Assessing climate change impacts on flood inundation patterns: a study of northern Pakistan Open Access

Floods have become the most frequently recurring natural events, causing immense damage to both the economy and human life. In recent years, floods have resulted in greater losses than any other natural disaster (Karim et al. 2015). The 2010 and 2022 floods in Pakistan were particularly catastrophic, leading to substantial losses in lives, property, and the agricultural sector (Ullah et al. 2024). The increasing frequency of floods is governed by changes in climate and land-use patterns. To address the challenges posed by floods and their associated impacts, there is a global push for more advanced flood warning and mapping systems (Merwade et al. 2018). Recently, the intensity of extreme precipitation events has increased, further exacerbating the frequency of floods, affecting people more than any other natural disaster worldwide (Dullo et al. 2021). Developing hydrological models of watersheds under changing climate conditions is important for effective flood risk mitigation and management, as well as for guiding future policy development (Syed et al. 2023).

Accurate information on flood events is vital for informed decision making, design, mitigation, and management strategies. Consequently, models developed for flood mapping and risk assessment are crucial for evaluating and managing flood-related risks (Xenarios et al. 2016). The primary factors contributing to flood events include urbanization, population growth, and climate change. Recently, shifts in rainfall patterns have led to an increase in the intensity and recurrence of floods (Dullo et al. 2021). The significant negative impacts caused by flooding demand comprehensive efforts to mitigate the effects. HEC-HMS and artificial neural networks (ANNs) are considered to be very efficient tools for predicting flood patterns and flow distribution in different riverine systems (Hassaan et al. 2024). Climate change continues to place the world in an increasingly vulnerable state with respect to flood inundation. It is of utmost importance to incorporate climate change factors into flood modeling to ensure accurate flood risk assessment (Bakhsh et al. 2011). Climate change alters precipitation and temperature patterns, leading to changes in flood regimes, such as discharge reduction in the Amazon river. However, precipitation is expected to increase in other regions, potentially causing variations in river flow and the extent of flooded areas (Langerwisch et al. 2013).

Floods are triggered by intense rainfall, glacial melts, and failure of natural dams. Hydro-meteorological factors are being altered by ongoing climate change, leading to an imbalance in the hydrological cycle within watersheds (Hassan & Khan 2022). The intensity and frequency of floods have increased because of these altered climate variables. Although multiple studies have been conducted on flood inundation and assessment, most of them do not consider climate change in their models (Edamo et al. 2022). Understanding changes in flow and flood patterns is essential to develop effective mitigation measures. Climate change has a significant impact on a basin's hydrology and flood characteristics. Rainfall–runoff modeling is commonly used to assess a basin's susceptibility to changing climate conditions. The Soil and Water Assessment Tool (SWAT) model is widely used for hydrological modeling due to its compatibility with complex systems (Phy et al. 2022). Studies evaluate future streamflows and flood inundation through ensemble hydrodynamic modeling. This approach is far better for the accurate and reliable integration of climate shift and potential flood scenarios (Dullo et al. 2021).

In Pakistan, the effects of climate change are very important due to the country's diverse weather conditions. River flows are high as a result of snow- and glacier-melt caused by rising temperature (Shakir et al. 2010). Information regarding future streamflows in rain- and snow-fed catchments like Chitral is crucial for communities and stakeholders to implement effective flood mitigation and management strategies (Burhan et al. 2020). It is essential to consider climate change scenarios when developing flood inundation maps to identify flood risk zones and devise appropriate flood mitigation and management plans. The generalized circulation models (GCMs), under different scenarios, are widely used to project future streamflows and assess the impacts of climate change (Jayasimha Reddy & Arunkumar 2023). Although GCMs are reliable for hydrological studies, bias correction is still needed. Among the various methods, quantile mapping (QM) is widely used to correct errors in precipitation and temperature inputs for hydrological models (Yamamoto et al. 2021).

The GCMs are dynamically downscaled and bias-corrected using various approaches to accurately represent regional watersheds and to assess the impact of climate within specific regions (Barnard et al. 2019). The HEC-RAS and GIS applications are widely used tools for generating flood inundation maps by incorporating future peak flows (Farooq et al. 2019). The HEC-RAS program evaluates water surface profiles and extreme flood levels (Ogras & Onen 2020). Future flood flows under different climate scenarios, like the RCP4.5 and 8.5 scenarios, are also modeled using HEC-RAS. The RCP4.5 scenario represents a moderate climate condition, while the RCP8.5 scenario reflects a worst-case scenario. As expected, future flows show higher intensity under the RCP8.5 scenario compared with RCP4.5 (Edamo et al. 2022).

Many studies have employed GCMs and projected that river flows will increase over time because of changing precipitation and temperature patterns (Baig & ul Hasson 2024). The SWAT model is widely used for hydrological modeling and it is efficient in projecting future streamflows by incorporating climate variables like precipitation and temperature under different scenarios (Edamo et al. 2022). The ecology of the Chitral River is significantly impacted by climate change. However, previous studies have primarily focused on hydrological changes based on mean and extreme flow values. A more detailed study of hydrological alterations that incorporates climate variables like precipitation and temperature is urgently required to comprehensively assess the impact of climate change on the riverine system (Usman et al. 2022).

The Chitral River, a tributary of the Kabul River, is primarily fed by the melting glaciers of the Hindukush mountains. Few studies have focused on the Chitral basin hydrology, neglecting climate change factors. The hydrological cycle of this basin is continuously being altered by rising temperatures, intense rainfall, and glacial melts. To accurately estimate future flood patterns, a more comprehensive approach to hydrological and hydraulic modeling that considers climate change is required (Syed et al. 2023). Thus, climate change has become a major cause of environmental transformations worldwide, significantly impacting riverine systems and their hydrological processes. Understanding these effects is crucial for adopting effective strategies to mitigate and adapt to the changing conditions in the Chitral River basin.

While several recent studies have explored flood modeling under climate change, this research stands out by employing bias-corrected, ensemble CMIP6 data (SSP2-4.5 and SSP5-8.5), QM, and multi-temporal projections. It uniquely integrates hydraulic modeling with land-use-specific flood damage zoning, which is absent in most comparative studies across similar mountainous contexts.

This study is distinct in its integration of ensemble CMIP6 climate projections with hydrological (ArcSWAT) and hydraulic (HEC-RAS) models to evaluate flood risks under changing climate conditions. It goes beyond peak flow simulation by assessing inundation extent, depth, and land-use-specific damage across multiple climate scenarios, providing a scalable framework for other snow-fed mountainous basins.

The methodology used for this study consists of three major steps. First, ArcSWAT was used to create a hydrological model that simulates the flow for the baseline period using observed data. SWAT is a semi-distributed model widely used to simulating streamflows. Model calibration was performed using SWAT-CUP by inputting observed flow data after specifying the model parameters and attributes. In the second step, climate projections developed by Khan (2021) from CMIP6 were used to represent future scenarios. Five GCMs under two scenarios, SSP2-4.5 and SSP5-8.5, were utilized to project streamflows in the Chitral River during the near (2025–2050), mid (2051–2075), and far (2076–2100) future periods. SSP2-4.5 represents moderate climate conditions, while SSP5-8.5 depicts an extreme scenario. These five GCMs represent various environmental conditions, including wet, dry, cold, hot, and average conditions. The five selected GCMs are listed in Table 1. The climate data were downscaled and bias-corrected through a two-step process. For downscaling, a patch interpolation method of re-gridding was used, followed by bias correction to address systematic errors. The quantile–quantile mapping (QQM) technique was employed for bias correction due to its ability to remove biases while preserving original climatological trends (Khan 2021).

For this study the daily historical river discharge (m³/s) from 1980 to 2010, recorded at the Chitral River gauge station (71° 47′ 15″E, 35° 51′ 47.9″N), was obtained from the WAPDA (Water and Power Development Authority) in Peshawar, Pakistan. The data were used for the calibration and validation of the flood model. Weather data including minimum and maximum temperature as well as precipitation were obtained from the Pakistan Meteorological Department for the same period (1980 to 2010). These weather data were recorded at two meteorological stations located in Chitral City (71° 50′ E, 35° 51′ N) and Drosh (71° 47′ E, 35° 34′ N). The land-use–land-cover (LULC) data were retrieved from the ESRI Sentinel-2 satellite, providing insights into current land cover and usage patterns within the river basin. Soil classification data were obtained from the Food and Agriculture Organization map catalog website, detailing soil types, texture, and properties within the river basin. The digital elevation model with a 30 m spatial resolution was acquired from the United States Geological Survey (Azzam et al. 2022). The Chitral basin covers an elevation from 1,043 to 7,701 m above mean sea level (AMSL). Additionally, the projection file for the study area was downloaded from the SpatialReference.org website. The projection system for the Chitral basin corresponds to WGS 1984, UTM Zone 43N, which is used for spatial data in HEC-RAS.

The SWAT model was selected because of its versatility and broad applicability across various hydrological contexts, making it highly suitable for the complex and diverse terrain of the Chitral River basin. The hydrological modeling process begins with the delineation of the watershed. This study employed a DEM of 30 m spatial resolution from the USGS to proceed with the delineation. The process was executed within an ArcSWAT project setup, a hydrological tool integrated within the ArcGIS environment. The delineated watershed covered an area of 12,128 km², with the longest stream measuring approximately 305 km. Through this process, 43 sub-basins were created. Key steps involved in this process included DEM processing, flow direction and accumulation analysis, stream network definition, outlet identification sub-basin delineation, and subsequent adjustment.

SWAT-CUP was utilized to perform the calibration and validation of the baseline model. For this purpose, observed flow data from 1981 to 2010 were used. Two-thirds of this data were allocated for calibration, covering a period of 20 years from 1981 to 2000. While the remaining one-third, spanning from 2001 to 2010, was used for validation. The calibration process was initiated with 22 different input parameters which are listed in Table 2 along with their initial minimum and maximum values.

Flood frequency analysis was carried out to estimate extreme flood frequency for the Chitral River. Return periods of 5, 10, 25, 50, 100, 200, 500, and 1,000 years were considered for the SSP2-4.5 and SSP5-8.5 scenarios during near (2025–2050), mid (2051–2075), and far (2076–2100) future periods. The EasyFit (5.6 Professional) tool was employed to identify the best distribution method for the frequency analysis. The lognormal (three-parameter) distribution method was found to be the most suitable fit for the specific river flow based on the Kolmogorov–Smirnov, Anderson–Darling, and chi-squared tests.

In the HEC-RAS project, the projection file was integrated into the RAS Mapper environment to ensure that the spatial reference system precisely depicts the geographic coordinates of the study area. In RAS Mapper, the river centerline, along with the left and right banks and flow paths, were delineated using the DEM. Flow paths were created 1.5 km from the centerline on both the left and right sides to capture the full extent of the floodplain area. Cross-sections were created using the automated cross-sections within the geometry layer for the river centerline at equal intervals of 1 km along the 57 km stretch, with each cross-section extending 0.8 km in length.

The incorporation of Manning's n values is an important step in hydraulic modeling, as it defines the roughness coefficients for the channel and banks. Manning's n values were taken from the widely known document ‘Manning's n for Channels’ by Krest Engineers (2021). Based on the channel type and its characteristics, the main channel was assigned a value of 0.035, while a value of 0.04 was used for the bank lines. These values are crucial for accurately simulating the resistance to flow within the river and along its banks, thus ensuring precision in modeling flow behavior. Peak flows against return periods (5, 50, 100, 500, and 1,000 years) were incorporated for different scenarios under historic, SSP2-4.5, and SSP5-8.5 scenarios during near-, mid-, and far-future spans. Peak flows are important for simulating the river's response to various flood events under different climatic scenarios. Boundary conditions were defined to accurately represent the hydraulic characteristics of the river system. For both downstream and upstream, boundary conditions used the normal depth criteria, with an upstream slope of 0.00788 and a downstream slope of 0.00583. The river gradient is essential for conducting an accurate analysis of water surface profiles and flow dynamics throughout the river.

Flood risk zoning plays a vital role in identifying areas susceptible to flooding and developing strategies for mitigation and management. Along the Chitral River, many reaches are home to urban and agricultural sectors, making it crucial to assess these areas' vulnerability to floods of varying return periods and climate scenarios. Flood risk zoning not only increases the resilience of communities to flood hazards but also informs sustainable development practices by highlighting areas that require special attention and protection. For this purpose, flood risk maps for different recurrence intervals governed by different scenarios were exported from RAS Mapper. These maps were then imported into ArcMap for further processing. In ArcGIS, these maps were overlaid on the ESRI Sentinel-10 LULC map of the Chitral basin to determine the types of land uses exposed to flood hazards. The inundated areas were from the LULC map and converted from raster to vector layers. Last, the areas of different land types within the inundation zones were calculated and imported to Excel for detailed analysis and representation.

These parameters, as shown in Table 3, collectively provide a comprehensive analysis of the model's effectiveness and reliability, making it suitable for simulating the hydrologic response of Chitral basin.

The maximum flood values obtained for different return periods using the lognormal 3P distribution method under the different scenarios are given in Table 4.

Table 5 presents the maximum depths for various return periods of 5, 50, and 100 years, during the near, mid, and far futures under the two SSPs. In the near future, a 1.88 m increase in depth was observed from SSP2-4.5 to SSP5-8.5 for the 100-year return period. SSP5-8.5 shows an increased depth of 11.06 and 3.71 m during the mid and far futures, respectively, compared with SSP2-4.5 for the 100-year return period.

This study is of significant importance compared with earlier research on the Chitral basin, addressing critical and practical gaps that earlier studies overlooked. While previous works, such as Shakir et al. (2010) and Khalid et al. (2013), focused primarily on historical flood patterns driven by glacial melts, they did not account for future climate scenarios, which are essential for long-term flood risk management. By integrating future projections through the SSP2-4.5 and SSP5-8.5 scenarios, this study provides a dynamic and forward-looking analysis of flood patterns across multiple return periods, making it a valuable tool for policymakers and management authorities.

In contrast to Usman et al. (2022), who used the SWAT model for hydrological modeling but did not explore the impact of flood inundation, this study leverages hydrological and hydraulic models in tandem, extending the analysis to forecast future flood risks under shifting climate conditions. Similarly, while Azzam et al. (2022) considered climate change impacts, their study lacked a comprehensive flood risk zoning and land-use analysis using the most recent LULC data, which are crucial for understanding how land-use analysis changes exacerbate or mitigate flood hazards. This study, by incorporating LULC data, not only maps out current vulnerabilities but also projects how land use and development might interact with future flood risks, offering a multi-faceted approach to flood risk management.

Furthermore, Syed et al. (2023) assessed hydrological changes but did not employ a detailed flood risk assessment, leaving a gap in understanding the spatial extent of flood impacts within the basin. This study fills this void by integrating the outputs from hydrological models with hydraulic modeling to produce flood inundation maps, essential for identifying flood-prone zones and planning appropriate mitigation strategies. This approach is critical, as it provides stakeholders with actionable information regarding which areas are most vulnerable to future floods, thus enabling targeted interventions.

This study incorporates the lognormal three-parameter distribution to estimate future floods as recommended by Burhan et al. (2020). However, they focused only on glacier contributions, while this study goes further by projecting streamflows under future climate conditions, thereby providing a more holistic understanding of flood dynamics in the region. Shah et al. (2020) conducted a hydrological assessment of the Upper Indus basin under future climate scenarios using SWAT. Their results, much like the findings of this study, indicated significant increases in streamflows due to rising temperatures and altered precipitation patterns, particularly under extreme scenarios like SSP5-8.5. This agreement underscores the broader regional impacts of climate change on river basins across Pakistan, emphasizing the need for integrated flood modeling and climate adaptation strategies in both studies.

The study conducted by Waheed et al. (2024) used CMIP6 projections for hydrological modeling in the Kunhar river basin and reported that future scenarios involving higher emissions (SSP5-8.5) will result in more frequent and severe flooding. This is consistent with the findings of this study, where future floods in the Chitral basin are projected to increase more under SSP5-8.5. Similarly, Mahato et al. (2022) conducted hydrodynamic modeling for the Brahmani River basin in India using HEC-HMS. Like this study, they found that flood magnitudes increased with future climate scenarios. Their work also emphasized the need to incorporate land-use changes and local adaptations, which aligns with the conclusions of this study, particularly on the integration of LULC data and its impact on flood risk zoning.

Ali et al. (2015) conducted hydrological assessment in the Upper Indus River basin using two different models which demonstrated a projected increase in temperature and precipitation throughout the 21st century. Their findings provided clear evidence of ongoing climate change in the region and concluded that climate shifts are likely to accelerate snowmelt in the Upper Indus basin, resulting in an increase in river flows over time. The study of Pokhrel et al. (2020) for Neuse River, North Carolina, found that the SSP5-8.5 scenario led to the highest projected streamflow after modeling through HEC-RAS which resulted in greater flood inundation and increased future flood risk. Like this study, they emphasized the importance of using climate projections to guide effective floodplain management and risk mitigation strategies.

Studies on the Amazon basin by Langerwisch et al. (2013) also support this study's findings by highlighting that global precipitation and temperature changes will cause variability in river flows and flooding. This reinforces the idea that climate change impacts on hydrological systems are not just local but have global relevance. In terms of method, Yamamoto et al. (2021) noted that the application of bias correction techniques like quantile–quantile is essential to improving the accuracy of climate projections in hydrological models. This supports the robustness of the methodology employed in this study, which used bias correction to address systematic errors in climate projections.

By comparing the findings of this study with those of other regional and global studies, it becomes evident that while there is a broach consensus on the significant impact of climate change on future flood patterns, variations across different basins highlight the importance of localized assessments. The study's approach of integrating future climate scenarios, hydrological modeling (SWAT), hydraulic modeling (HEC-RAS), and using a state-of-the-art calibration technique provides a comprehensive and robust framework that addresses regional characteristics and can serve as a model for other river basins facing similar challenges.

The ability to predict flood flows under both moderate (SSP2-4.5) and extreme (SSP5-8.5) climate change scenarios allows for a comparative analysis that highlights the growing urgency for adaptive flood management strategies in the Chitral basin. The scientific community and local stakeholders stand to benefit immensely from this research, as it not only deepens our understanding of how climate change will affect flood regimes but also equips decision-makers with the necessary tools to design adaptive and resilient infrastructure. The insight gained from this study can be applied to similar mountainous river basins across South Asia and other regions vulnerable to climate-induced floods, thus broadening its impact beyond the Chitral river basin.

The generated flood inundation maps can support policymakers in zoning regulations, infrastructure planning, and early warning systems. Implementation steps include digitization of high-risk zones, updating building codes, and training local planners on model use. Challenges may arise from limited data resolution, technical capacity gaps, and land-use enforcement constraints in vulnerable communities.

This study focused on flood inundation modeling of the Chitral River basin, incorporating the effects of climate change. Hydrological and hydraulic models were developed based on two climate scenarios: SSP2-4.5 and SSP5-8.5. The hydrological modeling was conducted using the SWAT model to project future streamflows under varying climatic conditions. Peak flows were estimated for multiple return periods through flood frequency analysis, using the lognormal three-parameter method. The hydraulic model was used by integrating the projected flood peaks with other key inputs. The SWAT model performed well during both the calibration and validation phases. Notably, SSP5-8.5 consistently showed higher peak flow values compared with SSP2-4.5 across the near, mid, and far futures.

The HEC-RAS model proved effective in simulating flood inundation patterns, providing valuable insights for flood risk assessment and management. The model successfully identified areas susceptible to flooding along the Chitral River, including Chitral Town, Ragh, Maroi, Kaghuzi, Mastuj, and Reshun reaches. A maximum water depth of 12 m was observed at the Chitral Town reach during the near-future scenario, with depths increasing to 14.79 m for both the SSP2-4.5 and SSP5-8.5 scenarios for a 100-year return period. Climate change significantly increases the frequency and intensity of floods in the Chitral basin. Modeling under future climate scenarios indicated a higher likelihood of extreme flood events under SSP5-8.5, which depicts extreme climate change conditions.

We confirmed that the current work is an original for the authors and has not been published or under review in any journals.

All authors have read and contributed to this manuscript and agreed for the submission.

All authors have read and contributed to this manuscript and agreed for the publication.

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

The authors declare there is no conflict.