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.
24 July 2025 in Communications Earth & Environment
Robust predictions of shoreline change are critical for sustainable coastal management. Despite advancements in shoreline models, objective benchmarking remains limited. Here we present results from ShoreShop2.0, an international collaborative benchmarking workshop, where 34 groups submitted shoreline change predictions in a blind competition. Subsets of shoreline observations at an undisclosed site (BeachX) over short (5-year) and medium (50-year) periods were withheld from modelers and used for model benchmarking. Using satellite-derived shoreline datasets for calibration and evaluation, the best performing models achieved prediction accuracies on the order of 10 m, comparable to the accuracy of the satellite shoreline data, indicating that certain beaches can be modelled nearly as well as they can be remotely observed. The outcomes from this collaborative benchmarking competition critically review the present state-of-the-art in shoreline change prediction as well as reveal model limitations, facilitate improvements, and offer insights for advancing shoreline-prediction capabilities.
17 June 2025 in Journal of Geophysical Research: Earth Surface
Yes. Equilibrium shoreline models, which simulate wave‐driven cross‐shore erosion and accretion, are mathematically equivalent to a discrete convolution (i.e., a weighted, moving average) of a time series of wave‐forcing conditions with a parameterized memory‐decay kernel function. The direct equivalence between equilibrium shoreline models and convolutions reveals key theoretical aspects of equilibrium behavior. Convolutions (representing quasi‐low‐pass filter operations) provide an intuitive theoretical description of shoreline erosion and accretion behavior in response to waves: that is, shoreline position often mirrors the weighted moving average of wave time series. Model‐convolution equivalence also provides a conceptual basis to interpret, evaluate, and construct data‐driven Machine‐Learning/Deep‐Learning (ML/DL) models that use convolutions to extract features from data and then apply them for prediction (e.g., Convolutional Neural Networks (CNNs)). Finally, our findings provide a methodological pathway (based on Fourier transforms) for future understanding of wave‐driven shoreline change, which can be used to interpret the coherence between the frequency spectrum of the processes of waves and shoreline change and construct more computationally efficient and effective shoreline‐modeling approaches.
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.
22 November 2025 in Science of The Total Environment
Understanding the time scales of water transport in semi-closed ecosystems is essential for assessing the behavior of pollutants in coastal lagoon environments. Terminos Lagoon, Mexico's largest lagoon-estuarine system, is under significant environmental stress. We used a water age tracer combined with a high-resolution 3D hydrodynamic model to evaluate the lagoon's zones of accumulation and exchange. Our findings revealed a year-round counterclockwise circulation in the lagoon, which allows seawater to enter through Puerto Real and exit via Carmen Inlet. The inner littoral of Carmen Island features relatively young water currents, with an average age of 60 days. In contrast, the lagoon's center contains older water parcels, averaging 130 days. The southern and southeastern littoral regions hold older water areas, with the Candelaria Estuary showing the oldest waters, reaching up to 250 days. This emphasizes the Candelaria Estuary as a crucial area at risk of environmental impacts due to contaminant accumulation in the lagoon. As the primary inflow of seawater occurs through the Puerto Real inlet, it is vital to establish management strategies to minimize potential pollutant releases in the area.
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.