A Prioritization Framework for Adaptation Responses for Climate Change-Induced Erosion in Island Beaches—Cases from the Aegean Islands, Greece

Abstract

This contribution presents a new approach for assessing/ranking the vulnerability of beaches to mean and extreme sea level rise at regional (island) scales. It combines socio-economic information with beach erosion projections from morphodynamic models to rank beach vulnerability in a structured, ‘holistic’ manner. It involves the collation of various beach geo-spatial environmental and socio-economic data, which are then combined with erosion projections under different climatic scenarios. A Strengths–Weaknesses–Opportunities–Threats (SWOT) framework is employed for the indicator selection, and multi-criteria methods (Analytical Hierarchy Process—AHP, Technique for Order of Preference by Similarity to Ideal Solution—TOPSIS, Preference Ranking Organization Method for Enrichment Evaluations—PROMETHEE II) are then used to optimize indicator weights and rank beach vulnerability. Framework implementation in Lesvos and Kos has shown that there will be significant effects of the mean and (particularly) of the extreme sea levels on the carrying capacity and the capability of the beaches to buffer backshore assets, in the absence of appropriate adaptation measures. As the proposed approach relies on widely available information on many of the socio-economic indicators required to assess the beach’s significance/criticality, it can provide a reproducible and transferable methodology that can be applied at different locations and spatial scales.

1. Introduction

‘Sandy’ shorelines (i.e., low-lying coasts built on unconsolidated sediments—beaches) form a large segment of the global coastline [1] and are important habitats [2] and natural buffers that protect backshore ecosystems, infrastructure, and assets from coastal flooding [3,4]. Furthermore, these shorelines have high hedonic and recreational values and contribute very significantly to economic development through the tourism sector and its currently dominant ‘Sun, Sea, and Sand—3S’ model [5,6], underscoring their increasing societal and economic relevance. Sandy beaches are especially important for small island nations; for example, the touristic revenue in Seychelles and Mauritius is estimated at 65% and 24% of the GDP, respectively [7]. In Jamaica, about 25% of the available tourist accommodation is concentrated at the pocket beaches of Negril, western Jamaica contributing about 5% of the country’s GDP [8], whereas tourism revenues in Antigua and Barbuda account for more than 50% of the national GDP [9]. In the Mediterranean region, there are numerous examples of widely known 3S tourist destinations, such as the Balearic Islands (Spain) and Mykonos (Greece), where their carrying capacity has been found to be frequently exceeded [10,11].

At the same time, sandy shorelines are under a growing threat of erosion/retreat, a process exacerbated by both anthropogenic changes and climate variability and change—CV and C [12,13,14]. Mean sea levels have been rising at an accelerating pace, reaching 3.7 mm per year [15]. Future sea-level rise will depend on the effectiveness of global mitigation efforts. Under a high-emission scenario, sea levels could rise by up to 1 m by 2100, while even in the most optimistic climate stabilization scenarios, the residual effects of historical emissions will sustain sea-level rise for several decades [15]. Episodic storm-induced extreme sea levels (ESLs) are also expected to intensify in the future [15,16] with devastating consequences for low-lying coastal communities [17].

Rising relative mean sea levels, negative sediment budgets, and intensified marine storm activity are among the primary drivers of the long- and short-term shoreline retreat [18,19]. Long-term erosion/retreat can be catastrophic to those beaches that do not have the necessary accommodation space to ‘roll-over’ due to natural or artificial backshore features (e.g., coastal cliffs, and seawalls). Regarding short-term erosion, although not all storm-induced changes will be permanent, they could still result in severe disruptions to coastal environments and economies. Island beaches, particularly in developed and tourism-dependent areas, face increased vulnerability due to their generally limited size, scarce sediment resources, and high exposure to infrastructure and services [20,21]. For these areas, erosion management is essential to safeguard these critical systems and ensure their sustainable development [22,23].

Assessing the risk and dynamics of sandy shoreline erosion is vital for developing efficient adaptation strategies. Such assessments have gained attention, particularly due to the pressing need for efficient coastal management under CV and C [24]. The approaches and criteria used in these assessments depend on their spatio-temporal scales and the available information and resources, and involve evaluations of the current trends as well as projections under the CV and C [25,26]. In regional (island) scales, the extensive scope of erosion and the high potential costs of adaptation [27] require a ranking of the needs for (technical) measures in order to prioritize responses and plan an efficient allocation of the (mostly) limited resources. Such ranking should be facilitated by fit-for-purpose frameworks which should be based on integrated assessments that consider both the environmental and socio-economic conditions of the individual beaches and projections of their erosion under CV and C.

Various studies have attempted to quantify the vulnerability of coastal systems during the recent decades by combining various sets of parameters through single- and multi-variate analysis [28,29]. Most studies follow some basic approaches including scaling, weighting, and combining elements from one or more categories of coastal vulnerability factors, with various methodologies proposed for the evaluation at different spatial scales [30,31,32] through the combination of physical/environmental parameters and different coastal vulnerability indexes (CVIs) [33,34]. Despite advances, few of these methods include dynamic simulations for the estimation of future coastal impacts and mostly focus on flood risk rather than beach erosion [34]. Composite indices including also socio-economic data are now more frequently used to assess relative coastal risk levels [35]. However, many existing methodologies, often still lack a holistic integration of physical, environmental, and socio-economic dimensions, particularly in touristic activity contexts or when addressing the complexities of coastal environments [36]; this gap is further emphasized under extreme climate scenarios, where assessing and adapting to coastal hazards requires approaches that account for localized challenges and broader socio-economic implications [37].

In order to combine different environmental and socio-economic indicators into a composite index that can describe and rank beach vulnerability, multi-criteria decision-making (MCDM) approaches are required. In recent years, these approaches have gained considerable traction. An Analytical Hierarchy Process (AHP) approach has been used to incorporate expert judgements to determine the weights of the criteria/indicators used, including mainly geomorphological (shoreline change rate, beach width/height, beach slope and sediments) and wave indicators [38]. Tahri et al. [39] presented a comprehensive method for assessing the coastal risk by applying a Fuzzy Analytical Prioritization Process (FAHP) and spatial analysis techniques in a GIS environment, whereas Maanan et al. [40] provided a framework to assess the risk in Nador Lagoon (Morocco) by combining indicators of human activity, population density, erosion and sea-level rise under CV and C using GIS and an MCDM approach. De Serio et al. [41] also combined GIS tools with the AHP method to rank coastal risk, using a combination of physical and socio-economic indicators. Sekovski et al. [42] designed/applied a physical vulnerability index to sea-level rise and marine flooding for the Ravenna coast (Italy), whereas Ahmed et al. [37] assessed thirteen spatial criteria/indicators in the context of two components of vulnerability, namely physical and socio-economic vulnerability; these criteria were weighted based on an AHP and then combined to create an individual vulnerability indicators approach.

This contribution introduces a framework for the systematic evaluation and ranking of ‘sandy’ shorelines at a regional (island) scale, in terms of their need for adapting to CV and C-driven erosion. The framework offers an approach that considers the socio-economic (touristic) and environmental significance of the individual island beaches together with their erosion risk, using readily accessible geo-spatial environmental and socio-economic data, beach erosion projections under the CV and C and an array of MCDM approaches; it is applied in the Greek islands of Lesvos and Kos (Figure 1). The framework employs a Strengths–Weaknesses–Opportunities–Threats (SWOT) framework and multi-criteria decision-making techniques (AHP, TOPSIS, PROMETHEE II) to provide a structured approach to rank beaches in terms of their risk and urgency of adaptation measures. Integrating both beach environmental and socio-economic data with erosion projections under sea-level rise offers a tool for prioritizing resource allocation and informing adaptive coastal management strategies under CV and C, particularly in island settings where such considerations could be critical.

2. Environmental Setting

The islands of Lesvos and Kos (Figure 1) are large islands of the northern and southern Aegean archipelago, respectively. Lesvos has an area of 1630 km², a population of about 83,755 and a coastline length of 320 km, whereas Kos has an area of about 295 km², a population of 37,100 and a coastline length of 112 km [43]. The two islands show different development models, with the island of Lesvos being much less dependent on tourism than Kos: in 2023, about 53.4 thousand international air passengers arrived at Lesvos and 1.3 million passengers at the island of Kos. Concerning the touristic infrastructure, Lesvos has 7150 hotel beds and low tourism density and intensity (about 347.3 tourist nights/km² and 17,336 tourist nights/1000 inhabitants, respectively); by comparison, there are 57,200 hotel beds in Kos, with tourism density and intensity being very high (28,917 tourist nights/km² and 230,000 tourist nights/1000 inhabitants, respectively [44,45]). There were also 4463 day-tourist arrivals in Lesvos and 6925 in Kos in 2023 via cruise ship calls.

Lesvos and Kos have a complex geological setting and history [46]. Eastern Lesvos is composed mainly of Palaeozoic/Triassic schists and marbles, whereas the island’s centre is dominated by ophiolites and ignimbrites. The western part of the island is covered by pyroclastics, lavas, and tuffs, mostly 16–18 million years old; here, fossil pine and sequoia trees have been preserved in volcanic ash forming the ‘Petrified Forest Geopark’ [47]. Sedimentary formations and alluvial deposits are found mostly in the large embayment (Kalloni Bay) of the island. This geological setting is not conducive for the building of extensive sandy shorelines; indeed, the large majority of the island’s beaches are ‘pocket’ beaches (<1.5 km long) with small areas (Figure 1). Kos forms part of the easternmost volcanic centre of the South Aegean volcanic arc. The northeastern part of the island is mostly formed on Plio-Quaternary sediments, whereas the southeastern highlands are made mostly of alpine and pre-alpine metamorphics and some Miocene volcanics. Volcanism, after a hiatus, resumed three million years ago (mostly) in the island’s west; the eruption of the Kos Plateau Tuff (KPT) volcano (161 ka BP), was the largest known Quaternary eruption in the Eastern Mediterranean, depositing pyroclastics up to 15 m thick that covered most of the western Kos and adjacent islands [48]. The majority of the large beaches are found in the low-relief, sedimentary northeastern Kos (Figure 1), indicating their significant geological control.

The Aegean Archipelago coasts are microtidal, with tidal ranges rarely exceeding 0.2 m, although interactions with the coastal morphology can locally enhance the tidal range and currents [49]. The complex physiography also controls the wave regime [48]. In order to obtain more detailed information on the wave climate of both islands, hourly wave time series data (2010–2023) were retrieved from the Marine Copernicus repository [50,51]. Time series of significant wave heights (H_s) and peak wave periods (T_p) were analyzed by directional sector, using 45° bins.

In Kos’s northern coast, the NNW waves dominate, showing mean offshore H_s and (T_p) of about 0.9 m and 4.5 s, respectively; the largest waves show H_s of 3.8 m and T_p of 7.5 s. On the southern side, the dominant wave direction is from the SSW, with an average H_s and T_p of 0.4 m and 5 s, respectively, while maximum values correspond to waves of 5 m and 7.5 s approaching from the south (180° N). Large waves are observed also on the western coast (mean significant wave heights and periods of 0.8–1.3 m and 5–6.5 s, respectively), with the largest waves having typical periods of 9–10 s and wave heights between 4 and 5 m. Due to the limited fetch, wind waves from the east (NE-E) are much lower (0.3–0.5 m), with periods of 3 s, while the maximum values do not exceed 2 m.

In Lesvos, the dominant wave direction at the northern coast is from NNW, with mean H_s between 0.4 and 0.7 m and periods of 4 s, while under extreme conditions, long waves (~8 s) between 3.5 and 4.5 m high are observed. The highest wave heights are found for the SSW direction, with average H_s close to 0.9 m, while under storm conditions extreme waves can exceed 4.5 m with typical swell periods (8–9 s). Waves approaching directly from the south (~180°) have an average offshore wave height of about 0.7 m and wave period of 3.5 s, and during storms, heights of 3 m. Wave intensity is decreased towards the east side of Lesvos, where short wind waves (T_p ~3 s) dominate, with average wave heights of 0.2–0.4 m, while under storm events, wave heights do not exceed 2.5 m.

The recent mean sea-level rise trend is about 4.6 and 4.3 mm/yr in the northern and southern Aegean, respectively [52]. Extreme sea levels (ESLs) at the coast show a seasonal footprint, with extreme positives occurring in winter and under certain North Atlantic Oscillation modulations [53]. Future ESLs are projected to show spatial variability, being also sensitive to the dynamics of the thermohaline circulation and the Black Sea buoyant inputs; significant changes in the ESL return periods are projected for the future [52]. In terms of future waves, small changes in wave heights are projected until the 2050s [54].

3. Methodology

The proposed framework for ranking sandy shorelines integrates readily available geo-spatial, environmental, and socio-economic data and projections of beach erosion/retreat under different scenarios of relative sea-level rise (RSLR) and extreme sea levels (ESLs). It consists of several stages, summarized in Figure 2, and aims to rank the criticality and vulnerability of beaches on an island scale.

3.1. Data Collation and Database

An inventory of the sandy shorelines of Lesvos and Kos was compiled, drawing geo-spatial data from the THALIS-ISLA database [55] that provides information on the dimensions, sediment types and the presence of coastal defences or artificial and natural features of all Aegean archipelago beaches. Using satellite images (acquisition dates in 2003–2011) available in the Google Earth Pro application, subaerial (dry) beaches were digitized as polygons based on clearly visible boundaries: natural features like vegetated dunes or cliffs and permanent artificial structures such as embankments, seawalls, or buildings defined the landward limits, whereas the seaward limits were determined by the shoreline. To ensure consistency, all digitization was conducted by a single analyst adhering to strict delimitation rules. The database was updated in 2023, through the digitization of beach polygons from different selected images obtained in the period 2011–2023, in order to identify historical changes and incorporate recent information on these characteristics. Additional information on socio-economic factors, such as accessibility, carrying capacity, backshore asset density, touristic development, and tourist visitation patterns was also collated, using Google Earth Pro imagery, other remote sensing information and in situ observations; this was stored in the database.

During the digitization process, the tilt of the Google Earth Pro images when zooming in was restrained to 0°. Furthermore, images were zoomed to an eye-altitude of about 60 m; at this zoom level, the imagery for the two islands is composed of images provided by CNES/Airbus and Maxar with submeter spatial resolution (0.3–0.5 m).

The accuracy of the results is constrained by the accuracy of the analyzed images. For example, the co-evaluation of the landward beach limits digitized from the available (2023) satellite images with concurrent ground observations (using RTK GPS) carried out as part of the present study showed an RMS error of 2.3 m. Moreover, although tidal effects on the shoreline position are considered very small at the microtidal coasts and care was taken to analyze satellite images from the same season and under low hydrodynamic conditions, the available images along the islands’ coasts have been collected at different times and under different preceding hydrodynamic conditions; thus, recorded beach dimensions may not represent synoptic conditions at the island scale. Such limitations, however, cannot be avoided in comparisons of historical satellite imagery at large spatial scales [12,26].

Following these procedures, the characteristics of 218 (Lesvos) and 78 (Kos) beaches (Figure 1) were identified and recorded.

3.2. Erosion Projections Under the Mean and Extreme Sea Levels

The assessment of the historical beach changes is based on the comparison of the images found in the Google Earth historical imagery tool. A total of 1033 images spanning the period 2003–2021 for Lesvos and 234 images spanning the period 2004–2023 for Kos were examined. The time span of the comparisons differs along the coast due to the lack of synoptic images of appropriate quality covering the coastlines of the islands. A margin of error of ±1.2 m was used in the analysis of the changes in the maximum width of the beaches, based on the estimated RMS error of 2.3 m.

Morphodynamic model ensembles were employed to project beach erosion on Lesvos and Kos for the years 2050 and 2100, under the IPCC RCP4.5 and RCP8.5 scenarios. These projections account for: (a) the relative sea-level rise (RSLR); and (b) the 1-in-100-year extreme sea levels (ESLs₁₀₀), which combine storm-induced coastal water levels (storm surges and wave set-ups) with RSLRs and potential changes in tidal levels. Data used in these projections were extracted from the EU-JRC (Joint Research Centre) dataset detailed in Vousdoukas et al. [25] for every 100 km along the coastline. RSLR projections were based on Jevrejeva et al. [56], while ESL estimates were derived using dynamic models. Storm surges were projected by the DFLOW-FM model [57], whereas offshore wave parameters (heights—H_s, periods, and directions) were simulated by the WAVEWATCH III model [58] (details on how the offshore wave characteristics were estimated can be found in [59]). Wave set-ups were calculated using the CEM’s [60] generic approximation (set-up = 0.2 × offshore H_s). The dataset integrated all ESL components, addressing uncertainties through probabilistic Monte Carlo simulations. Additionally, wave control on ESLs was assessed via bivariate copulas, which linked each projected storm surge with the most likely corresponding wave height, period, and direction (details in [26]). This integrated dataset provided the forcing for the morphodynamic models. For each beach, the closest available projections (from the JRC database) for RSLR, ESL, and wave conditions (

Table 1. It displays 1-in-100-year extreme sea levels and their components (RSLR, η_tide, and η_CE) and corresponding wave conditions along the coastline of Lesvos and Kos for different dates (baseline, 2050 and 2100) and under the IPCC RCP4.5 and RCP8.5 scenarios. Key: RSLR, relative sea-level rise; η_tide, max. tidal level; η_CE, storm sea level component induced by the combined effect of storm surge and wave set-up; H_s, significant wave height; T, wave period. Source of data: EU-JRC (Joint Research Centre) (https://data.jrc.ec.europa.eu/dataset/jrc-liscoast-10012, accessed on 15 May 2024).

		Baseline	RCP 4.5		RCP 8.5
			2050	2100	2050	2100
Lesvos 1	RSLR (m)	0	0.13	0.47	0.18	0.75–0.77
	ηtide (m)	0.08–0.09	0.08–0.09	0.07–0.09	0.07–0.09	0.07–0.09
	ηCE (m)	0.88–1.01	0.85–0.97	0.85–0.94	0.89–1.01	0.88–0.98
	ESL (m)	0.96–1.09	1.06–1.18	1.39–1.49	1.15–1.27	1.72–1.81
	Hs	1.47–1.67	1.48–1.67	1.47–1.63	1.46–1.74	1.51–1.68
	T	5.13–5.38	5.07–5.36	5.04–5.31	5.08–5.33	5.10–5.36
Kos 2	RSLR (m)	0	0.14	0.50	0.19	0.81
	ηtide (m)	0.06	0.06	0.06	0.06	0.06
	ηCE (m)	0.86	0.83	0.88	0.85	0.84
	ESL (m)	0.92	1.03	1.44	1.10	1.71
	Hs	1.77	1.77	1.81	1.75	1.80
	T	5.31	5.32	5.40	5.31	5.37

¹ In Lesvos, the ranges shown on the table represent the different values along the coastline. ² In Kos, available data were found for only one point along the coastline; thus, the same values were applied to all beaches.

) at a water depth of 20 m were used to simulate the local erosion. For the island of Kos, available data were found for only one point along the coastline, the same values were applied to all beaches, while in Levos the values vary along the coastline.

Beach retreat was estimated using two ensembles of 1-D cross-shore morphodynamic models, chosen to leverage the strengths and address the limitations of each approach. The first ensemble, comprising analytical models, included the Bruun [61], Dean [62], and Edelman [63] models and was used in the evaluation of beach retreat due to relative sea-level rise (RSLR). The second ensemble, which included the numerical models SBEACH [64], Leont’yev [65], Xbeach [66], and a Boussinesq model [67], was used to simulate beach retreat under extreme sea levels (ESLs), integrating storm surge and wave effects.

Each analytical ensemble model follows a different approach. The results of the Bruun model that is based on the ‘equilibrium profile’ concept, are controlled by the closure depth and the beach slope; this information was abstracted from the database of Athanasiou [68]. The Edelman model accommodates temporally variable sea level changes and realistic beach profiles, while the Dean model places a greater emphasis on wave hydrodynamics.

Concerning the numerical model ensemble, the Leont’yev model adopts an energetic approach, and estimates separately sediment transport in the surf and swash zones. The SBEACH model simulates wave transformations in shoaling waters and sediment transport through coastal wave energy fluxes. Xbeach employs a time-dependent wave action balance equation and uses a depth-averaged advection/diffusion expression for sediment transport estimations, whereas the Boussinesq model captures non-linear wave transformations using high-order equations and includes a sediment transport module to estimate sheet flow, bed load, and suspended load over irregular seabeds [69].

The numerical models were applied from a depth of 20 m to the shoreline (up to 10 m elevation above the mean sea level), with the duration of simulations determined by the stabilization of beach profiles. Validation exercises comparing the model results with physical experimental data [55] demonstrated good agreement. Detailed descriptions of the models and their validation are provided in previous studies [55,70], whereas a concise flow of the adopted methodology is presented in [26].

Given the scale of the application, seabed slope and sediment data could not be based on direct observations, or open-source nearshore bathymetric datasets (e.g., GEBCO or EMODnet) due to their low resolution. Instead, the models were set up using a range of linear beach profiles. This approach has been validated by physical experiments, which demonstrated that model ensemble projections based on equivalent linear profiles closely approximated the results of experiments using natural profiles under the tested conditions [55]. Additionally, due to the interdependence between beach face slopes and sediment size (e.g., [71]), a plausible range of slopes and median grain sizes was selected for each beach. The selection was informed by the slope-grain size relationship described in [72] and the recorded qualitative sediment texture, ensuring a representative configuration for the models. Beaches were categorized into three groups—sandy, mixed (sandy-gravel), and gravel—and the models were run using various combinations of bed slopes and median sediment sizes (d₅₀) to reflect these classifications. For sandy beaches, bed slopes between 1/20 and 1/30 and sediment sizes between 0.2 and 0.8 mm were used. For mixed-texture beaches bed slopes between 1/10 and 1/15 and sediment sizes between 0.8 and 1 mm were utilized, whereas for gravel beaches slopes of 1/10 and sediment sizes (d₅₀) between 2 and 5 mm were used.

Cross-shore beach retreat was evaluated for the 1-in-100-year extreme sea levels (ESLs) projected for 2050 and 2100 under the IPCC RCP4.5 (moderate) and RCP8.5 (extreme) scenarios. Each model within the ensemble, reflecting differing set-up conditions, generated a range of beach erosion projections. From these projections, the minimum (most conservative), median, and maximum values were derived to estimate the corresponding minimum, median, and maximum beach retreats for each location. The projected retreats were then compared to the recorded beach maximum widths (BMWs) to determine the extent of change (erosion) for each beach. This metric was chosen as a conservative indicator of the potential impacts on frontline backshore infrastructure and assets, as decreases in the BMW reflect not only erosion at the widest points but also impacts for the remaining beach areas. The modelling approach is more suitable for microtidal environments where the wave energy during storms focuses on a relatively narrow strip of the beachface. Additionally, it assumes no changes in sediment supply. Projections excluded beaches in Lesvos and Kos identified as ‘perched’ or protected by coastal defence structures. It is noted, however, that these beaches are also expected to be impacted, as many coastal defences (e.g., breakwaters) are not designed to address erosion by rising sea levels [73].

3.3. Prioritization Framework

From the geo-spatial environmental and socio-economic information recorded in the inventory, several indicators were selected for use together with the projections of beach erosion/retreat due to relative mean sea-level rise (RSLR) and extreme events.

Initially, a Strengths–Weaknesses–Opportunities–Threats (SWOT) analysis was performed to assess positive and negative aspects in relation to the beach management [74,75] at Lesvos and Kos, in order to assist in the selection of indicators. This analysis was based on the quantitative and qualitative information collated in the database (e.g., beach accessibility, carrying capacity and development, the density of backshore assets, and tourist visitations), as well as other available information.

In order to define the weights (or relative importance) of the indicators, the Analytical Hierarchy Process AHP method [76] was used, i.e., pairwise comparisons (using a 1 to 9 scale) in relation to the objective of the analysis and the alternative decisions. Comparisons were based on (32) stakeholders’ and experts’ judgments through interviews during workshops that were held within the MARICC project, and the results were compiled into matrices [77], the consistency of which was evaluated using eigenanalysis. Results are considered consistent when their consistency ratio—C.R. is less than 0.1, whereas, for C.R. > 0.1, the initial judgments on the criteria significance can be adjusted to improve consistency [76]; consistency testing is a key stage of AHP, which makes it one of the most used methods. Then, the framework for ranking the islands’ beaches in terms of their need for management responses (prioritization) was applied in 2 stages (Figure 2).

The first stage of the analysis concerns the selection of the most significant (i.e., the 15 highest ranking) beaches in terms of their socio-economic value. The Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) multi-criteria method [76] was used at this stage, due to the large numbers (218—Lesvos and 78—Kos) of ‘alternatives’ (beaches); this method is considered to be less affected by the number of alternatives and criteria (indicators) compared to other approaches [78].

The socio-economic and environmental indicators selected for this stage of the analysis are as follows: (1) beach carrying capacity, which is also a proxy for the beach area (22 m²/per person [79]); (2) the sediment type, as beaches made of sand-sized sediments have higher hedonic values and recreation potential; (3) Blue Flag awards which are perceived as markers of beach quality by users [80]; (4) special environmental protection regime (e.g., NATURA 2000 sites, wetlands and other sites enjoying environmental protection under Greek law for which there are implicit obligations to mitigate and manage anthropogenic and climatic threats); (5) beach accessibility, based on the state of the road (or the absence of, in the case of beaches that can only be accessed from the sea) that leads to the beach and the distance from the main road network; (6) beach development/usage, based on the presence of umbrellas, sun beds and touristic facilities during the high season (summer); (7) beach-visitation traffic [81], i.e., the number and frequency of the visitors at the beach recorded on the basis of the average annual number of geo-tagged photographs collected near each beach shared on Flickr for the years 2005–2022); (8) Touristic activity (qualitative, 1–4 range), based on the numbers of hotels/restaurants in an 1 km radius); (9) coverage of the backshore assets at the landward boundary of the beach (ratio of the length of the backshore asset coverage to the overall beach length), that can be used as a proxy of the damage risk due to beach erosion.

The second stage of the analysis involved the prioritization of the 15 selected beaches in terms not only of their environmental and socio-economic criticality but also of their projected erosion under CV and C through the pairwise multi-criteria approach PROMETHEE II (a simplified preference function was used as described in Maity and Chakraborty [82]). The indicators used in this stage are related to the erosion risk under the CV and C (see Section 3.2) and include the following: (10) the current trends of beach erosion/accretion identified through the comparison of historical images; (11) the current beach maximum width—BMW recorded; (12) the future cross-shore beach erosion (as a percentage of the recorded BMW); (13) the change in the beach carrying capacity (percentage of reduction); and (14) potential beach erosion impacts on the backshore assets (qualitative—yes or no).

Indicators (1), (7), (9), (10), and (11) are quantitative, whereas indicators (3) and (4) are qualitative, measured on a binary scale. Indicators (5), (6), and (8) are also qualitative, with values ranging from 1 to 4 or 5 (depending on the indicator). For the sediment texture indicator (2), quantitative data are not available and, thus, a weighted scale is used with the sandy beaches assigned the highest value, the mixed sediment beaches (sand and gravels) a moderate value, and the coarse sediment beaches (gravels and pebbles) the lowest value (

Table 2. Selected indicators for the analysis.

ID	Indicators/Criteria	Type of Data	Effect
1	Beach carrying capacity	Quantitative	Beneficial
2	Sediment type	Qualitative—3 classes: sand, mixed (sand and gravel) and coarse sediments (gravels and pebbles)	Beneficial
3	Blue Flag awards	Qualitative indicator—yes or no	Beneficial
4	Special environmental protection regime	Qualitative indicator—yes or no	Beneficial
5	Beach accessibility	Qualitative—1: None (Only marine access). 2: Difficult (unpaved road with many turns). 3: Moderate (road with many turns or/and long distance from the main road network). 4: Easy (paved road and easy access from the main road network)	Beneficial
6	Beach development	Qualitative—1: None (basic facilities, e.g., changing rooms, bins,). 2: Low (basic facilities and a few umbrellas). 3: Moderate (More facilities and umbrellas). 4: Good (well organized). 5: Intense (too many umbrellas, sea sports, lifeguards)	Beneficial
7	Beach-visitation traffic	Quantitative	Beneficial
8	Touristic activity	Qualitative—1: None. 2: Low (a few small hotels). 3: Moderate (more hotels/restaurants). 4: High (many hotels and restaurants)	Beneficial
9	Coverage of the backshore assets	Quantitative	Beneficial
10	Current trends of beach erosion (positive values for accretion, negative values for erosion)	Quantitative	Non-Beneficial
11	Beach maximum width -BMW	Quantitative	Non-Beneficial
12	Future % reduction in BMW	Quantitative	Beneficial
13	Change in beach carrying capacity (% reduction)	Quantitative	Beneficial
14	Potential beach erosion impacts on the backshore assets	Qualitative—yes or no	Beneficial

), reflecting their relative importance.

Since the indicators are not measured in the same units, normalization procedures to ensure uniformity and comparability were used. During normalization, the indicators were considered to be ‘beneficial’ to the ranking (i.e., the higher their value, the higher the ranking in terms of criticality/vulnerability), with the exception of the trends in erosion/accretion (10) and the maximum beach width—BMW (11), which were treated as ‘non-beneficial’ indicators; for the erosion/accretion trends, positive values indicate accretion, while negative values represent erosion, whereas for BMWs, lower values indicate greater erosion exposure and, thus, higher ranking.

4. Results

4.1. SWOT Analysis

The analysis identified several strengths for the resilience and sustainable development of the island’s beaches, with some differences between the two islands. Major strengths (with some differences between the islands) include: a generally low degree of beach development and coverage of backshore assets with 82% of the 218 and 45% of the 78 beaches of Lesvos and Kos, respectively, found that can be considered as natural (i.e., no or negligible development); the large number of ‘blue flag’ awards; the high coastal accessibility through good quality road networks; and the good national and international connectivity (much higher in Kos). Major weaknesses include the following: already high rates of beach erosion (in some areas) and lack of coordinated plans for its management; and decreasing coastal sediment supply due to reservoir construction and lack of well-designed and effective coastal protection schemes.

There are also opportunities, such as the high potential for sustainable development (many natural sandy shorelines, particularly in Lesvos), and a regulatory regime for coastal management and monitoring that could improve the resilience and manage the coastal development. For example, the emerging regulatory framework for the management of the European coastal zone involves a network of intercalated instruments that consider coastal erosion/flooding as a major risk and prescribe for its mitigation (e.g., the Water Framework (2000/60/EC), Flood Risk (2007/60/EC), Environmental Impact Assessment (2014/52/EU) Directives, and the 2008 Integrated Coastal Zone Management Protocol to the Barcelona Convention, which, however, has not yet been ratified by Greece). The beach erosion expected under the CV and C is considered the most challenging threat for the Lesvos and Kos beaches, particularly if any further coastal development is not regulated accordingly. Another noteworthy threat is that the socio-economic development of both islands, which has been already weakened by the 2010s economic crisis, the 2015–2016 refugee crisis, the COVID-19 pandemic and the cost of the energy crisis, will also be vulnerable to the increasing geopolitical tensions which, among others, will restrict economic development and activity, particularly in tourism-dependent islands (e.g., Kos). The SWOT analysis assisted in the selection of the indicators used for the beach vulnerability ranking.

4.2. Indicators

In this section, the spatial distributions of the indicators that describe the current environmental and socio-economic situation in Lesvos and Kos beaches are presented.

4.2.1. Lesvos

In Lesvos, 218 beaches were identified and recorded (Figure 1 and Figure 3), with a total area of approximately 1.9 km², indicating a total beach carrying capacity (i.e., the number of visitors that can be accommodated concurrently at a beach) of approximately 190,000–22 m²/per person (Figure 3b). Most of these beaches are ‘pocket beaches’, i.e., generally small beaches bounded by rocky promontories, with the majority showing small beaches maximum widths (BMWs); only 4.6% of Lesvos beaches were recorded as having BMWs > 50 m, with 64% of the beaches recording BMWs of less than 20 m (Figure 3a). Regarding the sediment type, 37% of the beaches are sandy, 30% consist of mixed sediments (sand, gravels/pebbles), with the rest formed on coarse sediments (gravels/pebbles) (Figure 3c). Eleven beaches have been awarded ‘blue flags’, whereas there are also several Special Protection (SPAs) and Conservation (SACs) areas (Figure 3d), including Natura 2000 sites (https://natura2000.eea.europa.eu/, accessed on 15 May 2024), small island coastal wetlands, and areas protected by the Wild Birds Directive 2009/147/EC and the Habitats Directive 92/43/EEC); there is also the UNESCO Geopark of ‘Petrified Forest’ in western Lesvos [47]. Beach accessibility appears high throughout the island with the exception of the western and southwestern coasts due to their sparser (rural) road networks. Beach development appears generally low (Figure 3f), with beach ‘traffic’ due to visitation (Figure 3g) and tourist activity, also showing similar patterns.

Generally, most beaches (82%) can be classified as ’natural’, i.e., with no or little human development (touristic facilities/services, coastal works, and/or increased asset coverage of the backshore) and only about 4% as ‘fully developed’; backshore infrastructure/assets are found in 45% of the beaches. It is noteworthy that the sandy beach distribution (Figure 3c) does not always match that of beach ‘traffic’ due to visitation, although sandy beaches are considered more popular [80].

4.2.2. Kos

In Kos, 78 beaches have been identified/recorded (Figure 1 and Figure 4). The majority of the beaches have small widths, with 87% of the recorded BMWs being less than 50 m. Greater widths and carrying capacities are observed at the north of the island as well as at the central section of the southern coast (Figure 4a,b). Beaches are mostly sandy (53%), with mixed sediment beaches (32%), and gravel beaches (15%) are found mostly at the southwest and southeastern coasts of the island (Figure 4c).

Regarding the quality of the Kos bathing waters and the protection status of its beaches, 11 beaches have been awarded ‘Blue Flags’, while 17 beaches (22%) are located within areas under environmental protection (Figure 4d). Coastal cliffs are found at the backshore of 50% of the beaches, while sand dunes are found in 21% of the beaches, mainly in the northern part of the island. In the majority of beaches (51%), rivers/streams are not present, while beach rocks are found in only nine beaches (12%). The majority of the beaches are characterized by good accessibility, with the exception of a few beaches at the southwestern and southeastern coasts (Figure 4e). High densities of backshore assets are found along the southern and northeastern coasts, mostly matching the patterns of tourist activity and visitation (Figure 4f,g); large numbers of hotels and catering businesses are found at 35% of the beaches, mainly at the northeastern and the southern coast (Figure 4f). It is noteworthy that although Kos is a 3S touristic hub (1.3 million international visitors per year), many of its beaches (44%) show very little or no development.

4.3. Beach Erosion: Trends and Projections

4.3.1. Trends of Beach Retreat/Erosion

The current trends of shoreline change for the Lesvos and Kos beaches are shown in Figure 3h and Figure 4h and

Table 3. Current trends in the status of the Lesvos and Kos beaches on the basis of the comparison of satellite imagery available in the Google Earth Pro application (see Section 3.1). N: number of beaches.

Status	Lesvos (2003–2021)		Kos (2004–2023)
	N	%	N	%
Erosion	96	44	54	69
Accretion	61	28	14	18
Stable	61	28	10	13

. It appears that the majority of the beaches of both islands are already under erosion, with the problem being greater in Kos where 69% of its beaches have been assessed to be under erosion compared with Lesvos (44%).

4.3.2. Projection of Future Beach Retreat/Erosion

The model ensembles were used to estimate the beach retreat/erosion of all Lesvos and Kos beaches. Based on the sediment texture, a range of bed slopes and sediment combinations were used in the models (Section 3.2), resulting in envelopes of projections by the ensemble models under the examined CV and C scenarios (

Table 4. Minimum, maximum, and median beach erosion/retreat estimates—R (in m) for Lesvos beaches under the mean sea-level rise (RSLR) and the 1–100 year extreme sea levels—ESL₁₀₀ for 2050 and 2100. Beach erosion is also projected as a percentage of the current BMWs; if erosion exceeds the BMW, then the beach is fully eroded and backshore assets (if present) will likely be damaged.

Sea-Level Rise				Retreat-R (m)		R of 20% of BMW (%)	R of 50% of BMW (%)	R Equal to BMW (%)	Beaches with Assets Affected
	Year	RCP	(m)						N	%
RSLR	2050	4.5	0.13	Min	1.3–2.6	15	1	0	0	0
				Max	2.6–7.8	73	20	4	5	2
				Median	2.3–5.2	48	10	1	2	1
		8.5	0.18	Min	1.8–3.6	31	3	0	0	0
				Max	2.8–8.6	75	25	5	6	3
				Median	2.6–5.7	55	11	1	3	1
	2100	4.5	0.47	Min	4.7–9.4	81	33	7	8	4
				Max	5.7–17.4	95	65	28	24	12
				Median	5.1–12.9	90	49	14	15	7
		8.5	0.77	Min	7.5–15.3	94	64	22	20	10
				Max	10.3–31.1	100	89	60	48	23
				Median	8.6–21.7	99	78	39	32	15
ESL100	2050	4.5	1.06	Min	10.8–24.4	99	85	46	35	17
				Max	12.4–37.9	100	94	74	59	29
				Median	11.1–29.4	100	90	62	50	24
		8.5	1.15	Min	11.6–26.1	99	86	52	42	20
				Max	13.2–40.5	100	94	75	59	29
				Median	11.9–31.6	100	91	67	54	26
	2100	4.5	1.39	Min	14.1–30.4	100	90	64	52	25
				Max	15.6–47.1	100	97	80	64	31
				Median	14.4–36.4	100	95	75	59	29
		8.5	1.72	Min	17.4–36.8	100	94	76	58	28
				Max	18.9–57.1	100	98	87	71	34
				Median	17.7–43.9	100	97	83	65	31

and

Table 5. Minimum, maximum, and median estimates of long-term and episodic erosion for Kos beaches. It shows the percentages of beaches that will retreat/erode by 20%, 50% and 100% of their maximum width, as well as the numbers (N) and percentages of beaches with backshore assets affected.

Sea-Level Rise				Retreat-R (m)		R by 20% of BMW (%)	R by 50% of BMW (%)	R Equal to BMW (%)	Beaches with Assets Affected
	Year	RCP	(m)						N	%
RSLR	2050	4.5	0.14	Min	1.4–2.8	2	0	0	0	0
				Max	2.6–7.7	53	2	0	0	0
				Median	2.4–5.2	23	0	0	0	0
		8.5	0.19	Min	1.9–3.8	8	0	0	0	0
				Max	2.9–8.6	61	6	0	0	0
				Median	2.6–5.8	27	0	0	0	0
	2100	4.5	0.50	Min	5.0–10.0	74	9	0	0	0
				Max	6.0–18.1	97	41	6	2	3
				Median	5.2–12.9	82	20	0	0	0
		8.5	0.81	Min	8.1–16.1	97	35	3	0	0
				Max	10.3–30.9	100	80	26	8	12
				Median	9.1–22.7	100	67	14	5	8
ESL100	2050	4.5	1.03	Min	10.5–21.8	100	71	17	5	8
				Max	11.9–33.5	100	86	36	10	15
				Median	11.5–25.9	100	80	24	8	12
		8.5	1.10	Min	11.2–23.1	100	73	20	6	9
				Max	12.6–35.5	100	88	42	12	18
				Median	12.2–27.5	100	82	27	8	12
	2100	4.5	1.44	Min	14.6–29.6	100	82	32	8	12
				Max	15.9–45.7	100	100	65	17	26
				Median	15.5–35.9	100	92	50	13	20
		8.5	1.71	Min	17.3–34.7	100	88	44	10	15
				Max	18.6–54.1	100	100	79	22	33
				Median	18.2–42.7	100	98	67	17	26

and Figure 5 and Figure 6). The minimum, maximum, and median values of the model results of each ensemble were then estimated to project the minimum (most conservative), median and maximum beach retreat, respectively, for each beach, separately for RSLR and ESL₁₀₀. These retreats were then compared to the recorded beach maximum widths (BMWs) to assess the future impacts on each beach.

Under an RSLR of 0.13 m projected for the year 2050 under the RCP4.5 scenario, there might be impacts even under the median projections (Table 4 and Figure 5a); greater impacts are projected under the RCP8.5 scenario (0.18 m RSLR), as 11% of Lesvos beaches will retreat by 50% of their maximum width (BMW). For 2100, much greater retreats are projected due to the projected RLSR of 0.47 m (RCP4.5) and 0.77 m (RCP8.5). Under RCP4.5, 49% of the tested Lesvos beaches (‘perched’ and defended beaches have been excluded) will retreat/erode by 50% of their BMW, whereas 14% will be completely eroded, affecting 15 beaches (7%) which currently host backshore assets; these beaches will likely ‘drown’ due to the lack of accommodation space. Under RCP8.5, 39% of all tested beaches will retreat by their entire maximum width under the median projections, whereas the maximum estimates by the analytical model ensemble (Figure 5b) indicate complete erosion of 60% of the beaches tested (48 beaches with backshore assets).

For 2050, under extreme sea levels of 1.06–1.15 m (return period 1–100 year—ESL₁₀₀), the numerical model ensemble projected (temporary) shoreline retreats of 10.8–40.5 m depending on the scenario and the estimates (min., median, max), resulting in significant beach retreats/erosion (Table 4); 62% and 67% of the beaches tested will erode by 100% of their current BMW according to the median projections under RCP4.5 and RCP8.5, respectively. In 2100, median projections indicate that 75% and 83% of the beaches tested will erode by more than 100% of their current BMW under the RCP4.5 and RCP8.5, respectively.

In Kos, no significant erosion is indicated under the RSLR projected for 2050 (Table 5), due to the relatively large beach widths. Greater impacts are projected for 2100 (Figure 6a,b): under the RCP8.5 scenario (0.81 m RSLR), 67% of the Kos beaches tested will retreat/erode by 50% and 14% by 100% of their current BMW according to the median estimates. For the same scenario and timeline, maximum estimates suggest very significant impacts, with 26% of beaches projected to retreat/erode by more than 100% of their current BMW, affecting eight beaches (12%) currently hosting backshore assets.

Under extreme sea levels, the numerical model ensemble showed that in 2050 extreme sea levels (ESL₁₀₀) of 1.03–1.10 m would lead to (temporary) shoreline retreats of 10.5–35.5 m depending on the scenario and the type of estimates (minimum, median, maximum), resulting in significant beach retreats/erosion (Table 5); 24% and 27% of the beaches tested will retreat by 100% of their current BMW according to the median projections under RCP4.5 and RCP8.5, respectively. In 2100, median projections indicate that 50% and 67% of the beaches tested will erode by more than their current maximum widths, under the RCP4.5 and RCP8.5, respectively. Erosion and backshore asset impacts will be significantly greater according to the maximum projections (Table 5 and Figure 6).

Generally, beach retreats/erosion are projected to be more severe for Lesvos beaches than those of Kos, due to the differences in beach dimensions/widths (see also Figure 3 and Figure 4).

4.4. Island Beach Prioritization

The AHP method was applied to assign weights (relative importance) to the indicators used in the two stages of the beach ranking (Figure 2). For the first stage, weights were assigned to the environmental and socio-economic indicators (1)–(9) selected for the TOPSIS application. In the second stage, the method was used twice to assign weights to indicators (10)–(14) utilized in the ranking of beach vulnerability to current and future (projected) erosion through the PROMETHEE II method (Section 3.3). In both stages, the expert judgement weights (in a 1–9 scale) formed matrices, the consistency of which was evaluated by eigenanalysis. When the initial judgement matrices showed consistency ratios—C.R. > 0.1, weights were adjusted until the matrices showed consistency ratios close to zero in the final iteration; the final weights are shown in

Table 6. The weights assigned to the indicators for the 2 stages of the prioritization analysis, using the AHP method. Key: (1) beach carrying capacity; (2) the sediment type; (3) Blue Flag awards; (4) environmental protection; (5) beach accessibility; (6) beach development/usage; (7) beach-visitation traffic; (8) touristic activity; (9) coverage of the backshore assets; (10) current trends of beach erosion/accretion identified through the comparison of historical images; (11) the current (recorded) BMW; (12) the future cross-shore beach erosion (as a percentage of the recorded BMW); (13) the change in beach carrying capacity; and (14) potential beach erosion impacts on the backshore assets.

Indicators	Weights 1st Stage	Weights 2nd Stage (Current)	Weights 2nd Stage (Future)
(1)	0.049	0.059	-
(2)	0.146	0.118	0.118
(3)	0.146	0.118	0.118
(4)	0.146	0.118	0.118
(5)	0.073	-	-
(6)	0.146	0.118	0.118
(7)	0.073	0.059	0.059
(8)	0.146	0.118	0.118
(9)	0.073	0.059	0.059
(10)	-	0.118	-
(11)	-	0.118	-
(12)	-	-	0.118
(13)	-	-	0.059
(14)	-	-	0.118

.

In the first stage of the framework application through the TOPSIS method and using appropriate weights (Table 6), ranking scores were assigned to all Lesvos and Kos beaches, so as to select the 15 highest-ranking beaches in each island in terms of their environmental and socio-economic significance. Full results of this analysis can be found in the Supplementary Material (Tables S1 and S2) and are shown in Figure 7a,b (Lesvos) and Figure 8a,b (Kos) with the 15 highest-ranking beaches identified in red. These beaches were then evaluated and ranked in terms of their current and future vulnerability to erosion using the PROMETHEE II multi-criteria approach and appropriate indicator weights (Table 6). The procedure was applied three times, considering the current vulnerability and two scenarios of future vulnerability to erosion. The results of this analysis are shown in Figure 7c and Figure 8c and

Table 7. Ranking scores (PROMETHEE II analysis) of the 15 selected beaches of Lesvos and Kos, regarding the beach vulnerability to CV and C. The highest-ranking beaches in each considered scenario are shown in bold. The positive score expresses how an alternative outranks all the others, while the negative score expresses how an alternative is being outranked by all the others.

Lesvos				Kos
Beach Name	Current	RCP4.5, 2100, Med.	RCP8.5, 2100, Max.	Beach Name	Current	RCP4.5, 2100, Med.	RCP8.5, 2100, Max.
27. Kanoni	−0.094	−0.029	−0.056	1. Faros, Ammoglossa	−0.058	0.047	0.031
39. Tsamakia	−0.060	−0.012	−0.040	2. Faros, Ammoglossa−2 (Kritika)	−0.058	−0.013	0.021
45. Kratigos_1	−0.193	−0.280	−0.308	5. Skala−2	0.002	−0.025	−0.035
48. Charamida	−0.099	−0.115	−0.038	7. Marina (Aktaion)	−0.012	0.006	−0.007
50. Ag. Ermogenis_2	−0.027	0.116	0.088	9. Psalidi−2	0.067	−0.024	−0.023
60. Evriaki_1	−0.113	0.058	0.031	11. Psalidi—Ydroviotopos	−0.028	−0.046	−0.045
115. Nyfida	0.041	0.107	0.079	12. Okeanis (Peukokefali)	−0.057	−0.001	−0.010
125. Skala_Kallonis	0.181	−0.024	−0.053	15. Dimitra	−0.064	−0.069	−0.074
142. Tavari_3	−0.182	−0.292	−0.320	32. Paradise (Kefalou)	−0.070	−0.055	−0.027
150. Eresos	0.385	0.139	0.111	34. Kamari, Agios Stefanos−1	−0.017	0.041	0.025
163. Sigri_1	−0.092	0.083	0.055	70. Troulos (Mastichari)	−0.021	−0.013	−0.015
164. Sigri_2	−0.117	0.167	0.139	73. Marmari	0.174	0.057	0.056
203. Anaxos	0.220	0.101	0.178	75. Tigaki	0.098	0.056	0.056
206. Petra_3	0.091	0.032	0.109	77. Aiolos	0.008	0.052	0.036
208. Molyvos	0.058	−0.053	0.024	78. Faros, Lambi	0.033	−0.013	0.011

.

The ranking exercise for Lesvos beaches showed some interesting results. Eresos beach (ID 150), shows the highest TOPSIS ranking score (0.84), followed by Anaxos (ID 203, 0.76), Petra_3 (ID 206, 0.73), Tsamakia (ID, 65, 0.65), and Kanoni (ID 27, 0.64) (Figure 7a); this ranking reflects the high environmental and socio-economic significance of these beaches. However, when vulnerability to erosion is considered, there are changes in the ranking. On the basis of the current vulnerability, which was identified through the PROMETHEE II method, and using indicators (10) and (11) in addition to the environmental and socio-economic indicators, the two highest-ranking beaches are also Eresos (0.39) and Anaxos (0.22); these beaches are now followed by the beaches of Skala Kallonis (ID 125, 0.18), Petra_3 (0.09) and Molyvos (ID 208, 0.06) (Figure 7c and Table 7). The ranking is further modified under CV and C. Under the scenarios considered for 2050 (RCP4.5, median projections) and 2100 (RCP8.5 max. projections), Sigri_2 (ID 164, 0.17) and Anaxos (0.18) score highest, with Eresos dropping to the second and third place, respectively. These results have implications for the requisite management responses. Eresos and Anaxos beaches are the most obvious candidates for prioritized responses, as they consistently rank very high in terms of their environmental and socio-economic significance and their vulnerability to CV and C; the other beaches, however, show varying rankings, indicating that prioritization of management responses would also require decisions on the basis of the CV and C scenario selected, cost estimations, and the available resources.

For Kos beaches, the 5 highest ranking beaches in terms of their environmental and socio-economic indicators (TOPSIS analysis) were found to be Marmari (ID 73, 0,63), Marina_Aktaion (ID 7, 0.63), Tigaki (ID 75), Kamari (ID 34, 0.51) and Paradeisos_Kefalou (ID 32, 0.49) (Figure 8a). When the vulnerability to erosion is considered, the ranking changes. On the basis of the current vulnerability to erosion, the highest-ranking beaches were found to be Marmari (0.17) and Tigaki (0.1), followed by the Psalidi_2 (ID 9, 0.07), Faros_Lampi (ID 78, 0.03) and Aiolos (ID 77, 0,01) beaches (Figure 8c and Table 7). Under the scenarios considered for 2050 (RCP4.5, median projections) and 2100 (RCP8.5 max. projections), Marmari and Tigaki beaches also appear as the most vulnerable beaches, followed in both cases by the Aiolos, Faros_Ammoglosa (ID 1) and Kamari (ID 34 beaches. The consistency of these results shows a clear pathway for decision-makers; these 5 beaches should be prioritized for management responses.

5. Discussion

The erosion of sandy-shoreline beaches, which is already prevalent along the global coastline, will certainly be exacerbated due to the mean sea-level rise and (mostly) increasing extreme sea levels and waves due to climate change [13,55], with many implications for the coastal natural and human systems. The scale of the problem, the significance/criticality of the beaches for coastal (island) communities and the high potential costs of adaptation [27] require the development of approaches that can rank the need for and prioritize management responses, in order to allocate efficiently the (commonly) limited financial and human resources. In this study, a new approach is presented for assessing and ranking the vulnerability of beaches to climate variability and change (CV and C) at regional (island) scales. The approach combines easily obtained (from open access sources) environmental and socio-economic information with projections of beach erosion under CV and C and is applied to the beaches of the Greek islands of Lesvos and Kos, with the aim of ranking their vulnerability in a structured manner; the integration of socio-economic data with the exposure to erosion signifies the significant potential impacts on beach assets and populations [83]. The proposed framework comprises different stages/steps (Figure 2), has a wide potential for application at regional (and not only) scales, and can be exploited in various ways as a useful tool for the management of beach erosion. Its implementation involved the collation of various geo-spatial environmental and socio-economic information on all beaches of the islands of Lesvos (218) and Kos (78), which was then combined with projections of beach retreat/erosion under different scenarios of relative sea-level rise (RSLR) and extreme sea levels (ESLs); the vulnerability ranking was facilitated by an array of fit-for-purpose multi-criteria analysis approaches.

A particular characteristic of the framework is its emphasis on the socio-economic factors of the exposure to erosion of beaches, particularly tourism. Traditional (‘top-down’) beach vulnerability studies focus more on biophysical factors and less on socio-economic factors (‘bottom up’); this is an important gap as a combination of both sets of factors often results in higher damages and vulnerability [28,29]. The proposed framework ties together ‘top-down’ and ‘bottom-up’ factors to develop an integrated vulnerability ‘index’, composed of a wide range of geomorphological, environmental, and socio-economic factors, on the basis of which the island beaches are ranked in terms of the need for adaptation measures. By prioritizing adaptation responses for beaches critical to (3S) tourism through the incorporation of indicators/factors—such as those describing tourist visitation, accessibility, or recreational infrastructure—the method aligns with the need to evaluate the risk to the socio-economic development of island communities dependent on tourism. This specificity sets it apart from other coastal vulnerability assessments.

The framework employs a combination of fit-for-purpose multi-criteria decision-making (MCDM) approaches (AHP, TOPSIS, and PROMETHEE II) to assess and rank beaches systematically. AHP deals with the indicator weights in both TOPSIS and PROMETHEE II analysis and is based on stakeholder and expert judgments, ensuring consistency in the evaluation process. The AHP was selected due to its ability to incorporate expert judgments in decision-making processes characterized by subjectivity. It facilitates pairwise comparisons to determine the relative importance of criteria. In this study, AHP is employed exclusively for weight assignment, thus avoiding potential ranking reversal issues [84]. To rank a large number of beaches based on their criticality, the TOPSIS method was chosen. Unlike many other multi-criteria decision-making (MCDM) methods, TOPSIS maintains stable performance regardless of the number of alternatives or criteria. It ranks alternatives based on their proximity to an ideal solution while maximizing their distance from the worst-case scenario. For a more detailed evaluation of the (15) most critical beaches on each island, the Preference Ranking Organization Method for Enrichment Evaluations (PROMETHEE II) was applied; this method is particularly suited for ranking a finite set of alternatives based on multiple criteria [82,84]. PROMETHEE II relies on pairwise comparisons to determine the relative preference of alternatives, making it a robust preference-based ranking approach. However, its key limitation is the absence of a built-in mechanism for assigning criterion weights, a gap effectively addressed in this study by integrating AHP-derived weights.

The approach allows for more nuanced evaluations, combining beach erosion projections with socio-economic criticality. Compared to standalone ranking methods, the present approach provides a more balanced ranking/prioritization of beaches requiring management interventions. It should be noted that as the AHP method used to identify the indicator/factor weights relies on expert or stakeholder experience-based judgements, it might be prone to ‘biases’ arising from individuals’ judgements. Using equal weights instead, i.e., similar level of importance of the indicators, is not useful, as they can create greater discrepancies. In the future, a customized sensitivity analysis with regard to the significance of the (socio-economic) indicators might be developed to assist in managing such uncertainties and address this challenge [28].

Unlike other ranking methods that rely on assessments of beach exposure to sea-level rise through a combination of geomorphological and hydrodynamic information/indices (e.g., [30,38,42]), the present framework employs simulations by validated [26] 1-D morphodynamic model ensembles [55], to assess the spatio-temporal variability of beach erosion along the islands’ coasts under different scenarios of relative sea-level rise (RSLR) and extreme sea levels (ESLs). The approach offers a robust assessment of potential beach retreat ranges driven by marine forcing (i.e., sea levels and waves), using openly accessible data. It requires minimal environmental inputs that can be easily obtained, making it highly practical and easy to replicate in other coastal regions. Its simplicity and efficiency enable rapid evaluations at large spatial scales (i.e., regional (island) scales).

These projections add predictive capability, enabling the ranking method to account for future slow-onset erosion and storm impacts. Its advantage lies in the integration of simulated erosion metrics directly into the beach ranking process, so its ranking results reflect not only the current vulnerability but also long-term risks, offering a forward-looking perspective critical for adaptive management. However, it is noteworthy that the assessment/ranking is based on the current socio-economic factors/indicators, which might not be representative after many decades. It is likely that as coastal development will likely increase in the future (except if it is effectively regulated), these findings represent a rather conservative projection of the future beach vulnerability.

Concerning beach erosion, our results show that the majority of the beaches of both islands are already under erosion, with the problem being more acute in Kos. However, future erosion will be more significant for Lesvos beaches, under both the mean sea-level rise (RSLR) and extreme sea levels. By 2050, up to 11% of Lesvos beaches will likely retreat by half of their current maximum width (BMW) due to the projected RLSR, whereas, in 2100, up to 39% of all beaches might retreat by (at least) their entire current BMW, spelling trouble for backshore ecosystems, infrastructure and assets. Beach erosion/retreat is projected to be worse under the extreme sea levels; in 2050, 62–67% of the Lesvos beaches tested will (temporarily) retreat by their entire current BMW under the median model ensemble projections depending on the scenario; in 2100, 75% and 83% of the Lesvos beaches tested will erode by (at least) their current BMW under the RCP4.5 and RCP8.5, respectively. In Kos, no significant RSLR-driven erosion is projected for 2050 due to the relatively large beach widths. By 2100, however, beach erosion will be significant; 14% of Kos beaches are projected to retreat by (at least) their current maximum width. Under extreme sea levels (ESL₁₀₀), significant (temporary) retreats are projected. In 2050, 24% (RCP4.5) and 27% (RCP8.5) of the Kos beaches tested will retreat by (at least) their current BMW, whereas in 2100, 50% (RCP4.5) and 67% (RCP8.5) of the tested beaches will retreat by (at least) their entire current BMWs (median model ensemble projections). Such projections not only indicate challenges for the carrying capacity and hedonic value of the islands’ beaches but also suggest large impacts for those beaches with high occurrence of backshore assets, particularly under extreme sea levels under the high-end climatic scenario.

These projections suggest significant impacts on the islands’ beaches and their backshore ecosystems and assets, although they are rather conservative as they do not take into account other erosion factors, such as the cumulative effects of storm events [85], the effects of human activities related to tourism [10,86]; the diminished land sediment supply to the beaches over the years due to, e.g., dam construction [87], or the projected decreases in precipitation [88,89]. In addition, potential future land vertical motions due to tectonics and anthropogenically induced subsidence have not been included in the RSLR projections abstracted from the JRC database due to the spatial resolution of the dataset (~100 km) [25]. It should also be noted that, although the study has focused on beach erosion, there could be other significant CV and C effects that might potentially impact the environmental and socio-economic sustainability of island beaches, such as the deterioration of the beach climatology and ecological status [90,91]. Thus, the proposed approach could benefit from the incorporation of indicators related to such effects.

The proposed framework offers a useful tool for assessing the condition of island beaches by integrating geo-spatial and socio-economic characteristics with projections of beach erosion and retreat under various CV and C scenarios. It provides coastal managers and governance institutions with a better understanding of the complex challenges posed by beach erosion, as well as a methodology to rank beach vulnerability, prioritize adaptation measures, and optimize the allocation of often limited resources; it may also facilitate compliance to and implementation of regulation relevant to climate change adaptation. It can also provide accessible, reproducible, and transferable methods and tools so that similar vulnerability assessments may be performed for other types of hazards at different locations.

6. Conclusions

This contribution introduces a framework for the systematic evaluation and ranking of ‘sandy’ shorelines—beaches at a regional (island) scale, in terms of both their socio-economic (touristic) significance/criticality and their exposure to and the impacts of sea-level rise under CV and C and, thus, their need for adaptation. It uses accessible geo-spatial environmental and socio-economic information, beach erosion projections under the CV and C and an array of MCDM approaches to provide a structured approach to rank beaches in terms of their vulnerability to beach erosion.

In addition to providing a more holistic approach to coastal vulnerability analysis, the proposed approach, unlike other ranking methods that rely on assessments of beach exposure to sea-level rise through a combination of geomorphological and hydrodynamic information/indices), employs simulations by morphodynamic model ensembles to assess beach erosion (at regional scale) under different scenarios of relative sea-level rise (RSLR) and extreme sea levels (ESLs). This approach can account for future slow-onset erosion and storm impacts by integrating simulated beach erosion metrics directly into the ranking process. Implementation of the approach at the islands of Lesvos and Kos has shown that there could be very significant effects of the mean sea-level rise and (particularly) of the extreme sea levels on the carrying capacity as well as the capability of the beaches to buffer backshore assets unless appropriate adaptation measures are taken.

Another interesting feature of the proposed approach is that it relies mostly on widely available information for the collation of the socio-economic indicators necessary for assessing the significance/criticality of the beach; thus, it can provide a reproducible and transferable methodology that can be applied at different locations and spatial scales and assist users, primarily coastal managers and relevant institutions of governance, in developing a better understanding of the challenge posed by the CV and C driven beach erosion. It is noted, however, that although care has been taken to identify reasonably well the indicator/factor weights used in the study, the reliance on expert/stakeholder experience-based judgements might still introduce biases; in future iterations, this challenge could be (to some extent) addressed by introducing additional components into the framework, such as customized sensitivity analysis and (ideally) validation procedures where possible.