Vertical profiles of 3-D velocity components and directional spectral parameters of waves impacting the study site were recorded with a Doppler acoustic wave and current profiler (AWAC, Nortek, Rud, Norway) deployed 800 m offshore on a 7 m deep bottom, approximately at the limit of the active beach (45∘30′28′′ N; 12∘41′33′′ E, see Fig. 1). The mooring structure consisted of an aluminium framework stabilised by concrete blocks, hosting a removable polyethylene case containing the instrument and its battery canister, with permeable bottom for allowing sediment outflow during retrieval for maintenance and data downloading. Besides the traditional computation of velocity via analyses of the echo of a known emitted signal (1 MHz frequency for this device), the configuration of the four acoustic transducers installed in this system, together with an embedded pressure gauge, allow the reconstruction of the spectral parameters from an estimate of free water level oscillations (acoustic surface tracking) and orbital velocity statistics. An integrated orientation sensor records the rotation components of the instrument around three dimensions, permitting a retrospective control on the geotechnical stability of the system and on the fulfillment of the tolerance requirements for an appropriate operation of the wave measurements.

Current velocities along the water column were measured every 600 s within 50 cm cells, with a precision of 0.011 and 0.034 ms-1 on the vertical and horizontal directions, respectively. In turn, waves were sampled hourly by collecting 2400 samples at a 2 Hz sampling rate. In this configuration, the theoretical duration of the alkaline battery was about 100 days, but the real bottleneck for system maintenance was given by severe bio-fouling occurring in the shallow working environment that required inspections at least every month, especially in the summer season (Bonaldo et al., 2014).

Wave conditions were measured between March 2013 and October 2014. The instrument worked properly for most of the time. Ancillary wind data were collected on the Acqua Alta platform, approximately 26 km SSW of the research area but still the closest sea observation point available for this purpose. An outlook on time series of wave height and period at Jesolo, together with wind records at Acqua Alta is provided in Fig. 2.

A video monitoring station, which constituted of two high-resolution digital camera systems, was specifically set up for the project. The station was designed to record, without gaps, the response of the beach encompassed in the study area to the different meteomarine events, described in detail on the ocean side by the acoustic hydrodynamic measurements illustrated above. The video monitoring station was installed in July 2013 on the roof of a hotel facing the test area (45∘31′04′′ N, 12∘41′08′′ E), which was chosen after a survey of the coastal tract, aiming to maximise the optical performance. It consists of two reflex digital cameras (10.1 megapixel, optical zoom), each hosted in a single waterproof case, and a management module (computer control unit of 1.6 MHz, HD 160 GB; network interface controller Ethernet, modem UMTS/GSM/GPRS) allowing online monitoring and configuration. The control software package is composed of three modules: the first hosts the system management, controlling both image acquisition from all video station devices and image processing and uploading on the Web server using an FTP protocol; the second handles the remote connection and online servicing (setting changes, etc.), which allows real-time monitoring of the camera/video camera status and images and permits online system configuration; the third module hosts the Web server, enables downloading images and supports relocation to the Web server. Angle of view of the two cameras is shown in Fig. 1.

Timex, or time exposure, images are produced by digitally averaging image intensity over a prescribed time interval of image acquisition (Fig. 3b) (Holman and Stanley, 2007) in order to remove random transient sea conditions and variability in run-up height. On the other hand, the detection interval needs to be sufficiently short to be referred to a contemporary water level. Images obtained by following this procedure are the basis for the identification of key topographic information, because the adopted technique leads to an increase in pixel colour intensity, enabling identification of peculiar beach features, such as sand bars morphology, shoreline positions, intertidal beach profiles and slopes, and morphological formations in the shoreface (Kroon et al., 2007). In this experiment, timex images are obtained by averaging 40 of the 15 s (exposure time) images over detection intervals of approximately 10 min. The same methodology to obtain timex images was applied with still cameras at Igea Marina (Archetti and Zanuttigh, 2010), and it was verified that timex images obtained with the proposed methodology and the traditional methodology are comparable.

Instead, snapshot images (Fig. 3a) can offer information on the activities ongoing on the site: for example, it is possible to see when tractors and trucks are working on beach nourishments or preparing beach protections. In the presented study, images taken by Camera 1, NE oriented, were considered in order to analyse the last 300 m in the northern subregion of Jesolo beach.

On 18 July 2013, and on 30 September 2014, two high-resolution morpho-bathymetric surveys were carried out throughout the whole study area, with the multiple purpose of providing a reference bathymetry for an “initial state” of all the work and for morphodynamic evaluations, showing inter-annual variations in the beach morphology and identifying couples of ground control points required for georeferenced image processing. During the surveys, the shoreline position was also detected and later used for validating its identification based on image processing. The subaerial and intertidal beach was surveyed with a Trimble (Sunnyvale, California) 5700 real-time kinematics (RTK) GPS system, collecting 27 transects with an average 50 m spacing. The sub-tidal beach, from the 1.5 m bathymetric contour line to the offshore limit of the domain, was surveyed by a Kongsberg (Norway) GeoSwath Plus 500 interferometric multibeam, with 500 kHz frequency, 240∘ view angle and horizontal coverage up to 12 times the local water depth. Shallower sub-tidal zones were measured with a Teledyne Odom (Baton Rouge, Louisiana) Hydrotrac echosounder mounting a 200 kHz transducer.

During the study period, surface sediment samples were collected with a van Veen grab sampler at different positions on the study site. Samples were subsequently dried and sieved obtaining granulometric curves within the 19–2000 µm grain size range, allowing to classify the sediment as a silty sand.

Individual storm events over the study period were identified from the recorded wave heights using the common methodology described by Boccotti (2000). The method is based on a preliminary identification of a wave height threshold, namely 1.5 times the annual average HS (that is 0.58 m for the available wave data set described in Sect. 3). This order of magnitude for the threshold value was the same used in many papers concerning the North Adriatic area, for instance, in Archetti and Romagnoli (2011). In this way, it was possible to create a 1-year long list of storms occurred. As an example of application of this methodology to identify storms, the seventh recorded storm is illustrated below in detail, as representative of all the others. The relative recorded time series is presented in Fig. 5. This event began on 1 December 2013, at 01:00 LT, when the wave height overcame the defined threshold with a value of 0.91 m, and it lasted until 3 December at 03:00 LT. After this time, the wave height decreased under the threshold value. The maximum wave height, 1.90 m, occurred on 2 December at 09:00 LT. The total duration of this storm was 50 h.

This procedure led to the identification of a total of 31 storm events in the period October 2013–October 2014, summarised in Table 1, as explained in the following section, not all these events, but only 29, could be analysed because of malfunctions of video image station.

The first step, after the identification of the relevant storms, consisted of selecting the most appropriate timex images, among those that passed the quality control phase, taken during calm days before and after the identified storms. In order to allow the evaluation of shoreline position net of sea level oscillations, images for every storm were selected in correspondence with the same values of sea surface elevation, retrieved a posteriori from the AWAC pressure gauge.

The video camera system did not save images always in the analysed period: for instance, from 14 to 25 February 2014 (28th storm) there are no images available. The same happened during the 30th and the 22nd storms, and therefore these events are not considered in the analysis.

Image distortion was corrected with a MATLAB based open source Camera Calibration Toolbox (http://www.vision.caltech.edu/bouguetj/calib_doc/, Zhang, 1999; Tsai, 1987). The processed images were then rectified (i. e. projected from the image reference system into a user-specified horizontal plane, in this case the sea surface plan in world coordinates) by applying the collinearity equations (Weng et al., 1992; Lenz and Tsai, 1988; Slama, 1980) on seven pairs of ground control points (GCPs), whose real-world coordinates were surveyed during a dedicated field campaign.

An example of plan view (rectified image) of the surveyed beach, where the swimming pool next to the installation hotel and the beach umbrella lines are easily recognisable, is shown in Fig. 6. Coordinates are given in in UTM system units (m).

Many automatic procedures are available for the identification of shoreline position on images. Recently, Garcia-Rubio et al. (2015) presented a method based on the energy reflected in the NIR (near infrared) wavelengths, assuming that higher and lower intensities are related to the inherent physical properties of sea and land, respectively.

For camera images several methodologies have been developed (Aarnikhof and Roelvink, 1999). In the current analysis the shorelines were automatically detected on the rectified image, through an image processing tool specifically developed in the MATLAB environment. The methodology was the same applied in Archetti and Zanuttigh, (2010), based on a sub-pixel extraction of the line between water and non-water zones (Viet et al., 2014, Carniel et al., 2011; Vousdoukas et al., 2011; Archetti, 2009). Detected shorelines were then interpolated at 5 m spatial resolution. An example of rectified timex image and shoreline detected is presented in Fig. 6., whereas the ground-truth data are constituted by the waterline acquired at the same time of video-derived shoreline by a team operating a DGPS at RTK (real-time kinematic) mode during the topographic surveys (see Sect. 3.3).

In the present paper, analyses relevant to the period October 2013–October 2014 focus on storm-induced short-term changes in the shoreline position. In detail, the procedure to quantify the storm-induced shoreline variation, described also in Aarninkhof and Roelvink (1999), Aarninkhof et al. (2003), and Archetti and Lamberti (2009), is generally based on three steps. First, timex images collected immediately before and after the storm, at time of comparable sea water level (measured by the AWAC pressure gauge) are selected. Second, the position of the shoreline is detected on rectified timex images. Last, the distances between the two shorelines is calculated. For this reason, and in order to avoid bias due to wave set up, it was decided to select and analyse only images taken during good meteorological conditions (calm sea). Dedicated routines allowed to associate sea-water elevation and shoreline position. Due to the strict criteria to be fulfilled (comparable sea level, calm sea state and good visibility) and the limited availability of timex images (as a compromise between time resolution and storage, only four timex images per day were available), the selection of the images was a demanding operation.

Along the beach, over short temporal scales, such as those typical of a storm, shoreline position changes are not always homogeneous, but it was possible to observe different behaviours and responses to the marine events. Indeed, some parts of the beach can advance and others can retreat, filtering out changes with the opposite pattern with respect to a central section and giving rise to apparent temporary rotation (Fig. 9), even though this is not a pocket beach (Ojeda and Guillén, 2008) and the magnitude of rotation is not particularly relevant. Often, short-term beach rotation is manifested as variation of beach volume and width due to the lateral movement of sand alongshore (Masselink and Short, 1993; Klein et al., 2002), but otherwise it can appear without any variation of volume and width. It is also possible that cross-shore movements due to particular wave conditions that occurred during the previous days result in a net progression of the whole beach (Fig. 10). Two mechanisms can cause wave-energy dissipation on a beach impacted by waves: (i) wave-energy position retrieval by submerged sand bar accretion, and, more efficiently, (ii) alongshore sediment redistribution (Alvarez et al., 2011).

In Figs. 8–10, the blue line represents the shoreline before the storm and the green line the shoreline after the storm. For almost all of the conditions considered, as for example in the case of storm S7, we notice a retreat of the beach, enhancing the described mechanism (Figs. 8, 9). On the contrary, Fig. 10 shows that storm S9 induced a slight beach enlargement in response to a quite small wave height, but it is also possible to see that even a low power storm can modify the shoreline position, as explained in Sect. 4.1. Figure 11, moreover, shows the comparison between shorelines before (blue line) and during the peak (red line) of the analysed storm S7.

More information on the storm conditions and beach behaviour is presented in the next section.

The surveyed bathymetry is presented in Fig. 4, with bathymetry and later shoreline detection provided in UTM coordinate system and units in metres. Both RTK GPS surveys indicate that the beach slope in the intertidal zone was approx. 1:15.

A further estimation of the intertidal beach slope was performed also by analysing video images taken during calm days. The procedure was the same usually adopted to identify the intertidal beach position, described in Archetti and Romagnoli (2011), that is generally based on two steps. First, the position of the shoreline is detected on rectified timex images and horizontal distances between them are calculated (Δx); and second, the associated vertical elevation (Δh) is estimated from the sea level data (measured by the ADCP – Acoustic Doppler Current Profile). In this way, the average intertidal slope was the ratio Δh/Δx. We found a mean slope of 1:16, confirming the measurements based on GPS surveys performed in summer.

The accuracy of the image-derived method was calculated comparing video-detected shorelines with the two GPS beach surveyed detected at the same time, hereinafter δ (surveys). This deviation was due to the ortho-rectification and the shoreline-detection processes.

In particular, the validation of the video-detected shoreline against the surveyed datum collected on 30 September 2014, showed a good agreement and the stability of the calibration carried out in July 2013 (Fig. 7). The mean distance, computed for shoreline located in the 300 m closer to the video station with reference to a set of 60 transects, was 1.59 m against data from the first survey, and 0.62 m for the second survey. The results indicated a mean value of 1.45 m as difference between the two lines, in the first 300 m. The δ increased after the first 300 m, as shown in Fig. 7. This error value was considered to be acceptable, since it was comparable to the excursion of the swash-zone during calm days and corresponded to 3–20 cm of vertical range for an average intertidal slope of 1:15. The evaluated error was on the same order of magnitude, or less, than those obtained using video system for shoreline detection, as reported in many works (Archetti and Romagnoli, 2011; Ruiz de Alegria-Arzaburu and Masselink, 2010; Ruggiero and List, 2009; Holman and Stanley, 2007; Siegle et al., 2007; Harley et al., 2007; Elko et al., 2005; Holland et al., 1997).

In order to describe how Jesolo beach reacts after significant storms, and to correlate this short-term beach evolution to the storm energy, the mean distance Δx (difference in the cross-shore direction between shoreline position after and before the storm) was calculated for each selected event, as indicated in the Table 1. The displacement of the shoreline during the storms was estimated for a set of n= 60 transects on a 300 m long stretch of the beach.

The mean distance Δx between the two shorelines was defined as Δx=Σidin, where i were the transects (i=1-60) and di the distances in the cross-shore direction at every transect. Negative Δx values indicated beach progradation, whilst positive values were associated with beach retreat.

In the analysed time frame (October 2013–October 2014), Δx values varied commonly between 0.62 and 4.87 m. Slightly higher positive values (from 5.20 to 9.60 m) were observed during winter storms, the most intense stormy periods. Only the 9th storm gave a negative Δx value (-1.55 m, see Fig. 10), possibly due to the supply of sand mobilised during the strong previous storm and redistributed alongshore by this event.

The global evolution of the beach showed how a large retreat of the shoreline was not generally associated with the highest waves and the longest storms, but mainly with the cumulative effect of successive storms (Pardo-Pascual et al., 2014). In fact, a number of factors may explain the correlation observed between shoreline displacements and the storm features for each event. As suggested by some authors (Ruiz de Alegria-Arzaburu and Masselink, 2010; Ojeda and Guillén, 2008; Ortega-Sànchez et al., 2008; Dail et al., 2000) the shoreline behaviour may depend on antecedent beach conditions, and the previous morphological state should be considered, together with the along-shore gradients in the factors controlling coastal sediment drift, in order to provide a detailed morphodynamic characterisation of the processes at the scale of a single event. In our case, a complete morphological description of the beach morphology was not available for each event; therefore upcoming evaluations will need to be considered in terms of statistical tendencies, rather than strictly prognostic of the details of morphodynamic processes undergone by the coastal tract in the presence of a given storm. Nevertheless, it was possible to calculate (and summarise in Table 1) the distance of each pre-storm shoreline from a baseline, identified as an interface between beach and hotel areas on an image collected on 30 August 2014. This quantity, though not adding any substantial information in terms of beach erosion, provides a record of shoreline displacements over the annual scale, not showing any strong seasonal behaviour and therefore suggesting that short-term dynamics may be dominating with respect to annual processes in this system.

The correlation between beach response and sea conditions was finally investigated based on a preliminary evaluation of the outcomes of the measurement system with reference to specific storms. Four indexes were considered in the analysis to predict, with an empirical relationship, the shoreline displacement due to a forecasted storm under different conditions.

The four synthetic parameters are listed below:

the maximum wave height during the storm: Hmax [m];

In December 2013, the area experienced a particularly strong storm event (storm 8) associated with a “storm power index” of 617 m2h given by a 51 h duration of the storm and a Hmax about 4 m. Storm 6 presented even larger value for Ps, up to 780 m2h, given by the remarkable duration of the storm (214 h), in spite of a Hmax of 1.9 m.

As summarised in Table 1, for all the storms we reported maximum Hs and period, duration, direction and wind direction, together with the corresponding Δx, the maximum displacement occurred during the storm peak and the indexes E, Etot, Ps and the distance of each pre-storm shoreline from a baseline.

Scatter plots representing the shoreline displacements as function of Hmax, E, Etot and Ps are shown, respectively, in Figs. 12–15, in order to highlight the main relationships between these indexes and the shoreline displacement. Empty bullets in the plots indicate the storms occurred when the beach was protected by an artificial dune, generally in the winter season. In all the presented figures, the coloured bullets are related to the main wind direction generating the considered wave conditions in Jesolo beach. This information was necessary to understand the different features of the storms, showing that Bora storms (strong wind coming from North-East sectors) were generally the most energetic ones, and capable of inducing larger displacements of the shoreline based on all indexes. In Figs. 13 and 14, in which Δx was related to the energy of the corresponding storm (respectively E and Etot), it was possible to identify a main trend, followed by most of the events. In particular, we can analyse in detail some of the results, in order to better understand the behaviour of Jesolo beach during the study period.

S8 was the most energetic storm recorded, in term of wave height; during the storm the beach was protected, so the displacement was less than the expected one.

Similar considerations can be drawn also looking the relationship between Hmax and Δx. Worth noting, S4 and S5 had a high Δx because they had not a protection, even though the wave height was not very big.

Moreover, it is possible to see that several highest energy storms, (namely S5, S6, S8, and a part of S11 that is characterised by the maximum duration), were generated by northeasterly winds (Bora).

Figure 15 provides a further evidence of the importance of the storm duration in determining the impact on the shore. In fact, as well as we already said about S8 and S11, also S6 had a high value of Ps, due to its duration, 214 h.

Based on these observations, we can confirm that the extensive adoption of artificial dunes (Fig. 16) in the study area was useful also to reduce shoreline retreat during the storm. This type of interventions can sometimes contribute to extend overall stability not only in the replenished zone but also in downdrift areas (Del Rio et al., 2013).

In the endeavour of identifying a suitable parameterisation for predicting the shoreline retreat Δx in response to a storm event, we performed a fitting of this quantity against the four parameters listed above. Due to the strong variability observed in the beach response, besides using the full set of available measurements, the fitting was further tested for a number of data subsets. These were obtained by classifying the events by wind direction and beach dune conditions (Bora vs. Scirocco and protected vs. not protected, summarised in Table 2) or by overall storm energy (Low and High, with reference to a threshold of 300 m2s2, summarised in Table 3). After some tests (not shown here) it was found that the power law of the form Δx=aYb , with Y representing a generic storm characteristic, tendentially provided the best fit of the data, hence we referred to this case evaluating the fitting coefficients a and b, the the root mean square error and the coefficient of determination R2.

As a general trend, fitting results highlight that the presence of an artificial beach protection seems to significantly hamper the capability of a power law formulation of properly reproducing shoreline erosion (Table 2). On the other hand, the behaviour of an unprotected beach can satisfactorily be described by this model, especially in the case of Bora storms. In particular, the total energy impacting the beach and the storm power index appear more robust with respect to the criterion adopted for the classification, giving rise to overall better quality of fit compared to maximum wave height and wave energy at the storm peak. Retaining the information about the wave period, these parameters allow a more complete characterisation of the storm, better capturing the different behaviour associated to Bora and Scirocco, which in turn is strongly related to a different development of the sea state.

The classification criterion based on the generating wind direction allows us to identify the differences in the beach response to the storm event (worth noting, fitting coefficients for Bora and Scirocco are rather different, likely reflecting the differences in the sea state characteristics). Nevertheless, the best fit is given based on an energetic criterion, focusing on the storms characterised by total energy values exceeding 300 m2s2 (Table 3). Indeed, in the face of best fits yielding Δx=0.84Etot0.29 for the whole data set, with R2= 0.40 and RMSE = 1.58 m, considering only the most energetic storms (generally the most harmful for coastal erosion concerns) the empirical relation becomes Δx=0.05Etot0.71.