WAVE ENERGY RESOURCES ALONG CALABRIAN COASTS ( ITALY )

The assessment of wave energy is fundamental to well evaluate potential wave energy at di↵erent sea locations and time scales in conjunction with the related occurrence of hot spots for an optimal installation of Wave Energy Converters (WECs). The present study has been performed o↵ the coasts of Calabria (Southern Italy), a Mediterranean region characterized by a mild wave climate and quite representative of mean sea states in the Mediterranean basin. The wave energy potential has been assessed in deep waters by means of ECMWF operational wave data validated against RON buoys and UKMO data. The wave power is calculated as a function of the energy wave period deduced from directional wave spectra and compared with widely adopted relationships based on the use of a standard JONSWAP spectrum. The mean yearly and seasonal wave energy is then assessed at selected hot spots for Tyrrhenian and Ionian Seas at a water deep of 100 m suitable for the installation of several o↵shore WECs.


INTRODUCTION
In the field of renewable energies, wind-generated waves represent a recent and interesting resource of energy option.The availability and persistency of wave data from di↵erent sources allows for an accurate assessment of their energy with respect to other renewable resources such as wind and sun whose production proves to be intermittent.With reference to the Europe, the greatest wave energy contribution refers to the countries located along the Atlantic coasts with respect to those referred to the Mediterranean Sea (e.g., Rusu and Onea, 2016).Nevertheless, the Mediterranean basin shows interesting energy potentialities for o↵shore and nearshore WEC installations (e.g., Liberti et al., 2013;Besio et al., 2016).For di↵erent areas in the Adriatic and Thyrrenian Sea o↵ the italian coasts, several studies have been carried out to assess the wave energy potential for electricity production (Vannucchi and Cappietti, 2016).Highest wave energy contribution in Italy have been individuated in the North-West of Sardina (Vicinanza et al., 2013) and in the West coast of Sicily (Monteforte et al., 2015) with a yearly mean wave power of about 10 kW/m and 4.5 kW/m, respectively.Moreover, the development of di↵erent kind of devices for wave energy has led to recent field installations as in the case of nearshore WECs like REWEC3 (Boccotti, 2007), DIMEMO (Contestabile et al., 2016) and SYNCRES (Sammarco et al., 2013) respectively installed at the port of Civitavecchia near Rome, Naples and Piombino in Southern Tuscany.The above coastal structures are also configured in the family of perforated-wall caisson breakwaters (e.g., Aristodemo et al., 2015;Meringolo et al., 2015) which allow for a defense of port areas with an additional benefit in terms of harbor tranquillity.In the case of o↵shore WECs, a farm of ISWEC devices has been installed o↵ Pantelleria Island close to Sicily (Bracco et al., 2015).
Owing to the lack of specific studies, the present analysis is performed o↵ the coast of Calabria region (Southern Italy) which presents a relevant coastline length, i.e. about 700 km, and a variable wave climate also due to the di↵erent exposition of the coasts to Tyrrhenian and Ionian Seas.The assessment of wave energy along the Calabrian coasts is carried out using ECMWF operational wave data which have been validated against RON buoys and UKMO data by means of geographical transposition methods (e.g., Contini and De Girolamo, 1998) because of their di↵erent spatial position.The mean yearly and seasonal wave power is then assessed as a function of the energy wave period through the calibration of the spectral parameters deduced from observed directional wave spectra and compared with largely adopted relationships based on the use of the standard JONSWAP spectrum (e.g., Gonçalves et al., 2014).On the basis of the resulting spatial distribution of the wave power evaluated at a water depth of 100 m where a large amount of WECs can be installed (e.g., Babarit et al., 2012), the mean yearly and seasonal wave energy is then evaluated at selected hot spots for Tyrrhenian and Ionian Seas of the involved Italian region.
The present work is organized as follows.Firstly, ECMWF wave data are processed and then subjected to a validation using other wave sources.Afterwards, the wave power formula has been assessed paying attention to the estimation of the energy period.The evaluation of mean yearly and seasonal wave power allowing for successive analysis of wave energy in selected hot spots is finally performed.Input wave data have been obtained through the European Center for Medium-Range Weather Forecasts (ECMWF) atmospheric operational model from 1992 to 2015.The considered synthetic parameters of sea states refer to significant wave height, H s , mean wave period, T m , peak period, T p , mean wave period from 2 nd moment, T m2 , and mean wave direction, ✓.Even if the global atmospheric reanalysis ERA-INTERIM by ECMWF covers a larger time window  and the data were well validated, the operational model has been here preferred for the higher spatial resolution of the nodes (0.125 x 0.125 ) since it operates at the Mediterranean scale.Wave measurements furnished by directional buoys at Cetraro (1999Cetraro ( -2008) ) and Crotone (1989Crotone ( -2007) ) of the Italian Wave measuring Network (RON) and some nodes of UK Met O ce (UKMO) dataset  have been also taken into account for the validation of ECMWF data.A plan view along the considered Calabrian coasts of the adopted ECMWF, RON and UKMO wave dataset is shown in Fig. 1.All wave data refer to the deep wave condition d/L m > 0.5, where d is the depth and L m is the mean wavelength.In particular, the considered ECMWF nodes are located in a depth ranging from 80 m to 700 m.It is worth noting that the wave parameters using ECMWF were measured every 6 h as well as UKMO, while RON buoys sampled with a time window of 3 h and of 30 minutes when sea storms occur.As a result, a general underestimation of wave peaks of ECMWF data with respect RON buoys was observed (e.g., Vicinanza et al., 2013) and this e↵ect tends to influence the analysis of extreme events.However, ECMWF data can provide a good and conservative assessment of mean seasonal and yearly quantities for the wave energy analysis, as adopted in several seas of the world (e.g., Reikard et al., 2011).Moreover, their better spatial resolution of ECMWF nodes with respect to the other considered wave data (RON and UKMO) along the involved italian region allows therefore for a more accurate evaluation of the wave energy potential.
Input wave data deduced from ECMWF, RON and UKMO datasets have been subjected to a processing in order to check their quality for successive analyses of wave energy.The following criterion has been adopted to eliminate inconsistent i-th wave parameters: • presence of NaN, zero and repeated values; • outlier events not compatible with their spatial position; • T p,i /T m,i > 2; • wave steepness in deep waters up to its breaking limit.
As a consequence, the mean e ciency, given by the ratio between filtered data and raw ones, is 89.5 % for ECMWF data, while for RON and UKMO data is respectively equal to 82.1 % and 94.7 %.

Calibration
The reliability of adopted ECMWF wave data have been assessed against RON buoys and UKMO nodes.By using the ECMWF nodes no.E4 and no.E35 for the Thyrrenian and Ionian Sea, the comparisons have been carried out considering the closest UKMO and buoy locations to them (see red and blue symbols in Fig. 1).As successively performed, the choice of the selected two ECWMF nodes is due to the highest wave power appearing along the Thyrrenian and Ionian Seas of Calabria region.Owing to the di↵erent spatial position, UKMO and RON have been subjected to geographical transposition methods, largely adopted for applications studies along the italian coasts, at the location of two selected ECWMF nodes.The empirical approach of Contini and De Girolamo (1998) allows to virtually transfer the values of H s and T p by means of the e↵ective fetches, F e .Under the hypothesis that the wind velocities are the same in the original and final locations, this geographical transposition method is derived by the wave forecasting given by the classical SPM approach for fetch limited conditions.The new quantities H si and T pi at the final points have been then calculated as: where the subscripts i and o indicate the final and initial points.The evaluation of other wave periods T mi and T m2i has been performed as a function of T pi and the statistical relationships between the wave periods at the original node as follows: Under the above hypothesis on wind velocities, the wave directions at the new sea points have been evaluated by calculating the angle, , between the mean wind direction and the modeled mean wave one through a geographical approach (e.g., Lo Re et al., 2016): where F i is the geographical fetch, N is the number of directions taken every 5 and i is the wind direction, while the neighbouring directions, w , range from i -✓ and i +✓.
On the basis of the exposition of the coasts of Calabria to the open sea, the highest agreement between the wave data (ECMWF vs. RON and UKMO) in the considered sea locations has been obtained through the evaluation of e↵ective fetches using the canonical Saville's formula by setting the angular sector across the mean wave direction, ✓ = 45 , and the directional spreading parameter, n = 2 in Eqs. 1, 2 and 3.
As applied by Liberti et al. (2013), several performance indices have been determined at the selected ECMWF points to check the suitability of ECMWF wave data.Specifically, the bias, the root mean square error, rmse, the slope of the regression line through the origin, the scatter index, si, and the Wilmott index of agreement, d.With reference to H s (i.e. the most weighting wave parameter in evaluating the wave power), Figs. 2 and 3 illustrate the performance indices of ECMWF data compared with RON and UKMO ones at the selected ECMWF nodes of Tyrrhenian and Ionian Sea, respectively.The indices in Figs. 2 and  3 show an overall agreement between ECMWF and satellite H s .The values of bias and rmse are enough small as well as d is greater than 0.9 for all cases.The slope is less than 1, indicating an underestimation of ECMWF dataset in modeling highest sea states as previously observed about the time sampling of wave data.It can be noticed an higher agreement between ECMWF and RON data manly due the closest spatial position of these points with respect to ECMWF and UKMO ones (see Fig. 1).

CHARACTERIZATION OF WAVE POWER
Since the involved ECMWF nodes are located in deep water conditions, the largely used expression in literature to estimate the wave power per unit of wave front length, P, is adopted: where ⇢ is the water density and g is the gravity acceleration.The energy wave period, T e , representing the variance-weighted mean period of the one-dimensional period density spectrum.It is evaluated as the ratio m 1 /m 0 , in which m n is the n-th spectral moment that is defined as: where f i is the i-th frequency, S i j is the density over the i-th frequency and the i-th direction, f i is the i-th frequency width of the density and ✓ j is the angular width of the density.
A widespread approach adopted in literature (e.g., Gonçalves et al., 2014;Monteforte et al., 2015) is to assess the value of T e in Eq. 4 by considering relationships deduced from the use of a mean JONSWAP unimodal spectrum with peak-enhancement factor = 3.3, energy scale parameter, ↵ = 8.1*10 3 , and spectral width parameters a = 0.07 (for f < f p ), and b = 0.09 (for f > f p ), where f p is the peak frequency.The expression of JONSWAP density spectrum reads as: where S PM represents the Pierson-Moskowitz (PM) spectral density given by: The related statistical relationships for the practical calculation of the wave energy period as a function of a JONSWAP spectrum with canonical shape parameters lead to: T e = 0.904T p or T e = 1.143T m2 , where T m2 = (m 0 /m 2 ) 0.5 .
For two ECMWF nodes o↵ Thyrrenian and Ionian coasts where successive energy analyses will be performed, Figs. 4 and 5 show the relative di↵erences in evaluating the mean monthly and yearly wave power, P m , as a function of T e = m 1 /m 0 against those referred to the use of T e as a function of T p and T m2 when a standard JONSWAP spectrum is adopted.A substantial underestimation of about 10 % can be observed when T m2 is adopted, while a general overestimation of about 5 % can be highlighted when T p is used.A quite constant seasonality variation is noticed, except for a slight di↵erent trend during the summer when these di↵erences increase due to a lowest wave climate in the year.The above di↵erences in evaluating P m are consistent with the milder wave climate of the involved Mediterranean location with respect to the severe sea states appearing in the Northern Sea where the JON-SWAP spectrum was calibrated.As a consequence, the use of approximate relationships can lead to a certain di↵erence when real sea states are taken into account.In order to deduce more correct statistical relationships between the involved wave parameters, directional wave spectra discretized by 24 f and 30 ✓ from an available ECMWF dataset (2001-2015) has been considered.The analysis has been limited for wind-driven seas by modeling the shape parameters of the JONSWAP spectral form.The procedure to fit the shape parameters of JONSWAP spectrum has consisted in the initial evaluation of the energy scale parameter, ↵.On basis of the spline interpolation to better discretize the measured frequency spectrum, ↵ has been calibrated by means of a non-linear least square fitting method in order to minimizing the difference between measured and simulated wave spectrum in the range f p < f < 4 f p .Afterwards, the peak enhancement factor has been determined as the ratio S ( f )/S PM ( f ) for f = f p .The peak width factors a and b have been respectively evaluated in the frequency band 0.9 f p < f < 1.1 f p , as performed by Piscopia et al. (2002).The monthly and yearly changes of JONSWAP shape parameters, , a and b , as well as the period ratios, ↵ 1 = T e /T p and ↵ 2 = T e /T m2 , are respectively illustrated in Figs. 6 and 7 for the hot spots at the Thyrrenian and Ionian Sea.As expected, the mean yearly shape parameters at the considered two nodes are lower that those set in the canonical JONSWAP spectrum.Higher values of , a and b occur at the ECMWF node for Tyrrhenian coast as well as during the winter months.The mean spectral parameters are quite close to those determined by Piscopia et al. (2002) by analyzing the wave spectra recorded by RON buoys (1989RON buoys ( -2001) ) of Crotone and Ponza, whose wave climate is quite similar to the Thyrrenian coast of Calabria.For what concerns the period ratios, improved relationships between T e against T and T m2 have been found: ↵ 1 = 0.88 and ↵ 2 = 1.17 for the node of the Thyrrenian Sea, and ↵ 1 = 0.89 and ↵ 2 = 1.16 for that related to the Thyrrenian Sea.By inspecting the recorded directional wave spectra by ECMWF, it has been moreover observed a certain percentage of bimodal sea states with low wave energy (up to 30 %) due to the occurrence of swell components which are usually linked to sea states produced by wind rotation along the generating area.

ANALYSIS OF WAVE POWER AND WAVE ENERGY
Starting from the ECMWF nodes, the geographical transposition methods expressed by Eqs. 1, 2 and 3 have been applied to transpose the initial values of H s , T m , T p , T m2 and ✓ to a water depth of 100 m.Each point has been transposed to the closest reference depth for a distance of order of few km.The selected water depth has been chosen as a suitable location to further studies on the performances of WECs like point absorbers, linear absorbers and terminators.For more details for more of them, the reader can refer to the comprehensive review of Babarit et al. (2012).The wave power has been then assessed at a mean yearly and seasonal scale by applying Eq. 4. Adopting Cartesian coordinates, Fig. 8 describes the spatial distribution along Calabrian coasts of the linearly interpolated values of mean yearly and seasonal wave power, P m , at a water depth of 100 m.The seasonal distribution has been organized intro groups of months as follows: Winter (December, January and February), Spring (March, April and May), Summer (June, July and August) and Autumn (September, October and November).The most relevant wave power potential appears generally along the Northern and Central part o↵ Thyrrenian coast of Calabria.In this area, the related mean annual values of P m range between 2.2 kW/m and 2.7 kW/m, while they oscillate between 0.4 kW/m and 2.5 kW/m for the Ionian coast.Lowest P m are noticeable near the Strait of Messina and along the Northern part of Ionian coast with a mean annual P m less than 1 kW/m.By a seasonal point of view, remarkable temporal variability of wave power can be observed in the considered four time windows.For winter, spring, summer and autumn, the maximum values of P m are respectively 4.7 kW/m, 2.5 kW/m, 0.8 kW/m and 2.3 kW/m.This inter-annual fluctuation was also highlighted by Liberti et al. (2013) about the analysis of Coe cient of Variation in the Italian Seas, showing a relevant variation in the southern zone of Tyrrhenian Sea.A hot spot area for Tyrrhenian coast has been individuated along a coastline length of about 50 km between the towns of Amantea and Cetraro, as highlighted in Fig. 8.The highest mean yearly wave power is 2.69 kW/m and it occur in front of the town of Fiumefreddo Bruzio.A more restricted zone (about 5 km) has been observed for the hot spot for Ionian coast, reaching a peak in the yearly wave power equal to 2.51 kW/m.The hot spot is located in front of the town of Isola Capo Rizzuto near Crotone.The obtained yearly and seasonal values of P m are quite in agreement with the spatial variations calculated at a Mediterranean scale by adopting di↵erent wave datasets (Liberti et al., 2013;Besio et al., 2016).
Paying attention to the hot spot locations of Fiumefreddo Bruzio for the Thyrrenian Sea and Isola Capo Rizzuto for the Ionian Sea, Figs. 9 and 10 show respectively the polar diagrams of mean yearly and seasonal wave power climate, P m , as a function of selected classes of H s and ✓.It is evident a di↵erent directional distribution of P m for the above two selected sites.For the Thyrrenian site, the main contribution of directional wave power is substantially restricted to the angular sector 250 -290 related to the Western direction.In case of Ionian site, significant values of P m are related to two di↵erent directional sectors associated to the longest fetches: 10 -40 and 170 -200 .The seasonal polar diagrams of P m reveal similar features at the Thyrrenian hot spot, except during the summer when a high percentage of sea states come from North-West.For the Ionian hot spot, highest seasonal values of P m are substantially associated to the above two main directional sector observed for the yearly assessment It can be generally observed that the large spreading of mean wave direction of sea states at the Ionian hot spot could lead to uncertainties in order to assess the best planimetric disposition of WEC fronts in terms of real electricity production.
For the selected two hot spots, the assessment of mean yearly and seasonal wave energy, E, has been represented in the scatter diagrams given by Figs.11 and 12 as a function of classes of H s every 0.5 m and T e every 0.5 s.In particular, the upper colour bar defines the level of energy per meter of wave front (in kWh/m), the numbers within the graph refer the occurrence of classes of sea states in terms of number of mean hours per mean year, and the isolines illustrate the wave power, P, expressed in kW/m.The values of E are determined by multiplying the classes of wave power to the corresponding time windows (6 h) occurring between successive sea states.Within a mean year, the highest E are associated to the bins with H s = 1.25 -1.75 m and T e = 6.25 -7.25 s at the Thyrrenian hot spot and to the bins with H s = 1.75 m

CONCLUSIONS
A wave energy assessment has been performed o↵ the coasts of Calabria region (Southern Italy).ECMWF wave data, processed and calibrated with other sources (RON buoys and UKMO nodes), have been adopted to estimate the wave power.Particular attention has been paid to the di↵erences in assessing correctly the energy wave period as a function of the real directional wave spectra against relationships based on the use of a standard JONSWAP spectrum.Owing to the di↵erent water depth of ECMWF nodes, the geographical transposition has been adopted to evaluate the wave power o↵ Calabrian coasts at -100 m.For the Thyrrenian Sea, a hot spot has been individuated in front of the town of Fiumefreddo Bruzio located with a yearly wave energy of 2900 kW/m.Across this location, a quite constant wave energy has been observed in this central part of Calabria for an extension of about 50 km between the town of Amantea and Cetraro and suitable for potential WEC installations.For the Ionian Sea, the highest yearly wave energy is just restricted in few km o↵ the town of Isola Capo Rizzuto with a resulting value of 2750 kW/m.By crossing the present wave energy with the power matrices of o↵shore WECs, further developments of the study will be addressed to evaluate the performance of the devices in terms of electricity production for domestic and public supplies for coastal towns located near the selected hot spot areas.

Figure 4 :
Figure 4: Relative di↵erences in assessing the mean yearly and monthly wave power, P m , as a function of T e , T p and T m2 at the ECMWF node no.E4 o↵ Tyrrhenian coast.

Figure 5 :
Figure 5: di↵erences in assessing the mean yearly and monthly wave power, P m , as a function of T e , T p and T m2 at the ECMWF node no.E35 o↵ Ionian coast.

Figure 6 :Figure 7 :
Figure 6: Mean yearly and monthly variation of JONSWAP shape parameters and period ratios at the ECMWF node no.E4 o↵ Tyrrhenian coast.(a) , (b) a and b , (c) ↵ a and ↵ b .

Figure 8 :
Figure 8: Spatial distribution of values of mean yearly and seasonal wave power, P m , at a depth of 100 m o↵ Calabrian coasts.(a) year, (b) winter, (c) spring, (d) summer (e) autumn.

Figure 9 :
Figure 9: Polar plot of mean yearly and seasonal wave power climate, P m , at the hot spot of Tyrrhenian Sea.(a) year, (b) winter, (c) spring, (d) summer, (e) autumn.

Figure 10 :
Figure 10: Polar plot of mean yearly and seasonal wave power climate, P m , at the hot of Ionian Sea.(a) year, (b) winter, (c) spring, (d) summer, (e) autumn.

Figure 11 :
Figure 11: Scatter diagrams of mean yearly and seasonal wave energy at hot spot of Tyrrhenian Sea.(a) year, (b) winter, (c) spring, (d) summer, (e) autumn.

Figure 12 :
Figure 12: Scatter diagrams of mean yearly and seasonal wave energy at the hot spot of Ionian Sea.(a) year, (b) winter, (c) spring, (d) summer, (e) autumn.