Flood risk assessment by analysing flood frequency and probable inundation maps for Ishwardi Upazila, Pabna, Bangladesh Open Access

Floods are among the most frequent and destructive natural disasters worldwide, occurring with varying magnitudes and intensities across different regions and time periods (Tsakiris 2014; Patra et al. 2016; Tansar et al. 2020). Over the past few decades, the frequency and severity of floods have escalated, driven by factors such as climate change, human activities, and, in some cases, shallow groundwater rise (Iwalewa et al. 2016; Mokhtar et al. 2018). These catastrophic events have caused a significant loss of life and extensive damage to economic infrastructure, public property, and even historical monuments. While floods are natural phenomena that cannot be entirely prevented, human actions and the effects of climate change have been shown to amplify both their likelihood and their detrimental impacts (Beilicci & Erika 2014).

Flooding often results from heavy rainfall, rapid snowmelt, or storm surge. Among the various types of floods, flash floods are particularly destructive due to their sudden onset and lack of warning, making them more damaging than slow-developing riverine floods (Chowdhury et al. 2020). Bangladesh is among the most disaster-prone nations globally, with its vulnerability to floods being exacerbated by factors such as flat terrain, shallow riverbeds, intense monsoonal rainfall, and significant sediment discharge (Hossain 2015; Rahman et al. 2019). Urbanization and climate change are projected to further intensify this vulnerability, potentially altering rainfall patterns and increasing the frequency and severity of floods (Yang et al. 2015). Major flood events in Bangladesh, including those in 1998, 2004, 2007, and 2017, have caused extensive devastation, underscoring the urgent need for effective flood risk management (Rahman et al. 2019; Zhao et al. 2021).

Effective flood management begins with identifying flood-prone areas and developing flood hazard and risk maps that provide critical insights for technical, financial, and policy decisions. These maps also highlight risks such as environmental pollution associated with flooding (Beilicci & Erika 2014). While traditional flood mitigation focuses on altering flood characteristics, technological advancements in technology have driven the need for innovative and efficient strategies (Fan et al. 2017).

Flood inundation modelling has emerged as a critical tool for addressing flood-related challenges. This provides detailed insights into the distribution, extent, and dynamics of flood events. This model is particularly valuable for urban stormwater management, assisting with the planning, design, and analysis of storm sewer systems (Henonin et al. 2013; Laouacheria et al. 2019). By simulating severe storm scenarios, flood models can assess the performance of stormwater sewer networks and evaluate the effectiveness of operational and structural solutions. These applications are facilitated through various commercial and open-source simulation tools, enhancing their accessibility and utility (Bulti & Abebe 2020). Flood management typically involves the following two main approaches: (1) structural and (2) non-structural measures. In both cases, determining the design flow and stage plays a critical role. These parameters are essential for designing hydraulic structures for flood management projects. In addition, the design of the flood stage is instrumental in creating flood extent maps, which help assess vulnerability to flood damage over varying return periods (Rahman et al. 2011).

Advancements in technology and high-resolution topographic data have made hydrologic and hydraulic models, such as HEC-RAS/HMS, MIKE, and SWMM increasingly essential for flood hazard assessment (Manandhar et al. 2023). MIKE powered by DHI offers advanced tools for modelling water-related challenges, and its effectiveness has been demonstrated in numerous studies across South Asia. For instance, the MIKE11 hydraulic models revealed that 12.7% of the land in Bharatpur, Nepal, and 22.3% in Sylhet, Bangladesh, are at flood risk. Improved drainage systems could reduce these risks to 5.5% and 3.6%, respectively, while poor waste management could increase the risk to 7.6% and 18.5%, respectively, within 5 years (Pervin et al. 2020).

In Dhaka, Bangladesh, MIKE tools have been employed to address waterlogging issues. Khan et al. (2018) projected that a recurrence of the 2004 flood in 2050 could result in greater damage using MIKE FLOOD. Mark et al. (2018) combination of MIKE URBAN and MIKE FLOOD to simulate rainfall–runoff (RR) and flood extents, achieving good alignment with historical flood data. Similarly, Dasgupta et al. (2015) generated inundation maps using a coupled MIKE URBAN 1D-2D model, and Chen et al. (2016) modelled storm sewer and river flows based on historical rainfall and water levels.

In Sri Lanka, MIKE FLOOD and MIKE21 models have been applied to simulate RR and overland flow in the Kolonnawa basin, with a slight underestimation of peak canal levels during the May 2016 flood (Wagenaar et al. 2019). Additionally, coupled MIKE models assessed canal capacity in the Metro Colombo region, revealing the system's ability to handle only 10-year return events and identifying a safe flood level of 2.0 m (Moufar & Perera 2018). These studies underscore the reliability of MIKE software in simulating flood risks and informing infrastructure planning.

The MIKE 11 NAM hydrological model, a conceptual tool for RR modelling, can simulate both overland and groundwater flow (Aredo et al. 2021). Its applications are particularly beneficial in data-scarce regions (Ghosh et al. 2022; Keerthy et al. 2024). Makungo et al. (2010) demonstrated the utility of the model model's utility in simulating runoff hydrographs for the ungauged Nzhelele River, thereby aiding water resource management. Similarly, Doulgeris et al. (2011) investigated RR relationships in Greece's Strymonas River watershed, and Singh et al. (2014) applied the model for runoff predictions in India's Vinayakpur catchment. Globally, the MIKE 11 NAM model has proven effective, particularly in data-scarce regions. Studies in various settings, including those by Hafezparast (2013), Kumar et al. (2019), and Ghebrehiwot & Kozlov (2020), highlight its reliability in RR modelling. This adaptability and precision make the MIKE 11 NAM model a valuable tool for addressing diverse hydrological challenges.

Despite its extensive use, the MIKE 11 NAM model has not yet been applied to simulate RR processes in the Padma River catchment in Ishwardi, Pabna, Bangladesh. Limited understanding exists regarding the hydrological processes in this region. This study aims to address this gap by modelling RR in the catchment using MIKE 11 NAM. The calibrated parameters are expected to aid in assessing future hydrological events, developing design norms, and managing the water resources in the area.

Although the proposed study area is not prone to natural drainage issues or frequent rainfall-induced flooding, recurring canal breaches highlight the need for flood hazard assessments. This study includes the following two main components: (1) analysing rainfall, discharge, water level, and evaporation data and (2) developing a flood inundation map using ArcGIS and a RR model with MIKE 11. These efforts aim to provide insights into regional and local flood hazards and contribute to improved water resource management.

Water level data from the Padma River at the Hardinge Bridge Pakshi Station (SW90) were collected from the Bangladesh Water Development Board (BWDB) for the period 2010–2020. These data were used for flood frequency analysis (FFA) and to determine the design flood levels. Rainfall data, including ground-observed and GSMap estimates, were bias-corrected to ensure their reliability and accuracy.

DEMs were acquired from the United States Geological Survey (USGS) and processed using ArcGIS 10.8. Missing data cells were corrected, and vector contours were reinterpolated into raster DEMs using the integrated land and water information system (ILWIS) contour interpolation tool. DEMs were employed to generate slope maps and classify land based on elevation for flood risk analysis.

Landsat 7 satellite imagery was downloaded from the Global Land Cover Facility and classified using supervised techniques. Ground-truthing points obtained from field surveys validated the classification. The resulting land-use maps delineated agricultural land, water bodies, settlements, and other features to evaluate flood impacts across different land-use categories.

All spatial datasets were geo-referenced using the WGS 1984 Geographic Coordinate System and projected onto the WGS 1984 UTM Zone 45N coordinate system. This ensured spatial accuracy and facilitated seamless integration of the datasets for analysis.

Land was categorised into five flood depth-based classes:

•  F0: High land (0–30 cm flood depth)

•  F1: Medium-high land (30–90 cm flood depth)

•  F2: Medium-low land (90–180 cm flood depth)

•  F3: Low land (180–300 cm flood depth)

•  F4: Very low land (>300 cm flood depth)

These classifications were used to assess flood risks and their impacts on agricultural and settlement areas following standardised land classification practices in Bangladesh.

FFA was conducted to evaluate flood risks in Ishwardi Upazila by analysing the annual maximum water level data from the Hardinge Bridge Pakshi Station (SW90) on the Padma River. The dataset spanning 2010–2020 was obtained from the BWDB. Several probability distributions were considered, including Log-Normal (LN2), Three-Parameter Log-Normal (LN₃), Pearson Type III (P₃), Log-Pearson Type III (LP₃), and Gumbel distribution. These distributions were applied to estimate the flood levels for return periods of 2.33, 5, 10, 20, 50, 100, and 200 years.

The Probability Plot Correlation Coefficient (PPCC) method was employed to identify the most suitable distribution. Among the tested distributions, the P₃ distribution emerged as the best fit, based on the PPCC rankings and graphical probability plots with a 90% confidence interval. This distribution provides accurate estimations of flood risk across varying return periods.

The following equations were used for calculation.

• (a) LN₂ distribution

In probability theory, a LN₂ distribution is a continuous probability distribution of a random variable whose logarithm is normally distributed. The transformed variable is denoted by ⁠, where ⁠.

Z_T is the standard normal variation, y is mean and S_Y standard deviation of the log transformed data.

• (b) LN₃ distribution

Flood inundation maps were developed using DEMs and the interpolated water levels for different return periods. Water levels estimated from the P₃ distribution were overlaid onto the DEMs to delineate flood extent and depth. The inundation maps were further analysed to classify areas into flood vulnerability zones to aid risk assessment.

The model outputs were validated using historical flood records and observed runoff data. Sensitivity analyses were performed to assess the influence of key input parameters, such as precipitation and evaporation, on model accuracy and outputs.

A comprehensive analysis of water levels at the Hardinge Bridge Pakshi station (SW90) on the Padma River was conducted to evaluate the flood frequency and temporal trends. The station's danger level, as determined by the BWDB, is 14.25 m. Long-term observations indicated that water levels were consistently the lowest in April and peaked in October across the years under study. Notably, the levels surpassed the danger threshold in 2013, 2016, and 2019, with 2019 witnessing the most significant breach.

The correlation coefficient (R²) value of 0.0339 reflects a weak relationship between time and maximum water level, implying high variability in the data. Despite this, the existence of a statistically significant trend (with a P-value of 0.0015 and > critical value) emphasises a consistent upward shift in the maximum water levels over the years.

Using the P₃ distribution, the water levels at Ishwardi for various return periods were calculated based on data from 2010 to 2020. The calculated water levels for the SW90 gauge station on the Padma River for return periods ranging from 2.33 to 200 years are presented in Table 3. The results indicated that the water levels increased progressively with the return period, reflecting higher flood risks for extreme events. For instance, while the 2.33-year return period corresponds to a relatively safe water level of 13.89 m, the 200-year return period projects a critical level of 16.74 m, emphasising the need for proactive flood risk management for high-return period events.

Land was classified by elevation and flood depth using a standardised system developed for water resource management in Bangladesh. Table 4 presents the classification system, which categorises the study area into F0 (0–30 cm), F1 (30–90 cm), F2 (90–180 cm), F3 (180–300 cm), and F4 (>300 cm) flood depth zones. The classification revealed that 51% of the study area is high land (F0), suitable for high-yielding rice cultivation during the rainy season, whereas 18% is medium-high land (F1), suitable for both Aus and Aman rice. Flood depth significantly impacted agricultural potential, with F2 and F3 supporting limited Aman cultivation and F4 areas unsuitable for rice farming due to prolonged inundation.

Flood inundation mapping was conducted to evaluate the extent and depth of flooding over various return periods (2.33, 5, 10, 20, 50, 100, and 200 years). Historical flood data and FFA were utilised to analyse the return periods and determine inundation zones.

A detailed analysis of the flood impacts across different return periods highlights the progressive nature of inundation, which is summarised in Table 5.

• (1) 2.33-year return period: The flood depth is expected to remain below 40 cm, with negligible inundation. Low-lying areas, such as the Lakhsmikandi Union, may experience localised waterlogging, but no significant impact is anticipated elsewhere.

• (2) 5-year return period: Flood depths are projected to rise to approximately 40 cm, inundating 32.58% of the area. Lakhsmikandi, Muladuli, and Dashuria unions are likely to be most affected, while higher elevations such as Pakshi will remain unaffected.

• (3) 10-year return period: Flood depths are anticipated to increase to around 90 cm, inundating 46.32% of the area. Lakhsmikandi, Muladuli, and Dashuria unions will experience significant impacts, with minor flooding extending to Sahapur.

• (4) 20-year return period: Flood depths are expected to reach approximately 134 cm, affecting the same 46.32% of the area as in the 10-year scenario. Vulnerability will expand, with Sahapur and Silimpur joining the list of impacted regions.

• (5) 50-year return period: Flood depths are forecasted to escalate to 184 cm, inundating 59.71% of the area. Lakhsmikandi, Muladuli, Dashuria, and Sahapur will face severe impacts, with considerable damage to both residential and agricultural land. Higher elevations, such as Pakshi and Ishwardi Pourashava, will remain less affected.

• (6) 100-year and 200-year return periods: Flood depths are projected to peak at 249 cm during the 200-year return period, inundating 60.02% of the area. Lakhsmikandi, Muladuli, Dashuria, Sahapur, and Silimpur will experience near-total flooding. However, the critical infrastructure in Pakshi and Ishwardi Pourashava, including the Ruppur Nuclear Power Plant, will likely remain largely safeguarded.

Due to the absence of discharge data at the catchment outlet and groundwater data, the NAM model was not calibrated. Despite this limitation, Figure 12 shows several key hydrological behaviours. High evaporation rates lead to reduced groundwater recharge, whereas reduced percolation due to actual evaporation (AE) results in increased inflow (IF) as evaporation decreases. In 2018, lower outflows (OF) were observed because of the reduced groundwater recharge and percolation. However, if outflow increases without adequate discharge mechanisms, excessive rainfall could lead to flooding. BF, influenced by the silty soil type of the catchment, retained moderate water, with higher BF observed in 2019 due to excess rainfall compared to other years.

This study presented a comprehensive analysis of flood risks in Ishwardi Upazila by integrating hydrological data, DEMs, and RR simulations. Key findings, including the increasing trend in water levels at the Hardinge Bridge Pakshi station and the identification of the P₃ distribution as the most suitable model for FFA, highlight the increasing flood vulnerability in this region. The results emphasise the disproportionate impact on low-lying areas (F3 and F4), which are completely inundated during extreme events, while the highlands (F0) remain relatively safe and critical as flood refuges.

The inundation mapping, validated against historical flood data, provides valuable insights into the extent and severity of flooding across different return periods. Approximately 60% of the region is projected to face inundation during a 200-year return period, underscoring the urgent need for proactive flood management strategies. Key infrastructure, including the Ruppur Nuclear Power Plant and Ishwardi EPZ, appears to be safeguarded because of higher elevation, but agricultural and residential zones in unions such as Lakhsmikandi and Muladuli are at significant risk.

Despite its strengths, this study was limited by the unavailability of high-resolution rainfall and groundwater data, which restricted the accuracy of RR modelling. The underestimation of rainfall by the GSMap data further impacted the reliability of the simulations, highlighting the need for enhanced data collection and validation in future studies.

These findings underscore the importance of addressing urbanisation-induced changes in land-use and drainage patterns, which exacerbate waterlogging risks. Integrating structural measures, such as embankment reinforcement, with non-structural approaches, such as improved land-use planning and community-based disaster preparedness, could enhance the region's resilience to flood hazards.

This study offers a comprehensive assessment of flood risks in Ishwardi Upazila, highlighting the critical vulnerabilities and increasing risk of flooding in low-lying areas. The key findings and conclusions are as follows.

1. A statistically significant upward trend in water levels at the Hardinge Bridge Pakshi station indicates a potential increase in flood risk, with up to 60.02% of the area inundated during a 200-year return period flood.

2. The P₃ distribution was identified as the most suitable model for FFA, providing reliable predictions of flood levels across various return periods.

3. Inundation mapping revealed that low-lying areas (F1, F2, F3, and F4 land types) are highly susceptible to flooding, with all these areas being fully inundated during a 100-year return period flood. In contrast, the highlands (F0) remain largely unaffected under normal monsoon conditions.

4. Urbanisation and infrastructure development exacerbate flood risks by reducing the soil infiltration capacity, which could lead to increased waterlogging, highlighting the need for improved drainage systems and sustainable land-use planning.

Overall, this study emphasises the importance of flood risk management, strategic infrastructure planning, and disaster preparedness in Ishwardi Upazila. These findings provide valuable data for decision-makers to enhance flood resilience and address the emerging risks associated with land-use changes and climate variability. Future research should focus on integrating advanced hydraulic models, such as HEC-RAS or coupled 1D-2D models, to enhance flood hazard assessments. Incorporating groundwater monitoring data will improve our understanding of the impact of urbanisation on infiltration and recharge rates, which is critical for assessing waterlogging risks. Additionally, analysing the effects of climate change on rainfall patterns and flood risks will inform long-term planning. These efforts will contribute to the development of more robust flood risk management strategies, ensuring sustainable development and resilience in Ishwardi Upazila and other similar flood-prone areas.

The authors would like to express their sincere gratitude to the Bangladesh Water Development Board, Pabna, for providing the data essential for this study. Special thanks are also extended to Pabna Sadar Upazila and the Bangladesh Agricultural Development Corporation, Pabna for their invaluable support throughout the research.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.