Assessment of flash flood risks in the desert cities: a case study on Sabah Al-Ahmad, Kuwait

Climate change-induced disasters such as heat waves, heavy rainfall, and desertification pose a serious threat to Kuwait and numerous other countries worldwide (Rajkhowa & Sarma 2020; Abbas et al. 2021). It is high time we acknowledge this reality and take swift action to mitigate the environmental impacts before it is too late. In the same context, extreme precipitation events are causing an increase in flash flooding worldwide, which is exacerbated by climate change (Hassan et al. 2024). Flash floods, particularly in arid regions, are among the deadliest and most damaging natural hazards on Earth (Wang & Vivoni 2022). Flooding in urban areas occurs when the sewage and drainage systems cannot absorb rainfall. Heavy rains cause water to gather and speed up before colliding with each other on steep slopes. As a result, water can carry away debris, rocks, trees, cars, and structures and cause serious vulnerability (KISR 2000; Misak et al. 2012; Al-Rukaibi et al. 2017; Elkhrachy et al. 2021). Conceptually, dry channel beds, known in Arabic as ‘wadis’, fill rapidly with rainwater, which flows down the slopes, through the wadi system, into the sea, or inside sabkha/playa (Goudie 2003).

To estimate flood hazard impacts on urban areas, infrastructure, agricultural areas, and land use/land cover changes, various researchers have employed state-of-the-art spatial techniques like remote sensing (RS) and geographic information system (e.g., Mason et al. 2009; Abdelkarim et al. 2019). Others have mapped flood hazard vulnerability using statistical, probabilistic, hydrologic, and stochastic neural networks and fuzzy logic (e.g., Ahmad & Simonovic 2011; Youssef et al. 2011). Representative examples include Essel (2017) and Portugués-Mollá et al. (2016), who indicated that, based on geospatial techniques, a hydrologic assessment could be conducted to identify different hydrologic components, prepare hydrologic designs, and develop possible scenarios for overcoming flash flood hazards.

Floods in Kuwait occur during heavy rainstorms, with rainfall amounting to 30–40 mm in one storm lasting 4–6 h (Misak 2015). Historical flash floods in Kuwait included those of December 1934, November 1954, February 1993, November 1997, January 2004, January 2007, April 2008, December 2009, November 2013, January 2014, and November 2018 (Misak 2015; KMD 2020). These events commonly induce intensive damage to physical infrastructures (Misak 2015; Hassan et al. 2024). On the other hand, rain in Kuwait usually drops between October and May, with an average annual rainfall of 116 mm (Al-Senafy & Al-Fahad 2003). The number of rainy days in Kuwait is limited (29 days/year on average). Kuwait's rainfall events occur with different intensities (Al-Qallaf et al. 2019). In recent decades, extreme rainfall events have shown an increase in floods, which caused damage to properties and loss of lives (Wegler & Kuenzer 2023; Ghaderpour et al. 2024). The highest extreme rainfall event ever recorded since establishing meteorological stations in Kuwait was in November 2018. These events caused floods in most of Kuwait (Al-Qallaf et al. 2019).

The primary use of flood inundation modelling is to forecast the consequences of heavy rainfall events by simulating the flow accumulation by integrating spatial techniques, geomorphological field observations, and drone surveying after rainstorms. The hydrological community has developed computer algorithms that can automatically extract drainage networks from the digital elevation model (DEM), which has reduced the need for manual techniques (Tucker et al. 2001; Lin & Oguchi 2004; Reddy et al. 2004; Gelabert et al. 2005). The comparison between automated network extraction and traditional methods has also been studied (García & Camarasa 1999; Ashmawy et al. 2018). To identify the watershed, drainage networks are required for water-harvesting activities (Al-Jabari et al. 2009). The HEC-RAS 2D software is used to analyse the flow and timing of runoff in particular areas. This analysis helps to determine water depth, velocity, and flood extent. To ensure accuracy, the simulated results must be calibrated using real observations. These calibrated results are then used to create flood hazard maps and assess flood vulnerability. Many studies have provided accurate explanations of flood modelling using HEC-RAS (Mashaly & Ghoneim 2018; Abdelkarim et al. 2019; Elkhrachy et al. 2021; Abdelkarim & Gaber 2023).

This research aims to (i) monitor the hydro-geomorphological impacts of the floods on Sabah Al-Ahmad City, (ii) assess the strategic plan to diminish flood risks, aiming to prevent a recurrence akin to catastrophic events, and (iii) proposing a protection plan for the city. The results will aid decision-making for future planning and sustainable cities to mitigate extreme rain events.

During summer, the temperature can exceed 50°C, while it drops to less than 8° in the winter. The northwest wind is dominant in the study area, where it is active in the summer and primarily dust-laden (Hassan et al. 2021). In winter, the southwest wind, which is most laden with water vapour, is active due to passing the Arabian Gulf waters before reaching Kuwait. Due to its environmental conditions, Kuwait's vegetation cover is feeble. It is limited to some small seasonal and permanent plants that can withstand desert conditions. It is primarily found in depressions and desert valley paths where some water flows when it rains (Albanai 2021a, 2022).

The surface geomorphology of the State of Kuwait (and also the study area) reflects two critical factors: (1) The fluvial deposition in the delta environment, which is caused by the Tigris and Euphrates rivers and the historical deposition from Al-Batin Wadi, which forms most of the vast deposits from the far southwest to the northeast. (2) The wind force, where aeolian landforms occur. The topography of the State of Kuwait ranges from the west to the east, reaching about 290 m above the sea level in the southwestern part (Al-Shiqaya region) (Albanai 2021b). Overall, the surface geomorphology of Kuwait is dominated by deserts, coastal areas, sabkhas, and occasional wadis (Hassan et al. 2021). The flat nature of the landscape with sandy and salty features is characteristic of the country's arid environment.

Landsat 5 TM and Landsat 8 OLI data were relied upon to classify the land in the study area and investigate the temporal changes in 2002, 2013, and 2019 using selected images within the same season. The study used unsupervised and supervised classification to categorize pixels in satellite imagery based on their spectral signatures. Initially, unsupervised classification was used to understand spatial variations in the imagery. Then, spectral signatures were captured, and pixel values were extracted using supervised classification from training areas that corresponded to different land cover types. Once spectral signatures for each class were established, pixels were digitally classified into the most similar class.

Hydrological assessment requires terrain data to determine the watershed in the study area. DEMs are commonly used for this purpose, and they are available from different sources with different raster resolutions depending on the required accuracy and level of detail. In this study, the Forest and Buildings Removed Copernicus DEM (FABDEM) is selected as one of the most popular DEM sources for such research studies (Li et al. 2023). FABDEM removes building and tree height biases from the Copernicus GLO 30 DEM and is available at 1 arc-s grid spacing (approximately 30 m at the equator) globally.

HEC-RAS 2D software has been used to simulate flash floods. To extract water extent, radar data from the Sentinel-1 mission was used, supported by the European Space Agency. Sentinel-1A images captured on 25 November 2018 and 13 November 2018 were used to monitor and detect flooded areas (flood and reference images, respectively). The Level-1 product (VV polarization) was downloaded freely from vertex.daac.asf.alaska.edu with a scale of 5 × 20 square metres. To extract the flooded areas, a speckle filter was applied to the images. This filter reduces the complex speckled scattering variance in radar images. After that, the images were converted to a decibel scale to distinguish the flooded and non-flooded pixels. Then, the images were corrected using the terrain correction aiming to correct the topographic difference, which can cause geometric distortions. Then, the band math tool was used to map the flooded areas. This was done by setting an empirical value to separate flooded and non-flooded areas, as the water and land have low and high backscatter values, respectively. These steps were applied using the Sentinel Application Platform. The two images were used as reference data for the HEC-RAS 2D calibration step. Hazard maps have been produced using the water velocity and depth data obtained from the validation step.

To create an accurate hydrologic model, gathering data about the surface topography, land use and cover, and soil properties is necessary. For this, the DEM has been processed using HEC-GeoHMS software to delineate the drainage basin and network. The direction is typically assigned based on the steepest downhill gradient, and it uses a system of encoding directions as powers of two, corresponding to the eight neighbouring cells (8D flow model). The flow direction is then used in subsequent analyses, such as calculating flow accumulation, watershed delineation, and stream network generation (USGS, 2019). The results have been validated through Arc-Hydro by comparing them to the existing drainage network using high-resolution photos and images of drones and Google Earth Pro (Mashaly & Ghoneim 2018). In addition, basin characteristics such as surface gradient, channel slope, and longest flow path have been created to predict how the basin will respond to floodwaters (Mashaly & Ghoneim 2018).

The CN is a function of land cover and soil conditions, including the following information: antecedent moisture condition, land cover/lithology, and hydrological soil type. The antecedent moisture condition, which measures soil moisture, is a function of the total rainfall in the 5 days preceding the storm. Recorded weather data (KMD 2020) indicate that about 251 mm fell in 5 days (9–11 November) and (14–15 November), so the soil was moist and saturated with water, allowing for high runoff rates. The CN is calculated based on the hydrological group of the soil. The soil groups in the study area are classified into three categories.

Two main storm events are selected in this study as they are considered the highest-ever recorded storms in Kuwait and the most critical to consider in such research. These storms occurred in a continuous pattern for longer than 1 day, the first storm occurred from 9 till 11 November 2018 (total accumulated depth of 112.6 mm), and the second occurred just 2 days after and was even more severe, and it occurred on 14 till 15 November 2018 (total accumulated depth of 138.4 mm). Table 2 shows the daily values of the storm's vs the date of their occurrence.

HEC-RAS 2D software created a full 2D hydrodynamic rainfall–runoff model for flood inundation analysis. The model identifies flood-prone areas using storm patterns and a resampled FABDEM to construct a 2D mesh for the study area. Characteristics of the 2D mesh geometry are as follows:

• 2D mesh resolution = 60 × 60 m

• 2D mesh type = volumetric difference and 2D cell faces.

• 2D mesh cells = 217,134 cells

• Manning's coefficient values = spatially distributed values layer is developed for the study area using a soil layer generated automatically inside the HEC-RAS 2D using satellite imagery classification, the values ranged from 0.022 to 0.03 (based on the soil cover).

• SCS-CN values = SCS-CN method was the adopted method to represent the infiltration process inside the model as it is the most reliable and mostly used approach in such studies (Abdelkarim et al. 2019; Abdelkarim & Gaber 2023). A spatially distributed value layer is developed for the study area using a soil layer generated automatically inside the HEC-RAS 2D using satellite imagery classification, and the values ranged from 63 to 83 (based on the soil cover), as shown in Supplementary file (Appendix 1).

• Upstream boundary conditions = distributed precipitation for the storm events (using storm profile distributions.

• Downstream boundary conditions = normal depth.

The study adopted two scenarios using the main storm distributions from Table 2: one for a 3-day storm event and the other for a 2-day storm event. The hyetographs of both storms are presented in Supplementary file (Appendix 2). The simulation setup options in Appendix 4 have been adjusted for both scenarios, including the simulation time step to ensure model stability and accuracy.

Model verification is a crucial stage in the simulation process (Abdelkarim et al. 2019; Elkhrachy et al. 2021; Abdelkarim & Gaber 2023). It helps determine the accuracy and level of confidence in the model results from a numerical perspective. In general, modelling is a vast field that requires a high degree of accuracy, which can be challenging to achieve. Unless the model is well developed, precise data are input, accurate geometry is built up, and well-verified parameters are applied to both geometry and simulating settings, and a lot of unrealistic results may follow; thus, model verification shall be carried out to investigate the accuracy of the numerical solution within the model (Ghoneim et al., 2012). In this study, the model results in excellent verification results having a score of numerical volumetric error equal to 0.053% for the 3-day storm scenario and 0.022% for the 2-day storm scenario (acceptable numerical error values can reach up to 2–4%), which indicates the high level of accuracy of this model and shows the high level of numerical confidence in the results of this model Supplementary file (Appendix 6).

Model calibration is a very important step in the modelling process in hydrological analysis, especially in regions known for high levels of data scarcity (Abdelkarim et al. 2019; Elkhrachy et al. 2021). This step is always difficult to achieve using the physical comparison between numerical models and the existing field conditions, as mostly such calibration cannot be performed on the huge domain (study area) and in the absence of recent flood water marks or discharge/water depth gauges on streams, as most of the main streams (wadis) in those regions are ungauged. However, simple calibration steps are done within our study by ropes and drone surveying. The first step was adjusting SCS-CN and Manning (n) values using satellite imagery in comparison with technical references for those values in adopted technical references. Meanwhile, the second step of the calibration was comparing the results of inundation (flood-prone areas) with locations of water accumulation using historical satellite imagery. Supplementary file (Appendix 7) has some images comparing flood-prone areas from the model results with flood-prone areas from actual satellite imagery, which shows a high degree of matching between both the model results and the actual conditions Supplementary file (Appendix 7).

After reviewing the report issued by those committees, revisions must be made to ensure the safety and protection of the city and its citizens. Protection work is concentrated into three steps or procedures, as shown in Figure 12 (Ministry of Public Works 2019): (i) earthen mounds covered with concrete, with an approximate length of 15 km, (ii) deepening two collection areas to accommodate storage space for flood water, and (iii) constructing water channels of 15 km long. The state budget to protect the city of Sabah Al-Ahmad is estimated at $4 million (Ministry of Public Works, 2019). On the other hand, human encroachments such as camping, random traffic, and flood protection plans on the sides of the city that are not well studied can affect the geomorphological systems in the city. Also, anthropogenic forms and deposits can enrich and impoverish cities' complex morphological landscape and natural geodiversity (Brandolini et al. 2021).

In a practical view, protecting urban areas from flash floods has been a pressing issue recently, especially in regions prone to climate extremes like Kuwait (Omran & Dandrawy 2022). The City of Sabah Al-Ahmad is one such region that is susceptible to floods, which pose a significant threat to the city's infrastructure and the safety of its inhabitants. A comprehensive strategy is required to address this issue, considering the city's unique geomorphology and climatic conditions. One such strategy involves implementing measures to control water flow, mitigate its intensity, and utilize it for the city's benefit. This can be achieved by planting afforestation belts surrounding the city, protecting it from floods and sandstorms, and purifying the air from pollution.

This approach can be applied to new urban areas in Kuwait, such as Al-Mutlaa City in northern Kuwait and South Sabah Al-Ahmad City. The latter city is at high risk of experiencing floods as it is crossed in the middle by wadis of the sixth rank. Therefore, it is crucial to take proactive measures to protect the city from such disasters, as failure to do so can have severe consequences for the safety and well-being of its inhabitants. In brief, the proposed strategy of implementing dams/dykes and water storage, along with afforestation belts, presents a comprehensive solution for protecting urban areas from natural disasters like floods and sandstorms. On the other hand, several land use types in Kuwait raise the risks of floods, for example, off-road cars and rangeland grazing. Each land use results in soil compaction and vegetation degradation (Al-Dousari et al. 2019). As mentioned in the relevant literature (Al Jassar et al. 2023), the infiltration rate of rainwater in compacted soils may reduce by 40–100% of its natural rate if the soil is not compared – consequently, the runoff water and the associated soil erosion increase. So, managing rangeland and off-road traffic is one of the critical approaches to reduce risks of floods in Kuwait (Al-Awadhi et al. 2023). In other words, strategic plans to mitigate flood risks and foster sustainable urban development represent a pivotal paradigm shift in contemporary city planning. Emphasising proactive measures like improved drainage systems, flood-resistant building designs, and decentralised water management solutions, these strategies aim to create cities that are not only safer but also more reliable and adaptable to the uncertainties brought by climate change. It could be a good suggestive added mention in the present write-up.

Kuwait has seen an increase in flash flood events due to climate change. From 1934 to 1993, only three flash floods occurred, with long intervals between them. However, from 1993 to 2018, eight flash flood events occurred at shorter intervals, about 4 years. The amount of rainfall causing floods has also increased, from 23.7 mm/day in March 2009 to 103 mm/day in November 2018. Sabah Al-Ahmad City faced an environmental disaster in November 2018 due to heavy rainfall, with 2.6 million m³ of water accumulating in the surrounding areas. It is crucial to base urban planning on environmental geomorphology and scientific principles to achieve sustainable urban development. Cities can make informed decisions that promote long-term sustainability by considering strategies for mitigating flood risks and utilizing scientific research and data. This approach ensures that urban areas are developed in a way that minimizes the effects of floods, preserves natural resources, and creates a high quality of life for residents.

The study deeply investigated the causes of flash floods and the dynamics of surface runoff. This study enhances flood management literature by using UAV/drones data, satellite imagery, and advanced modelling to analyse flash floods in Sabah Al-Ahmad City. It offers practical recommendations for urban planning and risk management, improving existing methodologies, and providing valuable insights for similar regions. Results indicated that the city is located through the surface streams of Arifjan Wadi. The terrain difference between the west and east of the city of about 40 m resulted in torrential waters flowing from the west and northwest, causing severe damage to the infrastructures of the city, as well as the city, which turned into water-inundated areas (12 large, inundated areas monitored). The results also clearly showed that the flooded areas in the eastern part of the city doubled (from 2017 to 2019). The total size of inundated areas was almost 956 km² in 2017, which increased to 1,713 km² in 2019. Furthermore, three main inundated areas were recorded in 2017, while the number increased to five in 2019.

The analysis reveals several key findings: Sabah Al-Ahmed City lies adjacent to a substantial drainage basin spanning 780 km², with streams of order 5 and 6 coursing through its vicinity. During storm events, peak upstream flows on the city's western side reach 54.83 m³/s for a 3-day storm and 117.31 m³/s for a 2-day storm, exhibiting a multi-peak hydrograph pattern due to consecutive storm days. Inundation depths range from 0.1 to 1.2 m and 0.15 to 1.45 m for the respective storm durations, while inundation velocities vary from 0.2 to 0.6 m/s and 0.3 to 1.00 m/s. Overall, the flood hazard degree within the city's vicinity is classified as low to moderate, underscoring the need for cautious approaches and moderate preventive measures. On the other hand, recent advancements in understanding flash flood impacts have seen a transformative shift towards integrating cutting-edge methodologies such as artificial intelligence, high-resolution RS, and real-time data analytics. These technologies not only enhance our ability to predict and map flood-prone areas with unprecedented accuracy but also enable swift response and adaptive urban planning in the face of increasingly erratic weather patterns due to climate change. However, as these methods evolve, challenges persist in refining predictive models, integrating diverse data sources, and ensuring accessibility in resource-constrained regions. Ultimately, the study recommends several recommendations as follows:

• The study's methodology identified the causes of flash floods in Sabah Al-Ahmed City through hydrologic modelling, geomorphology, and drone surveying. These methods can be replicated in six other flood-affected locations in Kuwait to formulate adaptation plans and mitigate rainstorm effects.

• To achieve sustainable urban development in Kuwait, a hazard risk management strategy should be applied to future residential areas, with early warning systems.

• A vast amount of water collected in flood storage can be used for landscaping, greening, and agricultural projects in Kuwait.

• An environmental impact assessment must be conducted for the areas of southern Sabah Al-Ahmad that are permanently inundated. It will face many hydro-geomorphological problems in the future if it is not protected from the floods of Arifjan Wadi.

This research has not received any financial support.

All authors participated in all steps of this paper.

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

The authors declare there is no conflict.