Assessing future flood hazards with high-resolution climate model projections: a case study in northeastern Sicily, Italy

Abstract

Flooding is one of the most frequent natural disasters with extensive impacts on society and the environment. Under projected climate conditions in the future, the occurrence of severe floods may increase. Therefore, assessing and increasing awareness of flood hazards is crucial for disaster risk reduction and adaptation planning. However, evaluating flood hazards under past and future extreme rainfall involves considerable uncertainties, as climate data at the required temporal and spatial resolution is most often missing. In this study, we use daily and sub-daily in situ observations (2001–2022) and projected hourly rainfall from an ensemble of high-resolution projections, exemplified for a coastal region in Sicily, Italy. We estimate the projected rainfall changes by using an ensemble of REMO2015 regional climate projections at a spatial resolution of 12.5 km and an hourly time-step, driven by 8 different CMIP5 general circulation models. We considered the historical climate (1981–2010) and the RCP8.5 scenario (2031–2060) to establish historical and near-future rainfall depth-duration-frequency (DDF) curves. With these data, we drive the hydrological model (HEC-HMS) for short-duration design rainfall events of 6 and 12 h in small urban coastal catchments in northeastern Sicily (Italy). The resulting flood depths, inundation area, and flow velocities were obtained from a 1D/2D hydrodynamic model (HEC-RAS) for 100-year design rainfalls. Our results show that future rainfall extremes are projected to be more frequent and severe in the study area, with the 100-year rainfall event increasing by 20% on average. This leads to increasing flood hazards compared to the baseline period, with inundation depth and velocity increasing by 9 and 12% on average in the study region. Our findings reveal that the regional climate model sub-daily rainfall data can reproduce extreme rainfall events obtained from observations despite some limitations, increasing confidence in future projections at this resolution.

Introduction

Flooding is one of the most frequent and destructive natural disasters that significantly impacts both the environment and the socio-economic well-being of society (Bouwer, 2019; CRED, 2020; Tellman et al., 2021; Kreibich et al., 2022; Jonkman et al., 2024). According to the CRED’s Emergency Events Database (EM-DAT), floods accounted for 44% of all natural disasters that occurred between 2000 and 2019 and affected 1.6 billion people around the world, being the and most destructive disaster (CRED, 2020). Flood disasters have intensified over the recent past (CRED, 2020; Shamsudduha, 2025) and are closely linked with climate change and increased human activities such as new settlements, infrastructure development, and urbanization (Han et al., 2020; Ngo et al., 2020; Janizadeh et al., 2021).

Based on current observations, the global climate is changing due to increasing greenhouse gas emissions, and under projected future climatic conditions, the occurrence of floods and their severity may increase, resulting in a myriad of possible impacts (Ranasinghe et al., 2023). In relation to flooding, changes in extreme rainfall are expected to lead to more frequent and more intense flooding (Atanga and Tankpa, 2021; Ghanbari et al., 2021; Hsiao et al., 2021; Habeeb and Bastidas-Arteaga, 2023; Liu et al., 2023). On the other hand, human activities (e.g., urbanization, industrialization, and population growth) have directly caused land cover changes, an increase in impermeable surfaces, and soil erosion and loss (Fernández and Lutz, 2010; Huong and Pathirana, 2013; Han et al., 2020; Ahmad and Afzal, 2022; Peker et al., 2024). Therefore, human activities exacerbate the above situation more than ever by aggravating the flood hazard, as well as increasing the exposure of assets and people. In this context, assessing and increasing awareness of flood hazards is crucial for disaster risk reduction and planning for adaptation measures.

Over the last few decades, the Mediterranean region has been experiencing adverse impacts of climate change and anthropogenic activities, being vulnerable to many extreme events such as extreme rainfall, flash floods, debris flow, and heat waves. Such events have led to considerable building and infrastructure damages, monetary losses, and loss of life (Aronica et al., 2012b). According to the Intergovernmental Panel on Climate Change (IPCC) report (Arias et al., 2021), an increase in high precipitation and resultant pluvial flood is projected in the Mediterranean regions for the mid-21st century with medium confidence. Mean precipitation, however, tends to decrease (high confidence), leading to water shortages. In contrast, there is high and medium confidence that hydrological and agricultural drought, respectively, have increased, with a clear upward trend observed in the region, attributed to human activities (Arias et al., 2021). The Mediterranean region will be hotter and drier, and some parts of the region will have wetter extremes. At the same time, rapid urbanization, deforestation, and increases in demand for natural resources have been observed and continue.

Italy is one of the hotspots for a growing number of extreme hydrological events. According to the European Severe Weather Database, in Italy, approximately 1,500 extreme weather events occurred in 2020 compared to 380 similar events in 2010 (Levantesi, 2021). Mazzoglio et al. (2022) summarized that the rainfall extremes of different durations (1 to 24 h) over at least 30 years in the past did not increase uniformly over Italy, and there are high variabilities of the trends. Furthermore, a recent study by Avino et al. (2024) shows an increased frequency of rainfall events with durations of less than 24 h over Southern Italy and particularly in the Calabria region from 1970 to 2020. This upward trend is inapplicable for longer rainfall durations. Similarly, a few studies carried out in Sicily have recognized the increases in upward trends in rainfall for durations of sub-hourly, while decreasing tendencies in longer durations (Arnone et al., 2013; Bonaccorso et al., 2020; Treppiedi et al., 2021). Therefore, assessment of future flooding due to wetter extremes, particularly at the sub-daily scale, is crucial in this region to reduce the risk and vulnerability to future climate impact drivers.

The translation of changes in rainfall patterns to alterations in flood hazards, including variations in frequency and intensity, and ultimately impact on flood risk, remains highly uncertain, given the uncertainties in rainfall projections, their scale, and detail, especially for shorter (sub-daily) durations. An important basis for many studies into changes in rainfall and hydrology is the use of climate projections from climate models. The best comprehensive information is resolution climate simulations from regional climate models that dynamically downscaled the climate from General Circulation Models (GCMs) under a range of emission scenarios. However, rainfall projections for the sub-daily scale are not readily available for most locations, mainly because of data access and storage limitations. It is, however, essential to study these high temporal and spatial resolution data for hydrological analysis, for instance in the form of Intensity-Duration-Frequency (IDF) curves. For Europe, the most comprehensive ensemble of climate projections is the EURO-CORDEX dataset at about 12.5 km x 12.5 km spatial resolution (Jacob et al., 2020). In recent years, convective permitting models at a very high resolution of 3 km are being used; however, this application also has its computational challenges, resulting in selective areas of study and limitations for hydrological studies (see, e.g., Wagner et al. (2025)) using projected temperature changes. Also, short-duration rainfall and related flooding require detailed, small-scale hydrological and hydraulic flood hazard assessments at the local scale, which is often not performed for most regions. Such basins are also generally ungauged, meaning that flood hazard analysis can only be performed by using models.

Flood risk assessments are important to prevent plausible cascading disasters and reduce the impact and damage. For this assessment, determining the hazards in terms of flood depth, velocities, and inundation areas is crucial. Therefore, determining these flood hazard indicators (viz., depth, velocities, and inundation areas) accurately and in a timely manner is essential. In recent years, the advancement in geographical information systems (GIS) and the integration with numerical hydrological and hydraulic modelling have significantly enhanced the accuracy of flood simulations, which is the basis for the evaluation of flood risk (Peker et al., 2024). Hydrological models simulate the flood discharge at the desired outlet(s) for rainfall events at shorter (a few hours 1 to 12 h) or longer time durations (1 day to a few weeks). A few examples are TOPMODEL (Beven and Freer, 2001), HBV (Bergström, 1976), MIKE-SHE (Refsgaard and Storm, 1995), RORB (Laurenson and Mein, 1983), and HEC-HMS (Hydrologic Engineering Center, 2023). Similarly, hydraulic models (1D, 2D, 1D/2D coupled, or 3D) provide detailed flood hazard indicators: water depth, levels, and velocities at a given location and inundation areas. A few examples are HEC-RAS (Hydrologic Engineering Center, 2020), FLO-2D (O’Brien et al., 1993), Delft3D (Kernkamp et al., 2011), MIKE 21 (DHI, 2017), and TELEMAC-2D (EDF-DRD, 2000). The selection of suitable modelling approaches depends on the aim of the study, data availability, and model structure (viz., resolution). Many successful research studies used these models in many diverse regions. However, many of the world’s river basins are ungauged or poorly gauged (Sivapalan et al., 2011); thus, flood modelling and risk assessments become challenging in many parts of the world.

Here, we evaluate flood hazard changes using high spatial and temporal resolution climate projection data from a single regional climate model, exemplified for a coastal location in Sicily, Italy. In this study, we aim to address the following questions: (1) How will shorter-duration extreme rainfall events change with respect to regional climate change projections in the future, and what are the uncertainties? and (2) How will these projected changes in extreme rainfall impact flood hazards in the same region? This study is a part of the “risk workflow for CAScading and COmpounding hazards in COastal urban areas” (CASCO) project, which enables the combination of the overall risk of several important natural hazards: floods, earthquakes, tsunamis, heat waves, and landslides in Sicily, Italy (https://www.climate-service-center.de/science/projects/detail/105032/index.php.en). Under this project, we tested the workflow for overall damage and losses due to the combined effects of different hazards.

Materials and methods

Study area

This study is focused on the two river basins (Mili and Santo Stefano Briga) belonging to the municipal territory of the city of Messina in northeastern Sicily, Italy (Figure 1). Many floods, subsequent impacts, and casualties have been recorded in this region recently. For example, in 2021, more than 300 mm of rain, which is nearly half of the average annual rainfall of the island, fell over Catania, East of Sicily, within a few hours (Levantesi, 2021; Vitanza et al., 2023). In 2009, approximately 225 mm of rainfall within nearly 8 h fell over eastern Sicily, causing lots of property damage and loss of lives in the Giampilieri Basin (Aronica et al., 2012b; Jalayer et al., 2018). In addition, these areas face other important natural hazards that are studied in the context of the same project mentioned before (CASCO project), namely earthquakes, tsunamis related to earthquakes, as well as submarine landslides, and heat wave events.

Figure 1

The Mili basin has a narrow and steep terrain covering an area of 5.88 km^2, and the main river, approximately 4.5 km long, drains to the Ionian Sea. The river passes through the villages of Mili San Pietro and Mili San Marco and the downstream town of Mili Marina (Figure 1). The average gradient of the Mili River is 11%, and the average slope of the basin is 52%. The average topography of the basin from the mean sea level (MSL) is 428 ± 262 m, and the highest altitude is 1,062 m. More than half of the basin area (~60%) is covered by sclerophyllous vegetation, followed by mixed forest (~ 20%) and mixed cropland (~15%). The major soil types found in the basin can be categorized as Cambisols and Leptosols (Fantappie et al., 2015).

The Santo Stefano Briga basin covers an area of 16.65 km². The main river (Santo Stefano) has a length of ~7 km and a gradient of 10% and drains to the Ionian Sea after passing through the major villages and towns of S. Stefano di Briga, S. Stefano Medio, Santa Margherita, and S. Stefano Marina (neighboring towns of S. Stefano Briga) (Figure 1). The basin slope varies between 0 and 1,081 m. MSL and average is 451 ± 268 m MSL. Approximately 37% of the land area is covered by sclerophyllous vegetation, followed by mixed croplands (25%) and mixed forests (10%). Similar soil types to those in Mili Basin are found in this region (Fantappie et al., 2015).

Both basins have a typical Mediterranean climate (Augustyn, 2025). Heavy rainfall events with shorter durations and high intensities occur during the wet season (October to April), and a few rainfall events during the dry season (May – September) (Aronica et al., 2012b). The mean annual rainfall is approximately 900 mm, and ~80% and ~20% in the wet and dry seasons, respectively. The average temperature is 19 °C, with the highest temperatures in August and the lowest in February.

Both rivers are mostly non-perennial and active only during the rainy season (October to April). Therefore, some river segments are used as parking lots, access roads, and for other activities, which we observed during a field visit in November 2023. A lot of solid materials can be seen in the riverbed, particularly the sediment carried by the river flow during heavy rainfall. This can cause serious problems, such as overflow due to a lack of room and space for river flow. Furthermore, both basins are ungauged, and therefore neither water level nor discharge data are available.

Data

The geospatial data used consists of (1) elevation data, based on the 10 m x 10 m resolution Digital Elevation Model [DEM (Tarquini et al., 2007)] and a 2 m resolution DEM based on Lidar (Aronica et al., 2012b); (2) land-use based on 100 m resolution land use data from Copernicus Service Information (2018) CORINE Land Cover (CLC) 2018, version 20; and (3) soil information based on information on Soil regions of Sicily (Fantappie et al., 2015).

The observed rainfall data comprise daily and sub-daily rainfall data at five gauging stations (Figure 1) from 2002 to 2022, obtained from the Sicilian Agrometeorological Information Service (SIAS) in Sicily, Italy, and Jalayer et al. (2018). As we have limited length of observed rainfall data (~ 20 years), we also compared the SIAS rainfall stations data with longer records from (~40 years) high-resolution (0.05° to 0.1°) gridded precipitation products from the Multi-Source Weighted-Ensemble Precipitation dataset [MSWEP, Beck et al. (2019)], the Climate Hazards Group InfraRed Precipitation with Station dataset [CHIRPS, Funk et al. (2015)], and the ENSEMBLES daily gridded observational dataset for precipitation, temperature, and sea level pressure in Europe dataset [E-OBS, Cornes et al. (2018)]. The detailed analysis is presented in Supplementary Figure S1 and Supplementary Table S1. In general, E-OBS and CHIRPS gridded precipitation values outperform MSWEP in the analysis period (2002–2022).

Finally, we used simulated hourly rainfall from a subset of the EURO-CORDEX (Jacob et al., 2014) ensemble. This subset comprises 8 simulations from the regional climate model REMO2015 (Remedio et al., 2019) at 0.11 × 0.11-degree (~ 12.5 km x 12.5 km) spatial resolution from 8 driving GCMs (Table 1) using the Representative Concentration Pathways 8.5 (RCP 8.5) scenario. Although CMIP6 (the latest climate projections) global climate simulations offer improvements, the dynamic downscaling of CMIP6 simulations using regional climate models, especially for the European domain, is still emerging (see: https://wcrp-cordex.github.io/simulation-status/CORDEX_CMIP6_status.html#EUR-12). For many applications, especially in regional impact assessments, CMIP5-driven Regional Climate Models (RCMs) are still being used and provide robust insights (see, e.g., Ruosteenoja and Räisänen, 2024; Akperov et al., 2025; Boeing et al., 2025). As input to the hydrological model, we used the available hourly precipitation data from REMO2015. This model has been applied to study extreme rainfall events, demonstrating the model’s ability to reproduce observed patterns and assess future extremes (Vautard et al., 2021). In this study, we considered the historical period (1981–2010) and the near-future period (2031–2060). Here, we thus used only a single RCM (here REMO) to test the usefulness of the hourly data from this dataset and include a range of driving GCMs to cover a range of uncertainties related to the boundary conditions of the regional climate model.

Derivation of depth-duration-frequency (DDF) curves

We extracted the maximum annual rainfall values for durations of 1, 3, 6, 12, and 24 h from the gauging station data and the ensemble of simulated data from the REMO2015 regional climate model. However, due to data inconsistencies at sub-daily scales in the gauging stations, we considered the maximum annual rainfall only for 24 h to find the best-fitting distribution for extreme values. Although the 24-h rainfall extremes may not represent the most suitable temporal resolution for flash floods in the basins under this study, it serves as an initial assessment of the performance and goodness of fit at the daily scale. We tested several extreme value distributions (viz., Generalized Extreme Value (GEV), Pareto, Exponential, Person 3, Gumbel, and Lognormal) at daily (24 h) rainfall events for the E-OBS dataset. Based on the two goodness-of-fit tests (Kolmogorov–Smirnov test (K-S test) and Chi-square test), the GEV was chosen as the best-fitted distribution for this dataset (Supplementary Figure S2). The GEV distribution was further confirmed in a previous study carried out for the same region (Aronica et al., 2012b). Therefore, we used the GEV distribution to derive the return period (i.e., 50, and 100 years) of rainfall values for different durations (1, 2, 3, 6, and 12 h) from the simulated rainfall based on the ensemble of 8 members for the historical period (1981–2010) and the future period (2031–2060).

Flood hazard simulation

Hydrological modelling

We used the Hydrological Engineering Center - Hydrological Modeling System (HEC-HMS) Version 4.11 (Hydrologic Engineering Center, 2023) to simulate the catchment hydrological processes. HEC-HMS is an event and continuous-based lumped and semi-distributed model and is widely used in hydrological simulation (El Alfy, 2016; Abdessamed and Abderrazak, 2019; Al-Hussein et al., 2022; Athira et al., 2023; Peker et al., 2024). It can simulate the rainfall-runoff process using various approaches with parameters. The Mili Basin was delineated with 8 sub-basins and 4 reaches, while the S. Stefano Briga Basin was delineated with 15 sub-basins and 8 reaches based on area characteristics and flood modelling requirements.

This model has three main components: the basin model, the meteorological model, and the control specification. Under the basin model, “Soil Conservation Service (SCS) Curve Number” as the loss method, “SCS Unit Hydrograph” as the transform method, and “Lag” as the reach routing method were used. For these methods, the calculated input parameters based on the land use, soil, and slope properties of each sub-basin and established methods are presented in Table 2. In the meteorological model, observed rainfall data were used at 10-min intervals. In the control specification, the start and end times of the simulation period and time steps were defined. The simulation time step was selected as 10 min based on the catchment responses and rainfall duration. Model simulations were carried out for the 2009 rainfall event that caused a disastrous flood in the adjacent Giampilieri Basin as reference extreme conditions. We assessed the model performances for the 2009 flood event qualitatively based on the previous study by Aronica et al. (2012b), where they simulated the flood hydrographs for the 2009 rainfall event at Giampilieri Basin (a nearby basin to our study area), because both study basins (Mili and S. Stefano Briga) are ungauged. Here, at first, the area-ratio method was applied to scale up and down the available flood hydrograph for Giampilieri Basin to fit the Mili and S. Stefano Briga Basin areas while preserving the temporal pattern. Then model performances are evaluated with parameter adjustment. Since these are ungauged basins, this evaluation is different from a calibration and validation exercise of hydrological modelling using observational discharge data. The simulated discharges at the basins’ outlet (duration, pattern, and magnitude based on area ratio) were compared (Figure 2). This calibrated model was then used with DDF-derived rainfall data for baseline and future periods to simulate the streamflow. The rainfall value derived from the DDF Curves (fixed return period events (i.e., 50 and 100 years) for different durations (i.e., 1, 2, 3, 6, and 12 h), based on baseline and future projections and extreme value analysis, was used in the meteorological model in the HEC-HMS. Model simulations were carried out in the same time step as before.

Figure 2

This (qualitatively) calibrated model was used to simulate the streamflow for the 100-year return period of rainfall events at 6 and 12 h durations. We selected the frequency storm method among several other meteorological models available as described in the HEC-HMS manual (Hydrologic Engineering Center, 2023) to represent the extreme rainfall event (i.e., 6-h and 12-h duration) of 100-year return period rainfall. Therefore, model simulations forced with baseline and future DDF data represent the event-based simulations in both basins. We extracted the flood hydrographs at different locations to be used as the boundary conditions for hydrodynamic modelling.

Hydrodynamic modelling and hazard assessment

The Hydrological Engineering Center – River Analysis System (HEC-RAS) Version 6.5 (Hydrologic Engineering Center, 2020), which is one of the most commonly used freely accessible flood modelling tools, was utilized in this study. HEC-RAS is capable of 1D, 2D, and 1D & 2D hydraulic simulations: flow, sediment transport, and water quality. We set up the 2D HEC-RAS models for both basins with a 2 m mesh for river and floodplain areas and a 50 m mesh for the rest of the basin considered in the hydraulic simulations. A high-resolution 2 m DEM was used for this modelling. The main hydraulic structures, such as bridges, were incorporated into the models. The surface Manning’s roughness (Manning’s n) values were determined based on the land use types and literature (details are in Supplementary Table S2). The upstream boundary condition was set as an inflow hydrograph derived from HEC-HMS modelling. The normal depth (e.g., 0.01 in the Mili River) obtained from the river slope was used as the downstream boundary. Catchment runoff was derived from the spatial distribution of rainfall time series given in the 2D modelling area. Model simulations were carried out for the required flooding periods at different rainfall events considered at 0.5 s time step. Initially, the 2009 flood event was simulated to verify the Manning roughness, as it is the only calibration parameter used in this study. The maximum flood and velocities for the 2009 flood event are shown in Figure 3. We analyzed the flood hazards in terms of flood depth, velocity, and areas inundated under different extreme rainfall events calculated for baseline and future periods using the REMO2015 ensemble dataset.

Figure 3

To provide information on flood hazards, one of the reasonable methods is the classification-based mapping approach (Aureli et al., 2008; Aronica et al., 2012b; Papaioannou et al., 2017). Therefore, we also used a classification of the flood hazard and provided flood hazard maps for the historical and future periods. We adopted a method described by Aureli et al. (2008), where the hazard is presented as total depth based on water depths and velocities (Equation 1). We categorized the hazard levels into 3 classes: (1) Low: D < 0.5, (2) Medium: 0.5 ≤ D ≤ 1.0, and (3) High: D > 1.0.

Where D is the total depth, h is the water depth, and Fr is the Froude number.

Results

The results section includes two sub-sections. First, we summarize the findings of historical and future extreme rainfalls over the study area. Subsequently, the results of the flood hazard simulations are presented for two basins.

Current and future extreme rainfall events

The extreme value analysis with gridded observations, gauge data, and RCM data was compared for the historical period at a daily scale due to the inconsistency of the temporal resolution of the datasets. Furthermore, it is noted that the datasets do not span the same period, with the hourly gauge data spanning a shorter and more recent period than the E-OBS gridded data and the RCM data. However, extreme rainfalls derived from the 8 members of the REMO2015 RCM ensemble show a reasonable agreement with the observations (Table 3), particularly for the gauge stations (i.e., S. Stefano Briga and Messina) near the study basins. E-OBS gridded data underestimates the extreme rainfall in the region. The possible reason is that none of the local rainfall gauge data are used in the E-OBS dataset, and only a few stations covering the whole of Sicily are used, which are outside of our study domain. However, it is important to note that a comparison of point rainfall at stations with a gridded dataset has limitations because of the spatial heterogeneity of rainfall. This analysis confirms the ability of the REMO2015 ensemble to reproduce climate extremes in the region reasonably well during the historical period.

We analyzed the different short-duration rainfall events for different return periods based on the REMO2015 1-h data of the historical or baseline period (1981–2010) and near-future (2031–2060) under the RCP8.5 scenario. We used 8 ensemble members of REMO2015 for this analysis, and results are presented for the extreme-value analysis for the 50 and 100-year return period rainfall events. Our results show higher rainfall along the coast (> 180 mm) than in the inland area and higher rainfall variability within a smaller region for the baseline period (Figure 4). Furthermore, there is no clear pattern identified among the rainfall durations, return periods, and extremes derived from different REMO2015 runs considered. For example, based on the baseline period, extremes derived from REMO2015 forced with the CNRM-CM5 GCM show higher rainfall in inland and coastal areas for a 100-year return period in a 6-h duration, while higher rainfall is more concentrated in the coastal area in a 12-h duration for the same return period. The output from MPI-ESM-LR shows more than 60 mm of extreme rainfall for both durations (i.e., 6 and 12 h) in the considered region.

Figure 4

The projected changes of extreme rainfall for both durations with respect to the baseline period are presented in Figure 5 and Supplementary Figure S3. In general, rainfall extremes are projected to increase for both rainfall durations in the study area, with some exceptions of reductions in extreme rainfall shown in REMO2015 forced with CNRM-CMS. For example, in a 6-h rainfall event, extreme rainfall is projected to decrease by about 30% over the land area. In contrast, extreme rainfall derived from MIROC5 data shows the highest increase (> 50%) in a considerable area within the same region. Furthermore, even though extremes forced with MPI-ESM-LR show the highest rainfall in both durations (Figure 4), a significant reduction in extremes along the coastline is shown in the 12-h duration. These summarized results show that most runs projected an increase in extreme rainfall rather than a decrease in this region.

Figure 5

To develop the DDF curves for our study region, we aggregated the rainfall data from the nearest 4 grids that cover the study area and the nearby region. Aggregation of the RCM grid data can better represent the study area rather than individual ones (Kotova et al., 2023) because rainfall may vary highly within smaller areas (Figure 4), and the projection uncertainties of each grid can be reduced, improving the extreme analysis. From this point onwards, we discuss the rainfall extremes over the entire study area as one unit. Extreme rainfall is projected to increase in the future (Figure 6; Supplementary Figure S4), resulting in more intense rainfall within the same duration. The projected increase in rainfall is approximately 20% (on average) for all the return periods considered and all rainfall durations considered. For instance, under the RCP8.5 scenario in the future (2031–2060), for the 100-year return period, the 6-h rainfall event is projected to increase by −26 - 77% (average of 27%, from 8 simulations). In general, larger uncertainties of extreme rainfall projections are shown beyond 5-year return period events. It is observed that the larger the extreme rainfall values, uncertainties are also higher.

Figure 6

Changes in flood hazards under a changing climate

To carry out a detailed flood hazards assessment under the changing climate, we selected a 6-h rainfall event in a 100-year return period. The flood event that occurred in 2009 in the study area was a result of an 8-h rainfall event with 225 mm of intense rainfall (Aronica et al., 2012b). Therefore, selecting a 6-h rainfall event for flood hazard analysis can be justified and reasonable for this study area, when an extreme event is considered. This event would be below the observed event from 2009 in terms of rainfall intensity. For comparison, we considered a 12-h event as well, which would have an equal/ or higher intensity than the 2009 event. DDF curves developed from both baseline data and future climate data under RCP8.5 scenarios were utilized for the flood hazards analysis.

Similar to the future changes in rainfall extremes discussed above, our hydrological model simulations show that the resultant flow at each basin outlet is also projected to increase in the future compared to the baseline period (Figure 7). In the Mili Basin, the average flows are projected to increase by approximately 39 and 42% (on average) under 6-h and 12-h rainfall events, respectively, compared to the simulations for the baseline period. In contrast, the peak flows will increase by approximately 32 and 30% (on average) for the same rainfall events, respectively. Similarly, compared to the baseline, the average flows will increase by approximately 39 and 41%, respectively, from the simulations forced under 6-h and 12-h rainfall events in the S. Stefano Briga Basin. Furthermore, 31 and 29% increases in peak flows, respectively, are projected for the two rainfall events.

Figure 7

Simulation forced with REMO2015 CNRM-CM5 shows a decrease in average and peak flows by approximately 32 and 15%, respectively, in both basins under a 6-h rainfall event compared to the baseline period (Supplementary Table S3). Similarly, under a 12-h rainfall event, both average and peak flows are projected to decrease by 21 and 9% in both basins. In contrast, for 6 and 12 rainfall durations, the highest increase (> 109%) in projected flows (both average and peak) of both basins is under simulations forced with REMO2015 NorESM1-M. More detailed results can be found in Supplementary Table S3.

Mili Basin

A detailed analysis of flood hazards was carried out only for a smaller area to avoid the depression zones in the upper basin area in the larger 2D model area. Accordingly, the maximum inundation areas of 0.217 km² and 0.22 km² were obtained with the simulation forced by CNRM-CM5 for the historical period under 6-h and 12-h rainfall events, respectively. However, for two rainfall events, there are no considerable changes in inundation area in the future (i.e., decrease in area of ~0.4%) for the simulation with the same RCM under the RCP8.5 scenario compared to the baseline period. On average (based on 8 simulations), the inundation area may increase by ~4% and ~3% under 6-h and 12-h rainfall events in the near future compared to the baseline period. The simulated 2009 flood event inundated a 0.227 km² area of the Mili Basin. The highest increase in inundation area (8.8%) was obtained in the simulation forced by NorESM1-M for both rainfall events. It confirms that changes in rainfall and resultant floods show a non-linear relationship, as the rainfall intensity for this event, according to this RCM-GCM combination, is projected to double (see above).

For the baseline simulations, the median maximum depth ranges between 0.27 m and 0.41 m, and 0.35 m and 0.43 m under 6- and 12-h rainfall events, respectively. Similarly, the median of maximum velocity varies between 0.8 m/s and 1.6 m/s, and between 1.2 m/s and 1.6 m/s under both rainfall events, respectively. The hydraulic simulations show an increase in maximum flood depth and velocities in the near future (Figure 8) under both rainfall events. On average, the median maximum flood depth in the near future increases by 12 and 7%, for the 6- and 12-h durations, respectively. Similarly, the median maximum velocity may increase by 20 and 11% (on average) for the two rainfall events, respectively. However, the increase in the higher percentile (> 90th) of maximum depth and velocities is less than that for the median (50th percentile) under a 6-h rainfall event (Figures 8a,b). The smaller maximum flood depths and velocities will gradually increase up to a certain level (nearly the 80th percentile), and then gradually decrease toward the highest values. In contrast, under a 12-h rainfall event, higher percentage changes are more toward the median values of maximum depth and velocities. This implies that average flood hazard conditions are projected to be elevated in the future compared to the baseline period. According to our analysis, the 2009 flood event is certainly beyond the simulation results based on baseline data, but within the future simulation results. For instance, the mean maximum flood depth and velocity are 0.4 m and 1.6 m/s. However, Aronica et al. (2012b) summarized that the 2009 flood event was equivalent to a 1 in 130 years return period annual probability in the Giampilieri Catchment outlet, which is the neighboring basin of the S. Stefano Briga basin.

Figure 8

Santo Stefano Briga Basin

Similar to the analysis for the Mili Basin, a detailed hazards analysis was carried out for a designated area. For the baseline period, the simulation forced by REMO2015 CNRM-CM5 inundates the largest area (0.275 km² and 0.277 km² for 6- and 12-h rainfall events, respectively) of the basin. However, the highest increase in inundation area (by ~21 and 19% for the two rainfall events, respectively) was obtained in the simulation forced by REMO2015 NorESM1-M. The average of 8 simulations under the RCP 8.5 scenario shows an increase of the inundation area by ~6% in the near future period for both rainfall events. In the 2009 flood event, the area of 0.268 km² in the Santo Stefano Briga Basin was inundated.

Furthermore, for the baseline period, simulations show that the median of maximum depth varies between 0.44 m and 0.54 m and between 0.46 m and 0.55 m for 6-h and 12-h rainfall events, respectively, while the median of maximum velocities varies from 1.97 to 2.56 m/s and 2.04 to 2.58 m/s. An increase in maximum flood depth and velocity is projected (by 7 and 9% (on average), respectively) in the near future for the two rainfall events (Figure 9). It is observed that the future changes in smaller maximum depths and velocities (< 50^th percentile/median) show larger uncertainties than higher percentiles of the variables (> 75th percentile). In contrast to the Mili Basin, median percentage changes are gradually decreasing towards higher depths and velocities. Our simulation results indicated that flood hazards will increase in the future compared to the current conditions. Both the maximum depth and velocity induced in the 2009 flood event are within the mean of each flood hazard in historical and future periods. The simulated mean maximum flood depth and velocity for the 2009 event are 0.51 m and 2.34 m/s, respectively.

Figure 9

Flood hazard mapping

We categorized the flood hazards into 3 main classes (low, medium, and high) for spatial visualization (see Method section). Figure 10 shows the hazards over the study region at a 6-h rainfall event, and Supplementary Figure S5 shows the same for 12 h. Here, we considered the median of hazards during the historical and future (under the RCP8.5 scenario) periods to assess the impact of climate change in the study area in terms of extreme rainfall-induced flooding. According to our findings, some parts of both basins may undergo an increase in flood hazards, particularly at the main confluence and downstream of the Santo Stefano Briga Basin and downstream of the Mili Basin (Figure 1; Supplementary Figure S5).

Figure 10

Discussions

Available data for extreme rainfall events

This study analyzed the flood hazards in the coastal region of Sicily, Italy, based on 100-year design rainfall extremes of 6-h and 12-h durations, using various observational datasets and historical and projected hourly rainfall from the REMO2015 regional climate model at a 12.5 km x 12.5 km spatial resolution. The simulated rainfall data for the historical period (1981–2010) show good agreement with the 50-year and 100-year extreme values derived from daily in-situ rainfall gauge observations and the other global rainfall products considered (viz., E-OBS, CHIRPS, and MSWEP). In the study area in northeastern Sicily, Italy, the most recent extreme rainfall event (225 mm of rainfall within 8 h) in 2009 was used as a reference event, highlighting the area’s vulnerability to high-intensity, short-duration rainfall events. However, other rainfall events with different return periods and longer or shorter durations may also be relevant to study.

Projected rainfall data at an hourly resolution are rare, and such analyses and applications are limited in the literature. This study underscores the importance of using high-resolution (spatial and temporal) historical and projected climate data in a detailed analysis of extreme flood events. Few global rainfall datasets are available at sub-daily resolutions, for example, MSWEAP at 3 h, NASA’s Integrated Multi-satellite Retrievals for GPM (IMERG) at 30 min, and the Tropical Rainfall Measuring Mission (TRMM) at 3 h. However, their long-term data records are limited, and projections are not available. Therefore, hourly rainfall data from regional climate model simulations in CORDEX, exemplified here with the REMO2015 simulations, serve as a high-resolution product for climate and hydrological applications.

Implications for future flood risk

Flood simulations forced with 100-year design rainfall values for 6- and 12-h durations resulted in an increase in hazards compared to the baseline period. Under the RCP8.5 scenario, flow velocities may increase by up to 62% (minimum −3%), and the inundation depths by up to 29% (minimum −3%). The two basins respond differently in terms of percentage increases of hazards in the future compared to the baseline period. An important reason could be their topographic features, channel geometry, and drainage characteristics. However, rapid human activities, such as river regulation, increasing soil sealing, and land-use changes in the basins, could also amplify flood hazards beyond those projected by climate models alone.

In addition, changes in exposure (e.g., expansion of human settlement in flood-prone areas) and vulnerability (e.g., aging infrastructure) can lead to higher flood risk by driving up potential damage and losses, independent of changes in flood hazards. A similar study carried out at the nearby basin (Giampilieri) by Aronica et al. (2012a) demonstrated four hazard classes based on flood depth and total hydrodynamic force per unit width, ultimately calculating the spatial flood risk variability. Our analysis can also be further extended to quantify the flood risk by combining hazard, exposure, and vulnerability components. Therefore, understanding the future risk induced by rainfall extremes and floods is crucial in the study region (Sirisena et al., 2024) as well as in other urbanized regions of the world.

Based on our findings, we highlighted that future rainfall extremes are likely to produce more intense flooding in the study region. Thus, increasing urban development in flood-prone valleys may escalate future risk more rapidly than climate change alone. Reducing risk requires both climate adaptation and area management by local authorities and communities. Our results highlighted that ensemble-based approaches should be expanded to better represent uncertainties in both climate and human-induced drivers. Further integration of hydrological, hydraulic, and socio-economic datasets is needed to move from hazard mapping toward full probabilistic risk modelling.

Limitations

Flood hazard modelling

The hydrological and hydrodynamic modelling undertaken in the present study contains several assumptions and limitations to represent the catchment and channel characteristics. For instance, the representation of channel geometries and hydraulic structures must be evaluated for current conditions and can potentially result in adjusted discharges and inundation levels. The study basins are ungauged. We verified the pattern of the simulated streamflow hydrograph (output of the hydrological model) based on available literature on the nearby Giampilieri basin that presented the 2009 flood event. This study assumed that future land use will remain the same as current conditions, which will add additional uncertainties to the simulated flood hazards. Therefore, incorporating dynamic land use changes would be useful. Furthermore, incorporating the dynamic geometry of the river, primarily accounting for spatial variations in river width, depth, slope, and cross-section, could further enhance the accuracy of hydrological and flood modelling.

Climate change scenarios and approaches

In the present study, we only considered the RCP8.5 scenario, where we tested a high-emissions and high-global warming scenario. Other scenarios, with less greenhouse gas emissions and warming, may result in more limited changes in rainfall extremes and a smaller increase in the flood hazard. The modelling of extreme events requires high-quality long-term hydroclimatic datasets. Furthermore, the use of other hourly regional climate model projection data, such as the full ensemble of EURO-CORDEX, could help address the uncertainties in flood simulations for small catchments. RCMs might not be able to well represent the convection processes that are key to sub-daily scale rainfall extremes, thus convection-permitting modelling is useful for shorter-duration higher extremes (Fosser et al., 2024). However, due to the high computational demand for convection-permitting modelling, RCMs with high temporal resolution (such as hourly rainfall) still serve the purpose of shorter-duration rainfall extreme analysis, especially when rare events are concerned (e.g.50- and 100-year return periods). Such rare event estimations require longer time series (30 years), which are rarely available from convection-permitting models, and therefore, currently available regional model projections from RCMs serve this purpose. Also, our study demonstrates that the sub-daily data from the RCMs align very well with the observational rainfall extremes at this sub-daily duration.

Conclusion

This study evaluated extreme rainfall and resultant flood hazards in terms of inundation area, maximum flood depth, and velocities in a coastal region, Messina, Sicily, Italy. We used high-resolution rainfall data from the regional climate model REMO2015 ensemble simulations at a sub-daily scale (one-hour) and hydrological (HEC-HMS) and hydrodynamic (HEC-RAS) modelling for flood simulations at short-duration rainfall events of 6 and 12 h at the 100-year return period in a case study area in Sicily, Italy.

Our results show that extreme rainfall values for the 50 and 100-year return period with a duration of 24 h derived from REMO2015 ensembles show a good agreement with the same design rainfall derived from in-situ observations, with some exceptions. Thus, REMO2015 ensembles serve as a high-resolution climate dataset that can be used for several applications in climate extremes and hydrology. On average, extreme rainfall will increase by approximately 20% in the study area in the near future (2031–2060) under the RCP8.5 scenario, compared to the baseline period (1981–2010). The resultant flood hazards (inundation area, depth, and velocity) are also projected to increase by approximately 3, 10, and 15% in the Mili Basin and 6, 7, and 9% in the Santo Stefano Briga Basin, respectively, under 6 and 12-h design rainfall events of RCP 8.5 projections, compared to the baseline period.

Our results show that flood severity and impacts could increase in the future in this region, according to the simulations from the REMO model. On the other hand, growing anthropogenic activities (e.g., urbanization and deforestation) may also exacerbate the above situation by aggravating the flood hazard, as well as increasing the exposure of assets and people. Therefore, assessing and increasing awareness of flood hazards is crucial for disaster risk reduction and planning for adaptation measures in the future.

References

• 1

  AbdessamedD.AbderrazakB. (2019). Coupling HEC-RAS and HEC-HMS in rainfall–runoff modeling and evaluating floodplain inundation maps in arid environments: case study of Ain Sefra city, Ksour Mountain. SW of Algeria. Environ. Earth Sci.78, 1–17. doi: 10.1007/s12665-019-8604-6

  • CrossRef
  • Google Scholar

• 2

  AhmadD.AfzalM. (2022). Flood hazards and livelihood vulnerability of flood-prone farm-dependent bait households in Punjab, Pakistan. Environ. Sci. Pollut. Res.29, 11553–11573. doi: 10.1007/s11356-021-16443-4,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 3

  AkperovM.ZhangW.KoenigkT.EliseevA.SemenovV. A.MokhovI. I. (2025). Projected changes in near-surface wind speed in the Arctic by a regional climate model. Polar Sci.43, 101162. doi: 10.1016/j.polar.2024.101162

  • CrossRef
  • Google Scholar

• 4

  Al-HusseinA. A. M.KhanS.NcibiK.HamdiN.HamedY. (2022). Flood analysis using HEC-RAS and HEC-HMS: a case study of Khazir River (Middle East—Northern Iraq). Water14. doi: 10.3390/w14223779

  • CrossRef
  • Google Scholar

• 5

  AriasP. A.BellouinN.CoppolaE.JonesR. G.KrinnerG.MarotzkeJ.et al. (2021). “2021: Technical Summary,” in Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, eds. B. Z. Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, et al. (Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press), 33–144. doi: 10.1017/9781009157896.002

  • CrossRef
  • Google Scholar

• 6

  ArnoneE.PumoD.ViolaF.NotoL. V.La LoggiaG. (2013). Rainfall statistics changes in Sicily. Hydrol. Earth Syst. Sci.17, 2449–2458. doi: 10.5194/hess-17-2449-2013

  • CrossRef
  • Google Scholar

• 7

  AronicaG. T.BiondiG.BrigandìG.CasconeE.LanzaS.RandazzoG. (2012a). Assessment and mapping of debris-flow risk in a small catchment in eastern Sicily through integrated numerical simulations and GIS. Phys. Chem. Earth49, 52–63. doi: 10.1016/j.pce.2012.04.002

  • CrossRef
  • Google Scholar

• 8

  AronicaG. T.MrigandíG.MoreyN. (2012b). Flash floods and debris flow in the city area of Messina, north-east part of Sicily, Italy in October 2009: the case of the Giampilieri catchment. Nat. Hazards Earth Syst. Sci.12, 1295–1309. doi: 10.5194/nhess-12-1295-2012

  • CrossRef
  • Google Scholar

• 9

  AtangaR. A.TankpaV. (2021). Climate change, flood disaster risk and food security nexus in northern Ghana. Front. Sustain. Food Syst.5. doi: 10.3389/fsufs.2021.706721

  • CrossRef
  • Google Scholar

• 10

  AthiraS.KatpatalY.B.LondheD.S. (2023) ‘Flood modelling and inundation mapping of Meenachil River using HEC-RAS and HEC-HMS software’, in HaddoutA.M.LakshmiKrishnamoorthyHoguaneP. (ed.) Springer proceedings in Earth and environmental sciences. CCORE 2022. Springer, Cham.

  • Google Scholar

• 11

  AugustynA. (2025) Mediterranean Climate, Britannica. Available online at: https://www.britannica.com/science/Mediterranean-climate (Accessed: 4 June 2025).

  • Google Scholar

• 12

  AureliF.MaranzoniA.MignosaP.ZiveriC. (2008). 2D numerical modelling for hydraulic hazard assessment: A dam-breakcase study, in:Proc., Int. Conf. on Fluvial Hydraulics, RiverFlow 2008, Kubaba Congress Dept. and Travel Services, Ankara,Turkey, 729–736.

  • Google Scholar

• 13

  AvinoA.CimorelliL.FurcoloP.NotoL. V.PelosiA.PianeseD.et al. (2024). Are rainfall extremes increasing in southern Italy?J. Hydrol.631:130684. doi: 10.1016/j.jhydrol.2024.130684

  • CrossRef
  • Google Scholar

• 14

  BeckH. E.WoodE. F.PanM.FisherC. K.MirallesD. G.van DijkA. I. J. M.et al. (2019). MSWep v2 global 3-hourly 0.1° precipitation: methodology and quantitative assessment. Bull. Am. Meteorol. Soc.100, 473–500. doi: 10.1175/BAMS-D-17-0138.1

  • CrossRef
  • Google Scholar

• 15

  BergströmS. (1976) Development and application of a conceptual runoff model for Scandinavian catchments, report RH07 Sweden. Swedish Meteorological and Hydrological Institute (SMHI), Norrköping, Sweden.

  • Google Scholar

• 16

  BevenK.FreerJ. (2001). A dynamic topmodel. Hydrol. Process.15, 1993–2011. doi: 10.1002/hyp.252

  • CrossRef
  • Google Scholar

• 17

  BoeingF.AttingerS.WagenerT.RakovecO.SamaniegoL.ThoberS.et al. (2025). Spatially and seasonally differentiated response of soil moisture droughts to climate change in Germany. Earths Future13. doi: 10.1029/2024EF005495

  • CrossRef
  • Google Scholar

• 18

  BonaccorsoB.BrigandìG.AronicaG. T. (2020). Regional sub-hourly extreme rainfall estimates in Sicily under a scale invariance framework. Water Resour. Manag.34, 4363–4380. doi: 10.1007/s11269-020-02667-5

  • CrossRef
  • Google Scholar

• 19

  BouwerL. M. (2019). “Observed and projected impacts from extreme weather events: implications for loss and damage” in Loss and damage from climate change. eds. MechlerR.et al. (Chambridge: Springer), 63–82.

  • Google Scholar

• 20

  CornesR. C.van der SchrierG.van den BesselaarE. J. M.JonesP. D. (2018). An ensemble version of the E-OBS temperature and precipitation data sets. J. Geophys. Res. Atmos.123, 9391–9409. doi: 10.1029/2017JD028200

  • CrossRef
  • Google Scholar

• 21

  CRED (2020) Human cost of disasters, human cost of disasters. CRED, Brussels, Belgium.

  • Google Scholar

• 22

  DHI (2017). MIKE 21 flow model & MIKE 21 flood screening tool - hydraulic model. Hørsholm: Denmark.

  • Google Scholar

• 23

  EDF-DRD (2000). TELEMAC Modelling System: TELEMAC-2D Validation Document. Version 5.0. Chatou, France: Électricité de France - Direction des Études et Recherches.

  • Google Scholar

• 24

  El AlfyM. (2016). Assessing the impact of arid area urbanization on flash floods using GIS, remote sensing, and HEC-HMS rainfall-runoff modeling. Hydrol. Res.47, 1142–1160. doi: 10.2166/nh.2016.133

  • CrossRef
  • Google Scholar

• 25

  FantappieM.PrioriS.CostantiniE. A. C. (2015). Soil erosion risk Sicilian Region. J. Maps11, 323–341. doi: 10.1080/17445647.2014.956349

  • CrossRef
  • Google Scholar

• 26

  FernándezD. S.LutzM. A. (2010). Urban flood hazard zoning in Tucumán Province, Argentina, using GIS and multicriteria decision analysis. Eng. Geol.111, 90–98. doi: 10.1016/j.enggeo.2009.12.006

  • CrossRef
  • Google Scholar

• 27

  FosserG.GaetaniM.KendonE. J.AdinolfiM.BanN.BelušićD.et al. (2024). Convection-permitting climate models offer more certain extreme rainfall projections. NPJ Clim. Atmos. Sci.7, 1–10. doi: 10.1038/s41612-024-00600-w

  • CrossRef
  • Google Scholar

• 28

  FunkC.PetersonP.LandsfeldM.PedrerosD.VerdinJ.ShuklaS.et al. (2015). The climate hazards infrared precipitation with stations - a new environmental record for monitoring extremes. Scientific Data2:150066. doi: 10.1038/sdata.2015.66,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 29

  GhanbariM.ArabiM.KaoS.‐. C.ObeysekeraJ.SweetW. (2021). Climate change and changes in compound coastal-riverine flooding Hazard along the U.S. coasts. Earths Future9. doi: 10.1029/2021EF002055

  • CrossRef
  • Google Scholar

• 30

  HabeebB.Bastidas-ArteagaE. (2023). Assessment of the impact of climate change and flooding on bridges and surrounding area. Front. Built Environ.9. doi: 10.3389/fbuil.2023.1268304

  • CrossRef
  • Google Scholar

• 31

  HanY.HuangQ.HeC.FangY.WenJ.GaoJ.et al. (2020). The growth mode of built-up land in floodplains and its impacts on flood vulnerability. Sci. Total Environ.700:134462. doi: 10.1016/j.scitotenv.2019.134462,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 32

  HsiaoS.-C.ChiangW. S.JangJ. H.WuH. L.LuW. S.ChenW. B.et al. (2021). Flood risk influenced by the compound effect of storm surge and rainfall under climate change for low-lying coastal areas. Sci. Total Environ.764:144439. doi: 10.1016/j.scitotenv.2020.144439,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 33

  HuongH. T. L.PathiranaA. (2013). Urbanization and climate change impacts on future urban flooding in Can Tho city, Vietnam. Hydrol. Earth Syst. Sci.17, 379–394. doi: 10.5194/hess-17-379-2013

  • CrossRef
  • Google Scholar

• 34

  Hydrologic Engineering Center (2020) HEC-RAS Hydraulic Reference Manual. Available online at: https://www.hec.usace.army.mil/confluence/rasdocs/ras1dtechref/6.3 (Accessed: 3 April 2023).

  • Google Scholar

• 35

  Hydrologic Engineering Center (2023) HEC-HMS technical reference manual. Available online at: https://www.hec.usace.army.mil/confluence/hmsdocs/hmstrm (Accessed: 16 February 2023).

  • Google Scholar

• 36

  JacobD.PetersenJ.EggertB.AliasA.ChristensenO. B.BouwerL. M.et al. (2014). EURO-CORDEX: new high-resolution climate change projections for European impact research. Reg. Environ. Chang.14, 563–578. doi: 10.1007/s10113-013-0499-2

  • CrossRef
  • Google Scholar

• 37

  JacobD.TeichmannC.SobolowskiS.KatragkouE.AndersI.BeldaM.et al. (2020). Regional climate downscaling over Europe: perspectives from the EURO-CORDEX community. Reg. Environ. Change20:51. doi: 10.1007/s10113-020-01606-9

  • CrossRef
  • Google Scholar

• 38

  JalayerF.AronicaG. T.RecuperoA.CarozzaS.ManfrediG. (2018). Debris flow damage incurred to buildings: an in situ back analysis. J. Flood Risk Manage.11, S646–S662. doi: 10.1111/jfr3.12238

  • CrossRef
  • Google Scholar

• 39

  JanizadehS.Chandra PalS.SahaA.ChowdhuriI.AhmadiK.MirzaeiS.et al. (2021). Mapping the spatial and temporal variability of flood hazard affected by climate and land-use changes in the future. J. Environ. Manag.298:113551. doi: 10.1016/j.jenvman.2021.113551,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 40

  JonkmanS. N.CurranA.BouwerL. M. (2024). Floods have become less deadly: an analysis of global flood fatalities 1975–2022. Nat. Hazards120, 6327–6342. doi: 10.1007/s11069-024-06444-0

  • CrossRef
  • Google Scholar

• 41

  KernkampH. W. J.Van DamA.StellingG. S.de GoeE. D. (2011). Efficient scheme for the shallow water equations on unstructured grids with application to the continental shelf. Ocean Dyn.61, 1175–1188. doi: 10.1007/s10236-011-0423-6

  • CrossRef
  • Google Scholar

• 42

  KotovaL.LeissnerJ.WinklerM.KilianR.BichlmairS.AntretterF.et al. (2023). Making use of climate information for sustainable preservation of cultural heritage: applications to the KERES project. Herit. Sci.11, 1–18. doi: 10.1186/s40494-022-00853-9

  • CrossRef
  • Google Scholar

• 43

  KreibichH.van LoonA.SchröterK.WardP. J.MazzoleniM.SairamN.et al. (2022). The challenge of unprecedented floods and droughts in risk management. Nature608, 80–86. doi: 10.1038/s41586-022-04917-5,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 44

  LaurensonE. M.MeinR. G. (1983). RORB-version 3: Runoff routing program — User manual. 2nd Editio Edn. Monash, Australia: Department of Civil Engineering, Monash University.

  • Google Scholar

• 45

  LevantesiS. (2021). Assessing Italy’s climate risk. Nature. doi: 10.1038/d43978-021-00136-0

  • CrossRef
  • Google Scholar

• 46

  LiuW.FengQ.EngelB. A.YuT.ZhangX.QianY. (2023). A probabilistic assessment of urban flood risk and impacts of future climate change. J. Hydrol.618:129267. doi: 10.1016/j.jhydrol.2023.129267

  • CrossRef
  • Google Scholar

• 47

  MazzoglioP.GanoraD.ClapsP. (2022). Long-term spatial and temporal rainfall trends over Italy †. Environ. Sci. Proc21, 2–8. doi: 10.3390/environsciproc2022021028

  • CrossRef
  • Google Scholar

• 48

  NgoH.RanasingheR.ZevenbergenC.KirezciE.MahengD.RadhakrishnanM.et al. (2020). An efficient modelling approach for probabilistic assessments of present-day and future fluvial flooding. Natural Hazards and Earth System Sciencetem Sciences Discussions2050, 1–42. doi: 10.3389/fclim.2022.798618

  • CrossRef
  • Google Scholar

• 49

  O’BrienJ. S.JulienP. Y.FullertonW. T. (1993). Two-dimensional water flood and mudflow simulation. J. Hydraul. Eng.119, 244–259. doi: 10.1061/(ASCE)0733-9429(1993)119:2(244)

  • CrossRef
  • Google Scholar

• 50

  PapaioannouG.VasiliadesL.LoukasA.AronicaG. T. (2017). Probabilistic flood inundation mapping at ungauged streams due to roughness coefficient uncertainty in hydraulic modelling. Adv. Geosci.44, 23–34. doi: 10.5194/adgeo-44-23-2017

  • CrossRef
  • Google Scholar

• 51

  Pekerİ. B.GülbazS.DemirV.OrhanO.BedenN. (2024). Integration of HEC-RAS and HEC-HMS with GIS in flood modeling and flood hazard mapping. Sustainability16. doi: 10.3390/su16031226

  • CrossRef
  • Google Scholar

• 52

  RanasingheR.RuaneA. C.VautardR.ArnellN.CoppolaE.CruzF. A.et al. (2023). “Climate Change Information for Regional Impact and for Risk Assessment,” in Climate Change 2021 – The Physical Science Basis: Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, (Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press), 1767–1926. doi: 10.1017/9781009157896.014

  • CrossRef
  • Google Scholar

• 53

  RefsgaardJ. C.StormB. (1995). “MIKE SHE” in Computer models of watershed hydrology. ed. SinghV. P. (Colorado, USA: Water Resource Publications LLC), 806–846.

  • Google Scholar

• 54

  RemedioA. R.TeichmannC.BuntemeyerL.SieckK.WeberT.RechidD.et al. (2019). Evaluation of new CORDEX simulations using an updated köppen-trewartha climate classification. Atmosphere (Basel).10, 1–25. doi: 10.3390/atmos10110726

  • CrossRef
  • Google Scholar

• 55

  RuosteenojaK.RäisänenJ. (2024). Incorporation of RCM-simulated spatial details into climate change projections derived from global climate models. Clim. Dyn. doi: 10.1007/s00382-024-07258-3

  • CrossRef
  • Google Scholar

• 56

  ShamsudduhaM. (2025). Redefining flood hazard and addressing emerging risks in an era of extremes. NPJ Nat. Hazards2:29. doi: 10.1038/s44304-025-00082-7

  • CrossRef
  • Google Scholar

• 57

  SirisenaJ.RemedioA.NievasC.AronicaG.BouwerL. (2024). Uncertainties in flood damage assessment under projected future extreme rainfall conditions : a case study in Northeastern Sicily., in EGU General Assembly 2024(Vienna, Austria: European Geoscience Union), EGU24-9721. doi: 10.5194/egusphere-egu24-9721

  • CrossRef
  • Google Scholar

• 58

  SivapalanM.SavenijeH. H. G.BlöschlG. (2011). Socio-hydrology : A new science of people and water.Hydrol. Process.26, 1270–1276. doi: 10.1002/hyp.8426

  • CrossRef
  • Google Scholar

• 59

  TarquiniS.IsolaI.FavalliM.MazzariniF.BissonM.PareschiM. T.et al. (2007). TINITALY/01: a new triangular irregular network of Italy. Ann. Geophys.50, 407–425. doi: 10.4401/ag-4424

  • CrossRef
  • Google Scholar

• 60

  TellmanB.SullivanJ. A.KuhnC.KettnerA. J.DoyleC. S.BrakenridgeG. R.et al. (2021). Satellite imaging reveals increased proportion of population exposed to floods. Nature596, 80–86. doi: 10.1038/s41586-021-03695-w,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 61

  TreppiediD.CipollaG.FrancipaneA.NotoL. V. (2021). Detecting precipitation trend using a multiscale approach based on quantile regression over a Mediterranean area. Int. J. Climatol.41, 5938–5955. doi: 10.1002/joc.7161

  • CrossRef
  • Google Scholar

• 62

  VautardR.KadygrovN.IlesC.BobergF.BuonomoE.BülowK.et al. (2021). Evaluation of the large EURO-CORDEX regional climate model ensemble. J. Geophys. Res. Atmos.126. doi: 10.1029/2019JD032344

  • CrossRef
  • Google Scholar

• 63

  VitanzaE.DimitriG. M.MocenniC. (2023). A multi-modal machine learning approach to detect extreme rainfall events in Sicily. Sci. Rep.13, 1–15. doi: 10.1038/s41598-023-33160-9

  • CrossRef
  • Google Scholar

• 64

  WagnerO.MaleskaV.BouwerL.M. (2025) ‘Do convection-permitting regional climate models have added value for hydroclimatic simulations? A test case over small and medium-sized catchments in Germany’, EGUsphere [preprint], doi: 10.5194/egusphere-2025-2943

  • CrossRef
  • Google Scholar