Introduction

Coastal habitats are widely recognized for their contribution to coastal protection against flood hazard. Sand dunes act as natural barriers against rising sea levels1,2,3,4,5, seagrass meadows effectively dissipate wave energy and entrap cross-shore transported sediment6,7,8,9 and saltmarsh help attenuate impacts of storm surge and flash floods10,11,12,13. Notwithstanding their importance, these habitats are currently experiencing an unprecedented, global-scale decline in both spatial extent and ecological value, driven both by anthropogenic and climatic stressors14,15,16. This decline is expected to be further exacerbated by climate change, which will intensify coastal erosion through sea level rise and more frequent and intense storms. This degradation trend not only implies a loss of the biodiversity these habitats host, but also undermines the stream of benefits they provide to humankind17.

In this regard, Nature-based Solutions (NbS) are rising as effective ecosystem-based measures to mitigate flood risk and adapt to climate change18,19,20,21,22,23,24. Findings increasingly support the idea that these solutions may offer benefits that surpass their associated costs and, in certain instances, can be more cost-effective than traditional hard engineering solutions25,26,27. Moreover, NbS are acknowledged as equitable solutions that ensure that the benefits of enhanced resilience are accessible to all communities, including those that are economically disadvantaged or disproportionately vulnerable to climate impacts28,29. Although some studies highlight potential, unintended disbenefits associated with their implementation30,31, numerous cases worldwide confirm these advantages13,24,32,33.

Evaluation of flood risk reduction of NbS is traditionally pursued through cost-benefit analyses or more in general monetary-based assessment34. However, these type of assessments inherently overlook the non-monetary benefits that NbS provide38,39. Ecosystem Service (ESS)-based frameworks address this limitation by encouraging an equitable treatment of both monetary and non-monetary ecosystem values, placing budgetable benefits (e.g., damage reduction) and non-budgetable benefits (e.g., human well-being, cultural, recreational values) on the same scale17. Such perspectives support decision-making processes that are more inclusive of ecological and societal needs, fostering planning that is more holistic, just, resilient, and sustainable37,38,39,40,41.

Since the ESS concept was made popular by the Millennium Ecosystem Assessment42, the demand for robust and applicable methodologies for ESS assessments has increased. A popular and well-established method is the ESS-matrix approach43,44. This allows to evaluate and quantify the delivery of ecosystem services across different land cover or habitat types. It involves creating a matrix where habitat types are paired with ecosystem services, and scores are assigned to reflect the capacity of each habitat to provide specific benefits. The approach is flexible and can be adapted to various scales and contexts, making it valuable for land use planning, conservation, environmental impact assessments, and policy development45. By visualizing the relationships between habitats and ecosystem services, the method helps identify trade-offs and synergies, guiding sustainable management decisions. However, existing ESS-matrix approaches are mostly based on qualitative scoring, that relies on expert-based ESS assessments or, more often than not, on direct literature transfer of non site-specific scores from previous studies. In this regard, physically-based evaluations, in which scores are based on metrics from field- or model-based data are rather scarce46. Unless limited to an initial “quick-scan” assessment, qualitative expert-based approaches introduce deep uncertainties due to inherent subjectivity of the input data. This is particularly true when considering future climate scenarios, which already involve considerable unknowns related to the uncertain prediction of future changes in relevant variables and propagation of errors47,48.

The present study addresses this gap by developing a model-based approach to assess the flood risk reduction ecosystem service (FRR-ESS) provided by NbS. We conduct a comparative assessment of individual and combined effects of various NbS on flood risk reduction under different climate change scenarios. NbS are evaluated singularly and combined as Building Blocks (BB) to assess their synergic efficacy. By utilizing eco-hydro-morphodynamic numerical modelling with detailed habitat maps, we simulate hydrodynamic processes, sediment transport, and flood inundation to quantify changes in FRR-ESS due to habitat and hydrodynamic conditions change. An FRR-ESS score is then quantified by means of a scorecard methodology.

This contribution aims to bridge the gap between qualitative ESS assessments and quantitative modelling, providing a more holistic approach to evaluate flood risk reduction. By integrating expert-based analyses with physically-based numerical models, the methodology encompasses both hydro-morphodynamic and eco-dynamic processes. In particular, the inclusion of eco-hydraulic integrated modelling of low-lying transitional areas allows for a more accurate representation of vegetation’s role in dissipating wave energy and mitigating flood impacts, by explicitly accounting for site-specific plants’ characteristics and spatial distribution. This enables the simulation of vegetation’s direct influence on flow resistance, wave energy dissipation, and sediment transport under varying hydrodynamic conditions. This approach intends to enhance the reliability of ESS evaluations under current and future climate scenarios, offering policymakers and stakeholders a robust, flexible tool to support ecological sustainability and disaster risk reduction.

The work is structured as follows: first a description of the overall methodology, the numerical modelling and the ecosystem service assessment framework is presented; then, the outcomes of the model and the ESS evaluation methodology are illustrated and discussed; Finally, a conclusive section closes the work with some final remarks.

Materials and methods

Case study

The “Pantani della Sicilia Sud-Orientale” is a saline lagoon and wetland system located in the South-East of Sicily (Italy) (Fig. 1a), internationally protected under RAMSAR and Natura 2000 network49.The Cuba-Longarini lagoons are the largest wetland area in the system, covering approximately 300 hectares, separated from the sea by a narrow coastal fringe composed of a dune strip and sandy beach. On the East side, the Cuba lagoon extends approximately for 80 ha, whereas on the West side, the Longarini lagoon extends for around 180 ha. These lagoons experience occasional connection to the sea through a small estuary channel, depending on the water level. They are generally shallow, less than one meter deep. The lagoons play a crucial role in biodiversity conservation, serving as a key habitat for endangered species and an important stopover for migratory birds50.

The construction of the village of Granelli in the 1970s, on the narrow fringe between the sea and the lagoons, has reduced the hydraulic connectivity between the wetlands and the sea, while also degrading the coastal dune belt over which the urbanized area is located, increasing flood risk and weakening the area’s natural defenses against coastal hazards (Fig. 1b). Since 1955, the Granelli coastline has experienced severe erosion, with the beach narrowing from 99 meters to just 11 meters51, particularly along a 2 km central stretch (Fig. 1c). Moreover, the area is increasingly vulnerable to climate-related hazards, with more frequent and intense wave storms causing damage, and a risk of permanent flooding due to projected sea level rise by 210052,53. The characteristics of this area are representative of many Mediterranean coastal zones, where intense urbanization and land use changes have squeezed wetlands, reducing their potential to provide protection from natural disasters54,55.

Fig. 1
figure 1

Satellite image and overview of the test site, with bathimetries and wave rose (a), aerial view of inundation in the residential area of Granelli due to Helios storm (11th February 2023) (b), coastal erosion in front of the Granelli village (c). Satellite image in subpanel (a) was obtained from Planet.com (\(\copyright\) Planet Labs PBC, CC BY-NC-SA 2.0) and the figure was created by QGIS v3.22 software (https://download.qgis.org/downloads/).

Full size image

The proposed Nature-based Solutions building blocks (NbS-BB)

With the aim of mitigating coastal risks and restoring biodiversity, a series of NbS were analyzed. Specifically:

  • Dune revegetation (DR) This measure seeks to enhance the system’s wave dissipation capacity while restoring the dune habitat. The endemic species Ammophila arenaria, already present in the area, was selected for the revegetation process. The proposed intervention involves expanding the dune habitat across approximately 4.3 hectares, stretching 3.3 km along the coastline with an average width of 13 metres.

  • Reconstruction of the Posidonia oceanica seagrass meadow (SR) This protected marine plant plays a critical role in reducing storm-induced coastal erosion and dissipating wave energy. Currently spanning 76.8 hectares, the meadow is planned to expand to 180.52 hectares, representing an increase of 103.72 hectares compared to its current coverage.

  • Beach nourishment (BN) This intervention addresses the severe erosion affecting the Granelli coastline, where the beach width has significantly decreased from 99 metres in 1955 to 11 metres along the central 2 km section. The restoration measure involves re-establishing 20 metres of beach width along a 3.3 km stretch by depositing 120,000 m3 of sand sourced from an offshore deposit. The design prioritizes natural restoration and avoids hard traditional coastal protection structures.

In addition, the proposed NbS were combined within a Building Block approach56 in order to investigate their joint contribution in reducing storm-induced flooded areas. Specifically, the following combinations are considered: (i) reconstruction of the Posidonia oceanica meadow and dune revegetation (DR+SR); (ii) dune revegetation and beach nourishment (DR+BN); (iii) reconstruction of seagrass meadow and beach nourishment (BN+SR). A map of present habitat conditions and of the proposed restoration interventions is reported in Fig. 2.

Fig. 2
figure 2

Map showing the present habitat conditions at the test site and the proposed NbS-BB. Satellite image was obtained from Planet.com (\(\copyright\)Planet Labs PBC, CC BY-NC-SA 2.0) and the figure was created by QGIS v3.22 software (https://download.qgis.org/downloads/).

Full size image

The numerical modelling chain

The modelling chain applied to our case study consists of the combination of two numerical models, SWAN and XBeach. SWAN57 (Simulating Waves Nearshore) is a third-generation spectral model, capable of generating random wind-generated short-crested waves in coastal regions and inland waters. It allows to simulate processes related to the propagation of wind waves, i.e. generation, refraction, dissipation, shoaling and wave-interaction processes. Furthermore, SWAN provides one- and two-dimensional wave spectra at any number of positions and a spatial distribution of various parameters, including: significant wave height, wave period, wave direction and strength, and directional spread. In the context of our work, SWAN is used for hydrodynamic modelling of wave propagation from offshore to nearshore.

Within the SWAN grid, a nested eco-hydro-morphodynamic XBeach58 model is setup. XBeach (eXtreme Beach behavior) is a numerical model designed to simulate nearshore hydrodynamic and morphodynamic processes. It is a two-dimensional depth-averaged model, which solves coupled cross-shore and long-shore equations for wave propagation, flow, sediment transport and bottom level changes during storms.

Wave dissipation due to the presence of aquatic vegetation is modelled in XBeach using the approach of Mendez and Losada (2004)59. It consists of an empirical model to predict wave height decay for both breaking and nonbreaking random waves propagating through a vegetation. The vegetation field is treated as a source of energy dissipation in the short-wave action balance equation, by representing vegetation elements as cylindrical obstacles that exert drag on the flow. The model calculates wave height attenuation by incrementally reducing wave energy across each computational step, with the drag contribution governed by vegetation characteristics (e.g., density, diameter, height) and the local wave conditions (e.g., water depth, wave height).

Howewer, representing vegetation elements as rigid cylinders is a simplifying assumption that does not account for complexities related to vegetation-induced turbulence, plant flexibility and reconfiguration under flow, or morphological heterogeneity60,61,62. More advanced modelling approaches do exist, which incorporate data-driven drag coefficient prediction63, more advanced wave-resolving frameworks64, complex vegetation morphologies63, partially vegetated channel flows65 and vegetation-induced turbulence in the presence of flexible canopies66. Notwithstanding these assumptions, the theoretical framework successfully represents wave height transformation over vegetated regions and captures the associated damping governed by the bulk drag coefficient with reasonable accuracy59.

In the present application, we employed the vegetation empirical model version adapted by Suzuki et al. (2012)67 to consider vertically heterogeneous vegetation , i.e. to take into account different morphological and hydrodynamic characteristics along the vegetation vertical section. A dissipation term is calculated as the sum of the dissipation per vegetation layer and takes into account parameters specific to the vegetation: number of vertical sections (\(n_v\)), vertical section height (\(\alpha h\)), drag coefficient (\(C_d\)), stem diameter (\(b_v\)), vegetation density (N). Specifically, this term is calculated through the following equation:

$$\begin{aligned} D_v = \sum (D_{v,i}) = \frac{1}{2 \sqrt{\pi }}\rho C_d b_v N (\frac{kg}{2 \omega })^3 \frac{(sin h^3 k \alpha _i h - sin h^3 k \alpha _{i-1} h) + 3(sinh k \alpha _i h - sinh k \alpha _{i-1} h)}{3k cos h^3 kh } H^3_{rms} \end{aligned}$$
(1)

where: \(D_{v,i}\) is the short wave energy dissipation due to i-th vegetation layer i, \(\rho\) is the water density, \(C_d\) is the drag coefficient, \(b_v\) is the stem diameter, N is the stem density per m2, k is the wave number, g is gravitational acceleration, \(\omega\) is angular frequency, h is the water level, \(\alpha _i\) is the relative vegetation height (= \(h_v/h\)) for the vegetation layer i, \(H_{rms}\) is the root-mean-square wave height.

Data description and investigated scenarios

In order to build the numerical grid for the modelling chain, bathimetric data from nautical charts68 and bathymetric data obtained through a multibeam survey conducted by the University of Catania in 2019 were combined with a 2x2 m digital elevation model obtained by a LIDAR flight carried out by the Sicilian Region69. The coastline position was traced using satellite images derived from PlanetScope70. For marine climate data, wave conditions were obtained from the MEDSEA MULTIYEAR WAV 006 012 reanalysis dataset71, for the years 1993-2022 with hourly temporal resolution. For future wave conditions, hourly wave climate projections72 for the period 2041-2100 with two emission scenarios proposed by the IPCC AR5: the RCP 4.5 scenario that considers a moderate decrease in greenhouse gases beyond 2040; and the RCP 8.5 scenario, a pessimistic scenario in which emissions will continue to increase until 2100. Both wave reanalysis and projection datasets are openly accessible through the Copernicus service website71,72.

For these present and future projection datasets, parameters were extracted, including: significant wave height (\(H_s\)), peak period (\(T_p\)) and mean wave direction (\(\theta\)). Since a prevalence of events from the western direction was observed, i.e. 255-270 \(^\circ\)N (see Fig. a), in terms both of significant wave height magnitude and frequency, we focus our analysis on that sector. An extreme event analysis was conducted to obtain \(H_s\) with return period \(T_r\) of 50 years, using a 30-year long timeseries for each scenario: 1993-2022 for the current scenario, 2041-2070 and 2071-2100 for the mid-term and long-term forecast scenarios, both for RCP 4.5 and RCP 8.5.

The local sea level rise projections, were based on the IPCC AR6 SSP (Socioeconomic Pathways) scenarios, and were obtained using the NASA Sea Level Change projection tool73. These correspond to the last year of the annual windows examined, 2070 and 2100, for the periods 2041-2070 and 2071-2100 respectively. Since the only wave climate projections available for the Mediterranean originate from AR5, while the sea-level rise projections are based on AR6, it was deemed appropriate to combine the RCP and SSP scenarios corresponding to the same radiative forcing values to ensure consistency in the analysis. Therefore, the wave conditions for the RCP4.5 and RCP8.5 scenarios were combined with the sea-level rise projections for the SSP2-4.5 and SSP5-8.5 scenarios (radiative forcing of 4.5 W m-2 and 8.5 W m-2 respectively). By combining extreme wave climate parameters and sea level rise, five hydrodynamic scenarios were studied: a present scenario (SP) and four climate change projection scenarios: S4.5-2070, S4.5-2100, S8.5-2070 and S8.5-2100. The parameters of the investigated hydrodynamic scenarios are reported in Table 1.

Table 1 Wave climate parameters and SLR for each hydrodynamic scenarios.
Full size table

The characteristics of the vegetation and its spatial distribution, used for the simulation of wave dissipation, were obtained through Carta Natura habitat maps74 supplemented with field surveys by experts at the study site. Furthermore, through various samples collected in the field, the grain size characteristics of the Granelli beach were obtained, which is characterised by fine sand, with grain size fractions of 50 and 90 percent (\(D_{50}\) and \(D_{90}\)) of 0.26 mm and 0.29 mm, respectively. While the particle size data used for the beach nourishment, with a \(D_{50}\) of 0.45 mm, were selected from the finest available from a submarine deposit located off the Sicilian coast, West of the port of Mazara del Vallo, in the Sicily strait.

Numerical modelling setup and validation

Figure 3a illustrates the flowchart of the numerical modelling chain. The SWAN computational domain, illustrated in Fig. 3b, covers 2068 km2, including approximately 90 km of coastline. The grid consists of 17,141 nodes with variable resolution based on bathymetry: for depths greater than 100 m, elements have a 1 km side; for depths between 100 m and 30 m, the size gradually decreases from 1 km to 400 m; and for depths less than 30 m, the mesh size is a constant 400 m. A finer 100 m grid is applied near the XBeach domain.

The XBeach computational domain, shown in Fig. 3c, is located within the SWAN domain and spans 3.36 km cross-shore and 3.69 km long-shore, covering a total area of 10.95 km2. It includes the Cuba-Longarini lagoon system, the town of Granelli, and its beach. The structured grid has variable cell sizes: 2.5-10 m cross-shore and 5-10 m long-shore, with 1,150 cross-shore and 544 long-shore cells. Each cell is assigned an elevation value based on topographic data.

Fig. 3
figure 3

Modelling chain applied to our case study (a): domain and grid of the SWAN model (b), domain and grid of the XBeach model (c), vegetation map for the No-NbS scenario (d).

Full size image

In order to take into account the vegetation within the study area, a representative vegetation species was assigned for the different habitats on the basis of a cross-comparison between EUNIS habitat mapping and field surveys. In particular, four representative species were considered for the four habitats: Ammophila arenaria, Phragmites australis, Posidonia oceanica, Spartina alterniflora. The respective values of their characteristics and their spatial distribution, implemented in the model, are reported in the table in the Supplementary materials. The Fig. 3d shows the distribution map of vegetation for the No-NbS scenario.

Both SWAN and XBeach models were validated to assess their hydrodynamic and morphodynamic accuracy. A real storm event, occurring from April 21, 2022, at 06:00 UTC+1 to April 22, 2022, at 07:00 UTC+1, was selected for validation. This 25-hour storm reached a peak significant wave height of 4.19 m, measured by a wave buoy installed by the University of Catania offshore of the Granelli Beach (latitude 36.63320, longitude 14.9348). The pre-storm shoreline from April 19, 2022, was used as the initial condition. The storm was simulated using the SWAN and XBeach modelling chain, and the post-storm shoreline was compared to satellite imagery from May 4, 2022. No other significant storm events occurred in the interim. Hourly wave data from April 20 to April 25, 2022, were extracted from the MEDSEA MULTIYEAR WAV 006 012 reanalysis dataset at the grid point closest to the SWAN offshore boundary. Each hourly sea state was simulated, totaling 120 SWAN simulations. The model’s performance was evaluated using Root Mean Square Error (RMSE), Bias (B), and the Willmott Skill Score75, which resulted in values of 0.37 m, 0.02 m, and 0.95, respectively, indicating strong agreement between the model and observed wave conditions. For morphodynamic validation, the modeled post-storm shoreline was compared with the observed shoreline from satellite imagery. Measurements were taken along 30 transects orthogonal to the pre-storm shoreline, assessing the distance between observed and modeled shoreline retreat. The Bias was 0.93 m (6.4% of the observed average retreat of 14.42 m), while the RMSE was 3.30 m. The Brier Skill Score76, a widely used coastal morphodynamic validation metric, was computed, resulting in a value of 0.95, confirming the model’s strong predictive skill in shoreline change. Further details on the validation of the proposed modelling chain can be found in Marino et al. (2025)77.

Regarding vegetation modelling, we relied on literature-derived values; however, an accurate estimation would ideally require calibration using pre- and post-storm beach profiles. To assess the sensitivity of the model, we varied the drag coefficient for Posidonia oceanica between 0.7 and 1.3, centered around a reference value of 1.0. The results indicated that increasing the drag coefficient to 1.3 led to a 10% reduction in the flooded area, while decreasing it to 0.7 resulted in a 6% reduction. Analogously, a sensitivity analysis was performed on the stem density of Posidonia oceanica, varying around the reference value of 300±100 stems/m2. The results show that increasing or decreasing the density by 100 stems has a limited impact on flooding, with the flooded area ranging from 0.065 km2 to 0.056 km2 and mitigation effectiveness varying from -6.5% to +7.5%, showing a limited sensitivity of the model to this parameter.

Flooding risk reduction ESS assessment methodology

In order to evaluate the change in flooding risk reduction (FRR) ecosystem service, the following methodology is employed. First, a benchmark habitat map is built, showing the habitat distribution at the current state, i.e. present scenario (SP) without any NbS-BB implemented (NN). For each habitat in the benchmark habitat map, FRR-ESS rank scores are obtained by means of a local expert-based survey approach. These scores provide a qualitative representation of the flood reduction ecosystem service provision for each habitat. The scores were assigned by a team of 10 experts with different expertise and backgrounds (hydraulic engineering, geology, botanics, biology and ecology). Each member of the team individually compiled a questionnaire assigning a score ranging from 0 to 5 for each habitat about their contribution to FRR-ESS. A mean score rounded to the nearest integer was calculated for each habitat to obtain the benchmark ESS rank score, or \(ESS_{rs}^0\) (Table 2).

Subsequently, habitat change due to sea level rise is evaluated for the 4 investigated climate change scenarios, thus generating 4 maps for each of SLR scenarios. Given the inherent complexity and uncertainties associated with habitat transitions under future sea-level rise scenarios, our approach employs a simplified framework based on key environmental drivers, i.e. salinity changes, hydraulic connectivity and elevation thresholds. While the model does not explicitly simulate intermediate processes (e.g. gradual shifts in salinity regimes) these factors are indirectly captured through clear, ecologically grounded transition criteria.

Three main habitat change criteria were considered applied in this framework: (i) Saline coastal lagoons transition to Mediterranean infralittoral sand when a permanent connection to the sea is established, driven by the loss of seasonal water level fluctuations and the influence of marine salinity, which favors nearshore ecotopes and reduce the natural capabilities of the lagoon to retain rainflood water; (ii) Reedbeds without free-standing water shift to Mediterranean littoral biogenic habitats, as increasing salinity levels inhibit freshwater species like Phragmites australis, allowing more salt-tolerant halophytic communities to dominate; (iii) a generic habitat transition to Mediterranean infralittoral sand when the local elevation falls below the projected mean sea level, following a traditional “bathtub” approach that reflects the inundation-driven loss of terrestrial habitats. This approach, though simplified, offers a robust means of assessing habitat shifts within the study’s spatial and data limitations, allowing for a meaningful evaluation of NbS effectiveness under evolving environmental conditions.

Table 2 FRR-ESS rank score \(ESS_{rs}^0\) for each habitat obtained by expert-based approach.
Full size table

The scores range from 0 to 5, where 0 indicates low or no habitat flood reduction potential and 5 indicates high flood reduction potential. These scores are considered our benchmark, i.e. they represent the contribution of each habitat to FRR-ESS at the current state, without any NbS-BB implemented. In scenarios where NbS-BB are implemented, the \(ESS_{rs}^0\) may change based on the FRR-ESS efficacy indicator \(E_{NbS}\). This term is computed by the following:

$$\begin{aligned} E_{NbS} = \dfrac{E_{fc}}{A_{hc}} \end{aligned}$$
(2)

where \(E_{fc}\) is the relative flooding reduction due to an NbS on a certain investigated scenario, whereas \(A_{hc}\) is the relative change of the habitat for that specific NbS-BB. The \(E_{fc}\) is computed by:

$$\begin{aligned} E_{fc} = \dfrac{A_{f,NN}-A_{f,NbS}}{A_{f,NN}} \end{aligned}$$
(3)

where \(A_{f,NN}\) is the flooded city area in the case of no-NbS compared with the flooded city area in the presence of a NbS-BB, \(A_{f,NbS}\), for a certain hydrodynamic scenario; whereas the relative change of habitat \(A_{hc}\) is computed as:

$$\begin{aligned} A_{hc} = \dfrac{A_{h,NbS}-A_{h,NN}}{A_{h,NN}} \end{aligned}$$
(4)

where \(A_{h,NbS}\) is the habitat surface that changed in the NbS intervention scenario, compared with the pre-NbS habitat surface \(A_{h,NN}\) for a specific hydrodynamic scenario. Thus, \(E_{NbS}\) is the ratio between the reduction of flooded area determined by the NbS-BB to the habitat restored surface. For instance, an \(E_{NbS}\) = 1 means that 1 m2 restored correspond to 1 m2 flooded area reduction.

This indicator is used to adjust the \(ESS_{rs}^0\) of each restored habitat, reflecting variations in flood risk reduction efficacy provided by the NbS. Specifically, the rank score increases linearly by 1 point for every increase in \(E_{NbS}\) by a factor \(f = 0.75\), up to a maximum score of 5. Beyond this limit, further improvements in \(E_{NbS}\) do not increase the score. The methodology also considers scenarios where the NbS-BB leads to a decline in FRR-ESS. In such cases, the rank score decreases linearly by 1 point for each decrease in \(E_{NbS}\) by a factor \(f = 0.75\). However, the minimum attainable score is 0, and further reductions in \(E_{NbS}\) will not lower the rank score below this value. This correction allows to obtain the actual ESS rank score, or \(ESS_{rs}\), which follows this formulation:

$$\begin{aligned} ESS_{rs} = {\left\{ \begin{array}{ll} \max \left( -5, \min \left( ESS_{rs}^0 + \frac{E_{NbS}}{f}, 5 \right) \right) & \text {if } E_{NbS} \ge 0, \\ \max \left( -5, \min \left( ESS_{rs}^0 - \frac{E_{NbS}}{f}, 5 \right) \right) & \text {if } E_{NbS} < 0. \end{array}\right. } \end{aligned}$$
(5)

Figure 4a illustrates three examples of \(ESS_{rs}\) computations due to NbS-BB interventions, corresponding to habitats with low/null FRR-ESS (red continuous line, benchmark rank score \(ESS_{rs}^0 = 0\)), intermediate FRR-ESS (orange dashed line, bench mark rank score \(ESS_{rs}^0 = 2\)), and high FRR-ESS (green dotted line, benchmark rank score \(ESS_{rs}^0 = 5\)).

These scores correction are computed in scenarios in which only one habitat is restored (the NbS-BB) and then used for the combined NbS-BB scenarios within the same hydrodynamic scenarios.

Fig. 4
figure 4

Methodology to correct variation of FRR-ESS rank scores \(ESS_{rs}\) due to NbS-BB implementation based on NbS efficacy \(E_{NbS}\) (a); computation of the ESS sigma score \(\sigma _{ESS}\) depending on the relative ESS score \(ESS_{rel}\) (b).

Full size image

Afterwards, the total ecosystem service score \(ESS_{tot}\) is computed by calculating the weighted contribution of each habitat to FRR:

$$\begin{aligned} ESS_{tot} = \sum _{i=1}^{n_h} A_{h,i} \cdot ESS_{rs,i} \end{aligned}$$
(6)

where \(n_h\) is the number of habitats, \(A_{h,i}\) is the surface of each habitat i and \(ESS_{rs,i}\) is the ESS rank score of the ecosystem service for EUNIS habitat i. This score represent the total contribution of the whole site to the FRR-ESS in a given scenario. Any variation in the surface area of habitats due to a new condition (e.g. sea level rise or implementation of an NbS-BB) results in a change in the \(ESS_{tot}\). This change is quantified as a relative difference:

$$\begin{aligned} ESS_{rel} = \frac{ESS_{tot,2}-ESS_{tot,1}}{ESS_{tot,1}} \end{aligned}$$
(7)

where \(ESS_{rel}\) is the relative change in the ESS total score at time 2 with respect to time 1, where time 1 is always the benchmark condition (SP, NN) and time 2 is a future condition, either with/out implementation of NbS-BB or future climate scenario. In this context, \(ESS_{tot,1}\) is the ESS total score at the time 1 and \(ESS_{tot,2}\) is the ESS total score at the time 2.

The relative change in the ESS score ranges from -1 to +1, reflecting the degree of improvement (+) or decline (-) in the provision of the ecosystem service. To address the issue of comparing small changes (a few percent) with large changes (tens of percent) in total ESS scores, a transfer function is applied. This function uses a sigmoid curve that outputs values between -5 and +5, corresponding to relative changes between -1 and +1.

$$\begin{aligned} \sigma _{ESS} = \frac{10}{1+e^{-k ESS_{rel}}}-5 \end{aligned}$$
(8)

here, \(\sigma _{ESS}\) is the sigma-score (ranging from -5 to 5) and k is the logistic function shape parameter. An optimal value of k = 10 was determined by sensitivity analysis. The sigma-scores are then grouped into eleven categories (as shown in Fig. 4b) to express the relative change in ESS in a consistent, uniform metric for both gains and losses.

Results

By combining 5 hydrodynamic scenarios (1 present, 4 future projections) with 7 NbS-BB scenarios (1 no-NbS scenario, 3 with single NbS-BB, 3 with combined NbS-BB), a total of 35 simulations were conducted to assess the efficacy of NbS-BB in mitigating coastal flooding due to a 50 years return period storm.

Figure 5 shows flooding maps related to S4.5-2100 hydrodynamic scenario (\(H_s\) = 7.21 m, \(T_p\) = 11.16 s, SLR = 0.57 m) for no-NbS (NN, Fig. 5a), seagrass meadow reconstruction (SR, Fig. 5b) and combination of seagrass reconstruction with beach nourishment (SR+BN, Fig. 5c). In the NN case, a significant extent of the flooded city area is observed, amounting to 0.061 km2. The SR intervention allows to reduce the flooded area to 0.028 km2, with a corresponding flooding reduction efficacy \(E_{fc}\) = 53%. Figure 5c illustrates a combination of seagrass reconstruction and beach nourishment (SR+BN). The advancement of the shoreline allows for a further reduction in comparison with the SR case, with a flooded city area of 0.017 km2, corresponding to \(E_{fc}\) = 72%.

The bar plots in Fig. 5d shows the city flooded surfaces \(A_{fc}\), whereas Fig. 5e reports the flooding reduction efficacy \(E_{fc}\), for all hydrodynamic scenarios (in shades of blue) and for each proposed NbS solution (along the x-axis).

The NN case results show that for the present scenario (SP) no flooding is observed, whereas significant flooded areas for future scenarios, with the most severe being the S8.5-2100 with a city flooded surface \(A_{fc}\) = 0.113 km2 (36% of the total city area 0.40 km2). The DR intervention generally resulted the less efficient among all the proposed solutions, with city flooded surface between 0.021 and 0.113 km2 (\(E_{fc}\) = 17\(\div\)24%). The SR intervention generally outperformed DR, with \(A_{fc}\) between 0.024 km2 and 0.064 km2 (corresponding to \(E_{fc}\) = 44 \(\div\) 55%); while the beach nourishment intervention (BN) brings the values of \(A_{fc}\) down to 0.012 km2 and 0.042 km2. The BN outperforms DR and SR across all scenarios, with city flooded surfaces between 0.010 and 0.031 km2, and \(E_{fc}\) ranging between 67 to 74%. The combination of DR+SR slightly increases performance of the seagrass meadow reconstruction alone, with flooded surfaces in the range 0.015 to 0.057 km2, corresponding to \(E_{fc}\) between 50 and 64%. Combining DR with the BN generally give very slight improvements, with city flooded surface from 0.009 km2 and 0.037 and efficacy \(E_{fc}\) ranging from 67 to 75%. Finally, BN+SR combination provides the greater performance, with flooded areas ranging from 0.010 and 0.0312, corresponding to flooding reduction efficacy of 61 to 78%.

Fig. 5
figure 5

Inundation maps for the hydrodynamic scenario S4.5-2100, with \(H_S\) = 7.21 m, \(T_p\) = 11.16 s, SLR = 0.57 m and \(\theta\) = 262.5\(^\circ\)N, for three different NbS-BB scenarios: no-NbS, NN (a); seagrass meadow reconstruction, SR (b); combination of seagrass meadow reconstruction and beach nourishment, SR+BN (c). City flooded surface \(A_{fc}\) (d) and flooding reduction efficacy \(E_{fc}\) for all investigated hydrodynamic scenarios (in shades of blue) (e).

Full size image

The findings regarding the assessment of FRR-ESS alterations due to sea-level rise, both with and without the implementation of NbS-BB, are presented. The proposed methodology was applied to evaluate FRR-ESS relative changes for each scenario, quantifying improvements or declines in ESS provisioning for each habitat.

Figure 6 shows the habitat maps obtained through the expert-based approach for SP (Fig. 6a), S4.5-2070 (Fig. 6b), and S8.5-2100 (Fig. 6c). All habitat maps represent scenarios without NbS implementation (NN).

The comparison between SP and S4.5-2070 habitat maps reveals that the projected sea-level rise (SLR = 0.37 m) leads to a narrowing of the already thin beach strip. Additionally, the sea-level increase causes the opening of a channel in the estuarine region of the Longarini lagoon, creating a permanent connection with the sea. This connection equalizes salinity levels between the sea and the lagoon, alters water levels, and transitions the habitat from saline lagoon (EUNIS code X02) to Mediterranean infralittoral sand (MB55), with part of the “Mediterranean littoral biogenic habitat” (MA25) permanently submerged. Saline lagoon habitat reduces by 63% (from 1.76 to 0.65 km2), while Mediterranean infralittoral sand increases by 53% (from 3.55 to 5.44 km2). These changes impair the lagoon’s natural flood mitigation capacity, as it can no longer function as an expansion reservoir, leading to a reduction in the flood risk reduction score from 4 to 1. Reedbeds habitat (D5.1-2012) is particularly affected, decreasing by 84% (from 0.54 to 0.09 km2) due to increased salinity, favoring coastal saltmarsh habitat, which itself decreases by 14% due to partial submersion. Beach retreat further reduces its habitat from 0.10 to 0.06 km2 (40%). As a result, the total contribution of all habitats to FRR-ESS (\(ESS_{tot}\) score) decreases by 33%, from 2.16 \(\cdot 10^7\) to 1.43 \(\cdot 10^7\).

Figure 6c depicts the habitat map for the S8.5-2100 scenario (SLR = 0.77 m). In this extreme case, the sea inundates the Cuba lagoon entirely, eliminating the coastal lagoon habitat, and extends into urban areas. Mediterranean infralittoral sand habitat increases by 97% compared to SP, while reedbeds are completely suppressed (100% reduction). Coastal saltmarsh habitat initially expands due to reedbed retreat but ultimately declines by 60% (from 1.13 to 0.45 km2) due to permanent submersion. Consequently, \(ESS_{tot}\) decreases by 51%, from 2.16 \(\cdot 10^7\) to 1.05 \(\cdot 10^7\).

Figure 6d presents the sigma-scores \(\sigma _{ESS}\) for all investigated scenarios. The left column shows \(\sigma _{ESS}\) values based on relative ESS changes (\(ESS_{rel}\)) when comparing SP with future scenarios without NbS implementation (NN). These results highlight substantial reductions in lagoon, beach, and saltmarsh habitat areas due to SLR, leading to a consistent \(\sigma _{ESS}\) decrease of -5 across all hydrodynamic scenarios.

The central and right columns in Fig. 6d display the \(\sigma _{ESS}\) for NbS-BB scenarios, calculated by comparing NN and NbS within the same hydrodynamic context. Across all scenarios, SR consistently outperforms other interventions in terms of \(\sigma _{ESS}\), with scores ranging from 3 to 5. This indicates that SR effectively offsets FRR-ESS losses caused by SLR. In contrast, BN achieves moderate \(\sigma _{ESS}\) scores, ranging from 2 to 3, while DR exhibits the lowest performance, maintaining a score of 1 in most scenarios. However, DR’s effectiveness improves to 3 when combined with BN, demonstrating the added benefits of combining interventions.

The SR+BN combination achieves the highest performance across all scenarios, with \(\sigma _{ESS}\) ranging from 4 in SP to 5 in all future scenarios. This result underscores the synergistic benefits of integrating NbS-BB measures, particularly under extreme climate conditions. These findings emphasize the potential of combining interventions to enhance FRR-ESS provisioning and address the challenges posed by SLR and intensified wave climates.

Fig. 6
figure 6

Habitat map for present state (a), S4.5-2070 SLR scenario (b) and S8.5-2100 scenario (c), corresponding ESS total scores; overview of sigma-scores \(\sigma _{ESS}\) for all investigated scenarios.

Full size image

Discussion

Our study demonstrated the individual and combined efficacy of NbS in mitigating flooding under present and future wave climate and sea level rise. Among individual interventions, beach nourishment (BN) stands out as the most effective measure in reducing the flooded city area. When building block combinations are considered, the synergistic effect of integrating multiple interventions slightly enhances flood reduction. Combinations including BN maintain high levels of effectiveness even under extreme conditions, such as the worst-case scenario (S8.5-2100). SR+BN achieves the best overall performance, reducing the flooded city area up to 78%.

Interestingly, despite BN achieving higher absolute flood reduction compared to SR, the sigma-ESS scores reveal that SR outperforms BN in relative terms. This seemingly counterintuitive result is explained by the differences in habitat expansion. BN expands the beach habitat by 2.2 times, whereas SR extends the seagrass meadow by only 1.55. While BN achieves greater flood reduction, this performance is partly due to the significantly larger increase in habitat area, highlighting the need to account for the relative scale of interventions when assessing their ecosystem service contributions. This nuanced perspective is an important advantage of the sigma-ESS indicator, as it allows for a more balanced evaluation that considers both the effectiveness and proportional habitat impact of each intervention.

Our methodology provides a simple approach to correcting expert-based scores for flood risk reduction ecosystem services (FRR-ESS) using physically-based model results. This approach offers a rapid and customizable framework that not only improves the accuracy of ESS evaluation but also allows for the integration of multiple ecosystem services, such as erosion control or water quality, by incorporating additional model outputs. The approach was tested on a coastal case study in the Mediterranean, but with tailored modifications to the formulation or the modelling framework, it can be modified to account for different environments, e.g. cohesive sediments78, macrotidal environments79, gravel beaches80.

While this study provides a robust framework for linking hydro-morphodynamic modelling with habitat transitions to assess flood risk reduction ecosystem services, certain methodological simplifications were necessary. Habitat shifts were estimated using a suitability-based approach rather than process-based modelling, as detailed predictions of salinity dynamics, nutrient influx, and long-term morphological evolution are difficult to include in the model without deep uncertainties. The employed methodology relies instead on simple rules based on elevation and connection with the sea to evaluate habitat shifts due to sea level rise, providing a practical yet simplified approach to account for environmental changes. These approximations, while reasonable given the study scale, data availability and aim of the work, should be considered when interpreting the results.

Another assumption to acknowledge is that our approach assesses the response of Nature-based Solutions under the assumption of full functionality, meaning that each intervention is analyzed as if it has reached its intended effectiveness before the storm occurs. This was necessary to isolate and quantify the maximum potential flood risk reduction benefits of NbS without the additional complexity of modelling transitional dynamics. However, we recognize that this approach does not capture important factors such as the slow growth rates and uncertain success of Posidonia oceanica transplantation, which requires specific environmental conditions and can take decades to fully establish. Similarly, while sand nourishment provides immediate protective benefits upon placement, its longevity is influenced by dynamic sediment redistribution processes, which require careful monitoring and periodic re-nourishment to maintain effectiveness. Given these recurrent maintenance needs, a robust cost-benefit analysis may be warranted in some contexts to ensure that such interventions remain both economically viable and effective over time. Generally, the expenses required to sustain beach ecosystems in optimal condition typically represent only a small proportion compared to the tourism-derived revenue and broader socioeconomic benefits they generate81. In contrast, dune revegetation using Ammophila arenaria presents a more realistic assumption of full functionality due to the species’ rapid growth, resilience to sand burial, and proven capacity to stabilize and enhance existing dunes over relatively short timescales. To overcome these limitations, future studies could incorporate transitional scenarios, modelling the progressive evolution of NbS over time. For seagrass restoration, this could involve simulating gradual patch expansion based on survival rates or habitat suitability modelling82 under varying environmental conditions. For sand nourishment, models better suited for accretive processes than XBeach83,84,85, could offer more accurate predictions of long-term sediment redistribution and beach profile changes.

Our approach relies on single-point extractions from extreme value analyses; however, we acknowledge that this simplification does not fully capture the range of uncertainties associated with wave statistics, storm surge interactions, and future morphological changes. One source of uncertainty arises from the Med-MFC reanalysis dataset69 used as input for SWAN, which, despite slight underestimations at low and intermediate wave heights, shows good agreement for higher waves, particularly during winter months when variability tends to be lower. For our study area in the Ionian region, the dataset reports a minimal bias of -0.018 m and a root mean square error (RMSE) of 0.200 m, suggesting limited impact on the morphodynamic outcomes for the wave heights considered (5.61-8.36 m). Additional uncertainty stems from sea-level rise projections, where likely ranges are influenced by both emission scenario variability and uncertainties in physical sea-level drivers86. In our analysis, we focused on the 50th percentile projections, acknowledging that higher uncertainties arise over longer time horizons, potentially influencing flood risk reduction outcomes. Although we did not explicitly assess uncertainty propagation within our modelling framework, the significant differences observed between NbS and no-NbS scenarios across all projections suggest that our qualitative conclusions regarding solution effectiveness remain robust. Future work will aim to incorporate a more comprehensive uncertainty analysis to enhance the robustness of our findings and better reflect the complexity of coastal processes.

Overall, this methodology represents a step forward in ecosystem service evaluation by bridging physically-based modelling with expert-based assessments, enabling a more accurate, scalable, and context-specific analysis. The ability to incorporate additional ecosystem services and adapt the model inputs to different scenarios positions this approach as a valuable tool for designing and optimizing NbS interventions that balance flood risk reduction with ecological sustainability.

Conclusions

This study presented a framework for quantifying flood risk reduction ecosystem services provided by coastal Nature-based Solutions, combining expert-based assessments with physically-based numerical modelling. Applied to a Mediterranean coastal lagoon in Sicily, the methodology demonstrates the potential of NbS, both individually and in combination, to mitigate storm-induced flooding under present and future climate scenarios. The results highlight the enhanced performance of combined NbS measures through a Building Block approach, emphasizing the importance of integrating diverse strategies for robust flood risk reduction.

By incorporating habitat maps and eco-hydro-morphodynamic modelling, this framework provides a more accurate and adaptable tool for evaluating the effectiveness of NbS across different environmental contexts. The findings reinforce the critical role of coastal habitats in disaster risk reduction and offer a scalable approach for policymakers and stakeholders to align flood mitigation efforts with ecological sustainability, fostering resilience in the face of climate change.

Future developments of this framework could integrate more advanced habitat modelling techniques87,88 to refine predictions of ecosystem transitions under changing environmental conditions. Additionally, increasing spatial resolution and incorporating more detailed ecological and physical processes could further enhance the applicability of the approach across diverse coastal settings. Expanding its use to other geographic contexts and coupling it with adaptive management strategies would provide actionable insights for policymakers and practitioners seeking to optimize Nature-based Solutions for coastal resilience.