1. Introduction
Flooding is a recurrent natural phenomenon characterized by the temporary inundation of land not typically submerged by water [
1]. Floods rank among the most recurrent and widespread natural disasters globally [
2,
3], constituting 44% of all disaster occurrences between 2000 and 2019, resulting in the loss of 1.23 million lives, affecting 4.2 billion people and leading to economic losses estimated at USD 2.97 trillion worldwide [
4]. In the Mediterranean region, a cyclone named ‘Medicane Daniel’ caused floods in Greece and neighboring countries, resulting in economic losses reaching USD 17 billion in Europe [
5].
Southern Europe and the Mediterranean region are at high vulnerability to climate change impacts [
5,
6,
7], suffering from temperature increases, considerable reductions in rainfall and water runoff, and extreme events (e.g., flash floods and heatwaves) [
5,
8,
9]. These changes manifest in various forms, including increased intensity and frequency of rainfall events, leading to heightened flood risk [
10]. Moreover, sea-level rise exacerbates coastal flooding, posing additional challenges to vulnerable coastal communities [
11]. However, effective adaptation strategies require a comprehensive understanding of the complex interactions between climate change, land use patterns, and hydrological processes [
7,
8,
12]. In the hydraulic simulation process, another critical parameter is the selection of the appropriate time series analysis method, which should be based on the geomorphological characteristics of the area under consideration [
13,
14].
Especially, Cyprus, an island located in the eastern Mediterranean basin, experiences a temperate climate with scorching, arid summers, and according to the Köppen–Geiger classification system, a portion of the island is deemed hot and arid [
15,
16]. A study spanning up to the year 2030 indicates that the annual cost of water scarcity in the domestic and industrial sectors could reach up to EUR 88 million. From this amount, EUR 16–32 million would constitute additional costs due to reduced water availability attributed to climate change [
17].
To manage and monitor floods within urban areas, the European Commission implemented Flood Directive 2007/60/EC. This framework primarily aims to mitigate the adverse impacts of floods on the environment, human health, cultural heritage, and economic activities closely tied to society [
12,
18]. Member states of the European Union were required to establish River Basin Flood Risk Management Plans, which identify vulnerable areas while considering long-term developments and future flood occurrences. The primary objective of flood risk assessment at the national level is the creation of flood hazard maps [
19,
20]. These maps depict flood extent, depth, flow speed, and probability. Furthermore, flood risk maps evaluate potential adverse consequences for the economic activity of the region and for potentially affected residents [
1].
Cyprus, as a member of the European Union, has achieved full compliance with the provisions of Directive 2007/60. Moreover, Cyprus harmonized with the provisions of the directive by transposing them into Cypriot law [
21]. Specifically, it enacted the laws of 2010 and 2012 [
22]. The legislation enacted in July 2010 designates the Tax Department as the Competent Authority for the implementation of the directive and the Ministry of Interior as the Coordinating Authority. Specifically, Cyprus has already completed the second implementation cycle of the directive, which is in force until 2027 [
22].
Flood warning and forecasting systems play a major role in flood risk management; specifically, the evaluation of flood hazard achieved by inundation mapping and identification of flood risk regions [
23,
24,
25]. Flood inundation modeling plays a crucial role in generating spatial distribution data pertaining to inundation patterns, including water depth and flow velocity. This information can provide insights into the severity of the hazard, potential risks to public safety, and potential financial implications [
2,
12,
26,
27,
28,
29].
Appropriate and effective tools for flood inundation modeling and mapping assist with flood risk assessment. Primarily, these requirements have been evaluated through the utilization of one-dimensional (1D) and two-dimensional (2D) hydraulic models (e.g., [
30,
31]. Hydraulic simulation, through advanced modeling techniques, offers a valuable tool for assessing and mitigating flood risks, particularly in the context of emergency preparedness. The integration of hydraulic simulation into emergency management frameworks enables authorities to anticipate and respond effectively to flood events, thus minimizing the potential impact on communities and infrastructure. By simulating various scenarios, policymakers can identify vulnerable areas, evaluate evacuation routes, and develop targeted response strategies [
27,
28,
29]. Moreover, hydraulic simulation facilitates the testing and optimization of flood protection measures, enhancing the resilience of communities to future disasters.
This research presents a practical framework for mapping flood inundation in ungauged urban areas, tailored for the context of the EU Floods Directive’s second implementation in Cyprus. The framework’s efficacy is demonstrated in Larnaca and Nicosia, regions prone to frequent flooding from intense storms. The comparison between different hydraulic simulation methods enhances our ability to accurately estimate and delineate flood-prone areas, assess the uncertainty inherent in flood mapping, offer guidance to flood management professionals, and inform the development of protective measures and policies safeguarding human life, property, and economic assets.
3. Results and Comparison
By simulating flow behavior under various scenarios, HEC-RAS enables the calculation of critical parameters such as water depths, flow velocities, and inundation areas. Results are presented per APSFR and per return period.
3.1. Hydrographs
HEC-RAS modeling generates detailed output hydrographs that are essential for understanding the temporal dynamics of flood events. These hydrographs depict how water flow parameters, such as discharge, water surface elevation, and velocity, change over time at specific locations within the study area.
3.1.1. APSFR 21–22
In the comparison of the results for each of the three return periods, T20, T100, and T500 years show a high identity of the flood peaks between full 2D and combined (1D/2D), which indicates the high convergence of the two methods.
According to
Figure 3, a simulation of the 20-year flow hydrograph, a high identification of the flood peak between 2D and combined 1D/2D flood peaks can be observed. Specifically, 2D simulation predicts a flood peak of 142.74 m
3/s on September 1st at 14:00, identical to the peak predicted by the combined model at the same time, demonstrating the accuracy and reliability of the 2D simulation in replicating flood peak values.
According to
Figure 4, a simulation of the 100-year flow hydrograph, a high identification of the flood peak between 2D and combined 1D/2D flood peaks can be observed. Specifically, 2D simulation predicts a flood peak of 451.62 m
3/s on September 1st at 14:15, identical to the peak predicted by the combined model at the same time, demonstrating the accuracy and reliability of the 2D simulation in replicating flood peak values.
According to
Figure 5, a simulation of the 500-year flow hydrograph, a high identification of the flood peak between 2D and combined 1D/2D flood peaks can be observed. Specifically, the 2D simulation predicts a flood peak of 687.53 m
3/s on September 1st at 15:45, identical to the peak predicted by the combined model at the same time, demonstrating the accuracy and reliability of the 2D simulation in replicating flood peak values.
Figure 6 presents the initial 1D/2D model and the examined 2D model hydrographs, respectively. The selected location for APSFR 21–22 is one of the reported flooding incident areas in the confluence of the rivers (Location 1 is shown in the relevant Figure in
Section 3.4.1). The following figure illustrates a high identification of the flood peak between 2D and combined 1D/2D flood peaks. Specifically, 1D/2D simulation predicts a slightly higher flood peak, which comes in agreement with the difference in the inundated area shown in
Table 2.
3.1.2. APSFR 29
According to
Figure 7, in the hydraulic simulation of a 20-year return period event, a notable agreement between the flood peaks of the full 2D and combined 1D/2D model was observed. The 2D simulation produced a flood peak of 56.12 m
3/s at 14 h and 45′ from the start of the event, while the combined model indicated a flood peak of 65.31 m
3/s at 14 h and 40′. The difference of 14% in flood peaks is deemed reasonable and aligns with the results presented in
Table 2, which recorded the largest runoff volume error at 5.2%. This discrepancy can be attributed to the retention of water approximately 580 m upstream of the outlet, which prevents this volume from reaching the outlet. This retention is likely due to the agricultural land in the area, which affects the flow dynamics. Additionally, the accuracy of the calculation and the detailed design of the computational grid, which can be adjusted by the user, contribute to this result.
According to
Figure 8, in the hydraulic simulation of the 100-year return period, a high match of the flood peak between full 2D and combined 1D/2D flood peaks is observed. The 2D simulation gives a flood peak at 14 h 30′ from the event occurrence and is equal to 110.02 m
3/s. The combined one at 14 h 20′ is equal to 115.44 m
3/s. Therefore, there is a high convergence between the peak flows and the occurrence of small discrepancies considered non-critical.
According to
Figure 9, in the hydraulic simulation of the 500-year return period, a high correlation between the flood peaks of the full 2D and combined 1D/2D models was observed. The 2D simulation indicated a flood peak of 198.2 m
3/s at 15 h and 15 min from the start of the event, while the combined model predicted a flood peak of 196.83 m
3/s at 15 h. The close alignment of peak flows, with a negligible time lag, underscores the reliability and accuracy of both modeling approaches.
Figure 10 presents the initial 1D/2D model and the examined 2D model hydrographs. The selected location for APSFR 29 is one of the reported flooding incident areas near the University of Cyprus campus affecting many facilities (Figure 20 is shown in Location 1). The following figure shows a high identification of the flood peak and the time from the event occurrence and between the 2D and combined 1D/2D flood peaks.
Regarding the results, there is an agreement between the flood peak values and their occurrence in the full two-dimensional and the combined 1D/2D hydraulic simulations. A higher convergence of hydrographs was observed in the hydraulic simulation of the APSFR 21–22, constituting an area of higher slopes compared to APSFR 29.
3.2. Inundation Maps
By simulating water surface profiles and flow patterns, HEC-RAS generates detailed maps that depict the extent and depth of flooding under various scenarios, such as different storm intensities or infrastructure failures. A flood field is considered to be a piece of land within the channel that is covered by water. These maps are instrumental for urban planning, emergency response, and floodplain management, as they provide clear, spatially resolved information on potential inundation areas, contributing to the identification of vulnerable regions, planning evacuation routes, and designing flood mitigation structures.
Thus, a criterion of comparison for this thesis is the flooding areas of the two different hydraulic methods (2D and combined 1D/2D). The extraction of the fields is carried out in HEC-RAS through the results for each return period T = 20, 100, and 500 years in each APSFR.
Table 2 illustrates inundated areas for each hydraulic model, scenario, and methodology.
3.2.1. APSFR 21–22
Comparing the inundation area maps produced by the 1D/2D combined hydraulic simulation method and the full 2D method, no notable differences were noticed. The following inundated maps (
Figure 11,
Figure 12 and
Figure 13) present the flooding area in APSFR 21–22 for each return period, T = 20, 100, and 500 years, respectively. These results came into agreement with the hydrographs that are explained in
Section 3.1. In smaller flooding areas, greater percentage differences are observed.
For comparative purposes, floodplain maps were generated for each method, 1D/2D and 2D, for each return period. These maps provide a visual representation of the estimated floods and allow for the analysis and evaluation of the differences between the modeling methods. This comparison is crucial for understanding the impact of each method on the accuracy and reliability of flood predictions. The following
Figure 14 presents inundation boundaries that are quite similar for both simulation methods, 1D/2D and 2D. Any minor areas of inundation are attributed to the presence of agricultural lands where water tends to accumulate rather than flow downstream. This similarity indicates that both methods produce comparable results in predicting flood extents, with variations primarily influenced by local land use and topography.
3.2.2. APSFR 29
Inundation maps showing the spatial distribution of accuracy and omission and commission errors are shown in
Figure 15,
Figure 16 and
Figure 17. In comparison to maps of APSFR 21–22, these maps show a slight deviation between the flood extents observed, as presented below. Such minor discrepancies are considered reasonable and are attributed to the computational accuracy selected for the specific models. Additionally, these discrepancies may be due to varying land uses (higher Manning’s coefficient), especially in agricultural areas that show areal water retention.
The full 2D model generally provides a more detailed representation of flood extents, capturing complex flow patterns and interactions with terrain features more accurately than the 1D/2D method. This results in a higher-resolution map with finer details in areas of variable topography. In contrast, the 1D/2D model, while computationally less intensive, may smooth out some of these details, leading to a less precise delineation of flood boundaries, especially in areas with abrupt changes in elevation or complex flow paths.
Figure 15.
Comparison of inundation maps in APSFR 29 for T = 20 years.
Figure 15.
Comparison of inundation maps in APSFR 29 for T = 20 years.
Figure 16.
Comparison of inundation maps in APSFR 29 for T = 100 years.
Figure 16.
Comparison of inundation maps in APSFR 29 for T = 100 years.
Figure 17.
Comparison of inundation maps in APSFR 29 for T = 500 years.
Figure 17.
Comparison of inundation maps in APSFR 29 for T = 500 years.
For comparative purposes, floodplain maps were generated for each method, 1D/2D and 2D, for each return period. These maps provide a visual representation of the estimated floods and allow for the analysis and evaluation of the differences between the modeling methods. This comparison is crucial for understanding the impact of each method on the accuracy and reliability of flood predictions.
Figure 18 presents inundation boundaries that are quite similar for both simulation methods, 1D/2D and 2D. Any minor areas of inundation are attributed to the presence of agricultural lands, where water tends to accumulate rather than flow downstream. This similarity indicates that both methods produce comparable results in predicting flood extents, with variations primarily influenced by local land use and topography.
3.3. Simulation Time
Another parameter taken into account was the simulation and development time of both hydraulic models. As supporting studies show [
22], 1D/2D models are time consuming for the building of the model, taking about 7–10 days for each model. As for the hydraulic simulation time, by comparison, it only takes 2 days for the 2D models. As the following
Table 3 and
Table 4 show, the 2D simulation time might take longer than 1D/2D, although in terms of total time, the 2D method is better.
3.4. Depth Maps
The extraction of depth results through HEC-RAS was a fundamental part of this study. HEC-RAS 6.3.1 software offers robust capabilities for displaying and analyzing water depths, which is essential for flood risk management and hydraulic modeling.
Table 5 and
Table 6 show statistics of depth outputs that are in agreement with the inundation maps and the flooding area differences between the two methods. In general, these differences between the methods are acceptable and not critical. Max values of depth (>6 m) are explained by a wide area of agricultural uses.
Using the hydraulic simulation capabilities of HEC-RAS, detailed maps were generated depicting the spatial distribution of water depths across the study area. These maps include specific locations, designated as Positions 1, 2, and 3, where flood events have been reported, as analyzed in
Section 2.1. Highlighting these areas with historical flood occurrences provides valuable insights into the flood behavior of the region and supports the development of effective flood risk management strategies.
3.4.1. Depth Maps of APSFR 21–22
The following
Figure 19 illustrates the max depths of APSFR 21–22 for each return period of 20, 100, and 500 years, respectively. Comprehensive maps that depict the maximum flood depths obtained from each method facilitate a direct visual comparison. Also, locations reported for flooding incidents are shown. A high degree of congruence between the two simulation methods is revealed, suggesting that both approaches are capable of producing reliable predictions of flood extents and depths. These findings underscore the robustness of the combined 1D/2D approach while highlighting the enhanced spatial resolution and detail achievable with fully 2D simulation.
3.4.2. Depth Maps of APSFR 29
Figure 20 illustrates the max depths of APSFR 29 for each return period of 20, 100, and 500 years, respectively. Comprehensive maps that depict the maximum flood depths obtained from each method facilitate a direct visual comparison. Also, locations reported for flooding incidents are shown. A high degree of congruence between the two simulation methods is revealed, suggesting that both approaches are capable of producing reliable predictions of flood extents and depths. These findings underscore the robustness of the combined 1D/2D approach while highlighting the enhanced spatial resolution and detail achievable with fully 2D simulation.
3.5. Velocity Maps
The extraction of velocity results through HEC-RAS was a crucial component of this study. Using the hydraulic simulation capabilities of HEC-RAS, detailed maps were generated depicting the spatial distribution of flow velocities across the study area. The identification of these high-velocity zones is essential for understanding the hydrodynamic behavior of the river system and for implementing effective flood risk management strategies. The following figures highlight specific locations for APSFR 21–22 and 29, respectively. Elevated velocity values were observed, particularly within the river channel.
3.5.1. Velocity Maps of APSFR 21–22
Figure 21 illustrates velocities across the study area for each hydraulic simulation method (combined 1D/2D and full 2D) and each return period, with comprehensive maps that depict the velocity distributions obtained from each method, facilitating a direct visual comparison of areas with elevated velocities as well as the entire region of interest. Figures (a–c) and (d–f) show return periods of 20, 100, and 500 years, respectively. Indicatively, areas with higher velocities in the riverbed are emphasized. Analysis revealed a high degree of congruence between the two simulation methods, both in the overall velocity distribution and in identifying specific locations of high velocities. These findings underscore the robustness of the 1D/2D hybrid approach while highlighting the enhanced spatial resolution and detail achievable with fully 2D simulations.
3.5.2. Velocity Maps of APSFR 29
Figure 22 illustrates velocities across the study area for each hydraulic simulation method (combined 1D/2D and full 2D) and each return period, with comprehensive maps that depict the velocity distributions obtained from each method, facilitating a direct visual comparison of areas with elevated velocities as well as the entire region of interest. Figures (a–c) and (d–f) show return periods of 20, 100, and 500 years, respectively. Indicatively, areas with higher velocities in the riverbed are emphasized. Analysis revealed a high degree of congruence between the two simulation methods, both in the overall velocity distribution and in identifying specific locations of high velocities. These findings underscore the robustness of the 1D/2D hybrid approach while highlighting the enhanced spatial resolution and detail achievable with fully 2D simulations.
4. Discussion and Conclusions
Floods can pose risks to both human life and the environment, necessitating robust measures for effective management and mitigation. The European Union’s Directive 2007/60/EC on the assessment and management of flood risks provides a framework to reduce and manage these risks through comprehensive planning and preventive actions [
1]. In Cyprus, rainfall patterns can lead to severe flooding events, particularly in urban areas with inadequate drainage systems. The implementation of Directive 2007/60/EC in Cyprus involves identifying areas at significant risk and developing Flood Risk Management Plans (FRMPs) tailored to these zones. Hydraulic simulation plays a crucial role in this process by providing detailed models of flood behavior, helping to predict flood extents, depths, and velocities. These simulations support the development of effective FRMPs, ensuring that appropriate measures are taken to minimize the impact of flooding on vulnerable areas. The accuracy of flood prediction is crucial for the proper mitigation of flooding effects on people, infrastructure, and the environment.
Hydraulic simulation plays a key role in this purpose. The present research successfully investigated the two different behaviors between combined 1D/2D and full 2D models applying HEC-RAS 2D in two existing designated areas of potential flooding in Cyprus.
Specifically, the behavior and the results of the two different hydraulic modeling approaches were compared. The main findings of this research are summarized below:
Two-dimensional simulation provides a more accurate representation of complex and high flow conditions and hydraulic behavior.
Hydraulic models based on a fully two-dimensional grid show higher capabilities of capturing inundation areas where high slopes are present or in cases of flooding in numerous directions.
The 2D hydraulic simulation offers greater flexibility in modeling various hydraulic scenarios, such as urban and riverine flooding and hydraulic structure adaptions, making it suitable for a wide range of applications, alternative solutions, and enhanced resilience planning.
The 2D model is found to be more stable compared to the 1D/2D approach, where significant instabilities and computational errors are common.
Hydraulic modeling using the 1D/2D approach remains highly effective in various scenarios, as stated by recent research [
35,
36,
37]. The combined 1D/2D approach excels in environments with complex geometries, such as urban areas and intricate river systems, where it combines the strengths of both 1D and 2D modeling techniques. Recent studies highlight its efficiency in large-scale watershed simulations, where the 1D component efficiently handles extensive river networks, while the 2D component accurately captures floodplain dynamics [
38]. In urban flood scenarios, the 1D/2D method effectively models both the drainage network and surface flooding, providing a detailed and accurate representation of urban hydrodynamics. This method is also advantageous in scenarios requiring rapid simulations, such as real-time flood forecasting, due to its balance between computational efficiency and spatial accuracy [
39]. Moreover, the 1D/2D approach facilitates the evaluation of flood mitigation measures by allowing efficient scenario analysis and detailed spatial representation [
40]. Consequently, the 1D/2D modeling method is considered suitable for complex river systems, large watersheds, urban flood scenarios, and comprehensive flood risk assessments.
However, the rapid development of home or office computer capabilities has pushed current practice toward the full 2D approach [
41,
42], whereas already noted by the present research, many advantages are attributed when compared to the interim 1D/2D approach. Contemporary research has already noted the efficiency of 2D HEC-RAS in modeling urban flooding and suggested its incorporation in early warning systems or other flood management practices [
43]. Additionally, 2D modeling flood extents have been compared to satellite imagery of severe flooding and validated accordingly [
41]. Similar research has additionally validated the performance of 2D HEC-RAS modeling regarding flood depths, discharge, and velocities against delicate and secluded modeling software [
44].
Returning to the findings of the present research, full 2D modeling showed high identification results, with the approval of the Cypriot authorities, for combined 1D/2D hydraulic simulations, as far as inundation, discharge, flood depths, and velocities are concerned [
45]. In conclusion, the 2D approach is found to be capable of successfully capturing the flooding incident of different return periods, from common to rare events. Moreover, its greater ease of use and the fact it is less prone to computational errors nature compared to the 1D/2D approach, contribute to the suggestion for broader usage of the 2D HEC-RAS methodology presented for flood management. Additionally, the authors highlight the importance of validation with ground measurements, which was not possible in the present case. It is suggested for future research to follow the presented methodology for 2D modeling in a country and a river system with sufficient in situ data in order to provide further evidence for the efficiency of 2D hydraulic modeling.