The basin of the Adriatic Sea is part of the Mediterranean Sea and is connected to it through the Strait of Otranto. Because the sea state conditions from the Mediterranean Sea influence the waves near the connection area between the two seas, a two-level system for the wave simulations in the Adriatic Sea is implemented. The entire basin of the Mediterranean Sea represents the first level of the wave modelling system, and it is used only to provide boundary conditions for the second level, which is focused on the Adriatic Sea (see Figure 1). The wave model considered is SWAN model (Simulating WAves Nearshore, version 41.31AB [42]), a state-of-the-art wave model based on the spectrum concept, which has been implemented with good results both in the Mediterranean Sea [43], [44] and in other basins comparable to the Adriatic Sea [45]. The computational domain covering the Mediterranean Sea has a spatial resolution of 0.2° in both directions, while the second domain has a resolution of 0.1°.

Figure 1.

The computational domains of the SWAN simulations

In all simulations, for the spectral space, 36 directions and 30 frequencies (from 0.04 Hz to 1 Hz with logarithmic increment) were used. The SWAN model was run in the non-stationary mode, with a 10 min for the time step and the iteration number was set to 4 (default is 1). A new source term formulation implemented recently in SWAN (denoted as ST6 [46]), together with the new option for the non-breaking dissipation of Ardhuin et al. [47] were used. This option was implemented with good results in semi-enclosed basins [48] and at the ocean scale [49].

For the historical reference period 1976-2005, hindcast simulations were first carried out, when ERA5 wind fields with a spatial resolution of 0.25° and temporal resolution of 3 hours (the last reanalysis data provided by the European Centre for Medium-range Weather Forecast - ECMWF) were used to force of the wave modelling system. These results delivered by the wave modelling system in the Mediterranean Sea were validated by performing a comparison of the simulated significant wave heights (Hs) with altimeter measurements and the statistical parameter values confirmed a very good agreement between the measured and simulated Hs values. The second set of simulations for the reference period 1976-2005 was made by forcing the wave modelling system with historical high-resolution wind fields (0.11°) simulated at SMHI (Swedish Meteorological and Hydrologic Institute) with the Rossby Centre regional atmospheric model (version RCA4 [50]) and provided for public access through EURO-CORDEX database [51].

For the near future period (from 2041 to 2070), the wind fields considered to drive the wave modelling system are those simulated by the same RCM, namely RCA4, under RCP4.5 emission scenario. All wind fields simulated by RCA4 and used in this study have a temporal resolution of 6 hours. Like all the wind fields provided by EURO-CORDEX, they also cover entire Europe, being used in various studies, e.g. [52].

For the evaluation of the wave climate and wave power corresponding to the recent past (historical period) and near future in the Adriatic Sea, the results provided as output at the second level of the SWAN model are used. Thus, at each point of the computational domain, 3-h resolution time series of the various wave parameters are generated and further, based on this information the wave climate and wave power assessment are designed.

In this study, the attention is directed especially to the analysis of the main wave parameter, namely Hs, and also to the wave power (or wave energy transport) Etr. SWAN provides as output the wave energy transport components (i.e. energy flux per unit of the wave crest length in kW/m). The relationships implemented in SWAN to compute the wave transport components are:

$\begin{array}{c}Et{r}_{\text{x}}=\rho g\iint c_{\text{x}}E\left(\sigma ,\theta \right)d\rho d\theta \\ Et{r}_{\text{y}}=\rho g\iint c_{\text{y}}E\left(\sigma ,\theta \right)d\rho d\theta \end{array},$ (1)

where E(σ,θ) is the directional wave energy density spectrum, x and y are the spherical grid coordinate system (longitude and latitude), and c_x, c_y are the propagation velocities of wave energy in the geographical space (absolute group velocity components).

The absolute value of the wave power in each point of the computational grid is calculated with the following relationship:

$Etr=\sqrt{Et{r}_{\text{x}}^2+Et{r}_{\text{y}}^2}$ (2)

The SWAN hindcast driven by the ERA5 wind reanalysis dataset is used as a reference. Then, they are compared with the SWAN historical results when the wave model is forced by the RCA4 historical wind dataset. This comparison assesses the capability of the SWAN historical data to reproduce the recent past wave climate and corresponding wave power fields. The results are presented as normalized differences (ND): the difference between the historical data (CD - when the RCA4 climate dataset is used) and the hindcast data (ER - when ERA5 reanalysis dataset is used) means, divided by the hindcast means (ER).

$ND=\frac{\overline{CD}-\overline{ER}}{\overline{ER}}\times 100$ (3)

In order to analyse the dynamics induced by climate changes on the significant wave height and wave power means along the 21^st century under RCP4.5 scenario, the relative change (RC) index was computed as the difference between the near future (2041-2070) and recent past (1976-2005) means, divided by recent past means, i.e.,

$RC=\frac{\overline{NF}-\overline{RP}}{\overline{RP}}\times 100$ (4)

where $\overline{RP}$ represents the mean wave power or mean Hs over the recent past period and $\overline{NF}$ is the mean wave power or mean Hs over the near future period. RC values are given as a percentage relative to the recent past period.

For the evaluation of the wind climate, it is well known that the wind fields distributed by the databases, which are used in forcing the wave models, are simulated at a typical height of 10 m above sea level. However, for the evaluation of the wind power potential in a certain area, it is necessary to use the wind speed at a height corresponding to that at which the wind turbines are installed. Nowadays, the tendency is to develop larger wind turbines installed at heights that reach 100 m and for this reason, this height will be considered as a reference for the analysis made in the present study regarding the wind climate and wind power. Thus, in each grid point, the logarithmic law is applied following the next relationship established by [53], [54]:

$U_{100}=U_{10}\frac{\ln z_{100}/z_0}{\ln z_{10}/z_0}$ (5)

where U₁₀₀ represents the wind speed at 100 m height, U₁₀ is the wind speed at 10 m height, z₁₀₀ is the operational height (100 m) and z₁₀ is the reference height of 10 m, respectively. The surface roughness length is denoted with z₀, and an average value of 0.0002 m is considered in the present work.

Another important parameter evaluated in this study is the wind power density P_w (W/m²) per unit of the swept area, computed based on the information regarding the wind speed at the installed height of the wind turbine (in this case 100 m) using the relationship:

$P_{\text{w}}=\frac{1}{2}{\rho }_{\text{air}}{U}_{100}^3,$ (6)

with ρ_air the air density has the value of 1.225 kg/m³.

As in the case of the waves, to assess the capability of the RCA4 model to reproduce the recent past wind fields and corresponding wind power, the normalized differences (ND) were calculated. Then, in order to assess the dynamics induced by the climate changes on the wind speed (U₁₀₀) and on the wind power density means along the 21^st century under the RCP4.5 scenario, the relative change (RC) index was also computed.

The mean significant wave height fields (Hs) resulting in the Adriatic Sea, based on the hindcast simulations, are illustrated in Figure 2a, while those for the historical simulations are presented in Figure 2b. As above mentioned, for the hindcast simulations, the ERA5 wind fields were used to force de SWAN model, while for the historical simulations, the RCA4 wind fields were considered. A visual evaluation of the two maps shows a similar pattern in both fields, with the highest values in the southern part of the basin, reaching maximum values of 0.8 m (located very close to the southern border of the domain where the connection is made with the Mediterranean Sea) and 0.87 m, respectively. In general, the enclosed and semi-enclosed seas are characterized by a limited fetch for wave development and these features prevent waves from travelling for long distances and generating swell. For this reason, in such basins, sea waves with lower wave periods characterise the wave climate. However, under the Bora wind, it is possible, and relatively frequent in the Adriatic Sea, to be generated waves with Hs higher than 10 m, as mentioned by Cavaleri et al. [55]. The presence of higher Hs means in the southern part of the Adriatic Sea is explained by the influence of the waves penetrating from the Mediterranean Sea.

The normalized differences of the two Hs fields, computed as indicated in eq. (3), are illustrated in Figure 2c. Over the entire basin, the historical Hs means present higher values, with extended areas showing a general rise of about 20%. There are a few small areas with excessive increases (reaching 40%) but they are located near the coast, and this is influenced by the different resolutions of the wind fields used to drive the wave model. The ERA5 wind fields, having a lower resolution, cannot capture accurately the wind speed changes (in magnitude and direction) in the coastal areas, this being also influenced by the complicated coastal orography that surrounds the Adriatic Sea [56]. In the coastal zone of Croatia, where many small islands are present, small decreases (up to 10%) of the Hs means are noticed. This can be caused by the fact that the bathymetry used is not fine enough to take into account all the small islands and then the effect induced by their presence is not represented by the model results. A new level of the wave modelling system should be added, using a very high-resolution bathymetry, if the interest is focused on that coastal zone.

Figure 2.

Average values of Hs for the recent past 1976-2005: a) SWAN forced with ERA5 wind fields; b) SWAN forced with RCA4 wind fields; c) normalized differences (ND) between the two Hs fields

The seasonal characteristics of the average Hs values are evaluated and presented in Figure 3, and the corresponding normalized differences are illustrated in Figure 4. From top to bottom, the panel rows from Figure 3 show the maps of the mean Hs spatial distribution in the Adriatic Sea for DJF, MAM, JJA and SON. On the left side of Figure 3, the results corresponding to the hindcast simulations are presented, while on the right side those for the historical simulations.

In all seasons the maximum values of the Hs means are higher in the case of the historical simulations, as can be seen in Figure 3. For each type of result (hindcast or historical), the geographical positions of the maximums do not change significantly compared to the corresponding ones illustrated for the average fields computed over the entire period. The seasonality of the wave climate in the Adriatic Sea is clear, showing in the transition seasons (spring - MAM and autumn - SON) similar average values for Hs (with maxima located in the range of 0.8-0.93 m), while the values in the winter are almost double compared to those simulated during the summer season (1.04 m and 1.14 m, compared with 0.54 m and 0.6 m respectively).

Regarding the normalized differences shown in Figure 3, they have higher values (both positive and negative) in the warmer season (summer) and decrease as the analysis of ND moves towards the colder seasons. In the northern part of the basin, the results of the historical simulations indicate increases in the Hs averages of approximately 40% compared to the results of the hindcast simulations, while in the summer similar increases are indicated along the eastern coast of the basin. In very small areas negative ND values are found, without exceeding decreases of about -14% encountered in summer.

Figure 3.

Seasonal average values of the significant wave height for the recent past (1976-2005): results from SWAN forced with ERA5 wind fields – left panels; results from SWAN forced with RCA4 wind fields – right panels

Figure 4.

Seasonal normalized differences (ND) between the average values of Hs for the recent past 1976-2005: a) DJF, b) MAM, c) JJA and d) SON

The next step is to evaluate the changes of the mean Hs values for the near future computed as a percentage relative to the historical period, denoted also in eq. (4) as the recent past (RP).

The relative change values in terms of mean Hs computed for the entire period are presented in Figure 5, while those for each season are illustrated in Figure 6. In the near future, the decreases in Hs means are envisaged in the northeastern part of the Adriatic Sea, the decrease being more pronounced in the northern part (about 4.8%). This pattern was identified also in a previous study [57] under the IPCC A1B emission scenario. In the southeastern side, the decrease is ranging from 0.5 to 2%, while in the centre part of the basin, very small changes are noticed (increases of about 0.5% in the west-central zone and decreases of about 0.5% in the east-central area).

Figure 5.

Changes of the mean Hs as a percentage relative to the recent past period for the near future corresponding to the RCP4.5 scenario

The pattern of the RC values shows some changes from one season to another (see Figure 6), but in the northern area of the basin, the average Hs values decrease always in the near future. These decreases are more accentuated and cover larger areas (north and southeast of the basin) during spring when they reach 6.5% in the north, followed by winter when negative RC values are noticed in the all-northern half of the basin (the highest decrease being around 5.6% in the north of the Adriatic Sea). An opposite RC’s trend is expected during the summer. Thus, in the southwest of the basin positive values of RC are found (reaching 5.8%) indicating an increase of the Hs average values in the near future, while in the northeastern side of the basin, RC negative values up to 18% are encountered, which indicates decreases of the Hs average values.

Figure 6.

Seasonal changes of the mean Hs as a percentage relative to the recent past period for the near future corresponding to the RCP4.5 scenario: a) DJF, b) MAM, c) JJA and d) SON

The mean wind climate over the Adriatic Sea at 100 m above the sea level, obtained by applying the relationship from eq. (5) to ERA5 and RCA4 datasets for the recent past period, is illustrated in Figure 7a and Figure 7b, respectively. The maximum values of both mean wind fields are 7.5 m/s for ERA5, and 7.4 m/s for RCA4. According to these results, the wind climate has a similar pattern for both datasets. Higher values of RCA4 data are noticed in the northwestern and southeastern parts of the basin, as indicated by the normalized differences (ND) between the two fields illustrated in Figure 7c.

Figure 7.

Average values of U₁₀₀ for the recent past 1976-2005: a) ERA5 wind fields; b) RCA4 wind fields; c) normalized differences (ND) between the two fields

The projections of the mean wind speed U₁₀₀ for the near future period are given in Figure 8a, while the changes in the mean U₁₀₀ values for the near future, computed as a percentage relative to the historical period, are illustrated in Figure 8b. For the near future, the expected RC values have small variations over the Adriatic Sea, ranging from -3.8% to 2.2%, with positive values in the southern part of the basin.

Figure 8.

a) Average values of U₁₀₀ for the near future 2041-2070; b) Changes of the mean U₁₀₀ as a percentage relative (RC) to the recent past period for the near future corresponding to the RCP4.5 scenario

The wave power (Etr) results obtained as an output of the SWAN model driven by ERA5 (see Figure 9a) and RCA4 wind fields (not shown here) for the recent past period have maximum values of 2.37 and 2.8 kW/m. These fields are compared using the normalized differences and the results are illustrated in Figure 9b. As expected, the historical wave power presents higher values than those simulated using ERA5 data. The normalized differences can reach values of almost 50% over extended areas.

Figure 9.

Average values of Etr for the recent past 1976-2005 based on SWAN results forced with ERA5 wind fields; b) normalized differences (ND) between the recent past data

From the average values of the wave energy calculated for each season, it can be seen that there is a drastic reduction in it during the summer (see Figure 10), compared with the other seasons. The wave power from the Adriatic Sea does not have high values (especially in the northern part of the basin) compared with wave energy resources encountered in the ocean coastal areas, but through joint exploitation with other renewable sources available in the marine environment, it can represent however a source of reducing the variability of the renewable energy for the integration to the power system [58].

Some projections regarding the dynamics of wave energy in the near future under the RCP4.5 emission scenario are presented in Figure 11, and the results indicate a tendency of reduction of the wave energy in the northwest part by up to 8%. In the south-eastern part of the sea, an increase in terms of wave power of up to 10% is estimated over extended areas.

Figure 10.

Seasonal average values of wave power for the recent past (1976-2005): results from SWAN forced with ERA5 wind fields – left panels; normalized differences (ND) between the recent data – right panels

Figure 11.

a) Average values of Etr for historical data, results from SWAN forced with RCA4 wind fields; b) Changes of the mean Etr as a percentage relative to the historical period for the near future corresponding to the RCP4.5 scenario

The seasonal analysis of the wave power dynamics in the near future is presented in Figure 12. Autumn appears to be the period when, in most of the basin, wave power increases of up to 8% are expected. In the other seasons, the areas where wave power growths occur are moving across the basin from season to season, but the increases are not found in the northern area.

The Adriatic Sea has many coastal areas where shallow water depths are found, which would facilitate the deployment of the offshore wind farm projects. For this reason, the assessment of these resources and their dynamics in the near future is important. Next, an analysis of the offshore wind power density (P_w) will be carried out to evaluate its potential. Thus, the average values of the wind power density are calculated based on the wind speed at 100 m and their spatial distribution over the Adriatic Sea is illustrated in Figure 13.

The maximum values calculated for P_w based on the two sources of wind speed (ERA5 and RCA4) for the recent past indicate values close to 500 W/m² over the southern part of the basin. The normalized differences (ND) between the results obtained using the two sources are small, showing a maximum of around 4% higher values for RCA4 in the southern part of the basin and around 6.5% lower values for RCA4 in the northern area of the basin.

Figure 12.

Seasonal average values of Etr for historical (1976-2005), results from SWAN forced with RCA4 wind fields – left panels; b) Seasonal changes in the mean Etr as a percentage relative to the historical period, for the near future under RCP4.5 – right panel

Figure 13.

Average values of P_w for the recent past 1976-2005: a) P_w ERA5 fields; b) RCA4 P_w fields; c) normalized differences (ND) between the two P_w fields

The wind power density for the near future, computed based on U₁₀₀ values, and its relative changes in comparison with the results from the RCA4 for the recent past are given in Figure 14. The images indicate an important increase of the wind power density almost all over the basin with RC values reaching 40% over extended zones.

Figure 14.

a) Average values of P_w for the near future 2041-2070; b) Changes of the mean P_w as a percentage relative to the recent past period for the near future corresponding to the RCP4.5 scenario