Search

The process of shoreline change is inherently complex, and reliable predictions of shoreline position remain a key challenge in coastal research. Predicting shoreline evolution could potentially benefit from deep learning (DL), which is a recently developed and widely successful data-driven methodology. However, so far its implementation for shoreline time series data has been limited. The aim of this contribution is to investigate the potential of DL algorithms to predict interannual shoreline position derived from camera system observations at a New Zealand study site. We investigate the application of convolutional neural networks (CNNs) and hybrid CNN-LSTM (Long Short-Term Memory) networks. We compare our results with two established models: a shoreline equilibrium model and a model that addresses timescales in shoreline drivers. Using a systematic search and different measures of fitness, we found DL models that outperformed the reference models when simulating the variability and distribution of the observations. Overall, these results indicate that DL models have potential to improve accuracy and reliability over current models. [ABSTRACT FROM AUTHOR]

The process of shoreline change is inherently complex and reliable predictions of shoreline position remain a key challenge in coastal research. Predicting shoreline evolution could potentially benefit from Deep Learning (DL), which is a recently developed and widely successful data-driven methodology. However, so far its implementation for shoreline time series data has been limited. The aim of this contribution is to investigate the potential of DL algorithms to predict interannual shoreline position derived from camera system observations at a New Zealand study site. We investigate the application of Convolutional Neural Networks (CNNs) and hybrid CNN - Long Short-Term Memory networks. We compare our results with two established models, a shoreline equilibrium model, and a model that addresses time scales in shoreline drivers. Using a systematic search and different measures of fitness we found DL models that outperformed the reference models when simulating the variability and distribution of the observations. Overall, these results indicate that DL models have potential to improve accuracy and reliability over current models. [ABSTRACT FROM AUTHOR]

Predicting shoreline change is a key issue in coastal research. Predictors, process-based or data-driven, tend to be developed and tested on high-frequency and high-quality data sets. Combining hydrodynamic and morphological variables extracted from video images and artificial neural network allows us to evaluate if sparse data could still provide physically-sound shoreline change predictions. The data set covered a 3-year period with shoreline position data (with an accuracy of ±5 m) available 73 % of the time and 66 % for the morphological parameters (beach state or bar location). The best configuration of the trained shallow (one hidden layer) Feedforward Artificial Neural Network (ANN), includes 10 input variables and 10 nodes allowing to capture the shoreline dynamic at different time scales, from the storm-event to the seasonal scale, and to predict the shoreline position on a 1-year period with a RMSE of about 6.7 m. Increasing the complexity of the architecture of the ANN by increasing the number of hidden layers did not improve the predictions. By modifying the number of input variables in the algorithm, the ANN also allows us to highlight the mitigation effect of the bar during the storm event and its role as sediment buffer during seasonal accretion. [ABSTRACT FROM AUTHOR]

Understanding the interactions between dune systems and beaches is critical to determining the short-term shoreline response and the long-term resilience. In this study, almost 15 years of monthly beach/dune measurements were analyzed for three different profiles at Vougot Beach, France to understand and predict shoreline changes from intra-annual to multi-annual time-scales. Four migration modes: advance/retreat (translation modes) and steepening/flattening (rotation modes) were identified through a centroid analysis. The analysis showed that translation and rotation can occur simultaneously, with long-term trends of beach retreat and profile steepening (lower beach retreating and upper beach advancing), which was interrupted by two energetic wave events causing profile flattening (lower beach advancing and upper beach retreating). These two observations are evidence of how the sediment contribution resulting from the dune erosion events temporarily caused a large advance in the shoreline position. A recent modeling approach that accounts for different time-scales is applied to predict the shoreline changes, showing significant improvements in comparison to a traditional shoreline equilibrium model when time-scales related with the dune erosion and recovery are considered. The results showed that the dune system affects the beach profile evolution both spatially, with different impacts at different elevations along the cross-shore profile, and temporally, by periodically redistributing the sediment in the system. [ABSTRACT FROM AUTHOR]

Reliable predictions and accompanying uncertainty estimates of coastal evolution on decadal to centennial time scales are increasingly sought. So far, most coastal change projections rely on a single, deterministic realization of the unknown future wave climate, often derived from a global climate model. Yet, deterministic projections do not account for the stochastic nature of future wave conditions across a variety of temporal scales (e.g., daily, weekly, seasonally, and interannually). Here, we present an ensemble Kalman filter shoreline change model to predict coastal erosion and uncertainty due to waves at a variety of time scales. We compare shoreline change projections, simulated with and without ensemble wave forcing conditions by applying ensemble wave time series produced by a computationally efficient statistical downscaling method. We demonstrate a sizable (site‐dependent) increase in model uncertainty compared with the unrealistic case of model projections based on a single, deterministic realization (e.g., a single time series) of the wave forcing. We support model‐derived uncertainty estimates with a novel mathematical analysis of ensembles of idealized process models. Here, the developed ensemble modeling approach is applied to a well‐monitored beach in Tairua, New Zealand. However, the model and uncertainty quantification techniques derived here are generally applicable to a variety of coastal settings around the world. Key Points: We assess uncertainty of ensemble Kalman filter data‐assimilated shoreline change predictions using the spread of model trajectoriesWe demonstrate an increase in modeled shoreline uncertainty due to ensemble wave conditions compared with single realization simulationsUncertainty in shoreline predictions can be quantified using a novel, straightforward relationship, derived analytically [ABSTRACT FROM AUTHOR]

Shorelines respond to a number of "drivers" operating on a variety of time‐scales. For some time‐scales (e.g., seasonal), the driver‐shoreline relationship is often evident; however, at longer time‐scales (e.g., multiannual), the shoreline changes may be superimposed on changes at shorter time‐scales and thus are difficult to identify. Here, we predict shoreline evolution from storm events to decadal time‐scales, using a novel approach based on the Complete Ensemble Empirical Mode Decomposition. This approach identifies and links the primary time‐scales in the model drivers (large‐scale sea level pressure [SLP] and/or waves) with the same time‐scales in the shoreline position. The multiscale approach reproduced shoreline changes at two beaches more skillfully than a common shoreline model when SLP and wave information were used in combination. In addition, the analysis can be applied to climate indices, providing the opportunity to link longer time‐scales with climate patterns (e.g., El Niño Southern Oscillation). Plain Language Summary: Beaches are changing constantly, advancing or retreating depending for instance, on the climate and ocean conditions. Beach retreat and advance may occur in cycles (seasonally, annually, or over several decades) or because of particular events such as storms. All these changes are superimposed and difficult to disentangle. Therefore, the same beach can look completely different in summer or winter, and the changes are not the same year after year. Therefore, predicting the beach state over the following months, years, or decades is a daunting task. Here, we introduce a new approach to the prediction of shoreline changes and test it at two beaches (one in New Zealand and the other in Australia). The new approach relates changes in shoreline position with "drivers" (waves and atmospheric patterns) decomposed into time‐scales (e.g., seasonal, annual, and bi‐annual) and uses these connections to predict shoreline changes. Key Points: A novel modeling approach to shoreline prediction is presented and compared with an established model at two study sitesA decomposition of different time‐scales (from storm to climate anomalies) in shoreline and drivers is used to predict shoreline changeThe addition of sea level pressure information to wave bulk parameters as model inputs improves shoreline predictions [ABSTRACT FROM AUTHOR]

Beaches around the world continuously adjust to daily and seasonal changes in wave and tide conditions, which are themselves changing over longer time-scales. Different approaches to predict multi-year shoreline evolution have been implemented; however, robust and reliable predictions of shoreline evolution are still problematic even in short-term scenarios (shorter than decadal). Here we show results of a modelling competition, where 19 numerical models (a mix of established shoreline models and machine learning techniques) were tested using data collected for Tairua beach, New Zealand with 18 years of daily averaged alongshore shoreline position and beach rotation (orientation) data obtained from a camera system. In general, traditional shoreline models and machine learning techniques were able to reproduce shoreline changes during the calibration period (1999–2014) for normal conditions but some of the model struggled to predict extreme and fast oscillations. During the forecast period (unseen data, 2014–2017), both approaches showed a decrease in models' capability to predict the shoreline position. This was more evident for some of the machine learning algorithms. A model ensemble performed better than individual models and enables assessment of uncertainties in model architecture. Research-coordinated approaches (e.g., modelling competitions) can fuel advances in predictive capabilities and provide a forum for the discussion about the advantages/disadvantages of available models. [ABSTRACT FROM AUTHOR]

A three-year dataset (2007–2010) of shoreline and sandbar positions derived from video observations of an open sandy beach (Biscarrosse, France) is analyzed, to explore the impact of storms on the seasonal shoreline dynamics. The results indicate that a clear seasonality is observed in the offshore significant wave height and in the occurrence of ‘winter storm’ events that are defined as periods with significant wave height greater than 4 m lasting at least 12 h. Data indicate that the range of variation of the inner sandbar positions (120 m) is two and a half times larger than the range of variation of the shoreline position. Correlation analyses suggest that the offshore conditions do not conclusively explain the position of the inner bar. Shoreline positions show a clear seasonal pattern, displaying a stable position occurring in February and July, providing evidence of seasonal-recovery although analysis of storm activity indicates that the number and intensity of storms is highly variable over the three years. The Biscarrosse data show that rapid erosion can be observed also under moderate conditions, and that beach recovery after the storm season can be relatively rapid. [ABSTRACT FROM AUTHOR]

Wave runup is one of the most critical parameters contributing to coastline flooding and shoreline change. Many formulas have been developed to empirically predict wave runup characteristics. However, large errors are still associated to these estimates, feeding a debate on the best approach to describe runup through simple parameterizations. In this work we present a comprehensive review on runup, setup and swash empirical models. We use a large database of field measurements to verify the predictive capability of recent formulas, addressing possible sources of variability. We identify the effect of the nearshore bathymetry as an important factor missing in empirical predictors and examine this issue through an analysis of runup simulated with the SWASH model over different barred and linear bathymetries. The details of the submerged beach profile affect estimates, especially with respect to setup values. The lack of standardization in measuring, post-processing and sharing runup data, limits the application of previous published databases on new analysis. We list a set of recommendations that can be used as guidelines for new measurements to broaden the applicability of the datasets in future studies. Finally, we discuss the remaining questions that still need to be further explored. [ABSTRACT FROM AUTHOR]

Abstract: We analyse a seven-year dataset (1999–2005) of shoreline and sandbar variations derived from video observations at the embayed Tairua Beach, New Zealand, to explore sandbar–shoreline coupling and to determine how this coupling is related to alongshore-averaged sandbar–shoreline separation and beach rotation. Our results quantify the coupling between the sandbar and the shoreline which is directly related to the reduced separation between the sandbar and the shoreline. Using a simple predictive shoreline model, we show that this behaviour is related to the energy of the incoming waves. Our observations are obtained from a headland-enclosed beach and show aspects of the morphological state that relate to the existence of headlands that are not considered in the Wright and Short beach state classification. Both the sandbar and shoreline at Tairua Beach rotate relatively quickly (<1month) during the winter storm events, when the storms are accompanied by strong alongshore components, gradually returning during the summer months. The sandbar–shoreline coupling length scale is controlled by this rotation, with large wavelengths occurring when the sandbar is strongly rotated. [Copyright &y& Elsevier]

Abstract: Alongshore variability in swash motions – shoreline oscillations about the mean water level on the beach face – were investigated using video images and a high-resolution morphology survey on an intermediate beach. Under mild, swell-dominated offshore wave conditions, alongshore variation of up to 78% in significant runup height Rs (defined as 4 times the standard deviation of the swash time series) was observed. This variation was predominantly driven by energy at the incident (>0.05Hz) frequencies (where most of the swash energy was observed), and, consistent with previous observations, was mainly controlled by changes in the slope of the beach face (measured at the mean swash location). However, alongshore patterning in wave breaking over the sandbar caused variation in the degree of wave dissipation along the beach and also resulted in alongshore changes to swash motions. Although alongshore changes in beach slope and wave breaking patterning over the bar were observed to be typically correlated, both were needed in a regression model to provide the best explaination of alongshore changes in Rs. At infragravity frequencies (<0.05Hz), alongshore variability was not well associated either with changes in beach slope or wave breaking patterning. Low-mode edge waves were observed in the swash measurements and their contribution to the total energy spectrum was greatest near the location where a shoal was observed, suggesting this shoal may play a role in forcing. The edge waves may have contributed to the swash variability observed at infragravity frequencies. However, in these reflective conditions, the infragravity band plays a secondary role in controlling alongshore variations to swash motions. [Copyright &y& Elsevier]

Abstract: Groundwater seepage is known to influence beach erosion and accretion processes. However, field measurements of the variation of the groundwater seepage line (GWSL) and the vertical elevation difference between the GWSL and the shoreline are limited. We developed a methodology to extract the temporal variability of the shoreline and the wet–dry boundary using video imagery, with the overarching aim to examine elevation differences between the wet–dry boundary and the shoreline position in relation to rainfall and wave characteristics, during a tidal cycle. The wet–dry boundary was detected from 10min time-averaged images collected at Ngaranui Beach, Raglan, New Zealand. An algorithm discriminated between the dry and wet cells using a threshold related to the maximum of the red, green, and blue intensities in Hue-Saturation-Value. Field measurements showed this corresponded to the location where the water table was within 2cm of the beachface surface. Time stacks and time series of pixels extracted from cross-shore transects in the video imagery, were used to determine the location of the shoreline by manually digitizing the maximum run-up and minimum run-down location for each swash cycle, and averaging the result. In our test data set of 14 days covering a range of wave and rainfall conditions, we found 6 days when the elevation difference between the wet–dry boundary and the shoreline remained approximately constant during the tidal cycle. For these days, the wet–dry boundary corresponded to the upper limit of the swash zone. On the other 8 days, the wet–dry boundary and the shoreline decoupled with falling tide, leading to elevation differences of up to 2.5m at low tide. Elevation differences between the GWSL and the shoreline at low tide were particularly large when the cumulative rainfall in the preceding month was greater than 200mm. This research shows that the wet–dry boundary (such as often used in video shoreline-finding algorithms) is related to groundwater seepage on low-sloped, medium to fine sand beaches such as Ngaranui Beach (mean grain size ∼0.27mm, beach slope ∼1:70) and may not be a good indicator of the position of the shoreline. [Copyright &y& Elsevier]

Sediment exchange between the shoreface and the foreshore is investigated through the analysis of sandbar and berm migration cycles at La Misión beach in Baja California, Mexico. In general, nearshore sandbars migrate onshore during low-energy wave conditions merging with the shoreline to become intertidal sandbars that later contribute to berm rebuilding. However, significant alongshore differences were observed in the sandbar welding and berm rebuilding process, which can be attributed to alongshore wave skewness and near-surfzone flow variations. The analysis of topographic and bathymetric surveys collected monthly over a 5-year period revealed larger shoreline retreat and longer recovery times in the southern beach compared to the north. This was related to the inability of sandbars to weld to the intertidal beach during spring and summer, occurring only during extended periods of low-energy waves. In the southern beach, the berm formation process exhibited less dependence on incoming wave energy, persisting during mild-energy winters, and reforming in the presence of sandbars. Hydrodynamic measurements indicated that the southern part of the beach experienced relatively more intense net offshore flows (up to −0.2 ms−1) due to undertow, and possibly related to the presence of persistent rip currents. In contrast, the berm in the northern beach followed a seasonal erosion and rebuilding process associated with sandbar migrations driven by variations in incoming wave energy and cross-shore sediment transport. This was characterized by larger near-bed wave velocity skewness compared to the south. Net flows in the north were predominantly onshore, reaching up to 0.1 ms−1 during lower-energy periods with oblique wave incidence. This study reveals the presence of considerably distinct sandbar and berm migration cycles along a relatively short beach of 2.2 km long. It also highlights how such variability can be related to the morphodynamic feedback with sandbars affecting hydrodynamics, particularly near-bed wave skewness and near-surfzone flow variations, which in turn influence sediment transport and morphological evolution. • Bar merging affects berm rebuilding, influenced by wave skewness and flow variations. • Net onshore flows and higher skewness allow seasonal berm rebuilding. • Net offshore flows and lower skewness kept the bar offshore but the berm recovered. [ABSTRACT FROM AUTHOR]

The use of variance images for runup and shoreline detection is analyzed in this work. A methodology based on the variance images is first motivated with synthetic time stacks and then confirmed with real images. A total of 55 time stack images from five different cameras at three video stations (in Spain and New Zealand) were digitized and used for the study. These images include different tide and wave conditions. From the time stack images we: 1) obtained the mean shoreline position and maximum runup manually digitizing the water/sand interface and, 2) computed the variance function. The variance function is shown to have a particular cross-shore shape, with two characteristic positions which allow us to satisfactorily recover the mean shoreline position and maximum runup obtained in the first step. Variance images were shown to be useful both for detecting mean shoreline, and for evaluating runup, providing a potential alternative for timex images. The method did not work well when runup occurred within the groundwater seepage face. [ABSTRACT FROM AUTHOR]