Assessing Coastal Flood Risk in a Changing Climate for Dublin, Ireland

Abstract

With increasing urban expansion and population growth, coastal urban areas will be increasingly affected by climate change impacts such as extreme storm events, sea level rise and coastal flooding. To address coastal inundation risk for impact studies, integrated approaches accounting for flood hazard modelling, exposure and vulnerability of human and environmental systems are crucial. In this study, we model the impacts of sea level rise on coastal inundation depth for County Dublin, the most extensively urbanized area in Ireland, for the current period and for 2100 under two Representative Concentration Pathways RCP 4.5 and 8.5. A risk-centred approach has been considered by linking the information on coastal flood-prone areas to the exposure of the urban environment, in terms of potential future land cover changes, and to the socioeconomic vulnerability of the population. The results suggest significant challenges for Dublin city and the surrounding coastal areas, with an increase of around 26% and 67% in the number of administrative units considered at very high risk by the end of the century under a RCP 4.5 and 8.5, respectively. This study aims to contribute to existing coastal inundation research undertaken for Ireland by (i) providing a first-level screening of flooding hazards in the study area, (ii) demonstrating how land cover changes and socioeconomic vulnerability can contribute to the level of experienced risk and (iii) informing local authorities and at-risk communities so as to support them in the development of plans for adaptation and resilience.

1. Introduction

Coastal areas have always attracted people because of the uniqueness of their resources and most of the world’s mega-cities are located on the coast. Indeed, coastal urban expansion has been significantly higher, historically, when compared with inland areas [1,2]; this trend of coastal migration is expected to increase in the future [3,4], driven by economic growth and national policy choices [5]. Coastal flooding is the cumulative result of a range of relevant factors including climate-induced sea level rise (SLR), non-periodic water movements such as storm surges, tidal variation, wave climate and ocean currents. Low-pressure systems along with wind speed and direction can force water into harbours and estuaries, causing surge effects and extreme wave conditions. In the mid to long term, the population potentially exposed to a 100-year coastal flood is expected to increase by around 20% in the case of a 0.15 m SLR compared to 2020 levels [6]. This will result in the increased exposure of these areas to risks associated with coastal inundation by the end of the century [7].

Understanding sea level trends at a regional and local scale is thus crucial for the development of effective climate adaptation plans. In Ireland, more than 50% of the population is concentrated in the coastal zone, with the majority located in a few major urban centres, i.e., Dublin, Cork, Limerick, Waterford and Galway (Figure 1). The largest inward migration has been seen in Dublin, with an increase of more than 46,500 people in 2022 if compared to the previous census in 2016 [8] and a predicted increase of approximately 32% by 2036 [9]. These areas of increasing urban sprawl are expected to be particularly at risk in the future. Climate Central, a US-non-profit research organization, developed a global-scale coastal risk screening tool showing the land projected to be below the annual flood level in 2050 at a global scale [4]. Based on the projections of coastal flood heights in 2050 under a global temperature rise of 2 °C rise, many urbanized areas of Ireland are expected to be below sea level by then. Low-lying coastal zones, such as the Irish east coast, are particularly exposed to coastal erosion with SLR already having a significant impact on the soft boulder clay coasts of such areas [10,11].

The importance of SLR and the impacts of storm surges and wave patterns in coastal zones in Ireland has been highlighted by a number of authors in recent years, from different perspectives. Devoy [10] discussed the risks of climate-related events for Ireland and how the country should respond. O’Brien et al. [12] performed a survey of extreme wave events and storm surges in Ireland as far back as 14,680 BP (Before Present) up to 2017. Cámaro García et al. [13] highlighted how, based on satellite observations, sea level around Ireland has risen by around 2–3 mm/year since the 1990s but the most recent estimates for the period 2006–2015 indicate a rise of 3.6 mm per year [14]. A recent study by Shoari Nejad et al. [14] investigated the sea level trend in Dublin Port by performing a careful quality control of the available tide gauge data. They concluded that the overall sea level rise is in line with expected trends, but higher rates have been observed since the 1990s. They point to large multidecadal variability as the reason for these large recent increases [15]. The recent 2019–2024 climate change action plan published by Dublin City Council [16] indicates an increase of approximately 6–7 mm per year between 2000 and 2016, which is almost twice the rate of global SLR. This is surprising, considering that studies over past decades (i.e., between the 1950s and 2000s) in Dublin revealed much lower rates than the global mean SLR [14]. This highlights Dublin as being at particular risk. The Dublin area experienced several coastal flooding events, which have caused severe impacts in the past [17]. The future sea level in the North Atlantic and along the Atlantic coasts of the UK and Ireland is projected to rise by nearly 1 m under RCP 8.5 [18] and, with a projected increase in the intensity and magnitude of extreme storms, this will increase coastal damage [19]. Predicted changes in SLR will be the primary factors in magnifying the impacts of wave patterns and changing storm surges in urban coastal areas [20,21].

In urbanized environments, such as the Dublin area, the capacity of urban drainage systems, the alteration of land cover and the fragmentation of water systems pose challenges for flood mitigation, especially as cities increase their footprint and encroach on flood plains. Last but not least, the impacts of climate change on critical assets (e.g., transport networks) in urban areas and their hinterlands pose a threat to the provision of basic services (e.g., food, water, energy), highlighting the intrinsic vulnerability of such systems and underlining the need for long-term planning strategies in a context of on-going and projected climate change [22,23,24]. Similar criticalities are common to other European islands. For example, the Canary Islands, where coastal tourism is a major economic activity, have seen an increase in exposure to marine storms due to growing anthropogenic pressure on the coastline [25]. In the Balearic Islands, sea level rise and extreme events lead to widespread economic and ecological impacts, in particular in highly exposed urban areas. Here, scenarios of sea level rise suggest a permanent flooding extension by the end of the century, with several losses in terms of basic and recreational services [26]. In Sicily, coastal flooding related to common storms and to medicane events produced extensive flooding and significant damage to public infrastructure, as well as erosion effects on beaches [27]. In Malta, recent studies performed an index-based assessment integrating physical exposure and social vulnerability into an overall index, to identify areas that can be negatively affected by coastal and marine-related processes [28].

Planning is not only a matter of the biophysical environment of a coastal area, but also relies on a range of socioeconomic, physical and political aspects that together determine the context in which possible actions can be taken [29,30]. Urban land-use planning can play a key role in the framework of disaster risk reduction (DRR) strategies. Planning regulations can help to reduce the exposure to hazards but are not designed to deal with emergency situations. In this case, the DRR plans should be integrated with structural measures. Nonetheless, the potential negative impacts and repercussions of these structural protection systems on ecosystems, property and human lives should be taken into account [31]. To implement adaptation measures aimed at coastal flood risk reduction, the adaptive capacity of the systems to respond to hazards is as crucial as their inherent susceptibility [32].

Globally, there is extensive literature on physical and socioeconomic vulnerability to coastal flooding risks [28,30,33,34,35,36,37,38,39,40,41,42,43]. In the Irish context, recently, some significant attempts have been made to assess the vulnerabilities and impacts of coastal flood hazards. A recent study by Caloca-Casado (2018) focused on the physical vulnerability of the coastline through a vulnerability index-based approach in County Dublin. The method focused only on physical parameters to explore the linkage between geological boundary conditions, coastal indicators (e.g., tidal range, wave height, long-term shoreline erosion, geomorphology), sea level rise and the coast’s response. Potential flooding areas, by the end of the century, have been mapped based on a 0.5 Annual Exceedance Probability (AEP) and 1.98 m sea level rise, under the worst-case scenario RCP 8.5. Flood and Sweeney (2012) explored the potential economic impacts induced by SLR on Irish coastal cities. The authors found that vulnerable lands would be around 350 and 600 km² under a 1 and 3 m SLR, respectively, with a related potential economic cost of around 1.1 and 2.1 billion euros. This study is one of the few highlighting, from an economic perspective, the need for planning actions and adaptive approaches addressing SLR impacts on the Irish coastline.

As defined by the International Panel on Climate Change [44], risk results from the interaction of hazards, exposure and vulnerability. A hazard is any physical phenomenon (e.g., increased coastal erosion, frequency and intensity of coastal storms, landslides, flooding, extreme air temperature, etc.) having the potential to cause damage and loss to natural and human systems. Exposure is the presence of people, animals, livelihoods, ecosystems, infrastructures and assets in general that could be adversely affected by climate-related natural hazards. The vulnerability concept encompasses the sensitivity and adaptive capacity aspects. Adaptive capacity entails the ability of the human, environmental, ecosystems and socioeconomic systems to cope with the impact of climate change, to adjust to its adverse consequences or take advantage of potential opportunities. Sensitivity is the susceptibility of a system to be adversely or beneficially affected by the effects of climate-related hazards. Thus, while assessing risk in flood-related studies, it is crucial to take a holistic approach considering all inherent components of risk, i.e., hazard exposure and, last but not least, vulnerability, driven by a range of socioeconomic and demographic factors as well.

People and areas experience different levels of impact based on their degrees of vulnerability [39]. The capacity of the population and related systems to respond to such hazardous events is important both for the assessment of the impacts and for the implementation of fit-for-purpose resilience strategies and policies [30]. A joint assessment of the socioeconomic and demographic features of the population along with the critical environmental exposure is crucial in determining the levels of current and future risks related to hazardous flood events, especially in the dense urbanized areas [40,45,46].

Comprehensive approaches considering all components of risk (i.e., hazard, vulnerability and exposure) to mitigate climate-related natural hazard risk show promise in the context of disaster risk reduction (DRR) strategies. Indeed, since the establishment of the International Strategy for Disaster Reduction (ISDR) in 2000 by the UN General Assembly (successor of the International Decade for Natural Disaster Reduction (IDNDR)) the need for a holistic risk reduction approach has been realized [47,48].

To the best of our knowledge, to date in Ireland, a comprehensive risk assessment, which takes into account the full range of factors that contribute to the definition of coastal flood risk, is still lacking. Therefore, the main aim of this paper is to determine levels of current and future risk of coastal flooding in the Dublin area based on an integrated assessment of hazard, exposure and socioeconomic vulnerability. We model the coastal flood risk for the current period and for the future under different Representative Concentration Pathways (RCPs). Future projections are computed up to 2100, under RCP 4.5 and 8.5 which are the most widely used scenarios informing planning [29]. We thus consider potential impacts from set SLR amounts covering different scenarios. Nevertheless, it has to be noted that the aim of this study is not to provide a detailed description of the inundation patterns, but to provide a picture of the flooding hazard across the study area. Nonetheless, the outcome of this risk-screening study provides information on the areas subject to coastal flooding that require a greater focus and, in turn, a more detailed and computationally expensive modelling task. In addition, the study aims to demonstrate how, in addition to projected climate change, modelled future land-use changes will also impact the level of risk experienced across the study area. Finally, we aim to provide stakeholders with tools and information to improve the management of the impacts of SLR in the region and in other coastal urban areas with similar risk profiles. The maps and information provided in this study are provided at the community level; thus, they should be informative and raise awareness among the local authorities and organizations of the likelihood of flooding in such areas and on the main adverse impacts of these predicted floods in each scenario.

The paper is organized as follows. After a brief presentation of the study area (Section 2), the main data and methods to assess the three components of coastal flood risk, i.e., hazard, exposure and vulnerability are presented (Section 3). The results are then presented (Section 4) and discussed (Section 5), along with the main conclusions (Section 6).

2. The Study Area

The focus of this study is County Dublin (Figure 1). County Dublin includes 4882 Small Areas (SAs) as defined by Ireland’s Central Statistics Office (CSO). SAs are the smallest administrative units in Ireland designed for the compilation of statistics and generally comprise either part or complete neighbourhoods or townlands (CSO, 2019b). Dublin, which is the capital of the Republic of Ireland, is the largest urban area of the region and of the country. The extent of the investigated area is approximately 700 km² (boundary coordinates: 53.29° N, 6.39° W: 53.41° N, 6.11° W, WGS 84).

County Dublin is generally a low-lying basin and many areas are at or near sea level. The river Liffey, which is 132 km in length, drains most part of the region and reaches the sea at Dublin Docklands. Dublin Bay features beaches while the remaining part of the coast is either muddy or rocky (e.g., Howth Head). The River Liffey and Dodder River meet in the heart of Dublin. This, combined with the presence of other numerous small rivers and canals, make the hydraulic and hydrologic situation in the Liffey estuary even more complex [23].

Dublin County’s coastline is particularly exposed to coastal flooding and erosion. Erosion rates exceed 3 m per year during severe extreme events [11]. Irish Sea surges are associated with the movement of Atlantic depressions over the Irish Sea basin. Models suggest that sea surge events will increase in the future in coastal areas with an increase in both the frequency and heights of extreme storm surges (over 1 m) [11]. Crossing the Atlantic, the storms pass over the warmer surface temperature of the Gulf Stream, which leads to a significant increase in wave heights in the Irish Sea. Extreme surges occur mainly in the winter period [11].

Maps showing current and future flood-prone areas based on the typology of flood in a given area for a flood event of a given probability are provided by the Office of Public Work (OPW). This is the government office that delivers public services for flood protection, and heads the National Catchment Flood Risk Assessment and Management (CFRAM) programme, [17]. Across County Dublin, OPW has identified areas from Skerries down to Swords, Malahide and Howth at risk due to many concurrent factors. In Dublin city, one of the most exposed areas is Clontarf and particularly the coastline [49]. This is due to high tides during storm events, leading to large waves overtopping the existing seawall, and coastal inundation. In the current scenarios, many roads, properties and amenities will be affected by coastal flooding. The Lower Liffey River is exposed to both fluvial and coastal flooding, with a number of business and residential properties and infrastructures located in the floodplain potentially at risk of flooding.

3. Data and Methods

As defined by the International Panel on Climate Change [44], risk results from the interaction of hazard, exposure, and vulnerability. Figure 2 shows the workflow adopted in this study. Hazard to coastal flood has been simulated by means of a 2D software package for hydrodynamic modelling (DHI Mike 21©) which requires a range of data inputs. One hazard indicator, i.e., the inundation depth, has been developed for coastal inundation risk. The model allows us to assess the likely inundation depth for the region of interest for the current period and for two different RCPs scenarios at the end of the century. Once coastal inundation has been simulated, the outputs have been compared against OPW inundation maps for visual comparison. For model evaluation, data from a past inundation event have been obtained. A single surge event in 2013 has been simulated and validated against past records. The same event has thus been simulated for a number of different RCPs scenarios (climate forcing) which are differentiated by mean sea water levels. Inundation depths values have been normalized to make them comparable across spatial and temporal scales and then categorized into five classes (very low, low, average, high, very high) to retrieve the Hazard index (H). Exposure to a coastal flood has been modelled in terms of urban extents across County Dublin based on Corine Land Cover (CLC) data. Classes of land cover have been related to a standard framework based on Local Climate Zones (LCZ) which describes urban morphology as a function of land cover changes, surface structure, surface material and human activities. Land cover changes have been assessed for the current period up to the end of the century. The percentage of non-urban and urban areas have been computed and, as for H, an Exposure index (E) has been computed. The vulnerability assessment has been performed only for the current period and is based on census socioeconomics data and open data retrieved by different national datasets and open free maps. Given the number of different variables involved, a methodology based on a Principal Component Analysis (PCA) has been used to reduce the complexity of the system by highlighting the most important factors driving vulnerability in the region of interest. As for H and E, the Vulnerability index (V) is then computed. The Risk Index (R) has been finally computed as the product of H, V and E across SAs for the current period and under RCP 4.5 and 8.5 in 2100, and categorized into five classes. Further details on how the different risk components have been addressed in this study, as well as their integration, are presented hereinafter.

3.1. Coastal Inundation Modelling

In this work, the inundation depth (expressed as the total water depth locally reached in one cell) has been used as the hazard indicator for coastal flood risk and has been simulated using the DHI Mike 21©, which is a leading software package for 2D modelling of hydrodynamics, waves, water quality, ecology and sediment dynamics in general. The Mike 21© includes cell wet and dry functionalities which allow the assessment of the propagation and dynamic of the inundation over the floodplain.

This model was employed to assess the likely depth of coastal inundation for County Dublin based on projecting a past surge event (specifically, Cyclone Xaver, 2013) under two mean Sea Level Rise (mean SLR) scenarios at 2100. The mean SLR estimated for the end of the century is 0.45 m and 0.81 m under RCP 4.5 and 8.5, respectively.

Cyclone Xaver was a winter storm that developed on 4 December 2013, characterised by high wind speeds and low atmospheric pressure which occurred during a spring tide phase. The storm persisted for around six days along the north coast of Europe, causing one of the worst storm surges for decades in the North Sea [50,51]. The storm’s low-pressure system formed off the coast of Greenland, passed northwest of Ireland, crossed the UK and moved to northern Europe resulting in a cold air outbreak southward of the North Sea and in strong northwest winds [52]. The convective process was associated with freezing precipitation and strong wind gusts all over the affected regions. The storm had several serious impacts on energy and coastal societal infrastructures in general across Ireland and other countries [51].

In order to simulate coastal inundation depths, a number of data inputs are required. These encompass topographic and metocean data and include, e.g., bathymetry, coastline, digital terrain model, seabed roughness, wave climate, tidal elevation, wind time series and measured water level time series [53]. The model was built by integrating the Digital Terrain Model (DTM) data retrieved by the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model (GDEM), which was then clipped around the study area (Figure 1), with the bathymetric data from Infomar (2021). The model computational domain is made on unstructured flex-mesh with triangular elements and with element size resolution increasing toward coastal areas, reaching a maximum spatial resolution of 45 m on the coastline, which is typically a sub-optimal resolution for a detailed coastal flooding assessment at a very local scale. Nevertheless, it has to be noted the coastal inundation model is just part of a more complex modelling framework, including heat stress and pluvial flooding risk assessment, which required a common lower-resolution DTM for the cross-calculation of hazard indices [53,54,55]. As such, the resolution of the model is coarser than what would be required for very detailed local studies; nevertheless, it allows to capture, in broader terms, flooding hazard patterns across the study area.

The model covers an area on the Irish East coast bounded by Warrenpoint (54°1′ N, 6°25′ W) and Ballymoney (52°68′ N, 6°22′ W) (Figure 1) and was forced at the two open boundaries, i.e., the North and South boundaries, by measured water level time series. The eastern-most side of the domain is treated as a closed boundary. The boundary conditions for the northern and southern open boundaries were generated by standard tidal-constituent analysis. The water level time series includes the storm surge generated by atmospheric forcing. More in detail, the storm surge was accounted for by the wind stress and atmospheric pressure corrections at the model open boundaries. The open boundary water level time series were generated by using 8 tidal constituents (semidiurnal: M2, S2, N2, K2, diurnal: K1, O1, P1, Q1) derived by the DTU10 Global Ocean Tide Model. The Mike21 flow model allows for wind stress and atmospheric pressure correction at the open boundary, hence to account for the meteorological-induced surge. Wind and atmospheric pressure time series were derived from the Dublin Airport met-station. Having used a single-point measurements station, the spatial variability of the wind and atmospheric pressure fields was neglected. Wave generation and propagation were not simulated. Given the large computation domain, a dynamically coupled flow-wave model was not performed due to computational resource incompatibility with the timeline of this study. Simulation and propagation of wind waves, which would require a detailed description of the defence structures along the coastline in contrast with the low resolution of the DTM, was out of the scope of this study. Precipitation was not considered as only the coastal inundation dynamics were addressed in this paper.

The model was calibrated for a period of about two weeks (1 December 2013–12 December 2013), i.e., including both a neap and a spring tide, against measured water levels at Dublin Port (53°20′44.5″ N, 6°13′18.1″ W), Skerries (53°35′06.0″ N, 6°06′29.2″ W) and Howth (53°23′28.7″ N, 6°04′05.2″ W) tide gauges. The model is calibrated and run for the same period and demonstrated good performance (

Table 1. Error metrics for Dublin Port, Howth and Skerries tide gauges. The model was calibrated and validated during the same period 1 December 2013–12 December 2013. MAE: mean absolute error; RMSE: root mean square error; ubRMSe: unbiased RMSE; r: Pearson’s Correlation Coefficient.

	MAE	RMSE	Bias	ubRMSE	Scatter Index	r
Dublin Port	0.1367	0.1793	0.0289	0.1770	0.146	0.9906
Howth	0.1087	0.1323	0.0507	0.1222	0.0978	0.9974
Skerries	0.0612	0.0797	0.0069	0.0795	0.0566	0.9985

). These values are in line with previous studies [56,57]. To obtain an optimal calibration, bottom roughness and wind friction parameters have been adjusted. Wind friction varies with wind speed. The wind friction coefficient (drag coefficient) was adjusted to achieve optimal calibration. The values range between 0.00125 and 0.0038 depending on wind speed. Bottom friction is determined by a classic quadratic friction law. For the drag coefficient values, Manning numbers ranging between 32 and 36 m^1/3/s were used. It is also worth noting that the Manning number used in the DHI Mike21 flow model is the reciprocal value of Manning’s number described in some textbooks.

SLR for the two RCP 4.5 and 8.5 scenarios in 2100 was considered by adding the level increase to the water level boundary conditions, i.e., by shifting up the mean water level value by a quantity equal to the SLR.

The predicted depth of flooding at a given location is indicated on inundation depth maps. Maps have been produced for the present-day situation (referred to as the Current Scenario) and for the future scenario 2100 under two different RCPs. These maps have been resampled in a GIS system at a linear resolution of 100 m in the form of a square cell to allow easier interpretability by the readers. Conservatively, the inundation depth is shown as the maximum value reached at any given time step and by any model output cell within a 100 m horizontal resolution square cell. As previously stated, due to the DTM resolution available for this study the model does not include any currently existing or future coastal hard defences that could mitigate flooding in specific exposed zones. The inundation depth values, as shown in Figure 3, have been mapped and then averaged values have been obtained for each SA.

3.2. Exposure to Coastal Flood

Corine Land Cover (CLC) data [58] have been used to model urban extents across County Dublin and represent a proxy of the influence of urban dynamics on exposure to climate change impacts.

Using a Land Change Modeler (LCM) module in a GIS environment [59], the CLC data have been modified to fit the modelling requirements. Land Cover data have been obtained at a resolution of 30 m based on the approach in Alexander and Mills [60].

In this work, the non-urban land cover classes have been simplified while the same level of resolution was kept for the urban land cover resulting in 4 urban and 7 non-urban classes (for a total of 11 land-use classes). These classes have been associated with the Local Climate Zones (LCZ) classification [61] as also detailed in previous work applied to Dublin city [55,60]. Although the LCZ classification is usually used in urban heat studies, the incorporation of a LCZ approach here allows important urban morphological features to be captured introducing heterogeneity in the morphological features of the urbanized areas, as also demonstrated in previous works on urban rainfall simulations and in environmental studies [62,63]. Indeed, differences in land cover and surface morphology drive intra-urban variations in local hydrology and air temperature, in terms of runoff and subsequent landscape response to extreme events. We refer to previous works for the outline of the LCZ present in Dublin city in the last decade and the main related properties [55]. This approach is valid for both the current situation and for generating future land-use scenarios in five-year cycles from the current period up to the end of the century [60]. Historical transitions (i.e., the Markov Chain and time-series of land-use maps) are used to model forward to user-specified dates based on transition potentials. These are informed by restrictor variables (e.g., proximity to special protected areas or areas designated as flood zones) and predictor variables (e.g., distance to roads, slope of the landscape).

For each land-use scenario, we computed the percentage of land area with natural spaces (defined as land with green-blue spaces from now on), i.e., green space (e.g., vegetation cover, parks, street trees, woodlands), coastal-blue spaces (e.g., beaches, promenades) and inland-blue spaces (e.g., lakes, rivers), in each SA and thus, the percentage of the impervious area accordingly.

3.3. Vulnerability to Coastal Flooding

To assess the sensitivity and adaptive capacity of Dublin County to climate impacts, the 2016 national census [64], land cover data and information from the Ordnance Survey Ireland [65], Ireland Open Data Initiative [66] and OpenStreetMaps (OSM) have been employed at the SA scale, i.e., the smallest available administrative unit [64].

To determine levels of vulnerability, a range of 14 possible meaningful variables were first considered (

Table 2. Variables used to assess vulnerability, related socioeconomic and physical aspects and rationale for inclusion, with references.

Dimension	Proxy Variables	Rationale and References
Urbanization	Population density per km2	Proxy of urbanicity [42]
Housing	% population without house ownership	Proxy of deprivation [55]
Housing	% of old buildings (<1980)	The quality of a home could determine the susceptibility of the dwellings to flood damage [40,74]
Social isolation	% population with one-person households	This is a proxy of social isolation [38]
Low income	% population with income below poverty level (unemployed residents having lost job or due to sickness)	Low income is a proxy of deprivation which could impact negatively on the capacity to adapt and react to hazardous events [67]
Disability	% population with disability	It hampers the ability to react and recover from a flood event [34,78]
Health status	% population with poor health (self-reported health status “not good”)	This is a proxy of deterioration of health and difficulties to recover from hazardous events [67,78]
Extreme ages	% population with age <15 or >65 years (extreme ages)	Mobility constraints hamper the ability to evacuate [35,67,69]
Low education	% population with less than a secondary school diploma (low-level education)	It affects the adaptive capacity of and ability to respond and mitigate [74]
Information access	% population with internet access	It influences the ability to access to hazard-related information [74]
Language isolation	% population with a low level of English	Language barriers could affect the ability to catch weather alerts [38,39]
Travel	% population without a car	Owning a car is indicative of mobility and easier access to services [79]
Network	Roads, Rails, Bicycle network density per km2	Transportation infrastructure has clear implication for susceptibility of the urban areas [42,46]
Critical facilities	No. of facilities: energy (e.g., power plants currently in operation), transport (e.g., airports, ports, bus stations), social services (e.g., nursing homes, children centres, etc.) water and wastewater infrastructure, education (e.g., schools, universities, etc.), hospitals and clinics, public administration, security services (e.g., police, firefighters’ stations), industrial and leisure facilities	These are key public and economic services and potentially susceptible to flooding [46,74]

), based on previous research identifying major factors that have implications for coastal inundation-related impacts [46] and on the availability of the CSO dataset.

The sensitivity and adaptive capacity of an area are dependent on a range of factors, including demographic and socioeconomic ones [30,39,67,68,69]. Socioeconomic disparities are linked to the differential capacity to react, cope and recover from such natural disasters [67,70]. To quantitatively model socioeconomic vulnerability, researchers have used empirical methods to extract associated sociodemographic variables and assemble them into indicators of socioeconomic vulnerability. In this study, the indicators of demographics and socioeconomics at the SA scale have been retrieved from the 2016 national census data (CSO, 2019).

Physical drivers of vulnerability should account not only for demographic and socioeconomic features of the population exposed to coastal flooding but also for the urban development and anthropic assets in the investigated area [71,72] in terms of critical facilities. Critical facilities are elements of the infrastructure that support essential services in a society. They include transport systems, industrial facilities, electricity, water and communications systems, hospitals and health clinics and centres for fire, police and public administration services. In this study, critical facilities refer to the following sectors and services: energy (e.g., power plants currently in operation), transport (e.g., airports, ports, bus stations), social services (e.g., nursing homes, children centres, etc.) water and wastewater infrastructure, education (e.g., schools, universities, etc.), hospitals and clinics, public administration, security services (e.g., police, firefighters’ stations) and industrial and leisure facilities close to the coast. The density of the transport network is another main factor of vulnerability. Road (e.g., primary, secondary roads, highway, motorway), rail and bicycle network have been extracted from OSM as well.

Table 2 presents the variables considered in this study based on the existing literature.

In order to detect the most important factors of vulnerability contributing to coastal flood risk, a multivariate analysis—Principal Component Analysis (PCA)—has been performed. This is a widely used method in many climate-related risk assessment studies to eliminate redundant data by revealing principal factors which best describe variation in the data [42,53,73,74]. Since vulnerability variables are measured in different units thus having different ranges and scales, a normalization procedure has been performed first. Values have been scaled between 0 and 1 based on minimum-maximum normalization. In order to ensure that high index values are indicative of higher levels of vulnerability, variables related to adaptive capacity thus potentially decreasing risk have been reversed. The computation of the Vulnerability index (V) has been based on the following steps.

(i)

Since it is likely that strong correlations exist among socioeconomic variables, a Pearson correlation matrix was computed to assess the degree of pair-wise association among the considered variables.

(ii)

In order to minimize the number of variables involved, an orthogonal varimax rotation method has been used. This enables a simpler structure to emerge where each variable is loading on as few components as possible and to maximize the correlation to one principal component at a time, increasing the interpretability.

(iii)

The final selection of the principal components has been based on a set of requirements, namely: (i) the Kaiser criterion retaining components with eigenvalues greater than one [75], (ii) total explained variance above 70 % and (iii) variables loading on multiple components have been removed from the analysis.

(iv)

The component scores are the scores of each SA on each principal component (Equation (1)). This is calculated by considering the SA’s standardized value on each variable, multiplying this by the corresponding loading of the variable for the given principal component factor and summing these products [76].

$$\begin{gathered} y_1=a_{11}{z}_1+a_{12}{z}_2+\cdots +a_{1p}{z}_p \\ {y}_2=a_{21}{z}_1+a_{22}{z}_2+\cdots +a_{2p}{z}_p \\ {y}_i=a_{i1}{z}_1+a_{i2}{z}_2+\cdots +a_{ip}{z}_p \\ {y}_p=a_{p1}{z}_1+a_{p2}{z}_2+\cdots +a_{pp}{z}_p \end{gathered}$$

(1)

where i = 1,…, p, being p the total number of variables involved, y_i (i = 1,…, p) is i-th the principal component score, z_i (i = 1,…, p) is the i-th standardized variable, a is the factor loading (or weight) of the i-th variable z for the given principal component. The first principal component accounts for the maximum proportion of the variance of the set of z, the second one for the maximum of the remaining variance and so on until the last principal component [76].

(v)

The composite vulnerability index score for each SA is computed by summing the products of each factor score weighted (Equation (1)) by the corresponding variance explained by each principal component. The weighted approach allows for the consideration of the different contributions of the principal components to explain the spatial patterns of vulnerability to coastal flooding [77].

3.4. Coastal Flood Risk Assessment

The final output of the hazard, vulnerability and exposure models have been scaled between 0 and 1 based on minimum–maximum normalization across all periods and scenarios to make them comparable (Equation (2)). Normalization allows for the adjustment of values measured at different ranges and prevents the generation of spatial biases caused by very large/small SAs.

$$z=\frac{\left(x-x_{min}\right)}{(x_{max}-x_{min})}$$

(2)

where x is the real value of the variables, x_min is the minimum value (for H, referring to the period current scenario), x_max the maximum value (for H, referring to period RCP 8.5 in 2100) and z the normalized value of the variable [80]. To compute the Hazard index (H), Exposure index (E) and Vulnerability index (V) for each scenario, normalized values have been grouped into five categories of increasing hazard/exposure/vulnerability (from very low to very high) using equal intervals to show the spatial distributions at the SA scale.

The Coastal Flood Risk Index is determined by multiplying the Hazard (H), Exposure (E) and Vulnerability (V) indexes in accordance with the three different scenarios considered namely, the current scenario, and 2100 under RCP 4.5 and RCP 8.5. Equal weights are adopted to produce the Coastal Flood Risk, thus giving the same importance to three aspects.

$$R=H\times E\times V$$

(3)

In the same way as for the other indices, the final Risk index (R) score is classified into five classes using equal intervals from very low to very high risk. Since vulnerability data refer only to the current situation, changes in risk in the future are only based on changes in the exposure and hazard components.

4. Results

Hereinafter the main results from the coastal flood model, outputs of the exposure and vulnerability analysis and the final coastal flood risk maps are presented. The indicators H, E, V and R are shown as one mean value within each SA; this does not imply that all of that SA will be equally affected, exposed or vulnerable to coastal flooding but that a mean value within the SA has been adopted.

4.1. Projections of Coastal Flood Hazard and Exposure

Figure 3 shows the modelled inundation depth for the current situation and under RCP 4.5 and 8.5 scenarios, with a mean SLR of 0.45 m and 0.81 m, respectively. The highest surge height recorded by tide gauges during the 2013 event was 3.4 m (with respect to Malin Head ordnance datum). Figure 3 is a GIS reprocessing of the model output, i.e., a resampling at a square cell resolution. This is thus an indirect output of the model. This is mainly for graphical reasons as an unstructured mesh with triangular elements would be difficult to interpret by the users. The single square cell value in Figure 3 is thus the maximum value reached at any time by triangular elements of the mesh (model output) within the square cell. Values of the unrealistic inundation depth close to model open boundaries and at harbour structures caused by boundary effects and low DTM resolution, respectively, were removed during the post-processing of the outputs. After the application of the procedure described in Section 3.1, the inundation depth is categorized into five different classes of H at the SA scale (Figure 4).

Figure 4 shows that, for the current period, in Dublin city areas along the banks of the Liffey river, as well as along the Monkstown and Dun Laoghaire coast, H ranges from very low to low H (classes 1 and 2). SAs in H class 3 is seen in the Clontarf area. Outside the core city up to the north, the inundation depth in class 3 are seen in the southern parts of Howth head, close to Malahide and in Baldoyle Estuary Nature Reserve. Areas of low H (class 2) are seen in Balbriggan, close to the northern county boundary. Here the modelled inundation depth is above 2 m in some areas (Figure 3).

Based on the outcomes of the model, beaches around Howth, Malahide, Portmarnock, Rogerstown and up north to Balbriggan will be particularly adversely affected by 2100 under both RCPs, with potential inundation depths in excess of 2 m (Figure 3). Much of the rest of the coastline, including Dublin city centre, will be affected as well, with projected inundation depths exceeding 1 m (Figure 3).

In terms of H, under a RCP4.5, and even more so under RCP8.5, flooding is expected to increase, both in extension and in depth. Out of 382 flood-prone SAs, those at very low H (class 1) decreased by approximately 16%, whereas there is a 9% increase in average and high H (classes 3 and 4). Under RCP 8.5, a decrease of more than 40 % in the areas at very low H (class 1) is seen with an increase of around 18% in the area at average (class 3), high (class 4) and very high (class 5) H (Figure 4).

The enhanced exposure due to land cover changes varies and increases gradually across decades. As stated in previous studies (Alexander et al., 2016), compact midrise (LCZ2), open low-rise (LCZ6) and large low-rise (LCZ8) zones capture together two-thirds of the urban types in Dublin City and neighbouring areas. Projections indicate that Dublin city centre will remain quite stable in terms of levels of urbanization, whereas suburban areas will expand (Figure 5). Exposure on the coastline appears stable except for the northern areas of the County around Rush, Rogerstown and Portmarnock.

4.2. Vulnerability to Coastal Flooding

Out of 14 variables, two (population density and access to communication media) were omitted during the PCA process since the loadings occurred on multiple factors thus complicating the interpretation. Based on the Kaiser rule, four components have been retained. These account for approximately 77.5 % of the total variance.

Table 3. Pearson’s correlation coefficient for final selected sensitivity and adaptive capacity variables in County Dublin. Values with asterisk show non-significant correlation with p-values > 0.05.

Variables	Low Income	No Car	Not a Property Owner	Social Isolation	Low Education	Social Language Isolation	People with Disability	Poor Health Status	Extreme Ages	Old Buildings	Transport Network	Critical Facilities
Low income	1	0.40	0.28	0.16	0.76	0.26	0.63	0.60	0.00 *	0.06 *	0.05 *	0.07 *
No car		1	0.79	0.53	0.17	0.05 *	0.14	0.32	−0.60	−0.16	−0.16	−0.10
Not a property owner			1	0.38	−0.10 *	0.05 *	−0.07	0.14	−0.70	−0.50	−0.16	−0.09 *
Social isolation				1	0.06 *	0.05 *	0.34	0.33	−0.15	0.16	−0.12	−0.07 *
Low education					1	0.18	0.68	0.56	0.30	0.33	0.14	0.11
Social language isolation						1	0.16	0.19	−0.02 *	−0.06 *	0.07 *	0.09 *
People with disability							1	0.68	0.37	0.41	0.08 *	0.10
Health status								1	0.15	0.21	0.03 *	0.07 *
Extreme ages									1	0.50	0.09 *	0.08 *
Old buildings										1	0.03 *	−0.03 *
Transport Network											1	0.79
Critical facilities												1

shows Pearson’s correlation coefficients above the selected variables and

Table 4. Factor loadings and variance explained from Principal Component Analysis (PCA) for the vulnerability assessment. Significant factor loadings for each PC are over |0.4| and with an asterisk.

Variables	PC1	PC2	PC3	PC4
Low income	−0.49 *	−0.12	0.02	−0.23
No car	−0.14	−0.50 *	0.03	0.18
Not a property owner	−0.02	−0.55 *	0.01	−0.02
Social isolation	−0.11	−0.27	−0.01	0.59 *
Low education	−0.48 *	0.09	−0.01	−0.12
Social language isolation	−0.24	−0.03	0.02	−0.51 *
People with disability	−0.46 *	0.08	−0.01	0.17
Poor health status	−0.43 *	−0.06	0.00	0.09
Extreme ages	−0.15	0.49 *	0.03	0.10
Old buildings	−0.15	0.31	0.04	0.48 *
Transport Network	0.00	0.02	−0.70 *	0.00
Critical facilities	−0.01	−0.02	−0.71 *	0.00
Variance explained	28.8%	24.9%	14.8%	9%
Eigenvalues	3.5	2.9	1.8	1.1

shows the results of the PCA for the current period. The first two components are indicative of the socioeconomic and health dimensions. The third is more of an urbanicity component and includes a transport network and the presence of critical facilities in the area. The last, even if less significant, is indicative of both isolation and deprivation, since it includes social and language isolation along with a variable indicative of the age and quality of the buildings.

Figure 6 shows the spatial distribution of the Vulnerability index for the current period across County Dublin. Most parts of coastal flood-prone areas are ranked in the highest classes of the vulnerability index V (classes 4 and 5). Specifically, 51 % and 27 % out of the total 382 coastal flood-prone areas are in the very high and high V classes (classes 4 and 5), respectively. This is particularly true for Dublin city centre, i.e., for the most urbanized area in the county. Here, a significant vulnerability is also seen along the River Liffey banks and connected canals. In the northern part (Skerries, Rush and Balbriggan), there are numerous adjacent areas of lower and higher V as well.

Disaggregating the components, it appears that the more disadvantaged areas from a socioeconomic point of view are those around the inner parts of the core city centre along the River Liffey Banks. This is also confirmed by the Pobal HP Deprivation Index which measures the relative affluence or disadvantage of all SAs around Ireland based on socioeconomic variables [81]. Areas close to Dublin Port and River Liffey and Tolka Estuaries (e.g., Clontarf) up north to Howth and down to Sandymount, Booterstown up to Dun Laoghaire are seen as vulnerable as well. These are areas of high-density population, which are not classified as socioeconomically disadvantaged but rather considered critical because of the presence of many residential areas, business activities, several social amenity sites, cultural heritage, public, social and transport infrastructure assets. Similarly, outside the capital, Skerries and Rush urban areas are highly vulnerable (classes 4 and 5). Balbriggan (north) and Shankill area are seen as particularly vulnerable (classes 4 and 5) as well. Here, the presence of critical facilities and coastal transport plays the most important role in determining high V (classes 4 and 5). Areas around beaches and natural reserves (e.g., Dollymount strand, Rogerstown Estuary Nature Park, Howth and Portmarnock), which are critical leisure facilities, are considered particularly vulnerable to coastal flooding too.

4.3. Coastal Flood Risk Projections

Three different R maps have been generated (Figure 7): (i) present-day inundation based on the 2013 surge event, (ii) future inundation for a similar event in 2100 due to SLR under RCP 4.5 and (iii) future inundation for a similar event in 2100 due to SLR under RCP 8.5. As can be seen, coastal flood risk continues to increase both spatially and temporally. Between the current and future period, areas at high and very high R (classes 4 and 5) are expected to double, and the same occurs between RCP 4.5 and 8.5. More in detail projections suggest 67% more SAs in class 5 under RCP 8.5 compared to RCP 4.5 and a decrease of around 34% of SAs included in the very low and low R classes (classes 1 and 2). A similar trend is also seen for the same classes between the current period and RCP 4.5 (approximately 26%). The number of SAs included in the lowest (1 to 2)/highest (4 to 5) classes is shown in

Table 5. Number of SAs included in each class 1 to 5 (1 = very low, 2 = low, 3 = average, 4 = high, 5 = very high) of Hazard (H), Vulnerability (V), Exposure (E) and Risk indexes (R) classes across 2020s (current scenario) and 2100 under RCP 4.5 and 8.5. V is available only for the current scenario.

	Current Scenario				2100—RCP 4.5			2100—RCP 8.5
	V	H	E	R	H	E	R	H	E	R
1	8	332	23	68	277	19	60	196	19	59
2	21	45	20	279	80	22	240	135	22	164
3	53	4	16	5	17	17	10	42	17	22
4	105	1	9	13	7	9	12	8	9	35
5	195	0	314	17	1	315	48	1	315	80

.

In Dublin city for the end of the century, under a RCP 4.5, the areas at higher R (class 5) will be around the northern part of Dublin port, Clontarf, Baldoyle, Sutton (Howth area) in the north and Blackrock, Sandymount and Sandycove in the southern part. Further north in the County, Skerries, Malahide and Portmarnock will be at high and very high risk too (classes 4 and 5). For the same period under RCP8.5, also the Liffey inland will be potentially at high risk (class 4) and specifically the area around Heuston station. Just west of that, there is the Kilmainham weir, which is designed to separate fresh from salt water. If the inundation were to exceed the height of the weir, then much of this area would be flooded. In general, R intensifies around the areas previously identified (i.e., under RCP4.5) and along all the coastline. Around the city, new areas will be included in high and very high R (classes 4 and 5) flood-prone areas, such as beaches around Dollymount Strand, Kilbarrack and Bayside (north to Howth) and further south to Booterstown.

5. Discussion

Ireland is particularly exposed to Atlantic storms, and coastal flooding and erosion are issues that have serious economic and social impacts. Indeed, the rate of SLR in the Dublin Bay area is considered twice the global average [14], which is unprecedented over the last century and faster than expected.

The current study simulated the 2013 Storm Xaver event on the basis of a projected increase in mean sea levels of 0.45 and 0.81 m. Under both RCP scenarios, results indicate an increase in areas at risk of coastal inundation. Inundation is not restricted to the coast. A number of areas bordering the river Liffey within Dublin city centre are also extremely exposed and vulnerable and will require tailored adaptation measures to reduce risks whilst being sympathetic to the existing cityscape. It should be noted that the indicators shown in this study are presented at the SA scale as a mean value within each area. As such, this does not mean that all SAs will be equally affected, exposed or vulnerable to coastal flooding. Thus, in terms of index, it is not possible to distinguish between SAs that only undergo partial inundation impacts (i.e., covering only a part of the SA) and SAs that have greater fractional area impacted by inundation. Nevertheless, Figure 3 does allow variations in inundation depts to be distinguished at the 100 m grid resolution scale.

The current risk of coastal flood hazards shows significant challenges for Dublin and coastal towns such as Balbriggan, Malahide and Howth. This has been exacerbated over recent decades when there has been a rise in the number and range of new coastal building developments thereby increasing the stress on coastal systems and potentially reducing their sustainability and ability to mitigate flooding events [10]. These challenges are likely to intensify in the future as climate change will bring increases in SLR, and changing precipitation and storm patterns [12,82].

The CFRAM study undertaken by the OPW has recently detailed the methodology used to update national scale coastal flood depth and mapped the existing flood extents, hazards and risk for a range of flood events under different exceedance probabilities; it is also developing a plan for a long-term management strategy to cope with risks arising in the future. In the study presented here, the areas identified that are subject to the increased hazard of inundation are in line with the flood depth forecast under both the mid-range and high-end scenarios of the CFRAM programme [49]. Nevertheless, we should consider some important differences between our study and the OPW maps. In this study, unlike in CFRAM, existing or planned coastal defences have not been accounted for, which means that some areas identified here as exposed to inundation risk may not flood due to the presence of hard defences. However, this is not particularly limiting in this context, since less than 4% of the Irish coast is protected by defence structures [10]. From 2009 on, large-scale coastal protection projects have been very few, while minor funding for small-scale defence works has been promoted through the Minor Flood Mitigation and Coastal Protection Scheme of the OPW [83]. Differently from OPW maps, this study accounts for potential impacts from a set of SLR values related to the medium (RCP 4.5) and high emission (RCP 8.5) scenarios, and thus, includes a consideration of changes in SLR.

The methodology proposed in this study is scalable and can be applied at a finer scale provided that higher-resolution data and models are considered. The authors acknowledge that local impact studies would require a far higher and detailed scale than the resolution adopted in this study. We underline again that the results of this coastal hazard study are intended to be used as a screening tool at the county scale showing the lands which are currently or projected to be likely affected by flooding; these areas should thus be the object of higher-resolution modelling efforts. Indeed, these areas would require, in turn, a more detailed underlying DTM dataset and a finer resolution model capable of reproducing in detail the local hydrodynamics as carried out in the framework of very recent and ongoing projects [84].

A number of potential uncertainties should be noted, arising from, e.g., estimated extreme water levels, topographic and other survey data and from some assumptions in the model. No coupled flow-wave model has been run; in this regard, the model cannot account for locally generated seas but can reproduce the wind and atmospheric pressure-generated surge. A dynamically coupled flow-wave model was computationally unfeasible at the time of this study for such a large computational domain. Moreover, the coastal inundation study accounting for wave overtopping would require detailed information on the coastal defence structures, which is typically carried out at much smaller scale study areas. In this regard, a nested model (with finer mesh) would be required for small-size target coastal areas, which was out of the scope of this study.

In terms of uncertainty, the discrepancies in the methodology used to derive flood depths and the reliability of the input data impact the level of confidence assigned to the flood depths. Furthermore, most of the accuracy of the flood maps depends largely on the accuracy of the Digital Terrain Model, which is 30 m. As also stated by OPW [56], the acquisition of finer scale data (e.g., such as recent LiDAR data acquisition) would significantly improve accuracy levels and the confidence in the flood outputs. Model calibration metrics show slightly positive bias values (in the range of 1–5 cm) at the three reference tide gauges (i.e., Dublin Port, Skerries and Howth) indicating a trend in the model of overestimating water levels within the computational domain; the model is thus quite conservative. Indeed, looking at the bias and RMSE (including bias therein) together, it can be noted that higher RMSE and bias values are related to the overestimation of the simulated signal (i.e., of high waters). Last but not least, the highest unbiased RMSE (i.e., approx. 18 cm), with respect to the increase in SLR in the RCPs and with respect to the DEM resolution (30 m), could be considered acceptable and in line with similar studies carried out in the same area [57]. In fact, considering the tide range in the Dublin area, this RMSE is equal to approximately 5%.

Similarly to the OPW Flood Map and given the uncertainties raised above, the information presented in this paper should not be interpreted to mean that areas will flood but rather that there is the possibility that they may flood in the future. In this context, the integration of new data retrieved by climate services could be of help in improving coastal flooding models and local hydrology models [85,86].

This work employs the IPCC AR5 risk-centred approach to determine spatial and temporal variation in the inundation risk along the coastline. Indeed, this approach takes into account all three different aspects contributing to the definition of the coastal flood risk, i.e., hazard, exposure of the region and vulnerability of the population [44]. In terms of implication, this approach can be of help in providing clearer information on the risks in terms of vulnerable infrastructure but also vulnerable populations, hence possibly facilitating the development of much more nuanced and targeted (community level) responses to the hazard and risk. Although there are uncertainties in the flood depth calculations, by integrating both exposure and vulnerability in the index calculations and presenting these at the SA scale, this allows the identification of the most at-risk SAs and therefore, can support an adaptation prioritisation exercise.

Indeed, differently from previous works on SLR related-impacts in Ireland, the key added value of this study is in the socioeconomic vulnerability and CLC/LCZ-based exposure mapping, therefore showing which already highly vulnerable areas are prone to flooding and how future changes in land cover could affect risk levels. From a socioeconomic standpoint, this information is very important in terms of putting policies and measures in place to try and protect or mitigate the impacts on such populations. Identifying socioeconomic vulnerable groups can be of help for policy makers in taking the most effective mitigation actions in the most susceptible coastal-flood-prone areas. Areas with a high vulnerability could potentially be more affected by coastal floods compared to lower vulnerability zones. As an example, extreme-aged people (elderly and very young) are less able to react promptly to flood events. Similarly, people with low incomes will need financial assistance to protect their homes. People with a low-education level could be less aware of the risk of coastal flooding in their area and be less prepared for floods [40]. Thus, an indicative idea of the spatial patterns of vulnerability across the region could help policy and decision makers to tailor policies and actions to local needs and priorities. This is crucial if we think that, based on the last 2022 census [8], County Dublin has a population of about 1.45 million people, with a predicted inward migration towards the capital of more than 30 % by 2036. This will put even more pressure on infrastructure, and social and environmental services supplies. The inclusion of more socioeconomic, demographic and environmental factors in such studies in the future could thus help to significantly improve vulnerability-related studies like this and to achieve a more comprehensive risk assessment by assessing implications of the interactions between climate risks and external climate forcing like, e.g., human migration, management of supply chains and economic markets [20,87]. This information is crucial for researchers to understand how people and systems perceive the risk and for policymakers to develop risk mitigation strategies and adaption to climate change risks. An example of a study based on other meaningful variables contributing to vulnerability definition, is that of post-Katrina New Orleans [88] where the contradiction of the environmental management of flood risk was analysed, finding that the socioeconomic status of the residents and the level of perceived risk, based on past experience, are determinant in the definition of vulnerability and finally in flood insurance purchase as a mitigation risk measure. In this context, another study in the same area analysed demographic, socioeconomic, geophysical and experiential variables in flood risk perception. Flood risk perception is positively associated with lower income, female gender and direct flood experiences [89]. Nevertheless, it is challenging to identify the complete set of the most thorough indicators to assess vulnerability to flooding since a priori, not all factors have an influence on socioeconomic vulnerability in all contexts [90].

This analysis has accounted for land-use change through the development of a land-use model informed by local and regional planners. The introduction of an exposure index based on changing land-use conditions in time is, in turn, a proxy of the influence of urban dynamics on environmental exposure to coastal inundation. In this framework, the use of a LCZ-based classification, which is commonly used in urban heat studies, represents an added value to further assess the nature of urbanization and its likely environmental impact on flood risk. Indeed this allows us to introduce additional heterogeneity in the urban morphological and environmental characteristics of the area, thus improving flood exposure assessment [62,63,91]. To determine the exposure, we are aware of the limitation of the method in terms of future scenarios of changes in the biophysical components of risk. We took into account a certain proportion of uncertainty by considering changes from 2020 to 2100 on a five-year cycle. Projected changes in the urban development and the replacement of natural areas with artificial landscapes would lead to a reduction in the percentage of land with green and blue spaces especially in the suburbs with potential impacts on, e.g., surface runoff, obstruction of natural drainage and decreased infiltration. This study demonstrates that the impacts of changing land cover in County Dublin will be more accentuated in the peri-urban areas, where high development is expected in the coming years, with the city centres and coastal areas being quite stable in terms of exposure across decades, except for northern areas of the County. Obviously, a more detailed study at a higher spatial resolution would allow us to identify finer patterns and heterogeneities at the very local scale as in the framework of very recent studies in Dublin [84]. Nevertheless, in this work, in terms of H, we aimed to provide an overview of the potential flood-prone areas, i.e., areas which are likely to be exposed to flood.

In this study, despite the added value of the indicator-based approach for local urban scale impact studies in Ireland, some limitations still remain. To assess the vulnerability component, we adopted a PCA-based method, which is a standard procedure in climate-related vulnerability mapping and assessment [40,42,87]. On the one hand, it allows the reduction in the dimensionality of the system and the highlighting of the main variables affecting the vulnerability. On the other hand, the considered variables could not fully explain the spatial vulnerability pattern across the region and the weights of indicators identified by the PCA might not necessarily reflect the actual significance of a specific indicator to coastal flood vulnerability [42]. To address these shortcomings, one possible way could be to integrate PCA and an expert-judgement weighting approach by, e.g., analyzing PCA results and impact correlations with local planners and decision makers to derive more thorough weightings [92]. Discussion with stakeholders is also useful to obtain more updated information on the socioeconomic situation across the region of interest to continually improve the indicator-based system [42].

We are aware that the vulnerability component is only a snapshot of the present situation, being based on 2016 data. As such, it does not include potential changing-age profiles or specific plans for the development and modification of the socioeconomic situations across the study area in the coming decades. This leads to a possible underestimation of the future risk of coastal flooding in the region. The results presented here can be thus considered as a best-case scenario since the demographic and socioeconomics are assumed as static, whereas the population will increase across Dublin County [8]. Nonetheless, this methodology is in line with previous recent studies, where future socioeconomic and demographic projections are not included [30,43]. Recently some progress has been made in this direction: demographic projections and shared socioeconomic pathways [93] are one of the emerging approaches which will allow better assess changing conditions in future studies [37,94,95].

Considering all the above-mentioned aspects, we thus expect an even-growing anthropic pressure on economic, environmental and societal supplies in the future, with potential impacts on vulnerability patterns, in particular affecting negatively the susceptibility of the system and making the adaptive capacity more challenging. In terms of hazards, as previously discussed, our coastal inundation model is quite open; this is reflected in a potential worst-case scenario representation of the hazard of coastal flooding in the study area. Thus, for the same worst-case hazard but assuming a likely increase in vulnerability and exposure patterns, we could reasonably hypothesize that coastal flood impacts in the study area will be greater than shown.

Until recently, policy and practice at both local and national levels of governance were heavily influenced by a largely reactive decision-making process based on the retrospective analysis of past events [10,96]. The OPW’s climate change flood risk management plan [97] represents a coherent approach to addressing flood risk and adaptation in light of projected climate change. It sets out a series of potential actions to implement the adaptation policy and importantly it states as an objective the need to “promote a joined-up approach to adaptation”. Currently, 29 flood risk management plans are being published by the OPW for different catchments across Ireland [17]. Of particular relevance here is the plan for the Liffey and Dublin Bay [49]. It proposes a series of measures to be implemented to improve the management of floods in terms of risk prevention, protection, preparedness and monitoring. The risk projections presented in this study could represent a significant added value to this framework, since they account for urban expansion in the coming decades and for the current susceptibility and adaptive capacity of the population living in flood-prone areas.

There is an ongoing need to raise awareness for decision makers and the general public of the coastal risk under different climate and environmental scenarios in order to be able to improve management of the impacts of SLR, as also recommended in the climate change flood risk management plan [49]. The publication by the OPW of historical flood maps and projected flood extents under climate change scenarios is a practical step in awareness raising [17]. Dublin City Council has increased the coastal defences in the Dublin area based on a SLR between 0.4 and 0.65 m. Future SLRs depend significantly on GHG trajectories and this study contributes to moving a step forward in this direction by considering this.

The coastal zone represents an area of diverse activities operating at different scales which have cumulative effects. As a result, the challenges posed by climate change need to be addressed through integrated and ecosystem approaches and instruments. Following the transposition of the Marine Strategy Framework Directive into law, the Irish Government published a roadmap for the implementation of the Directive [98]. The Directive addresses coastal zones, in addition to marine waters. The roadmap highlights the need to take climate change and related impacts into account in the development of a national marine spatial plan. Information on the risk of coastal inundation such as those highlighted in this study should be considered in the development of any such plan. Dublin City Council has already undertaken several flood alleviation projects and flood plans to cope with pluvial and tidal flooding [99].

Despite some of the stated limitations, the maps and results presented in this work are suitable to inform community-level strategies and planning and to provide an overview of coastal flood hazards in County Dublin; thus, it can be considered appropriate for resilience and adaptive capacity assessment as well. As previously stated, there are a range of inherent uncertainties within the process of preparing the inundation depth maps. As such, they should not be used to assess the flood hazard and risk associated with individual point locations, or to replace a detailed flood risk assessment but can support existing flood maps. Communities can use these maps to provide an overview of hazards at the community scale (SA) to help inform plans, but the maps cannot be considered on their own as they do not account for small-scale features or impacts. They can thus represent a screening tool providing estimates and indications for an open discussion with different stakeholders.

6. Conclusions

This work has provided valuable information and insights on the evolution of the risk of coastal flooding to climate change across County Dublin for the end of the century. It uses internationally accepted IPCC future climate scenarios (RCP 4.5 and RCP 8.5) to assess a range of possible climate change trajectories. Indeed, to support the assessment of coastal flood risk in the study area, it is essential to consider how sea level rise would change under greenhouse gas emissions scenarios while considering urban sprawl and land cover changes in the future too. These data have been integrated with a range of factors describing the current level of vulnerability of the population and of the built environment across the region at the SA scale. The consideration of such information in coastal inundation risk management planning can be of help in (i) guiding policy and decision makers in the development of climate resilience measures, (ii) supporting prioritization and targeting of adaptation measures and (iii) raising awareness among different stakeholders and the general public of coastal flood risk under different climate and environmental scenarios in order to be able to improve management of the impacts of SLR and changes in storm patterns.