Assessing the vulnerability of flash floods to climate change in arid zones: Amman–Zarqa Basin, Jordan

Based on the IPCC's Sixth Assessment Report (AR6) released in 2021, it is expected that unless there is a substantial reduction in carbon dioxide (CO₂) and other greenhouse gas (GHG) emissions in the coming decades, the global warming targets of 1.5 and 2 °C set in the Paris Agreement will be surpassed within this century (IPCC 2021). This increase in the average global temperature will lead to various climate changes, resulting in a range of impacts. Some regions may experience an increase in intense rainfall storms, while others may face more frequent and severe droughts (IPCC 2007).

Arid areas are characterized by limited rainfall and water resources. Higher temperatures accelerate evaporation, causing a decrease in surface moisture and amplifying the drying of the atmosphere. Climate change and population growth will increase the water demand in arid regions, worsening the issue of water scarcity (Lian et al. 2021). In contrast, a study conducted by Tabari (2020) examined the connection between variations in extreme precipitation and flood intensities using projections from global climate models (GCMs). The findings of this study revealed a significant rise in extreme precipitation events and a potential decrease in overall precipitation in arid regions. Accordingly, flood intensity in semi-arid regions is expected to increase by approximately 3.12% by the end of 21st century. These findings are in line with the conclusions drawn in the IPCC's AR6, which indicated that extreme precipitation events will likely become more frequent and intense over most mid-latitude land masses in a warmer world. Specifically, the report indicates that the current 1-in-20-year annual maximum 24-h precipitation rate is projected to become a 1-in-5 to 1-in-15-year event by the end of the 21st century (Seneviratne et al. 2021). Thus, arid areas are expected to face an elevated risk of flash floods under future climate change conditions. It is important to consider the effects of climate change on hydrologic systems while also considering human interventions in watersheds, such as reservoir construction and changes in land use. Urbanization can further exacerbate the effects of climate change, increase the risks associated with flooding, and degrade water quality (Bates et al. 2008; Chung et al. 2011).

Assessing the vulnerability of hydrologic systems to climate change involves considering both climatic and non-climatic factors. The IPCC defines vulnerability as the susceptibility of a system to cope with and adapt to the adverse effects of climate change, encompassing variability and extremes. The level of vulnerability depends on the system's sensitivity, adaptive capacity, and the nature, extent, and pace of climate variations it experiences (IPCC 2007). Numerous studies explored climate change vulnerability to identify regions highly prone to extreme weather events, including floods. These studies utilized indicators to assess hydrologic vulnerability across three main components. Exposure is typically assessed by analyzing the risks associated with floods and droughts, as well as changes in precipitation, temperatures, and evaporation. Sensitivity is evaluated by examining system characteristics such as population density, forest cover, and reliance on rain-fed agriculture. On the other hand, adaptive capacity is determined by assessing a system's ability to respond to climate change, considering factors such as sanitation services, electricity supply, water accessibility, and poverty levels within the population (Füssel & Klein 2006; Yusuf & Francisco 2009; Rivas et al. 2011; Boori & Voženílek 2014; Binita et al. 2015). Previous research has focused primarily on examining vulnerability to climate change on larger scales, such as large watersheds, countries governorates, and counties. However, there is an urgent need for more localized case studies, particularly in urban regions susceptible to flooding, to assess the social and environmental response to climate change. This need is particularly critical in low-income arid regions with rainfall variability and limited stormwater infrastructure such as the case of Jordan (Mirzabaev et al. 2022).

Jordan, as part of the eastern Mediterranean region, is predominantly arid to semi-arid, experiencing significant rainfall variability (Tarawneh & Kadıoğlu 2003). According to recent climate change projections by Abdulla (2020), Jordan is predicted to witness a temperature increase of 2.5–5 °C and a reduction in annual precipitation of 10–37% by the end of the century. The Amman–Zarqa Basin (AZB), located in the north central part of Jordan, encompasses the two largest cities: Amman and Zarqa (Al Kuisi et al. 2014). Over the past few decades, urbanization in the AZB has been driven by increased migration rates, an influx of refugees, and population growth. This urban expansion resulted in the conversion of over 70% of agricultural land into urban areas (Gharaibeh et al. 2019). The changes in land use, including the conversion of streams into roadways to accommodate traffic, have led to flash floods triggered by intense rainfall storms with high precipitation rates within a short period. Furthermore, the AZB presents an ideal study area for assessing flash flood vulnerability due to its high population density, rainfall variability, topography, and urban coverage. Various approaches, such as morphometric analysis, hydrological modeling, and unit hydrograph, have been employed to assess the risk of flash floods in Jordan (Alhasanat 2014; Farhan & Ayed 2017; Al Azzam & Al Kuisi 2021).

The AZB heavily relies on groundwater as a primary water source. However, the basin's aquifer has faced significant depletion and deterioration in water quality, reaching critical levels due to excessive groundwater extraction (MWI 2017). Previous studies on the impact of climate change on streamflow and rainfall patterns in the AZB have solely focused on the emission scenarios outlined in the Special Report on Emissions Scenarios and the representative concentration pathways (RCPs) (Hammouri et al. 2015; Al-Hasani et al. 2023). However, a comprehensive vulnerability assessment that considers the climatic, social, and physical characteristics of the hydrologic system and following the newest climate change scenarios has not been conducted in Jordan.

Based on the preceding discussion, this study aims to provide a localized and comprehensive assessment of rainfall patterns and related runoff in the southwestern part of the AZB, which is the most urbanized area within the basin, from 1970 to 2018, by considering changes in land-use patterns. For this, data from a regional climate model (RCM) based on the Shared Socioeconomic Pathway 5 (SSP5) scenario were used to examine future rainfall and temperature patterns for the period 2019–2060 and simulated the projected runoff. Additionally, a spatial vulnerability assessment was conducted to measure the risk of flash floods in an urban watershed, using exposure, sensitivity, and adaptive capacity indicators. This study aims to answer three questions: What are the trends in rainfall? What is the likelihood of flash flood occurrences based on climate change projections? And where are the regions, most vulnerable to climate change impacts? As Jordan ranks among the top 10 water-scarce countries globally (UNFCCC 2014), this study is essential for projecting likely water scenarios under a changing climate, enabling policymakers to assess and implement effective environmental protection and adaptation measures. Moreover, the methodologies employed in this study can be transferable to other arid regions facing similar climate change challenges. This study was conducted using the Watershed Modeling System (WMS 11.0) and Geographic Information System (ArcGIS 10.8.2) software packages.

Daily rainfall for the period of 1970–2018 was obtained for 11 rainfall gauge stations from the Ministry of Water and Irrigation (MWI) (Figure 3(c)). Mean, maximum, and minimum temperatures were acquired for Amman Airport station (AL0019) and Wadi Es-Sir (NRA) station (AL0057) for the period (1988–2016). Data related to population per locality were obtained from the Department of Statistics. The stormwater drainage network was obtained from Greater Amman Municipality and Zarqa Municipality. Table 1 summarizes the sources and other information for the datasets used in this study.

Since morphometric parameters offer valuable insight into the geometry and hydrological characteristics of a watershed, morphometric analysis was carried out to provide information about the basin's slope, topography, soil condition, runoff characteristics, and surface water potential. In this study, six basic morphometric parameters, seven derived parameters, and three shape parameters were calculated for the study area watershed using equations outlined by Farhan & Ayed (2017) with the aid of ArcMap 10.8.2 and WMS 11.0 software (Table 2).

Based on the morphometric parameter definitions provided by Sukristiyanti et al. (2018), the study area showed certain characteristics. It has a drainage density (Dd) of 1.38, indicating poor drainage, slow hydrologic response, and high susceptibility to flooding. The drainage texture (Dt) is 3.38, classifying it as coarse and in the initial stage of erosion. The form factor ratio (Rf) ranges from 0 for a long, elongated basin shape to 1 for a perfectly circular analog. The area has a form factor (Rf) of 0.78, suggesting a circular shape with high peak flows, strong floods, higher velocities, and greater erosion and transport capacity.

The inverse distance approach was used to calculate missing rainfall data (Musy et al. 2014). The Mann–Kendall test has been employed extensively in research on hydroclimatic series' spatial variation and temporal trends. This test compares the alternative hypothesis (H₁), which states that there is a trend, to the null hypothesis (H₀), which contends that there is no trend (the data are independent and randomly arranged). In this context, Sen's slope, a non-parametric measure of trend slope, represents the strength of a diminishing or growing trend by estimating the slope of the trend using a linear model (Da Silva et al. 2015). In this study, the Mann–Kendall test and Sen's slope tests were applied to examine the total annual rainfall and temperature trends for the current and upcoming periods. The null hypothesis was tested at the 95% confidence level (α = 0.05).

The intensity–duration–frequency (IDF) curve illustrates the relationship between rainfall intensity, rainfall duration, and return period. IDF curves are obtained through frequency analysis of rainfall observations (Mahdi & Mohamedmeki 2020). Frequency analysis is used to predict rainfall events that may occur in the future from available historical data. In this study, IDF curves were built for all stations to simulate the runoff and peak discharge resulting from different rainfall storms. Gumbel distribution was the selected statistical method to perform the frequency analysis on the maximum rainfall data due to its effectiveness in modeling maximum values and focusing on extreme events, such as peak rainfalls. Rainfall intensity was calculated for different durations (5, 10, 20, and 30 min and 1, 2, 3, 6, 12, and 24 h) and different return periods (2, 5, 10, 25, 50, 100, and 1,000 years).

To detect changes in land use and estimate runoff in the study area, three land-use maps were generated using satellite images from 2001, 2011, and 2021. Due to the limited availability of satellite data before 2,000, the selection of satellite missions was constrained to those with data at the desired time. The ArcMap 10.8.2 image classification tool was used for supervised classification, employing the maximum likelihood algorithm. This process involves comparing pixels in remotely sensed data to create classes that align with the user's specified categories (Perumal & Bhaskaran 2010). The classification accuracy was assessed by utilizing reference land-use data from the Jordan – Land Cover Atlas (Franceschini et al. 2019). The accuracy of the classification was ∼ 77%. To address the different resolutions of the satellite images, the resample tool in ArcGIS software was used for harmonization.

To estimate runoff under future climate change, a predicted land-use pattern is essential. For this purpose, the MOLUSCE plug-in in QGIS 2.16 software was used to predict the future land use for the year 2041. The MOLUSCE plug-in was designed to model and simulate land-use/cover changes based on the Monte Carlo Cellular-Automata modeling approach. In this approach, the system is represented as a grid of cells, and each cell has a state that changes over time according to specific rules. These rules are defined based on a combination of factors and considerations, including neighboring cells' influence, socioeconomic factors, and environmental factors. These rules decide how each cell's state is updated based on its neighboring cells' states, which can include randomness. By repeatedly applying these rules and updating the cell states, the model can simulate how the system behaves over time (Li & Li 2015). The model used land-use maps from 2001 and 2021, as well as DEM and slope raster, as input data. MOLUSCE provides four methods to model potential land-use transition: artificial neural networks (ANNs), logistic regression (LR), multi-criteria evaluation, and weights of evidence. In this study, ANN was chosen as it provided the highest correctness percentage among the other methods. To validate the accuracy of the modeling process, land-use maps for 2001 and 2011 were loaded into the model to simulate the land-use map for 2021. The simulated land-use map was then compared to the actual land-use map of 2021, generated from image classification. The simulation accuracy was 68%.

The CN of the different cover types in the study area was determined based on the SCS urban hydrology for the small watershed (TR-55) (SCS 1986).

Rainfall amounts obtained from IDF curves, HSG, and land cover CN were then entered into the HEC-1 simulation model in WMS 11.0 to calculate the runoff volume and peak discharge for each sub-catchment in different duration and return periods. To estimate the runoff amount under future climate change, we utilized rainfall data derived from IDF curves based on GCM data, in addition to land characteristics extracted from the projected land-use map of 2041.

The GCM provides reliable global climate information for historical and future periods. The RCM is applied with higher spatial resolution over a limited area and is driven by GCMs. In this study, climate change data were obtained from the Regional Initiative for the Assessment of Climate Change Impacts on Water Resources and Socio-Economic Vulnerability in the Arab Region (RICCAR). RICCAR established the Mashreq domain where they downscaled climatic data from the Mediterranean (MENA) CORDEX domain. Mashreq domain used six GCMs and one RCM (Table 3) based on a 0.1-degree grid (∼10 km) spatial resolution. The collected data were the predicted daily precipitation and mean temperature for the period of 1970–2060.

The data obtained were then subjected to bias correction and represent the latest SSP5 scenario, a part of the Coupled Model Intercomparison Project Phase 6. The selection of the SSP scenarios is crucial for the vulnerability assessment, as it considers non-climatic factors and necessitates the evaluation of both climate change trends and socioeconomic trends. The SSP scenarios provide a framework for exploring how global society, demographics, and economy are likely to evolve over the next century, and how societal actions will impact GHG emissions. The SSP5 scenario represents a fossil-fueled development pathway, characterized by high challenges to mitigation and low challenges to adaptation. In this scenario, radiative forcing reaches levels comparable to those in the RCP 8.5 scenarios (8.5–8.7 W/m²) (8.5–8.7 W/m²) (Riahi et al. 2017).

To determine the most effective model for representing the data, we calculated the root mean square error (RMSE) in annual maximum rainfall between the observed and projected data over the same period (1970–2018). RMSE measures the average magnitude of differences between actual (observed) and predicted (model-generated) data points. A lower RMSE suggests a better match between the model and observed data. After calculating RMSE, we found that the CMCC-CM2-SR5 and EC-Earth3-veg models had the lowest RMSE values. To determine the best model for predicting maximum rainfall between these two, we compared their error distribution curves. The error distribution curve provides valuable insights into how the model's predictions deviate from the observed data and show error distributions. After reviewing the RMSE values and the error distribution curves, we determined that the CMCC-CM2-SR5 model was the most suitable for our study. This model consistently aligned more closely with the observed data, indicating its high performance in predicting rainfall patterns.

The following are more details about indicators:

Exposure: This component evaluated changes in various meteorological and hydrological factors associated with flash flood risk. We examined changes in the total annual rainfall, peak runoff flow, and mean temperature between the historical and future periods.

Sensitivity: For sensitivity, we addressed four key indicators:

• 1.

  Urban extent: This indicator measured the impact of urbanization on runoff patterns, recognizing its significance in flash flood risk.

• 2.

  Population density: This indicator indicates the potential risk to human lives and property.

• 3.

  Distance to roads: This indicator measured the proximity of roads to drainage streams to assess the susceptibility of road infrastructure to flooding during rainfall events.

• 4.

  Slope: The mean slope was calculated as it influenced the flow of floodwater.

Adaptive capacity: The assessment of adaptive capacity considered the following indicators:

• 1.

  Vegetation extent: This indicator served as a natural proxy for resilience, reflecting the area's ability to adapt to flash flood risks naturally.

• 2.

  Stormwater drainage network: This indicator highlighted the role of the stormwater drainage network in managing excess water during rainfall events, providing insights into the area's capacity to adapt to flash flood risks.

The analytical hierarchy process (AHP) was employed to determine the weights of the indicators used in the vulnerability assessment. In the context of multi-criteria analysis, AHP is a commonly used decision-support model that was first introduced by Saaty (1980). The rationale for using the AHP approach to determine indicator weights lies in its ability to provide a systematic and mathematically rigorous way to assess the relative importance of various factors or criteria in a complex decision-making process. The AHP approach involves breaking down the problem into a hierarchical structure and then ranking the different alternatives based on user judgments. In this study, AHP was implemented as follows:

Step 1: Design paired comparison matrices for sensitivity, exposure, adaptive capacity, and overall matrix as listed in Tables 4 and 5.

Step 2: Design pairwise comparisons to determine the relative importance of each indicator over one another, as ‘extreme’ (9), ‘very strong’ (7), ‘essential’ (5), ‘moderate’ (3), or ‘equal’ (1). The relative importance of each pair of indicators was assessed by answering the question: ‘How much more important is indicator A compared to indictor B in causing/contributing to flash flood events?’.

Step 3: Pairwise comparisons should be normalized by adding the values from each matched pair matrix column, then dividing each value by the total of the corresponding columns to get the normalized matrix.

Step 4: Calculate the relative weight of each indicator by taking the average of the normalized values for each row.

Step 5: Calculate the weighted sum value by taking the sum of the product of the degree of importance and relative weight.

Step 6: Calculate the importance of each indicator by dividing the weighted sum value by relative weight.

Step 7: Calculate the maximum eigenvalue (λ) by dividing the total number of importance by the number of indicators.

The judgment matrix for the indicators that affect flash flood occurrence is listed in Table 4, whereas Table 5 shows the judgment matrices of sensitivity, exposure, and adaptive capacity for weight assigning. The resulting weights of the vulnerability assessment indicators are listed in Table 6.

To assess the likelihood of flash flood occurrences, evaluate flash flood risk, and devise damage prevention strategies, we adopted the methodologies introduced by Sakib (2023). These methodologies involve three fundamental formulas: Sakib temporal flood intensity ratio (STFIR), Sakib flood damage avoidance potential (SFDAP), and Sakib flash flood risk index (SFFRI). These formulas were applied to all sub-catchments to compute flood potential, potential damage avoidance, and overall risk.

The number of yearly rainy days was found to vary from decreasing, increasing, or relatively stable among gauge stations during the past decades. This indicates that the frequency of rainy days has not remained constant, but rather, it has shown fluctuations over time. It is noticed that the 11-year moving average mostly behaved as the long-term average line. This finding was observed in all stations.

Analyzing the total annual rainfall data for the period from 1970 to 2018 at 10 out of 11 rainfall gauge stations revealed a decreasing trend. Table 7 provides details of the maximum, mean, and minimum rainfall recorded at all stations. The results of the Mann–Kendall test and Sen's slope are presented in Table 7 and indicate that there is no statistically significant trend in the total annual rainfall across all stations. Nevertheless, a significant decreasing trend was observed in the total monthly rainfall for March at six stations. The Amman Hussein station followed by the Sweileh station has the largest decrease in total annual rainfall by –30.6% and –29.7%, respectively. Overall, the study area experienced a reduction of 14.61% in total annual rainfall during the historical period from 1970 to 2018.

Regarding temperature trends, July and August have the highest mean temperatures, whereas January has the lowest. The mean temperature is about 18.4 °C, the average maximum temperature is 23.9 °C, and the average minimum temperature is 12.8 °C. The minimum temperature has the highest increase trend. According to the results from the Sen's slope test, the increasing percentages for the mean, maximum, and minimum temperatures are 7.3%, 5.8%, and 10.9%, respectively.

On the other hand, sub-catchments 2, 5, 8, and 9 have the lowest sensitivity scores. Although sub-catchment 2 possesses the steepest mean slope, its score was offset by its low urban extent. Sub-catchments 3 and 7 have a moderate sensitivity to flash floods.

The hydrologic vulnerability map combines the potential impact and adaptive capacity indicators. In this context, sub-catchments 8 and 9 exhibit a moderate vulnerability with a score of 6. This vulnerability is attributed to their high exposure, characterized by a higher projected rate of rainfall and runoff than the current rate. However, these sub-catchments have low sensitivity due to their low population density and limited urban extent. Limited stormwater network systems contribute to their moderate vulnerability. Sub-catchment 1 has a vulnerability score of 5, indicating a moderate vulnerability. On the other hand, sub-catchment 6, which experienced previous flash flood events, is expected to have a low to moderate vulnerability with a score of 4. Its vulnerability is mitigated by its lower exposure to increased rainfall, while its adaptive capacity helps in reducing vulnerability. Sub-catchment 2 has the lowest vulnerability score followed by sub-catchments 3, 4, and, 5.

Table 10 presents the scores for exposure, sensitivity, adaptive capacity, vulnerability, and potential impact within each sub-catchment. As depicted in the table, exposure scores are closely tied to adaptive capacity scores, shaping the overall vulnerability scores. For instance, sub-catchment 2 moderate exposure is effectively offset by its high adaptive capacity, resulting in a vulnerability score of 2. Conversely, in sub-catchments 8 and 9, their high exposure is further exacerbated due to the lack of adequate adaptive capacity measures, resulting in a vulnerability score of 6. This highlights the importance of implementing appropriate adaptive strategies to mitigate the effects of climate change.

Furthermore, sensitivity scores have a direct impact on potential impact, as in the case of sub-catchment 6. This underscores the role of socioeconomic factors and physical basin attributes in either intensifying or mitigating climate change impacts.

Table 11 shows the calculated STFIR, SFDAP, and SFFRI indices using the formula mentioned for all sub-catchments.

Based on these results, some observations can be drawn:

• The STFIR values indicate that all sub-catchments except sub-catchments 1, 2, 8, and 9 are expected to experience a decrease in flood intensity in the future due to climate change impacts. Sub-catchment 9 has the highest STFIR value of 1.9, which means its flood intensity is projected to increase compared to historical levels.

• The UH values reflect the extent of urbanization within each sub-catchment. Sub-catchment 4 has the highest UH value of 0.94, whereas sub-catchment 8 has the lowest urbanization ratio.

• The RD values represent the ratio of road density in each sub-catchment, which has implications for drainage capacity and the speed of flood during flash flood events. Sub-catchment 5 has the highest RD value of 2.31 which can increase its flash flood risk. It can be noted that sub-catchments 8 and 9 have the lowest RD.

• The SFDAP values indicate the percentage of potential flood damage that can be averted in the future through the implementation of adaptation strategies that account for the impacts of climate change and urbanization growth. Generally, all sub-catchments show high potential for reducing flood damage through adaption measures. Sub-catchments 6 and 5 have the lowest SFDAP value of 31%, indicating a pressing need for increased attention and focused efforts in implementing flash flood adaptation measures.

• The SFFRI values provide an overall weighted index of flash flood risk. Sub-catchment 9 has the highest SFFRI value of 1.27, which means it has the highest flash flood risk among all sub-catchments.

Morphometric analysis revealed that the study area watershed exhibits physical characteristics that increase its vulnerability to flash floods. With a circular shape and low drainage density, the watershed is susceptible to experiencing high peak flows and more intense flooding. After analyzing the rainfall trends and characteristics in the study area the average total annual rainfall in the study area is approximately 254 mm. High elevations show high levels of rainfall whereas low elevations are associated with low levels of rainfall. Analysis of rainfall data from 11 rainfall gauge stations indicated the highest total annual rainfall at 517 mm for the Wadi Es-Sir station and the lowest at 127 mm for Zarqa stations. Additionally, the spatial rainfall distribution shows that the rainfall increases towards the west and decreases towards the east. Our results show a decrease of 14.61% in total annual rainfall within the study area between 1970 and 2018 and the maximum daily rainfall and number of rainy days tend to decrease across all stations; this indicates that climate change has affected the area. Furthermore, the average temperatures showed an overall increasing trend in the mean, maximum, and minimum values of 7.3, 5.8, and 10.9%, respectively. In this respect, the average minimum temperature demonstrated the most significant rate of increase, a finding that agrees with a separate study on climate change trends in Jordan (Hamdi et al. 2009).

To assess the characteristics of the study area that contribute to its flash flood vulnerability, the results showed that land-use patterns have changed in the past twenty years. Urban areas expanded by 14.76%, resulting in a net increase of 97.25 km² during this period. This growth came at the expense of reductions in bare soil and cultivated land cover, as more land was transformed into urban areas. The rise in urbanization was primarily driven by a rapid population increase due to an influx of refugees and internal migration from other Jordanian cities and towns. The projected land use in 2041 shows a dramatic increase in urban land use with 51% of the study area expected to be covered by infrastructure, because of projected population growth. As the study area undergoes more urbanization, the demand for a robust stormwater drainage system becomes increasingly critical to manage the runoff that flows through streets. However, it also poses a challenge to groundwater recharge, as increased impervious surfaces hinder water infiltration. Consequently, the shift towards urbanization could exacerbate the impacts of climate change in the study area, particularly concerning flash floods, as its main drainage streams are covered by roadways, thus increasing the area's susceptibility.

The GCM data were used to analyze the future climate data to reveal the expected trends in rainfall and mean temperature. The GCM results indicate a slight increase in total annual rainfall of 5.26%, although the rate is projected to decline over time. Specifically, the amount of rainfall is expected to increase in 3 of the 11 stations and decrease in the other 8. However, since the stations with increased rainfall are located downstream and only a few in number, their overall impact on runoff is limited. Despite the projected increase in urban cover, which can affect the land's capacity to absorb runoff, the runoff simulation results indicate a notable decrease in both runoff volume and peak discharge across most sub-catchments in the study area due to projected lower rainfall. This situation holds significant implications for water availability in an area already facing water scarcity. The mean temperature is expected to increase from between 18.4° and 19.4° representing a 11.7% increase.

According to the AHP, rainfall is the primary controlling factor influencing flash flood incidents. It carries the highest weight as compared to other factors (runoff, population density, urban extent, distance to roads, and mean slope). Thus, the flash flood vulnerability is highly related to the rainfall rates in the observed and future periods.

We assessed the flash flood vulnerability and located the most vulnerable site in the study area. The assessment of the degree of exposure to climate changes indicated that sub-catchments 9 and 1 are the most exposed to flooding given the expected increase in rainfall and runoff. In contrast, sub-catchments 4 and 5, located upstream, are expected to experience a dramatic drop in rainfall and runoff rates, leaving them susceptible to water shortage with important implications for agriculture activities.

Sub-catchment 6 ranked the highest in terms of sensitivity to flash floods because it has the highest population density. However, given this catchment's higher level of adaptive capacity attributed to the type of infrastructure present and low exposure due to lower rainfall, it is not considered vulnerable to flooding. After combining the exposure and sensitivity, the potential impact map was generated. Sub-catchment 1 scored highest in terms of the potential impact of flash floods. This was attributed to higher recorded levels of rainfall and runoff in a highly urbanized context where there are many roads built along the drainage stream and a high mean slope which impacts the area's capacity to drain water. Further, the adaptive capacity map showed that sub-catchments 8 and 9 have the least capacity to cope with flash floods given the low level of vegetation to absorb excessive runoff and insufficient stormwater drainage infrastructure.

Overall, the vulnerability of flash flood occurrence under future climate change is either low or moderate. Sub-catchments 8 and 9 are the most vulnerable sites because of lack of adaptive capacity due to low-level vegetation cover and poor stormwater drainage, as well as high exposure to flooding due to the expected increase in rainfall and runoff. The index findings indicate that sub-catchment 9 has the highest risk of flash floods. However, when we considered additional indicators beyond just urbanization ratio and road density, such as population density, slope, and adaptive capacity indicators, we found that sub-catchment 8 also experiences a high potential of flash flood risk. Sub-catchments 2, 3, and 4 have low vulnerability as they are projected to be less exposed to flooding over time due to the expected decrease in rainfall. These results are consistent across both the CMCC-CM2-SR5 model and the EC-Earth3-Veg model. The results of this study indicate that while climate change has already affected the study area, heavy rainfall events that lead to flash floods are expected to occur less in the future. At the same time, the study area will suffer greatly from lack of water with ongoing implications for social, economic, and environmental well-being.

The findings of our study were compared to previous studies conducted in the same region and other regions in Jordan. In alignment with prior research, our study confirmed that the southwestern part of the AZB is expected to experience a slight increase in precipitation and a rise in minimum temperatures. Additionally, western upstream sub-catchments will have a noticeable rainfall reduction. Similar conclusions were drawn by Al-Hasani et al. (2023) in their study to assess climate change's impact on surface water in the AZB. However, regional differences emerged within Jordan. While Al-Taani et al. (2023) highlighted the importance of slope in flash flood vulnerability in Aqaba City, our research emphasized the role of rainfall intensity as the primary factor. These differences emphasize that flash flood vulnerability is context-dependent, requiring tailored mitigation strategies.

The methodology utilized in this study can be transferred and applied to evaluate hydrological responses to climate change in urban watersheds within arid regions, particularly in Jordan. It's worth noting that limitations in data availability may hinder a comprehensive evaluation of flood vulnerability. To enhance future assessments, we recommend the inclusion of socioeconomic data and community resilience evaluations. These additions will contribute to a more comprehensive analysis, ultimately improving decision-making regarding site-specific adaptation measures.

This study aimed to assess the hydrologic vulnerability of the AZB, as it represents an important social and economic region in Jordan, to climate change. The study area includes 6 million inhabitants and most of the country's industries; it has also been suffering from flash floods for years. The study was carried out by analyzing climatic data and using GCM to predict climatic conditions. In summary, results from this study indicated a decrease in the overall rainfall for the period 1970–2018 and an increase in temperature for 1988–2016. Land use has also changed in the study area, becoming increasingly urbanized over the past 20 years; this is expected to continue, although at a reduced rate, with approximately half of the study area projected to be urban land by the year 2041.

Our results also revealed that the projected data derived from the CMCC-CM2-SR5 model show that the study area is expected to witness an increase in rainfall of about 5.26%. However, given the expected increase in rainfall will only take place in three stations that are located downstream, whereas the upstream stations will witness a decrease in rainfall and peak runoff discharge is expected to decrease in most of the study area sub-catchments. Moreover, sub-catchments 9 and 8 are the most vulnerable sites to flash floods; however, the highest vulnerability is considered moderate as the most exposed sub-catchments have a low sensitivity given their low population density and urbanization. To mitigate flash flood vulnerability in these two sub-catchments, we recommend enhancing the infrastructure of the stormwater drainage network and constructing streamways to manage excess floodwater.

While the vulnerability to future flash floods in the study area is assessed as moderate, it remains highly susceptible to the effects of climate change. The main populated sub-catchments in the study area are projected to experience a large drop in rainfall and runoff rates, leading to reduced water availability and groundwater recharge. To address these challenges, implementation of adaptation strategies is crucial. This may include the adoption of water harvesting systems, particularly in sub-catchments with the highest rainfall rates, promoting water-saving innovations in irrigation and wastewater treatment, and raising public awareness about rationing water use. It is recommended to take this study's findings into consideration in the national climate change adaptation plans and strategies.

The authors would like to thank the anonymous reviewers for their useful suggestions. The authors thank the Ministry of Water and Irrigation and Greater Amman Municipality for providing the necessary climatic data. We would like to thank RICCAR for providing the climate data. We would like to show our gratitude to Dr Mohmmad Al-Qinna for his valuable suggestions. Special thanks are going to Prof. Dr Ghazi Saffarini and Prof. Dr Jason Rech from Miami University at Oxford for reading and improving the manuscript. This research has been accomplished during the sabbatical leave offered to Prof. Dr Mustafa Al Kuisi from the University of Jordan from October 2021 to September 2022.

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

The authors declare there is no conflict.