The above questions are explored through hydrological modeling of Ramot Menashe, a rural region in the eastern Mediterranean, characterized by hilly terrain that gradually rises from west to east (20–400 ma.s.l. over a 13 km distance), with moderate to steep slopes of 3–45°. The region is drained by small to medium-sized basins; four of these, with areas of 18–69 km2 are gauged: Upper and Lower Dalya, Upper Taninim, and Ada streams (Fig. 1; Table 1). The lithology of the basins is dominated by the chalks of the lower Eocene Adulam formation. In the southeastern part of the region, chalks and marls of the Paleocene Taqiye formation are exposed (Grodek et al., 2012; Peleg et al., 2015; Sneh et al., 1998). Rendzina soils (Singer, 2007) of average depth of 30±10 cm cover the area, except at distinct ridge-top locations where soil thins or is completely absent, exposing patches of bedrock (Dan et al., 1975; Grodek et al., 2012). Channel width range from 2.3 m at the stream heads to 12.2 m at the outlets, with maximum channel depths reaching 3.6 m (Rinat et al., 2018).

Land use in the Ramot Menashe region remained relatively stable until the 1990s. The landscape is composed of uncultivated (39.2 %) and cultivated (13.4 %) fields, groves and forests (26.3 %), plantations (12.2 %), agricultural villages (6.2 %) characterized by a rural residential area of small, sparsely distributed family houses and a local industrial zone, urban/paved section (1.4 %), and others (1.3 %) (Rinat et al., 2018). The settlements in the region expanded in recent decades by approximately 20 %. In 1995 a power station of 0.5 km2 was established. In 2015 a new urban neighborhood was established at the north eastern part of the region and is expected to double its size in the future. In 2017 an industrial region of 0.5 km2 was established at the north eastern part of the region and is also expected to double its size in the future.

Rainfall occurs in the region mainly between October and May. Most of this rain (60 %–70 %) falls during the winter, from December to February. Mean annual rainfall increases from the lower western side of the region (586 mm) to the higher eastern side (774 mm) (1991–2020, IMS, 2025). The mean annual potential evapotranspiration is ∼1900 mm, where during the winter, potential evapotranspiration is approximately 70 mm per month (Goldreich, 2003). Rainstorms are mainly (94 %) of Mediterranean origin (Peleg and Morin, 2012), namely Cyprus or Mediterranean lows, and usually last several hours to a few days (Goldreich, 2003). The study area is characterized by ephemeral streams that experience, on average, 7 flood events per year. These floods are typically initiated within hours of intense rainfall and last for approximately 3 d. The maximum recorded specific peak discharge (1950–2024) is 3.3 m3s-1km-2, and extreme floods result in damage and casualties, notably in 2001, 2006 and 2020 (Grodek et al., 2012; Inbar, 2019; Morin et al., 2007). Outlet discharge from the Upper Taninim and Ada reservoirs is directed to an infiltration recharge basin, resulting in an estimated annual contribution of 10×106 m3 to groundwater recharge (Kurtzman and Guttman, 2020).

The event initial soil moisture conditions were computed using the DREAM water-balance model (Sheffer et al., 2010) similar to Rinat et al. (2018). DREAM is a continuous daily model that was originally designed and calibrated for water recharge estimation in the Israel Western Mountain Aquifer, which covers the study area. Its inputs include daily rain data and daily potential evapotranspiration and its outputs are daily recharge, runoff, actual evapotranspiration, and soil water storage. Here we used the soil water storage output for the day preceding the beginning of each event to set the initial conditions.

For the historical scenario, the DREAM model was applied similarly to Rinat et al. (2018), using measured average daily rainfall from 25 nearby rain gauges and daily climatological potential evapotranspiration derived from monthly data for a nearby location (Goldreich, 2003). For the future scenario, we estimated the mean difference in the climatological soil water storage between the future and historical conditions for every day in the year and added this difference to the historical initial soil water storage. This is done using the following steps: (1) A two parameter Weibull distribution was fitted to a long daily rainfall record (1938–2011; IMS, 2025) from a rain gauge within the study area, at Ein Hashofet (32.5941°, 35.0995°). We then randomly sampled the acquired distribution to produce 1000 years of synthetic daily rainfall data for the historical climate. (2) We altered the Weibull parameters to describe the expected future reduction in annual rainfall depth (33 %) and in the number of rainy days (20 %) to the end of the century (Hochman et al., 2018; Zappa et al., 2015), following the approach by Marra et al. (2021). (3) Using the altered distribution, we generated 1000 years of synthetic daily future rainfall data. (4) We then used the synthetic historical and future rainfall data and potential evapotranspiration data as an input to the DREAM model to calculate two synthetic series of 1000 years of historical and future daily soil water content. (5) We calculated the climatological daily soil water storage for every day in the year for the future and historical conditions and then subtracted the two. (6) Finally, for each event, we added the difference to correct the initial soil water storage value from the historical value to the future one. On average, future values of soil water storage showed a 17 % decrease relative to historical values.

To simulate the hydrological response to climate change and urbanization we applied the GB-HYDRA, a high-resolution (100 m2, <60 s) event and grid-based hydrological model that was validated against measurements and used in a previous research in the study area (Rinat et al., 2018). The model simulates various hillslope and stream runoff processes.

Model inputs include, for each event, rainfall data at 10 min and 1 km2 resolution, basin-averaged daily potential evapotranspiration and basin-averaged initial soil water storage. The model outputs used in this study include: flood discharge at the basin outlet and at each stream grid cell, and basin-averaged soil water storage. For further information on the model and its application refer to Rinat et al. (2018). Climate change is simulated in the model by altering the model inputs between historical and future conditions. Urbanization is represented by converting parts of the catchment area to urban land use (Fig. 1b and c; Sect. 3.1).

To analyze the combined spatial hydrological dynamics for all rainfall events and shifts we calculated the normalized change of each scenario relative to the Historical benchmark (Sect. 4.4.3). The following procedure was applied for each stream pixel: (1) For each event we randomly sampled a pair of simulated peak discharge values from the 43 shifts. One from the examined scenario and one from the Historical scenario (not necessarily from the same shift). (2) To avoid division by near-zero values we defined a discharge threshold of 0.01 m3s-1. When both values were below this threshold, the relative change was set to zero. If only the historical value was below the threshold, the sample was excluded. (3) When both values exceeded the threshold, we calculated the relative change between the two. (4) This procedure was repeated 5000 times, and per-pixel median was taken for each of the 32 rainstorm events. (5) Finally, for each pixel we calculated the percentiles of all relative changes to obtain the 10th, 50th, and 90th percentiles (Fig. 6). The median (50th percentile) represents the typical relative change between the tested scenario and the Historical one. The 10th and 90th percentiles are, by definition, biased towards cases of lower and higher discharge in the tested scenarios compared to the Historical one, respectively. We present these two percentiles here to emphasize such cases.

Similar to other Mediterranean regions, Ramot Menashe is expected to experience land use transformation, shifting from natural and uncultivated areas towards more intensive agricultural and urban development. To assess these changes we utilized the 2040 strategic development plan of Israel Planning Administration (IPA, 2025) to create a future land use map. While the region is projected to experience some urbanization until 2040, it is still one of the most naturally-conserved regions in Israel, with most of the land use changes expected near or within existing villages. However, given the region's population increase and to demonstrate the potential flood changes related to urbanization even after 2040, we incorporated a hypothetical, but realistic, urban area (5.3 km2) within the Upper Dalya basin, closely mimicking Harish, a small town established in 2010 in a nearby region (Fig. 1; approximately 6 km south to Ada basin). This town is considered as the fastest growing town in Israel, extending from just 0.1 km2 in 2016 to 3.5 km2 in 2019 (Schwake, 2020). Urban sections are represented as homogeneous regions with high curve number parameter values (Table 1), without an urban drainage network or any runoff mitigation measures. These changes resulted in an overall increase in the urban area coverage across all basins with a considerable increase from 0.42–8.4 km2 in the Upper Dalya basin (Table 1; Fig. 1c).

Different rainfall properties are associated with the magnitude and extent of flash floods (Doswell et al., 1996). Here, we identified changes in four flood-related rainfall properties, roughly following the properties identified in Armon et al. (2022). These changes were computed by comparing the change in the spatial ensemble medians of the 32 analyzed rainstorm events over the four studied basins. Given that results are similar between the four basins, we present here results for the Upper Dalya basin (Fig. 2); results for the other three basins are described in the Supplement.

Median ensemble values for accumulated precipitation decreased in 29 of the 32 rainstorms (90.6 %) with a significant average decrease of 24.5 % (Fig. 2a). Significant reductions were also observed in the median values of the rainfall area above 0.1 mmh-1 for 10 min (38.1 %, Fig. 2b) and of storm duration (16.9 %, Fig. 2c), where 96.9 % and 81.2 % of the events, respectively, exhibited a reduction. Conversely, maximum 30 min rain rates exhibited a non-significant increase of 13.1 %, where increases were present in 65.6 % of the events (Fig. 2d). The contrasting trends of decreased accumulated precipitation, rainfall area and storm duration compared to increased maximal rain rate make it hard to foretell the impact of changing rainfall properties on flood characteristics.

Different scenarios were used to assess the impact of individual and joint climate and land use changes on future hydrological properties (Fig. 3):

Urban: in this scenario we evaluate the isolated impact of future urbanization by representing late 21st century land use projections while maintaining late 20th century conditions for rainstorms, initial soil water storage, and potential evapotranspiration.

Projected changes in outlet peak discharge, outlet flood volume, and soil water storage during peak discharge (peak soil water storage) were evaluated by comparing the four scenarios – Urban, Future-rain, Future-rain + Urban, and Future to the Historical scenario (Fig. 4). For each of the discrete 32 events we use the median value among all 43 spatial ensemble shifts and utilize the t-test and Wilcoxon signed rank test (both with significance level of 0.05) to examine whether the hydrological properties under each of the four scenarios significantly differ from the Historical scenario (Table 2; Fig. 4). It is important to emphasize that these event-based results do not capture potential shifts in flood frequency or seasonal dynamics.

To better understand why future floods in eastern Mediterranean regions can exhibit contrasting trends under different scenarios, we analyze the temporal dynamics of two representative flood response types, named here: undersaturated and saturated (Fig. 5). In both types, future rainfall depth decreases while maximal rainfall intensity increases. The main difference between the two types lies in the Future basin-averaged soil water storage for the shift that resulted in median outlet peak discharge (bold orange line in Fig. 5). In undersaturated events (peak soil water storage in Future rain scenario ≤90 %), Future-rain soil water storage remains undersaturated throughout the storm, while in saturated events (peak soil water storage in Future-rain scenario >90 %) soil water storage reaches high saturation rate before the peak discharge in all scenarios. This difference results in different hydrological response. Undersaturated events account for most of analyzed storms – approximately 62 %, and saturated events represent approximately 28 % of the storms.

To further understand the hydrological dynamics within the basin, we calculated the normalized change of each scenario: Future-rain, Urban, and Future-rain + Urban, relative to the Historical benchmark. Results from the Future scenario are not presented as they were not significantly different from those of the Future-rain + Urban one. We first compare results from undersaturated and saturated events (see Sect. 4.3) for a single storm-shifting for each scenario corresponding to the median outlet peak discharge (bold lines in Fig. 5). This approach enables a spatial comparison between two individual rainstorms.

The Mediterranean region is known as a climate change hot-spot where annual precipitation decreases but rainfall intensities increase. As a result, high-impact floods are often reported in the media as climate change related (Al-Saqaf and Berglez, 2019; Cortiços and Duarte, 2025; Cremonini et al., 2024; Llasat et al., 2009). However, this is not always the case (e.g., Armon et al., 2025), in fact, recent studies around the world (Ionno et al., 2024; Meresa et al., 2022; Pal et al., 2023; Taniguchi et al., 2022; Wasko et al., 2021) including in the Mediterranean (Alfieri et al., 2015; Blöschl et al., 2019; Moore et al., 2006; Poncet et al., 2024; Di Sante et al., 2021; Wasko and Sharma, 2017) show current and future contradictory flood trends.

These contradictory trends are a direct result of changes in future properties that act in opposite ways to counterbalance or support flood intensification. Some trends in rainfall properties act to suppress flood peaks and volumes, these include: decreased rainstorm frequency, shortening of the rainfall season and event duration, reduction of annual and event rainfall depth, reduced rainfall area, and prolongation of the inter-event time (e.g., Alfieri et al., 2015; De Girolamo et al., 2022; Peleg et al., 2015). Soil water storage levels are reduced due to the same rainfall trends as well as the increased evapotranspiration rates and add to this damping (e.g., Mediero et al., 2014; Tramblay et al., 2019, 2023; Wasko et al., 2021). In contrast to these suppressing effects, rain rate intensification, urbanization, and soil erosion can act to increase flood magnitudes (e.g., Faccini et al., 2021; López-Moreno et al., 2006; Wasko et al., 2021). In urban regions rain intensity becomes the dominant factor; driving an increase in runoff magnitude (Yang et al., 2016). Finally, agricultural expansion and poor cultivation practices degrade soil (Amundson et al., 2015), reduce water storage and contribute to runoff intensification (Hillel, 1998).

The regional and local effects of these counterbalancing factors on flood trends are reported in recent studies. Blöschl et al. (2019) and Tramblay et al. (2023) found a decrease in observed flood intensities due to decrease in soil water storage, mainly for large agricultural basins with low urbanization (Tramblay et al., 2019). Bertola et al. (2020), analyzed European flood records and found that flood trends depend on the region, magnitude, and basin-size. They found that, in general, Mediterranean floods peak discharges are decreasing, while the trend is smaller at extreme floods or small basins. Poncet et al. (2024) studied projected floods in a medium size basin at southern France and found intensification of large floods but magnitude decrease in smaller ones as a result of low soil water storage and base flow reduction. Di Sante et al. (2021) found that daily peak flows are mostly decreasing, although some Mediterranean regions, such as: southern Spain, southern Italy, and Israel show a non-significant increasing signal. They also state that extreme floods generally increase in medium (100–1000 km2) basins and decrease for larger ones.

In this study, we find that, in the eastern Mediterranean, future rainfall tends to reduce flood peak discharge magnitude and volume and that urbanization acts to increase it. While both factors act in opposite directions the influence of urbanization in our research setting is higher even when initial soil water reduction and increased potential evapotranspiration are accounted for (Future scenario in Fig. 4). This is supported by various studies, for example: Ramezani et al. (2023) studied the effect of climate change and urbanization in 6 medium basins in eastern Australia and showed that climate change mostly results in runoff reduction, while urbanization causes runoff to considerably increase. When both scenarios are combined climate change only moderates the urbanization impact (Kalantari et al., 2017). However, in less urbanized basins, when different runoff management actions are taken (Ahammed, 2017; Chen et al., 2009), or in other regions climate change effects might become more pronounced. For example, Arnone et al. (2018) found that runoff extremes are dominated by variations in climate change even when urban regions increase from 1.5 % to 25 % basin coverage.

This interaction between local rainfall projections and environmental change can lead to different trends between and within regions, between neighboring basins (Tramblay et al., 2019), or even within a basin (Sect. 4.4; Figs. 6 and 7). Further caution is needed as these complex behaviors must be taken under consideration when preparing for future flood hazards (Llasat, 2021). In this study we used a unique high-resolution distributed hydrological model that uncovers inter- and intra-basin flood response variability. We find that as future rainfall is projected to become even more localized, intense basin headwater sections can experience increased local runoff. At the same time, such a scenario does not necessarily translate to increased outlet discharge. However, it can cause local flooding, erosion at headwater sections, debris flow and landslides (Grodek et al., 2012; Shmilovitz et al., 2023). When intense localized rainfall occurs in urbanized areas, or when preferable settings, such as high soil water storage (Saturated events, Figs. 5 and 6) prevail it can result in extreme hydrological response (Morin and Yakir, 2014). We therefore encourage researchers, decision makers and stakeholders to consider the different and combined effects climate change and land use change have on floods within different parts of the basin when preparing for future challenges.

Recent studies in the Mediterranean have analyzed observed data to identify flood trends. In the present study, such trends could not be reliably determined due to limited, non-homogeneous flood measurements and the influence of ongoing land use changes. Moreover, observed flood trends do not necessarily reflect future conditions (Blöschl et al., 2019) as these require the use of meteorological and hydrological models (Hall et al., 2014). Future flood projections are subject to significant uncertainties due to long-term simulation periods and complex modeling chains (Noto et al., 2023b). In addition, the coarse spatiotemporal resolution of global or regional meteorological models fails to accurately represent convective rainfall and the impact of mountain ranges and coastlines that are typical to the Mediterranean (Armon et al., 2022; Berthou et al., 2020; Marra et al., 2022; Noto et al., 2023b; Prein et al., 2015). Similarly, large-scale or even higher resolution semi-distributed hydrological models struggle to capture the Mediterranean's intra-basin spatiotemporal variability, particularly in small to medium-sized basins (Fatichi et al., 2016; Merheb et al., 2016).

To mitigate these limitations high-resolution convection permitting models and detailed hydrological models are employed. Convection permitting weather models better represent extreme short-duration rainfall and climate change scenarios (Poncet et al., 2024) which are crucial for hydrological studies (Ascott et al., 2023) particularly at small to medium basins and urban regions (Arnaud et al., 2011; Cristiano et al., 2017; Dayan and Morin, 2006; Paschalis et al., 2014; Singh, 1997; Yakir and Morin, 2011; Zoccatelli et al., 2011). Subsequently, high-resolution hydrological models are utilized to account for complex hydrological processes driven by high spatiotemporal rainfall and basin characteristics (Downer et al., 2002; Fatichi et al., 2016; Merheb et al., 2016; Moretti and Montanari, 2007; Rinat et al., 2018). However, despite their proven ability to accurately simulate intense rainfall and heavy precipitation events, convection permitting models remain underutilized in hydrological studies, especially in the Mediterranean region (Peleg et al., 2020; Poncet et al., 2024).

Long-term (∼10 years) high-resolution simulations of climate change are becoming available for limited regions (e.g., Berthou et al., 2020; Liu et al., 2017). However, their use remains limited because they are available for only a few regions and typically cover short time periods. As a result, large parts of the world, including the eastern Mediterranean, are still excluded from such simulations. We thus employed an event-based pseudo-global warming methodology, which offers cost-effective, high-resolution simulation of climate change impact on precipitation at a relatively low-cost, but contains only a limited number of events and does not allow for frequency analysis. In addition, the present study relies on a single high-emissions (RCP8.5) climate forcing based on the CMIP5 ensemble mean. Consequently, the results should not be interpreted as probabilistic or precise quantitative projections of future flood response, but rather as an assessment of the direction and relative hydrological impacts of climate and land-use changes under a strong warming scenario (e.g., Van Vuuren et al., 2026). Lower-emission scenarios would likely result in weaker climate-change signals. The use of multiple emission scenarios and climate forcings would enable probabilistic assessment of future flood response, but such convection-permitting ensemble simulations remain computationally demanding and are currently unavailable for the eastern Mediterranean region.

Moreover, events were identified based on heavy precipitation across the broader eastern Mediterranean region, not the specific study site. Consequently, the tested extreme rainfall series is incomplete and does not fully represent the range of potential scenarios. Comprehensive long-term, high-resolution simulations are essential for a full representation of the hydrometeorological processes, including: multi-year, annual, seasonal, and extreme/low-value analyses. Such analyses are crucial for flood impact analysis, mitigation, and prevention (Llasat, 2021). We intend to conduct complementary research in the eastern Mediterranean to study the projected flood frequencies due to climate and environmental changes in the region based on future availability of longer-term model output.

Water resources in the Mediterranean are limited (Milano et al., 2013). Agriculture dominates water consumption, accounting for 64 % of usage, followed by domestic (20 %) and industrial needs (16 %) (Burak and Margat, 2016). Projected increases in water demand are attributed to population growth, rising living standards, expanded irrigation practices, rapid urbanization, and even increased tourism (García-Ruiz et al., 2011; Lana-Renault et al., 2020; Milano et al., 2013; Zdruli, 2014).

Even though most of the Mediterranean streams are seasonal or ephemeral (Bonada and Resh, 2013; Skoulikidis et al., 2017) they play a significant role in contributing to renewable water resources (Masseroni et al., 2021). In particular, heavy precipitation events play a crucial role enriching groundwater, reservoirs, and lakes (Armon et al., 2024; Milewski et al., 2009; Moawad et al., 2016; Nasta et al., 2018; Samuels et al., 2009).

In recent decades annual stream volumes in the Mediterranean have decreased in a rate of approximately 1×103 m3km-2yr-2 (Masseroni et al., 2021), leading to reduced lake levels and diminished renewable water resources (Myronidis et al., 2012). Projections indicate that this trend will persist, with decreasing annual river runoff at large basins and reduced low-flows (De Girolamo et al., 2022; Marx et al., 2018; Schneider et al., 2013; Sellami et al., 2016). Such changes are likely to negatively affect groundwater recharge (Braca et al., 2019) and result in further declines in lake and reservoir water levels (Bucak et al., 2017; Zabalza-Martínez et al., 2018).

To address increasing water demand and scarcity, numerous reservoirs have been constructed across the Mediterranean in recent years to capture stream runoff and mitigate groundwater recharge reductions (Skoulikidis et al., 2017). Since the late 1960s, at the Ramot Menashe study site stream flow is diverted to several small reservoirs for irrigation. Additionally, approximately 10×106 m3 is annually diverted from the downstream sections of Upper Taninim and Ada basins to the Menashe streams managed aquifer recharge system. While this volume constitutes only ∼0.6 % of Israel's total estimated natural water resources, these readily available waters are critical to the region's water resilience (Kurtzman and Guttman, 2020).

Our findings reveal that in small to medium-sized Mediterranean basins future rainfall scenarios translate to a significant decrease in flood volumes (averaging -30 %). However, given our experiment design, substantial urban development is projected to result in a 41 % increase in flood volumes (mean increase; Table 2, Fig. 4). When the combined effects of future rainfall and urbanization are considered, mean flood volumes remain rather unchanged. When accounting for all factors – urbanization, future rainfall, future initial soil water storage, and future potential evapotranspiration – flood volumes are expected to decline slightly (averaging -8 %).

Our results highlight the critical need for adaptive water management strategies. In rural regions, projected declines in stream volumes necessitate the adoption of sustainable and water-efficient agricultural practices and the development of alternative water sources such as wastewater treatment (Ait-Mouheb et al., 2020) and seawater desalination (Curto et al., 2021). Conversely, urban areas face heightened vulnerability due to increased flood magnitudes, particularly during extreme precipitation events (Nussbaum et al., 2026). This urban hydrological pressure is exemplified in Israel which has experienced rapid population growth over the last eight decades (Orenstein and Hamburg, 2010; Tal, 2016), that is projected to persist (Phren and Klinger, 2017).

Historically, urban flood hazard mitigation has exclusively relied on the rapid centralized evacuation of stormwater. However, during intense rainfall, these drainage systems often contribute to urban flooding and increase downstream peak flows (Fletcher et al., 2024; Xu et al., 2023). To mitigate these risks, other measures have been developed and implemented in recent decades. Nature-based solutions such as: urban green parks, urban forests, green roofs, and green walls attenuate runoff, while detention and retention ponds store stormwater to reduce peak discharges and volumes and increase hydrological lag times (Ben Dor et al., 2025; Esraz-Ul-Zannat et al., 2024; Fletcher et al., 2013, 2024; Xu et al., 2023). Infiltration and groundwater recharge can be enhanced by using permeable pavements and rooftop rainwater harvesting (e.g., Esraz-Ul-Zannat et al., 2024; Nachshon et al., 2016; Netzer et al., 2024; Saha et al., 2024). However, the efficacy of these infiltration-based interventions is highly dependent on local geology; for example, they may be ineffective in our study site, which is characterized by thick, impermeable bedrock (Kurtzman and Guttman, 2020).