provides a comprehensive understanding of the basin's terrain, emphasizing the correlation between elevation levels and flood susceptibility.

The LULC classification system is a means of categorizing both natural and human-caused characteristics on Earth's surface. Different land cover types (e.g., urban areas, forests, agricultural lands) were mapped to assess their impact on surface runoff and water infiltration. Urbanization tends to increase runoff due to impervious surfaces. This system helps in planning, managing, monitoring, and preserving natural resources and ecosystems in the long term. LULC are categorized using supervised classification. LULC map Figure 5(e) illustrates a diversified terrain in the Western Ghats region. The forested area, characterized by extensive tree cover, accounts for approximately 2.60% of the study area. Savannas, which consist of grasslands with scattered trees, cover 7.01% of the territory, while grasslands in the center section account for 7.49%. Wetlands in the Western Ghats account for 0.42% of the study area.

Agricultural activities dominate the terrain, accounting for 79.74% of the studied area. This is especially noticeable in districts like Kolhapur, Satara, and Sangali, where built-up areas account for 1.33% of the region. Natural vegetation coexists with this primarily farmed landscape accounting 0.87% of the study area. Barren land makes up only 0.04% of the study area. Water bodies account for around 0.47% of the region, concentrated in Karnataka's Bagalkot district as shown in Table 5. This comprehensive overview highlights the diverse land cover and land use patterns in the Western Ghats and the surrounding areas.

Table 5

Study area classification based on LULC, TWI, distance from river, and distance from road

View Large

Figure 5(f) shows the TWI maps for identifying flood-prone areas, offering an alternative to traditional contour-based methods. TWI is one of the variables that indicate the possibility of base flow, intermediate flow, and surface runoff accumulation. TWI was calculated to understand the spatial distribution of soil moisture and potential zones of water accumulation. Higher TWI values indicate areas that are more likely to be wet and susceptible to flooding. Flood-prone locations using the TWI value work as a guide in many higher runoff indicators. On level terrain, there is a greater accumulation of water flow than on steep inclines. Dust content, organic matter content, and soil horizon depth are among the soil properties that have a strong correlation with TWI. From Table 5, the very high TWI value is 12.01–16.2, which has a very high inundation hazard, and the low TWI value, which is less than 1, has very low inundation susceptibility and is situated mostly in Western Ghats, whereas the high (8.01–12.00) high inundation susceptibility is mainly seen in the main Krishna River.

As shown in Figure 5 (g), the flood risk in the research area has been systematically analyzed using a categorization system based on river distance. The distance to rivers was evaluated to determine areas that are more susceptible to flooding due to their location relative to water flow paths and potential blockages. The diverse zones with separate levels of flood danger have been identified. From Table 5, the study region has a significant percentage (31.41%) in the very high-risk region, which is indicated by being 0–0.1 meters from the river. It suggests that certain places are more vulnerable to flooding. With 41.74% of the study area falling within the high-risk region and the distance from the river ranging from 0.01–0.04 meters, flooding is a significant risk. A moderate level of flood risk is found when one moves near the Moderate Risk Zone, which makes up 18.41% of the research area and corresponds to 0.04–0.08 meters from the river. With a distance from the river ranging from 0.08 to 0.12 meters, the Low Flood Region, which makes up 6.29% of the research area, indicates a lesser but still significant danger of flooding. Lastly, 2.14% of the study area is classified as having an insignificant risk of flooding because it is located between 0.12 and 0.21 meters from the river. This thorough analysis offers a thorough grasp of the distribution and gradation of flood risk according to river proximity throughout the research region.

Figure 5(h) comprehending and managing flood susceptibility by the utilization of ‘distance from road’ identification within land use planning and flood risk assessment. The distance from roads was assessed to identify locations that are more vulnerable to flooding due to their location relative to river flow pathways and possible obstructions. Greater distances from roads signify a lower risk of flooding, indicating areas with reduced infrastructure damage, fewer road access challenges, and diminished drainage issues. From Table 5, the moderate-risk areas, covering about 9.79% of the study area and characterized by an intermediate distance from roads, may still encounter flood-related challenges such as potential drainage issues and delayed emergency response times. On the other hand, very high-risk areas, covering approximately 44.07 and 42.81% of the study area and situated near roads, face a heightened likelihood of infrastructure damage, difficulties in emergency response and evacuation, and the risk of being cut off during flood events. The low and very low-risk zones cover about 2.74 and 0.58% of the study region, respectively. This categorization approach may be adapted for floodplain management by considering the distance from streams or rivers in addition to roads and recognizing the unique dynamics of these areas. In such high-risk floodplain zones, the implementation of stringent land use restrictions and comprehensive flood risk mitigation strategies becomes imperative to enhance overall resilience and safeguard against the potential impacts of flooding.

The definition of flood-prone locations depends critically on the kind of soil, which has a substantial impact on precipitated water infiltration and water-holding capacity. Soil types and their infiltration capacities were analyzed to understand how different soils contribute to water retention and runoff. Areas with low infiltration capacity are more prone to flooding. The larger pore space between particles in sandy soils leads to higher saturated hydraulic conductivities. Infiltration affects the amount and availability of surface runoff, and various soil textures absorb water in separate ways. Sandier soil absorbs more quickly than clay soil does. Table 6 identify places that are vulnerable to flooding, the soil texture factor considers the physical properties of the soil, specifically its texture. From Figure 5(i), the study area is covered by clay (Eutric Vertisols) and clay loam soil, which make up 2.77 and 0.93 km², respectively. Loam (Chromic Luvisols and Humic Nitisols) and sandy loam (Lithic Leptosols), which make up 0.43 and 0.69 km², respectively, are the next most common soil types.

Table 6

Study area classification based on soil type and NDVI

View Large

Figure 5(j) shows the indicator to designate the volume and quality of vegetation in the study area. The NDVI was used to assess vegetation cover and health. Dense vegetation can reduce runoff by increasing water infiltration and evapotranspiration, while sparse vegetation can contribute to higher runoff. NDVI is a RS technique that evaluates light reflection at visible and near-infrared (NIR) wavelengths. Tracking the growth and health of vegetation as well as identifying stressed or damaged regions are common uses of NDVI in forestry, ecology, and agriculture. The NDVI measurements have mapped and categorized different plant types as well as observed changes in vegetation cover over time. With values close to zero coming from rocks and barren soil and negative values coming from clouds, water, and snow, this index depicts green. From Table 6, it goes from −1.0 to 1.0. The extremely low values of the NDVI function (−0.36–0.16) correspond to empty patches of rocks, sand, or snow and cover around 14.85% of the study region. Large values (0.59–0.99) cover about 8.8% of the study area and depict temperate and tropical woods, whereas medium values (0.29–0.42) cover about 29.09% of the study region and show shrubs and meadows.

In Kolhapur and Satara districts, a ‘very low’ flood hazard zone covering 0.22% of the area has been identified. This designation has been attributed to several factors, including the steep terrain, high drainage density, and permeable soil composition. These characteristics contribute to reduced surface runoff and enhanced water infiltration, mitigating flood risk in these regions. The presence of major dams like Koyna and Dhom plays a significant role in regulating river flows and minimizing flood impacts. Moderate flood zones encompassing Satara, Kolhapur, and Belagavi were delineated, covering 22.90% of the study area. These areas are distinguished by their mountainous topography, high precipitation, and significant surface runoff. The presence of multiple dams, such as Dhom, Kanher, and Koyna, influences flood dynamics, although the strategic location of dams in Kolhapur's western areas helps manage river flows, albeit with potential risks associated with excess water storage.

The most significant flood hazard has been observed in parts of Kolhapur, Belagavi, and Satara districts in Karnataka, covering 68.27% of the study area. These areas are characterized by low elevation, flat slopes, and impermeable soil, exacerbating flood susceptibility. The limited surface runoff and reduced percolation capacity contribute to heightened flood risk compounded by the presence of major dams such as Bhivargi, Chandoli, and Morna. Floods originating from the Krishna River, along with contributions from its tributaries and reservoirs, such as Almatti and Narayanapura, amplify the vulnerability of these regions. In the eastern part of the study area, Sangli, Bagalkote, and Vijayapura districts have a high flood hazard zone covering 5.24% of the region. This designation is attributed to factors such as low elevation, agricultural land cover, and proximity to major rivers. The presence of significant dams like Almatti and Tungabhadra exacerbates flood risks, highlighting the interconnected nature of flood hazards across administrative boundaries. The assessment underscores the intricate relationship between natural and anthropogenic factors in shaping flood vulnerability in the UKB. By integrating diverse data sources and analytical techniques, researchers have produced a comprehensive understanding of flood risk dynamics, providing valuable insights for informed decision-making and proactive FSM in the region.

Flood risk (FR) has been assessed for each flood conditioning factor by calculating the ratio of flood occurrence to the area ratio. The derived FR results for each factor are presented in Table 3. A higher FR weight in a class indicates a greater likelihood of flooding, suggesting a stronger correlation between that class and flood frequency. Values greater or less than 1 signify a strong or weak relationship, respectively. From Table 2, the analysis shows that rainfall ranges with higher FR values (1.253–1,923, 1,924–2,594, 2,595–3,264, and 3,265–3,935) have a higher frequency of flooding, but this doesn't necessarily mean increased rainfall, as flooding usually occurs at lower elevations. The study found that slopes between 35.99–53.99 and 53.99–71.99 have FR values of 2.04 and 2.95, respectively, with 78.89% of floods occurring in these areas. Lower slopes are more susceptible to flooding, while risk decreases with slope degree. The basin is divided into five drainage densities: 726–1,535, 1,535–2,998, 2,998–4,461, 4,461–5,924, and 5,294–7,387. The 1,535–2,998 and 2,998–4,461 classes have strong correlations with flooding, with FR values of 1.52 and 2.36, respectively, and respective basin occupancy rates of 5.83 and 30.12% (Sarkar et al. 2022).

Elevation significantly influences floods, with the lowest elevation class (389–594 m) having the highest FR (1.28), with 55.56% of floods occurring in this range. As elevation increases, the FR decreases, indicating a stronger influence on flood occurrences. The study found that the TWI conditioning factor had the highest FR values (7.41) and lowest ratio values (0.69) in classes (8.01–12.00) and less than 1, while subsequent classes had higher frequencies (FR) and fewer floods in the initial TWI class (Tariq et al. 2022). The distance to the road and river significantly influences flood potential, with the 0.08–0.13 class showing the highest value (1.92), indicating increased flood risk. The study found that the most influential distance classes for floods were 0.01–0.04 and 0–0.01 m, with frequency ratios of 1.33 and 1.27, respectively, while other distance classes had negligible effects. The study region identifies four soil texture classifications: sandy clay, loamy soil, sandy soil, clayey soil, and loamy soil. Clay soil covers 53.33% of the area and has the highest FR value (1.33), while clay loam has 1.19 (Abdo 2022).

The LULC map identifies nine classes, with agricultural lands and urban and built-up areas being the most correlated with flooding, with FRs of 2.65 and 1.02, respectively, and the other four classes having FRs less than 1. The NDVI, a significant factor with a −1 to +1 value range, has a negative correlation with flood occurrence, with flood susceptibility increasing with decreasing values, with the highest three classes having the lowest FR (Shafapour Tehrany et al. 2019). Once the flood probability index obtained from FR analysis is classified, five classifications (very low, low, moderate, high, and very high) comprise the flood hazard map, as shown in Figure 6(b) by using the quantile method. As shown in Table 7, the ‘very high’ zone encompasses 42.90% of the study area. The flood susceptibility map reveals that 42.21% of the region is classified as high, while 10.50, 3.76, and 0.63% of the study area are classified as moderate, low, and very low, respectively.

The higher the SI value for each class of every conditioning component, the more likely it is that a flood will occur within it. A negative SI value suggests a negative correlation between the class and flood occurrences. As shown in Figure 6(c) and according to Table 2, the study found that increased rainfall led to a decrease in floods, but the sloping terrain in the study area's elevated regions contributed less to floods. Positive weights were found in four rainfall classes, while negative values were recorded in the 582–1,253 mm class. The study found that increasing the slope degree decreases the likelihood of flooding, with positive SI values observed in the last three slope classes. The study found that drainage density varies across different density ranges, with some classes showing negative impacts on flood occurrence (Rai et al. 2022). The elevation parameter's Class 389–594 m had a positive SI value of 0.25, but SI values decreased as heights increased above 594 m. The TWI parameter showed negative SI values for classes less than 1, with the highest value (2.00) observed for classes 8.01–12.00, indicating a continuous SI increase (Abdo 2022).

The SI value for distance from the road is negative for the class (0.02–0.04 m), while three classes (0–0.02, 0.04–0.08, and 0.08–0.13 m) have positive SI values. The study focuses on the seventh factor, distance from the river, which is considered one of the most significant causes of floods. The first and second classes (0–0.01 and 0.01–0.04 m) showed positive values, while other classes were negatively valued, suggesting that flooding decreases with distance from the river. The soil texture was analyzed, revealing that Sandy Clay Loam (A & B), Loam (B), and Clay Loam (B & C) had negative SI values, while Clay (A & B) had positive values (Khosravi et al. 2016b).

Positive SI values were found in croplands, agricultural lands, built-up areas, natural vegetation, barren land, and water bodies, while negative SI values were found in forests, savannas, and wetlands. The NDVI component, ranging from −0.69 to −0.58, showed no correlation with positive or negative weights. The second class had a positive weight among the five related factors, while the first and next three classes had negative weights (Khosravi et al. 2016a). The ArcGIS Spatial Analyst Tool's weighted sum option was used to calculate the final flood probability index, which was then divided into five susceptibility classes, similar to the FR analysis. As shown in Table 7, The areas of the very high, high, moderate, low, and very low flood-susceptibility zones (FSZs) were 27,054.80, 16,540.50, 2,590.01, 486.21, and 79.48 km², respectively, with corresponding area percentages of 57.87, 35.38, 5.44, 1.04, and 0.17%.

As shown in Table 7, the percentages for each model represent the proportion of the total area classified under each flood hazard class. These variations can occur due to differences in the methodologies and criteria used by each model to assess flood risk. The AHP (Analytic Hierarchy Process) model uses a multi-criteria decision-making approach, weighing various factors according to their importance. The resulting areas show a higher concentration in the high and moderate classes, indicating a balanced but slightly conservative assessment of flood risk. The FR model bases its classifications on the historical frequency of flooding events. This model shows a significant area under very high risk, suggesting a focus on historical flood data, which may highlight areas that have been repeatedly affected in the past. The SI model employs statistical correlations between flood occurrences and various influencing factors. The highest percentage of areas falls under the very high class, implying a data-driven approach that may be identifying areas with a strong statistical likelihood of flooding. These variations imply that each model offers a unique perspective on flood risk. The AHP model may be better suited for strategic planning due to its balanced approach. In contrast, the FR model could be useful for understanding historical trends, while the SI model might provide the most aggressive forecast of high-risk areas based on statistical data. Combining insights from all three models can lead to a more comprehensive FSM strategy, ensuring that different aspects of flood risk are adequately addressed.

Floods are a significant natural disaster that human societies face, and creating FSM stratergies is crucial for mitigating their impact. The AHP model reveals that 0.22% of the region has very low potential for flooding, while 3.37% has low potential. A significant portion, 22.90%, experiences moderate flood hazard potential. The FR model shows a very high potential for flooding across 42.90%, with 42.21% classified as having high potential. The Susceptibility Index (SI) model indicates that 1.04% of the region has low flood hazard potential, with 0.17% classified as extremely low. A segment, 5.54%, exhibits moderate risk for flooding hazards. 35.38% of the region demonstrates high potential for flooding, while 57.87% is classified as having a very high potential. The ROC-AUC verification of the Flood Hazard Map, produced by integrating the AHP, FR, and SI models, shows that the SI model has a greater reliability in identifying flood hazards and mitigating their impact compared to the FR and AHP models.

While the study demonstrates promising results through ROC-AUC validation for assessing flood hazard susceptibility using the AHP, FR, and SI models, it's important to consider several limitations of this approach. ROC-AUC, though commonly used, may oversimplify the evaluation by focusing on overall predictive performance without addressing class imbalance issues inherent in flood hazard mapping. Given the diverse nature of flood risk levels across different regions, the reliance on ROC-AUC alone may obscure nuanced differences in model performance, especially in scenarios where certain hazard levels are rare or predominant. Moreover, ROC-AUC does not provide direct insights into optimal decision thresholds, crucial for practical applications such as FSM. Future studies could benefit from complementing ROC-AUC with metrics like precision-recall curves or exploring alternative validation strategies that account for real-world scenarios and model uncertainties. Incorporating historical flood data or cross-validation techniques could enhance the robustness of findings and provide a more comprehensive assessment of model reliability in flood hazard mapping.

This research demonstrates that integrating the Analytical Hierarchy Process (AHP), Frequency Ratio (FR), and Statistical Index (SI) models within a Remote Sensing-Geographic Information System (RS-GIS) framework is effective for realistic and cost-effective flood risk management (FSM). This integrated approach identifies optimal flood risk models and recharge zones, allowing for more accurate delineation of potential risks. The FSZs are significantly influenced by factors such as rainfall, slope, elevation, drainage density, land use, landscape development, soil geology, and geomorphology. The final FSZ map, which categorizes areas into very high, high, moderate, low, or very low risk zones, is a crucial tool for informed decision-making in FSM.

The AHP model indicates that a substantial portion of the area falls under the high hazard class (68.27%), with moderate amounts in the moderate (22.90%) category and smaller areas in the very high (5.24%), low (3.37%), and very low (0.22%) categories. The FR model shows a more even distribution, with 42.90% of the area classified as very high, 42.21% as high, 10.50% as moderate, 3.76% as low, and 0.63% as very low. In contrast, the SI model classifies the majority of the area (57.87%) as very high risk, with 35.38% as high, 5.54% as moderate, 1.04% as low, and 0.17% as very low. These differences underscore the importance of utilizing multiple models to capture a comprehensive view of flood risk, as each model offers unique insights based on different criteria and methodologies. The AHP model's balanced approach, the FR model's historical perspective, and the SI model's statistical analysis collectively contribute to a robust FSM strategy.

Flood hazard mapping has the potential to improve spatial resolution and enhance accuracy. The assessment of flood risk can be enhanced using advanced RS techniques and a variety of Geospatial Data sources. The complexity of flood-related features at smaller scales can be captured using technologies such as Hyperspectral Imaging and LiDAR data fusion, which will improve the model's performance. Understanding climate change and land use dynamics is essential for developing robust forecasting models of flood dynamics. Future research should integrate climate projections and land use scenarios into flood hazard mapping frameworks to anticipate changes in flood risk under varying environmental conditions. Comprehensive validation studies are necessary to account for the temporal variability of flood conditions and to detect trends and anomalies affecting model accuracy.

Integrating water geosciences with social factors through interdisciplinary collaboration and data exchange initiatives can enhance the effectiveness of flood risk mapping. Sustainable flood management practices will benefit from promoting research focused on data exchange, interdisciplinary collaboration, and technological innovation. To promote sustainable flood management practices, it is essential to prioritize research initiatives that foster active data exchange among stakeholders, encourage interdisciplinary collaboration between water management and the social sciences, and drive technological innovation in flood forecasting and early warning systems. This approach will not only enhance the accuracy and reliability of flood risk assessments but also improve preparedness and response strategies in vulnerable regions.

A multidisciplinary approach to flood dynamics, which includes promoting open-access data repositories and cooperative research networks, can enhance our understanding and help develop comprehensive flood management plans. Integrating hydrological, geophysical, and socio-economic domains is crucial for informed decisions and resilient flood control techniques. Transparency and reproducibility in flood hazard mapping research are essential. A culture of cooperation and knowledge sharing can help communities adapt to changing environmental conditions. This research emphasizes the technical aspects of flood detection, providing detailed and accurate mapping of flood-prone areas through rigorous geospatial analysis, decision frameworks, and statistical measures. The outputs are region-specific, offering precise data for developing effective flood management strategies. Future research can explore policy implications through interdisciplinary collaboration with social scientists, economists, and policymakers, ensuring that the study provides a solid technical foundation for broader policy-oriented discussions.

This study on flood detection in the UKB using integrated geospatial analysis aims to identify and accurately map flood-prone areas, focusing on a comprehensive analysis of geospatial data with decision frameworks and statistical measures to ensure precise identification of flood hazard zones. While the alignment between identified flood hazard zones and current flood management strategies, as well as the socio-economic impacts, are significant topics, they extend beyond the specific focus and purpose of this study The research provides a strong foundation for future studies to explore these broader implications in greater detail. A high degree of technical accuracy is ensured by the specific focus on geospatial analysis, which significantly aids in flood detection. Discussing management strategies and socio-economic impacts would require additional data and expertise, leading to a separate study. This approach benefits researchers and practitioners in geospatial analysis and flood detection by providing detailed technical insights. Acknowledging the broader implications, interdisciplinary studies with social scientists, economists, and policymakers can best explore them, ensuring the study contributes precise and critical information for developing effective flood management strategies in future research or applications.

The corresponding author takes responsibility on behalf of all authors for ethical approval and permissions related to this research work.

The corresponding author takes responsibility on behalf of all authors for consent to participate related.

All the parties gave their written permission for the article to be published. The corresponding author takes responsibility on behalf of all authors for consent to consent to publish.

K.V.S. did the conceptualization and methodology. P.J. and P.D. worked on the figures and software work. The original draft preparation was done by V.K. and K.S. P.M. did the investigation work.

The corresponding author on behalf of all authors declares that they did not receive any funds or any other grant during the preparation of this manuscript.

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

The authors declare there is no conflict.