The rapid urbanization under the Kingdom’s Vision 2023 will see many developments along Saudi Arabia’s coastal areas. These areas, along the western coast of Saudi Arabia, are highly vulnerable to natural disasters such as flooding (Youssef et al., 2016). Therefore, an accurate flood hazard assessment is paramount for ensuring the safety of the proposed developments. NEOM, a futuristic city intended to drive the future of urban planning poses a prime example of a city prone to potential flooding as a consequence of its geographical location and ambitious development plans. This research paper aims to provide an enhanced flood susceptibility assessment by utilizing advanced geospatial techniques. Moreover, this study endeavors to provide a better understanding of flood susceptibility and potential impacts in the hopes of supporting urban planning of NEOM that is disaster-prepared.

The significance of flood hazard assessments cannot be overstated. With NEOM being strategically placed and its expectation to accommodate a high population, it is expected to become a global tourism and economic hub that will support the kingdom’s efforts in diversifying the economy. This entails a massive urbanization undertaking. Urbanization, which will trigger changes in Landuse and Landcover (LULC), is a prime culprit in changing the hydrologic processes as it increases runoff as well as peak water discharge (Sanyal et al., 2014; Apollonio et al., 2016; Khosravi et al., 2016; Zope et al., 2016; Yousuf Gazi et al., 2019; Tellman et al., 2021; Hagos et al., 2022). Coupled with climate change, which is leading to variable changes in weather/climate trends as floods are expected to vary in terms of magnitude and frequency (Cheng et al., 2007; Parry, 2007; Arnell and Gosling, 2016; Abdulalim and El Damaty, 2022), necessitates the development of accurate flood susceptibility maps.

The development of accurate and up-to-date flood maps is not an easy endeavor (Lin and Billa, 2021). This is generally attributed to three main factors. First, the varying nature of flood hazards is a consequence of climate, land use, and developing flood-control projects (Parvaiz et al., 2011; Kusler, 2016). Second, the availability of data and their accurate reflection of up-to-date disaster management practices (Tomaszewski et al., 2015). Finally, a lack of consensus on the selection of parameters for flood susceptibility studies (Alam et al., 2021). There are currently four existing methods to research flood susceptibility which are: (a) hydrological modeling, (b) machine learning models, (c) qualitative models, and (d) quantitative models. Hydrological models, such as SWAT (Soil and Water Assessment Tool) or HEC-RAS (Hydrological Engineering Centre – River Analysis System), are simple nonlinear models that are generally less effective when attempting to integrate complex features (Shafapour Tehrany et al., 2019). Elsebaie et al. (2023) developed both an AHP-GIS model as well as a 100-year return period scenario model using HEC-RAS for Wadi Al-Lith basin in Saudi Arabia and showed that both models produced comparable results. Machine learning integrates artificial intelligence but is generally very complex, requires high computational performance, and involves a relatively extended processing time making it less practical (Maier and Dandy, 2000; Ghalkhani et al., 2013). Al-Areeq et al. (2024) examined various Deep Neural Networks (DNN) algorithms for flash-flood prediction in Jeddah, Saudi Arabia, and have listed, in order, the various models that showed the highest accuracy. Qualitative strategies are able to solve complex decision-making problems, but they generally rely on expert knowledge in the required field exposing them to potential biases (Shafapour Tehrany et al., 2019). The incorporation of AHP, a tool widely utilized for flood susceptibility mapping, is heavily reliant on expert judgment for prioritizing criteria associated with flooding. For example, Dano (2022) used AHP to develop a flood susceptibility map for Dammam, Saudi Arabia, and pointed out that the qualitative nature of AHP can tend to lead to biases. Finally, Quantitative approaches encompass numerical representations of the connections between influencing factors and the incidence of disasters (Ayalew and Yamagishi, 2005). For example, Theilen-Willige and Wenzel (2019) utilized a technique that combines GIS and frequency ratio method to assess the natural hazards assessment along western Saudi Arabia and determined that it is most susceptible to floods.

Traditional flood susceptibility mapping methods have been limited by the availability of data (Alharbi, 2023). DEM and LULC, for example, are important for generating accurate flood susceptibility maps. DEMs provide information on the study area’s topography. Topography, namely slope, is a major factor in floods as areas with steeper slopes tend to be at higher risk of flooding than those with gentle ones. Therefore, a high-resolution DEM is ideal. It should be noted that higher resolution does not necessarily guarantee higher accuracy, particularly in areas with dense vegetation (Xafoulis et al., 2023). The study by Xafoulis et al. (2023) compared several DEM resolutions, namely 5 m, 2 m, and 0.5 m DEM, and noticed that despite various over-and-underestimations, the DEM resolution of 5 m managed to capture the true geometry of the site’s cross-section while supposedly being at a lower resolution than a DEM of 2 m. On the other hand, LULC provides information on the type of land cover of the study area. Most land use activities involve the conversion of the natural landscape, whether it be deforestation or simply converting large, vegetated expanses to areas suitable for human use. Impermeable surfaces such as roads, make a higher contribution to floods than areas overlain with vegetation. Land use changes have been deemed a prime cause of severe flood events (Owrangi et al., 2014) and deemed by some as potentially having irreversible effects (DeFries et al., 2004).

Pallard et al. (2009) investigated the link between drainage density and floods. The authors highlighted that drainage density controls flood peaks directly and indirectly. Control over stream network length and hillslope routes is a major direct effect. Drainage density impacts concentration time and peak flow magnitude in river networks due to higher flow velocity. Increasing drainage density increases flood peaks. Longer concentration times allow more water infiltration. Thus, decreasing drainage density reduces flood volumes. Drainage density as a geological measure has indirect impacts. Impervious rocky hillslopes limit drainage density, storage volumes, and flood peaks. Karstic regions, highly worn bedrock, and/or highly permeable fluvial deposits in valley floors may also cause low drainage density due to huge storage volumes and reaction times and modest flood peaks and volumes. Landform evolution, soil formation, erosion, and floods (driven by climate and regulated by geology) may also indirectly control. Large floods may modify catchments over generations by increasing topographic gradients, drainage density, and storage volumes, which may raise flood peaks and volumes. Finally, drainage density’s relationship with vegetation cover can provide indirect management in semiarid environments. Due to soil erosion, bare soils have high drainage density and runoff output, resulting in huge flood peaks and volumes.

Another key parameter affecting floods is precipitation. Ghoneim and Foody (2013) investigated flash floods in arid mountainous regions. The aim of the research was to determine the impact of rainfall depth, areal coverage and the locations of flash floods. They found that the study basin’s hydrological response to rainfall is determined by its inherent geomorphometry and land cover types. Furthermore, the basin’s response exhibits a high degree of sensitivity to the depth, size, and spatial distribution of the downpour. It should also be noted that surface run-off over extended periods of times (i.e., centuries) changes the morphology of basins and ultimately affects the drainage density which, in turns has an effect on the sites vulnerability to flooding (Pallard et al., 2009).

In this context, the Analytical Hierarchy Process (AHP), can be a very useful tool. First proposed by Saaty (1988), AHP has been widely utilized in the field of flood hazard mapping and risk assessment due to its ability to handle complex multi-criteria decision-making. Researchers have applied AHP to systematically evaluate and prioritize various factors that contribute to flood hazards, such as terrain characteristics, land use, climate data, and more. One of its most notable strengths is that it can be used to handle poor-quality data (Chen et al., 2014). AHP achieves this by assigning relative weights to these factors, enabling a comprehensive assessment of flood susceptibility, and providing valuable insights for flood management and mitigation strategies (Dodgson et al., 2009). This allows for AHP to be effective in handling both qualitative and quantitative datasets (Vinod Kumar and Ganesh, 1996). While AHP has been praised for this ability, it is often criticized for not being able to handle uncertainties (Büyüközkan et al., 2011). As a result, research has evolved to incorporate the Fuzzy Analytical Hierarchy Process (FAHP) to better enhance the generated weights from AHP with the aim of expanding the use, so it is not only limited to crisp values but also accounts for vague ones (Kerkez et al., 2017; Baalousha et al., 2023). Through fuzzification, this modified approach can convert crisp values into vague ones by employing a fuzzy judgment matrix, incorporating fuzzy numbers, to generate definitive weights through both consistent and inconsistent fuzzy comparison matrices (Baalousha et al., 2023). Another utilization of Fuzzy theory is to employ fuzzy logic. Fuzzy logic was first proposed Zadeh (1965) and involves the usage of partial truths instead of definitive values. This method allows for converting a definitive value ranging from 0 to 1, to a variant in-between them to represent the partial truths. A previous study by Baalousha et al. (2023) compared Fuzzy logic and FAHP, for flood exposure in arid regions, and found that Fuzzy logic is generally more conservative, especially in the high-risk zones, and that ultimately FAHP was likely more accurate.

The study area was investigated in three previous studies. The first by Abdulalim and El Damaty (2022) divided NEOM into eight (8) basins and utilized AHP to generate a flood risk map per stream for each basin through a morphometric analysis. For the assessment, the authors used the topographic wetness index, elevation, slope, stream power index, topographic roughness index, normalized difference vegetation index, sediment transport index, stream order, flow accumulation, and geological formation as the assessment criteria and have ultimately generated only flood-hazard maps for streams within Wadi Surr and Wadi Zhawi. The second study was done by Alharbi (2023) and utilized a weighted overall function in GIS software to map groundwater, flood, and drought zones. For this study, the author used drainage density, precipitation, elevation, slope, and soil type as the assessment criteria and divided NEOM into only four hazard classes ranging from Very Low to High with the majority of the study area falling under moderate risk. The third and latest study was undertaken by Bashir and Alsalman (2023) where they employed a morphometric analysis and followed the Approach set out by El-Shamy in order to determine Flash-flood vulnerability in NEOM and concluded that severe flash-flooding events in NEOM are highly improbable. It can be noted that two of the aforementioned studies relied largely on morphometric analysis, with only one incorporating environmental factors, namely precipitation and Soil Type. Our aim is to optimize and distill the relevant criteria in order to include relevant morphometric and environmental parameters for our assessment.

With the scarcity of data, namely in-situ data, the present investigation seeks to enhance the precision of the flood susceptibility map for NEOM. This is achieved by employing remote sensing techniques and integrating the Fuzzy Analytic Hierarchy Process (FAHP) with spatial modeling in ArcGIS Pro using open-sourced data to refine the analysis by allowing the assessment to account for uncertain or vague values within the data. The main starting point will be the development of adjusted assessment criteria based on the three prior studies to include eight criteria that combine morphometric and environmental factors affecting flooding. The ultimate goal is to deliver a more nuanced flood susceptibility map by utilizing and dividing the study area into seven categories of susceptibility, as opposed to the conventional four or five, which would maximize the potential of the final map to capture the wide diversity of the topographic and geomorphological features of the study area in an effort to avoid potential over or under-estimations. This refined map is intended to serve as a valuable tool for guiding master planning and facilitating decision-making for both present and future developments in NEOM.

The present study employed the FAHP methodology in GIS to produce a flood susceptibility map for the NEOM region. This methodology is extensively employed on a global scale and is commonly regarded as one of the most effective methods for this objective due to its ability to integrate spatial data from numerous sources at a consistent scale. To date, there have been no prior investigations that have employed this particular methodology to evaluate floods within the designated study area. The FAHP methodology was employed to integrate the existing criteria into a weighted overlay model implemented in ArcGIS Pro. The findings of the investigation indicate that NEOM, in general, is dominated by areas with varying degrees of moderate susceptibility to flooding. In general, areas with lower elevation and a steeper slope, commonly referred to as lowlands, exhibit the highest susceptibility to flooding. Conversely, regions characterized by higher elevation and a steep slope, known as highlands, demonstrate the lowest levels of flood hazard. A notable limitation of this study is the unavailability of in-situ and historical flood inventory data for the sake of validation. Another limitation is that this study did not incorporate existing flood-mitigation measures already available within the study area (i.e., culverts). Future research could be refined by incorporating in-situ data whenever it becomes available. Moreover, research can be expanded to model future climate change scenarios and incorporate future LULC to predict the associated socio-economic impact.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

BB: Conceptualization, Formal analysis, Methodology, Writing – original draft, Writing – review & editing. BT: Writing – review & editing, Conceptualization, Supervision. HB: Methodology, Writing – review & editing, Conceptualization, Investigation, Supervision.