Machine learning application in modelling marine and coastal phenomena: a critical review

Abstract

This study provides an extensive review of over 200 journal papers focusing on Machine Learning (ML) algorithms’ use for promoting a sustainable management of the marine and coastal environments. The research covers various facets of ML algorithms, including data preprocessing and handling, modeling algorithms for distinct phenomena, model evaluation, and use of dynamic and integrated models. Given that machine learning modeling relies on experience or trial-and-error, examining previous applications in marine and coastal modeling is proven to be beneficial. The performance of different ML methods used to predict wave heights was analyzed to ascertain which method was superior with various datasets. The analysis of these papers revealed that properly developed ML methods could successfully be applied to multiple aspects. Areas of application include data collection and analysis, pollutant and sediment transport, image processing and deep learning, and identification of potential regions for aquaculture and wave energy activities. Additionally, ML methods aid in structural design and optimization and in the prediction and classification of oceanographic parameters. However, despite their potential advantages, dynamic and integrated ML models remain underutilized in marine projects. This research provides insights into ML’s application and invites future investigations to exploit ML’s untapped potential in marine and coastal sustainability.

1 Introduction

Coastal areas are of vital significance due to their crucial role in supporting aspects such as biodiversity, economic activity, cultural heritage, climate regulation, food security, recreational opportunities, and strategic importance (Neumann et al., 2017). Ensuring their sustainability, however, is a challenge that requires addressing various factors, among which, climate change adaptation, beach protection and water quality management. One approach to ensuring the sustainability of coastal areas involves conducting a thorough examination of each contributing factor by employing data analysis and suitable methods. Effective data analysis and augmentation is, therefore, essential for informed decision-making and sustainable management of coastal areas.

The amount of data related to coastal systems has dramatically increased recently (Goldstein et al., 2019). This data, which often covers large areas and spans long periods of time, is now available in high resolution and can be accessed quickly. This has led to more opportunities for research on the sustainability of activities evolving in coastal areas. However, handling large and complex datasets, as well as identifying their patterns and trends, is not a convenient task. Despite their widespread use and mathematical rigor, conventional statistical techniques, including descriptive statistics (Emmanouil et al., 2020), inferential statistics (Agarwal and Manuel, 2008), regression analysis (Davidson et al., 1996; Hall et al., 2002), correlation analysis (Szmytkiewicz et al., 2000; Kroon et al., 2008; Ruiz de Alegría-Arzaburu et al., 2010), Analysis of Variance (ANOVA) (Martins et al., 2010), and Principal Component Analysis (PCA) (Hua et al., 2007; Miller and Dean, 2007), have limitations when processing large and complex data sets, and can present challenges in terms of interpretability. This has prompted researchers to explore alternative, more sophisticated approaches such as ML, which enables researchers to draw insights from data in a more efficient, accurate and automated way.

ML is a rapidly growing field that has the potential to make significant contributions to the sustainable use and management of marine and coastal environments. This by helping to better understand and predict the impacts of human activities and natural phenomena on coastal ecosystems and identify potential threats. The link between using machine learning to simulate coastal and marine events and sustainability revolves around creating models and taking action. Machine learning employs large amounts of data to create simulations for different scenarios, such as wave propagation or water quality management. These simulations help us fine-tune our actions, like improving wave energy converters or changing shipping paths to avoid pollution. Moreover, these simulations can guide our work towards adapting and mitigating the effects of environmental changes, like coastal erosion caused by rising sea levels. In essence, machine learning offers crucial insights that contribute to improved, sustainable care of our coastal and marine environments.

Typically, the primary input of ML algorithms consists of a data set in various forms such as numeric, image, DEMs collected by Lidar (light detection and ranging), video, and geographic information systems (GIS) data, which are mapped and visualized using GIS. The main output of ML algorithms in coastal engineering can vary depending on the specific application and dataset being used, which includes prediction [e.g., coastal flooding risk (Park and Lee, 2020), storm surge (Sajjad et al., 2020), wave height (Dogan et al., 2021), sediment transport (Pourzangbar et al., 2017b; 2017c; 2017a) and beach erosion (Beuzen et al., 2019)], image processing using satellite imagery data (Agrafiotis et al., 2019) or drone footage (Provost et al., 2020), pattern recognition [e.g., patterns of sediment transport (Liu et al., 2021)], placement optimisation (Cuadra et al., 2016; Sarkar et al., 2016; Neshat et al., 2019), optimization by identifying the most efficient and cost-effective solutions for protecting the coast from erosion and flooding, monitoring (e.g., using sensor data to detect erosion or changes in water quality), anomaly detection (e.g., unusual changes in water quality), and decision making (Lazuardi et al., 2021) by providing decision support to coastal managers and engineers. However, the applicability of ML approaches in coastal engineering is influenced by various factors such as data quality, computational resources, the complexity of the coastal system, and the choice of appropriate algorithms.

Several ML methods have been used to study the sustainable use of coastal areas, including: Artificial Neural Networks (ANNs) used for predictions such as water quality (Chen and Ma, 2010), river classification based on the water quality index (Wong et al., 2021), wave height (Rao and Mandal, 2005; Günaydin, 2008) and beach erosion (Hashemi et al., 2010) and tidal prediction; Decision Trees (DTs) used for classifying the dominating environmental factors; Random Forests (RFs) used for regression and classification tasks, such as predicting the effect of human activities on the coastal environment and water quality index modelling (Sakaa et al., 2022); Support Vector Machines (SVMs) used for solving classification and regression problems, such as identifying the most vulnerable areas in coastal zones; K-Nearest Neighbors (KNN) used for clustering and classification tasks, such as grouping coastal regions based on their sustainability indicators; Ensemble Methods used for improving the accuracy of predictions and classifications, such as predicting the impact of climate change on sustainability of coastal activities, among others.

ML has been widely used in numerous research studies, but there still exists a knowledge gap regarding the selection of parameters, choice of predictive models (be they dynamic or static), domain adaptation, and use of integrated models for analyzing complex systems and evaluating the effects of multiple factors. In relation to data treatment, many existing works have relied on simple heuristic methods or rules of thumb; however, there are more solid mathematical and metaheuristic methods for data preprocessing and parameter identification, highlighted in this paper. Choosing the correct model can be challenging and there is not a definitive method to identify the most suitable ML model for a given problem. In general, the ML approach used to solve a specific issue is selected through a process of trial and error. However, comparing how models perform under different conditions can aid in selecting the most suitable one for a specific issue. To the best of authors’ knowledge, there is not one single paper that offers comprehensive information about the data preprocessing and preparation phase. This paper provides an extensive review of various methodologies employed in coastal engineering to handle datasets. The main focus of this paper is to understand how ML models contribute to the sustainable use and management of marine and coastal environments, rather than the technical intricacies of their setup. The primary goal is to provide a critical review of literature that utilized ML approaches to manage marine phenomena. This review sheds some light on how to prepare parameters and datasets for input into the ML model, the pros and cons of various models, the suitability of ML methods for certain conditions, and their shortcomings and deficiencies.

Although numerous papers have discussed modeling coastal phenomena using experimental, numerical, and mathematical methodologies, the focus of the current paper is exclusively on literature that implemented ML techniques for modeling coastal and marine events. The selected literature spans a broad range of topics from data preprocessing and parameter considerations to different kinds of ML models used for various purposes. Due to the large amount of published papers, the focus of our contribution was directed towards resources published in reputable international journals such as Elsevier, Springer, IWA, Taylor and Francis, Wiley, ASCE, among others. The papers were chosen based on their publication in reputable international journals and were retrieved through online searches using relevant keywords. Among the publications, Coastal Engineering (Elsevier) with 18 papers and Ocean Engineering (Elsevier) with 17 papers, had the most papers in this area. The majority of the sources are fairly recent, predominantly within the past 10 years. Nevertheless, this paper includes some older references that established the groundwork for newer methods. Roughly, fewer than 5% of the literature we reviewed was published before 2000, about 14% between 2000 and 2010, 22% between 2010 and 2015, and over 60% in the last 10 years.

While ML has been implemented in numerous studies, knowledge gaps exist in areas such as parameter selection, choice of models for making predictions (dynamic or static), domain adaptation, and the use of integrated models for modeling complex systems. The emphasis of the paper is on the contribution of ML models to the sustainable use and management of the marine and coastal environment, rather than on the technical details of their configuration.

The paper is structured as follows: Section 2 discusses the key components of data analysis and preprocessing, including data collection and preparation for the modeling process. Section 3 focuses on studies that have applied AI to coastal engineering for sustainable outcomes. The paper also evaluates the accuracy and robustness of the different models in Section 4. Finally, the paper summarizes all the information presented and concludes with a list of references.

2 Data preparation (preprocessing)

Data preparation involves transforming raw data into a format that can be used by ML algorithms for extracting insights or predicting outcomes. This process is vital in ML as it considerably affects the performance of the model (Kelleher et al., 2015). In the event of missing or invalid data, the algorithm either cannot process it or yields less precise, possibly erroneous results. This procedure starts with the acquisition of raw data (refer to Section 2.2), followed by data integration, which entails consolidating data from various sources into a unified dataset. This is succeeded by data cleansing to rectify missing values and outliers (refer to Section 2.4), and then selecting the most pertinent features from the input parameters (feature selection or dimensionality reduction) (see Section 2.5). Subsequently, feature engineering is undertaken, which involves generating new variables from existing parameters using dimensional analysis (DA). Lastly, data transformation is carried out, which involves altering the scale or distribution of variables, such as through data normalization. Figure 1 depicts the multiple phases required for data preprocessing and the methods linked with each step. The upcoming sections provide a detailed explanation of these methods.

FIGURE 1

2.1 Marine data types

The following are the most well-known methods for collecting the data mentioned above: field observations, remote sensing measurements, experimental studies, numerical and mathematical models. Both the availability of equipment and the objective of the study influence the selection of the data collection medium (Prata et al., 2019).

2.2 Marine data resources

Data collection within the realm of marine sciences principally relies on three distinctive methods: in-situ observations, remote sensing techniques, and the use of mathematical and numerical models, as outlined by Verwega et al. (2021). In-situ data collection encompasses ship-based measurements, the deployment of moorings, gliders, autonomous underwater vehicles, drifters and floats, the use of sea-floor optic cables, and laboratory analyses. Field observations remain essential for the collection of real-world data on coastal processes, such as wave heights and tidal levels. In-situ instruments are highly accurate with proper maintenance but may have low-time frequency data for large areas. They offer historical climate trend insights not available from remote sensing and are less affected by atmospheric conditions. These observations serve to validate numerical models that simulate coastal processes and predict the behavior of the coastal system, including wave patterns, tidal currents, and shoreline evolution.

Remote sensing involves acquiring data on coastal topography, bathymetry, and other significant parameters through satellite and airborne platforms. Remote sensing technologies are divided into three categories: satellite, ground-based, and drones (Elsayed et al., 2021). The data thus collected enable the generation of high-resolution coastal environmental maps. Although satellites are powerful tools, they face limitations in obtaining high-resolution regional-scale imagery. Clouds can hinder data capture, and high-resolution imagery can be challenging to interpret (Elsayed et al., 2021). A combination of satellite- and ground-based remote sensing and drones could be effective in future marine engineering evaluations. Economically, combining these tools may be comparable to in-situ techniques in terms of overall cost. Such technology could enable rapid, high-resolution water condition assessments and enhance our understanding of water resource processes. Mathematical and numerical models generate data by simulating real-life systems or processes using mathematical equations and algorithms (Xie and Arkin, 1996). They provide the capability to extend observational data, even to the point of simulating future climate scenarios (Eyring et al., 2016). Nonetheless, it is crucial to understand that these models only approximate real-world scenarios and can encompass spatial and temporal scales that exceed the scope of observational data (Matthes et al., 2020). The outputs from these models are typically available on a unique grid, contingent on the specific simulation. For instance, climate models customarily provide a four-dimensional space-time grid. Consequently, the comparison of model outputs with measurements invariably necessitates interpolation or data aggregation. Table 1 provides a detailed summary of the advantages and disadvantages associated with these diverse data collection methodologies.

2.3 Data cleaning: outlier detection

Several factors can influence the quality of observational data. These include inaccuracies in the instruments, malfunctions of the equipment, disruptions from external sources, mistakes during data conversion, communication mishaps, and significant unforeseen errors (Yu et al., 2022). Such anomalies can pose major threats to operational functionality, downstream operations, system resilience, and cleaner production (Ba-Alawi et al., 2021). Therefore, these should be detected promptly and their data rectified to ensure more realistic measurements.

Anomaly detection methods are generally categorized into various types (see Figure 2) such as Statistical Methods, that utilize the properties of the underlying data distribution to identify anomalies (Chandola et al., 2009); Distance-based Methods, which calculate the distance between data points and identify the outliers based on a certain distance threshold (Ramaswamy et al., 2000); Density-based Methods, which estimate the density of data points and identify outliers as those points that reside in low-density regions (Ester et al., 1996); Machine Learning-based Methods, which employ supervised, unsupervised, or semi-supervised ML algorithms to detect outliers (Pimentel et al., 2014); and Ensemble Methods, which combine multiple outlier detection algorithms to improve the overall performance (Zimek et al., 2012). The choice of method, or combination of methods for better results, depends on the nature of the data and the specific problem being addressed.

FIGURE 2

Mahmoodi and Ghassemi (2018) used outlier detection algorithms to improve wave height predictions, while Oehmcke et al. (2015) demonstrated the effectiveness of ML for identifying significant events in marine long-term data. Daranda and Dzemyda (2020) developed a method combining the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering algorithm and k-nearest neighbors analysis for detecting marine traffic anomalies. These studies highlight the potential of leveraging advanced algorithms and ML in marine data analysis and decision-making. This section aims to provide a survey of contemporary outlier detection techniques, comparing their motivations, advantages, and disadvantages. Outliers can significantly impact the results, which makes addressing or eliminating them before analysis and model development crucial.

Considering the learning algorithm, three main methodologies exist for outlier detection (Hodge and Austin, 2004): 1) unsupervised approach, which uses a learning technique to identify outliers without prior knowledge of the data. The data is treated as a static distribution, and the most distant points are flagged as potential outliers; 2) supervised classification method, which requires pre-labeled data. It allows for online classification, where the classifier continuously learns the model and classifies new data as normal or abnormal, and finally 3) semi-supervised recognition technique, which only learns the normal class, using pre-classified data. It can distinguish new data as normal or novel based on its proximity to the boundary of normality. The choice of an outlier detection method depends on the data type, the number of vectors and attributes, speed and accuracy requirements, and the ability to accurately identify outliers. The key factors in choosing a method are selecting an algorithm that can handle the data and defining a suitable neighborhood for the outlier.

2.4 Dimensionality reduction

Incorporating parameters that are not relevant can result in intricate models that pose significant challenges in interpretation and execution compared to the models developed using the most crucial parameters (Pourzangbar, 2012). That is the reason why the focus is placed on building ML models using the most crucial parameters. These parameters are not only essential for the model’s output, but also are unconnected with other input parameters. To derive the most important dimensions (parameters) in the input space, there are several methods including min/max autocorrelation factor analysis (MAFA), dynamic factor analysis (DFA), Least Absolute Shrinkage and Selection Operator (LASSO), Independent Component Analysis (ICA), multicollinearity test and PCA. Table 2 summarizes some famous dimensionality reduction approaches used in marine engineering. The latter two methods are explained below.

2.4.1 Multicollinearity

It is important to note that none of these tests can definitively prove the presence of multicollinearity, but rather provide evidence that it may be present in a model. Therefore, it is important to use multiple tests and to interpret the results in the context of the specific research question and data being analyzed.

2.4.2 Principle component analysis

PCA can be utilized for dimensionality reduction (Pearson, 1901). PCA reduces the dimensions of datasets in a way that their interpretability increases. To achieve this, PCA maximizes the variance of datasets by mapping them in a new coordinate (new uncorrelated variables). The most correlated parameters are deleted while information loss is minimum. The initially proposed method was limited to up to three parameters; however, Harold Hotelling has described methods for computing multivariate PCA since 1933 (Hotelling, 1933).

In the mathematical description, it is assumed that the input environment contains parameters and measurements for each parameter. Hence, the input matrix has components. The input environment can be transformed into a feature environment whose dimensions are not dependent on each other. Accordingly, the feature environment can be represented by a matrix, i.e., . The transformation can be done using a whitening or sphering transformation matrix () as follows:

The primary goal of PCA is to identify the components of the transformation matrix in such a way that the new variables exhibit maximum discrepancy (represented by variance). With some mathematical manipulation, the following equation for the transformation matrix can be derived:where is the covariance matrix of the input environment (), is a diagonal matrix whose components are the eigenvalues () of the matrix , and is a matrix that its components are the eigenvectors of .

2.5 Dimensional analysis

Although numerous methods exist for DA, the majority of studies employ the Buckingham π Theorem to render the parameters dimensionless. Table 3 summarizes some of the studies used DA before feeding their ML models.

2.6 Normalization

Normalizing data helps to ensure comparability by transforming it into a common scale, avoiding bias in statistical analyses and allowing for accurate and meaningful results by removing the impact of unit differences, especially when comparing data from different sources. Normalization plays a crucial role in efficient machine and deep learning by ensuring that large numerical inputs are processed effectively (Van Komen et al., 2022). The choice of normalization method depends on the specific requirements of the data and the problem being solved. Some of the famous methods for data normalization are summarized in Table 4.

In Table 4, the transformed data, referred to as , is obtained by normalizing the original data () in a new range. The original data is contained within a vector, denoted as , and its minimum and maximum values are represented as and , respectively. The chosen minimum and maximum values for the transformed range are and , which are typically set to zero and one, respectively. is the mean of the data, and is the standard deviation of the data.

Min-Max normalization is a technique used to rescale a feature to a specific range, usually between 0 and 1. However, to avoid having zero data in the model, an alternative approach is to expand the range to include values between 0.1 and 0.9. It is a commonly used method for transforming variables so that they are comparable, as it scales the data linearly to a specific range. Through this normalization process, the values in are transformed such that the minimum value of is mapped to 0, the maximum value to 1, and intermediate values are mapped to corresponding values between 0 and 1. The Z-score normalization, also known as standardization, is a method of transforming data to a standard normal distribution with a mean of 0 and a standard deviation of 1. This normalization process rescales the data and centers it around the mean, allowing for easier comparison of values. It is commonly used in various fields, such as statistics, ML and data analysis. Sigmoid normalization uses a sigmoid function to transform the data, proving useful in instances where the data distribution is asymmetrical. The sigmoid function maps any input value to a value between 0 and 1 and it is commonly used in ML and ANN models to represent a probability or to rescale data. Additionally, the sigmoid function is differentiable, which makes it useful in optimization problems and backpropagation in neural networks.

In coastal phenomena, the relationship between inputs and outputs typically displays nonlinearity, but certain models, such as the M5 model tree, are unable to handle nonlinearity. To address this limitation, M5 models have been implemented using a logarithmic form for both inputs and outputs (i.e., the natural logarithm of inputs and outputs). This logarithmic form is more accurate than a linear formulation because it better captures the nonlinear nature of the contributing parameters (Pourzangbar et al., 2017a; Afsarian et al., 2018). Log scaling entails transforming data points through the application of a logarithmic function. The logarithm maps large values to smaller ones and vice versa, helping to make skewed data more symmetrical and manageable for analysis. The selection of a specific logarithmic function depends on the needs of the data and the analysis to be performed, such as log base 10, log base 2, or natural logarithm. Despite its advantages, normalization may result in a loss of interpretability, increased sensitivity to outliers (as seen in techniques like min-max scaling and z-score), loss of information, dependence on the entire dataset, impacts on categorical features, and varying sensitivity across algorithms.

3 AI learning algorithms and their application in marine/coastal engineering

3.1 Supervised-based ML methods

Supervised ML presents a powerful approach, necessitating labeled data for model training. Its versatility permits its usage across a variety of applications, such as image and speech recognition, natural language processing, and predictive analytics. Common algorithms used in supervised learning encompass linear regression, logistic regression (LR), decision trees, random forests, support vector machines, and neural networks. A key advantage of supervised learning is its capacity to generate precise predictions for novel and unseen data (Jiang et al., 2020). However, it also has certain drawbacks, including the requirement for labeled data, the quality and quantity of the training data, and the potential for overfitting. Despite these challenges, supervised learning is seen as an essential method in ML and data science, demonstrating high accuracy and less computational time compared to physical models. Despite the inherent complexity of marine processes, supervised-based ML models have demonstrated benefits in understanding coastal phenomena, thereby finding extensive application in coastal engineering to drive innovative models and solve intricate problems (as summarized in Table 5). Supervised ML models have been employed to predict wave parameters like significant wave height and period, wave reflection and transmission coefficients (van Gent et al., 2007; Gandomi et al., 2020; Kuntoji et al., 2020), tide levels (Lee, 2004), ocean currents and wind files (James et al., 2018; Shamshirband et al., 2020), prediction of wind Characteristics under future Climate Change scenarios (Yeganeh-Bakhtiary et al., 2022), flood inundation using Gaussian process model (Donnelly et al., 2022) and breakwater stability number and wave overtopping discharge, among others. Various ML models, such as ANN and SVM, can be employed to do these predictions. ML models have also found application in morphological and morphodynamic predictions, including profile elevation, area, and length, based on parameters like wind speed, direction, wave height, and beach angle (Hashemi et al., 2010).

3.2 Unsupervised-based ML methods

Unsupervised learning is a form of ML that functions without predefined labels or target outcomes (Bishop and Nasrabadi, 2006). Its main purpose is to independently discover patterns, structures, and relationships in data. Common applications include clustering, anomaly detection, and dimensionality reduction. Clustering groups similar data points, anomaly detection spotlights unusual patterns (as detailed in Section 2.3), and dimensionality reduction simplifies the number of features while preserving essential information (as seen in Section 2.4). Algorithms like k-means clustering, hierarchical clustering, PCA, and autoencoders are frequently used in unsupervised learning to identify patterns in data. While unsupervised learning can pose challenges due to the lack of a distinct optimization goal, it still holds a vital position in ML, contributing to advancements in fields such as computer vision, natural language processing, and recommendation systems. In the context of coastal engineering, k-means clustering can be used to classify centroid values for data like the maximum oceanic wind. Average centroid clustering can be obtained from both the previously chosen values and the currently selected clustering data (Baboo and Tajudin, 2013). PCA can be employed in coastal engineering to examine correlation matrices (Roseman et al., 2005) and pinpoint major changes in beach profiles and sand grain distributions (Tsujimoto et al., 2012). Moreover, PCA and hierarchical clustering can help characterize coastal plane shape and hydrodynamics. For instance, the form of arc-shaped coasts, largely influenced by geological structure, can be divided into four broad categories that reflect actual conditions using clustering (Scott et al., 2011). By identifying key data components, PCA can aid in elucidating the underlying patterns and structures of the data.

3.3 Reinforcement-based ML methods

Reinforcement learning (RL) is a type of ML where a program, known as an agent, learns to perform tasks by getting feedback from its environment in the form of rewards or penalties (Rengarajan et al., 2022). The agent executes a series of decisions in a mutable environment, aiming to learn the optimal way (or policy) to maximize rewards over time. This process is typically structured as a Markov Decision Process (MDP), encompassing states, actions, transition functions, and reward functions. There are two main types of reinforcement learning algorithms: model-based and model-free. Model-based RL is like making a map to understand the surroundings. On the other hand, model-free RL does not make a map; it just figures out what to do based on where it is at the moment. So, model-based RL is more about planning ahead, while model-free RL is more about learning on the go (Plaat et al., 2023). Model-free methods, like Q-learning, do not need a model of the environment and calculate the expected total of future rewards for each possible action at each state using the so-called the Bellman equation. Q-learning has been used successfully in many different tasks, which is why it is one of the most commonly used model-free RL algorithms. In coastal engineering, RL can be used to develop control policies to reduce the risk of flooding (Bowes et al., 2021). Deep reinforcement learning (DRL), an advanced form of RL, can be used to control devices that convert wave energy, and has been found to work better than traditional control methods (Anderlini et al., 2020). DRL can also adjust itself to changes in system dynamics, allowing for control even when faults occur. Moreover, RL has been used to maximize the electricity produced by wave energy converters (Zou et al., 2022). In addition, a type of RL called multiagent reinforcement learning can simulate the social and economic effects of sea level rise. This can be a useful tool for planning scenarios, analyzing costs and benefits, and optimizing strategies to adapt to changes (Shuvo et al., 2022).

Table 4 summarizes the various ML learning approaches and their corresponding model types. Each model type utilizes a unique set of algorithms. For instance, in the case of classification tasks, ANNs or SVMs may be utilized. The final column of the table highlights the research studies focusing on each specific learning approach, targeting the investigation of a specific coastal process or event.

3.4 AI contribution to the sustainability of marine environments

Predictive models, such as statistical, numerical, or ML models, play a vital role in marine and coastal engineering to safeguard structures from natural forces. Statistical models use past data to forecast future conditions, while numerical models simulate the event using mathematical equations and formulas. ML models, using artificial intelligence (AI), learn from past data for prediction purposes. Each model has its unique approach and is chosen based on data availability and specific project needs.

Various ML techniques have been implemented in the study of coastal and marine environments. Figure 3, sourced from Scopus, provides a visual representation of the percentage of published papers that used different ML methods since the year 2000. Upon reviewing this figure, it is evident that Principal Components Regression (PCR), Linear Model (LM), Regression Tree (RT), and ANN are the most frequently employed ML algorithms for analyzing coastal and marine phenomena. However, certain ML techniques, such as General Regression Neural Networks (GRNN), M5 model tree, Bayesian Model Averaging method (BMA), Generalized Boosted Regression (GBM), and Extreme Gradient Lift (Xgboost) have been applied less frequently in the investigation of coastal and marine events.

FIGURE 3

Figure 4 shows the application trend of different ML approaches for coastal and marine phenomena. Previously, techniques such as Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), and RF were sparingly employed in diverse studies. However, there has been a significant increase in their use over the past 5 years, demonstrating a growing reliance on these methods in recent research.

FIGURE 4

Figure 5 illustrates the trend of various ML algorithms since 2008 in coastal and marine applications. Some methods are not frequently used which are colored in red (these low-important approaches are not reported in Figure 4).

FIGURE 5

3.4.1 Prediction of oceanographic and morphologic parameters

Researchers use ML algorithms and soft computing techniques to predict oceanographic and morphological parameters, as shown in Figure 6. These methods include ANNs, SVMs, Support Vector Regression (SVR), Fuzzy Logic (FL), evolutionary algorithms, such as Genetic Programming (GP) and DTs, among others. Predictive models are widely used in oceanography and coastal management. Their accuracy critically depends on several factors. These include the dataset used for training, the type and configuration of the ML model, tuning parameters, termination condition, and input and output parameters. It is important to note that specific algorithms, with carefully adjusted parameters, are particularly valuable in various research endeavors, depending on the problem being addressed.

FIGURE 6

ANNs, SVRs, M5 decision tree algorithm, and Recurrent Neural Networks (RNNs) including Long-Short-Term Memory (LSTM) models are used to predict wave heights, as per studies by Duong et al. (2023) and Rizianiza and Aisjah (2015). These ML techniques have shown reliable wave prediction capabilities, maintaining accuracy up to 72 h ahead (Jain and Deo, 2008). The use of intact structural data for predicting significant wave heights has been explored, with emphasis on the critical role of data quality in training ANNs for wave height predictions (Ciortan and Rusu, 2018; Demetriou et al., 2021). ANNs have also been implemented to estimate wave breaking heights considering various factors like seabed slope, water depth, and deep-sea wavelength (Duong et al., 2023). In the field of marine energy forecasting, researchers have used multi-class classification methods with ordinal classifiers, such as SVOREX and SVORIM yielding precise results (Fernández et al., 2015). RNN, especially LSTM models, have been employed to predict motion responses in irregular wave patterns (Kagemoto, 2020). Table 6 provides a summary of the top 10 highly-cited papers focused on predicting significant wave height using ML algorithms. The majority of these studies used meteorological data and past wave height as input parameters. The results demonstrate that LSTM neural networks, ANN, kernel-based predictors like SVM and SVR, as well as decision trees, are capable of accurately predicting wave height.

FIGURE 7

FIGURE 8

The M5 decision tree algorithm, ANNs, and gradient boosting decision trees serve as robust tools for predicting wave overtopping discharge on coastal infrastructure such as breakwaters. When focusing on wave overtopping and runup, the M5 decision tree algorithm exhibits promising capabilities for predicting runup waves, taking into account laboratory data and multiple parameters (Abolfathi et al., 2016). ANNs are also used to predict wave reflection and transmission coefficients (Zanuttigh et al., 2016; Formentin et al., 2017). It has been proven that gradient boosting decision trees, as a novel ML technique, has improved the accuracy of predicting average wave overtopping discharges by nearly threefold in comparison to traditional neural networks (den Bieman et al., 2020). Kernel-based approaches, such as Gaussian Process Regression (GPR) and SVR, have also been utilized in predicting wave overtopping, with GPR showing superior performance over ANNs and empirical formulas (Hosseinzadeh et al., 2021).

The measurement of Sea Surface Temperature (SST) is vital for understanding the global climate. It significantly contributes to climate modeling, weather forecasting, and studies on marine ecosystems. Accurately predicting SST can aid in mitigating the environmental harm resulting from rising water temperatures due to human-induced climate change. This prediction not only benefits marine ecosystems but also preserves coastal economies and the broader coastal environment (Choi et al., 2023). LSTM neural networks have proven effective in forecasting SST, showing enhanced performances when the right amount of input data is used (Xu et al., 2020). Multivariate LSTM models, which take into account factors such as wind speed and sea-level air pressure alongside SST, have demonstrated superior results compared to univariate models that only factor in SST (Balogun and Adebisi, 2021). Traditional ML models have been studied for spatio-temporal time series prediction, highlighting the importance of spatial data. Among these, the LSTM model emerged as the most efficient, showing a 25% improvement in forecasting performance (based on RMSE) when spatial information was incorporated (Kartal, 2023). Research indicates that LSTMs, whether using single or multiple variables, surpass other ML models in predicting SST (Xu et al., 2020; Kartal, 2023).

Moreover, accurate predictions of coastal sediment transport are crucial for managing coastal erosion and development, with researchers traditionally estimating sediment transport using experimental methods. Artificial intelligence-based methods potentially improve decision-making for managing coastal erosion and development (Bakhtyar et al., 2008; Kabiri-Samani et al., 2011), given the importance of selecting valid input data and appropriate activation functions (Pourzangbar, 2012; Yeganeh-bakhtiary et al., 2012). Artificial intelligence and ML methods, such as Adaptive Network Based Fuzzy Inference Systems (ANFIS), Fuzzy Inference System (FIS), CERC (Coastal Engineering Research Center), Walton-Bruno (WB), Van Ridge (VR), and ANNs, have been employed to model sediment transport, with ANFIS showing higher accuracy and reliability for estimating longshore sediment transport rates (LSTR) (Bakhtyar et al., 2008; Hashemi et al., 2010). SVR has also been employed, demonstrating superiority over neural networks when the dataset is small or the relationships are linear or non-linear but with a clear margin (Dezvareh and Shafaghat, 2020). Deep learning models, like ANNs, have been developed to address the shortcomings of numerical models in analyzing simultaneous sand and sediment transport (Kim and Aoki, 2021).

3.4.2 Classification models

KNN classifier has been used in various marine-related projects. For the design of marine hydrokinetic turbines, KNN was used to identify and categorize the severity of the rotor blade pitch imbalance encountered by marine current turbines. This approach was found useful for fault detection and severity classification (Freeman et al., 2021). In ocean surface current forecasting, KNN was used as an alternate method (Jirakittayakorn et al., 2017). The KNN algorithm proved capable of forecasting future surface currents up to 24 h in advance. The KNN approach was compared with other prediction techniques such as ARIMA, exponential smoothing, and LSTM, and it was found that the KNN model had the highest accuracy. KNN was one of the six ML classifiers used to generate precise geographic estimates of seabed substrate and seabed habitat mapping (Diesing and Stephens, 2015; Leon et al., 2020). The accuracy of the predictions was evaluated using ground-truth sample data segmented into classes of seabed substrate. In coastal hazards projection, KNN was used to project dangers using several representative concentration route climate change scenarios, regional climate models, and sea level rise ratios (Park and Lee, 2020). Seafloor classification is another marine-related project where KNN was used along with ANN to class the structure of the seafloor and to pinpoint potential anthropogenic effects on delicate benthic assemblages (Gauci et al., 2016). Finally, in sea-land segmentation, KNN was used to produce a pixel-level, sea-land segmentation of the scene based on the Doppler bandwidth of a returns vector in maritime surveillance radars (Shui et al., 2020).

The Naive Bayes classifier is a machine learning algorithm commonly utilized in various applications to enhance model accuracy. A prominent application of the Naive Bayes classifier involves predicting water quality classes utilizing seven popular Water Quality Index (WQI) models (Uddin et al., 2023). There is some confusion about the proper classification of water quality due to differing techniques used in current WQI models. To address this, the Naive Bayes was compared with other ML classifiers. These included SVM, Random Forest, K-Nearest Neighbor, and Gradient Boosting. The goal was to determine the best classifier for evaluating water quality. Another application of the Naive Bayes classifier is in detecting small-scale assemblages of drifting vegetation and beach cast in Germany’s Baltic coast (Uhl et al., 2022). To obtain the best classification results, the classifier was used as part of an ensemble of five classifiers, including a RF, CART, SVM, and stochastic gradient boosting classifier to predict tropical Cyclone based on multi-model fusion across Indian coastal region (Varalakshmi et al., 2021). In all applications, the Naive Bayes classifier was effective in improving the accuracy of the models, particularly in predicting the quality of coastal water and detecting small-scale assemblages of drifting vegetation and beach cast. Its versatility and usefulness in different domains make it a popular choice for improving the accuracy of models in various applications.

Given coastaline extraction from satellite images, three well known methods including image processing techniques, unsupervised classifiers and supervised classifiers have been implemented. Shenbagaraj et al. (2014) employed visual interpretation and ISODATA (Iterative Self-Organizing Data Analysis Technique) classification techniques to extract shorelines from Landsat Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+) sensor images, Toposheet and Google Earth Images spanning a 60-year period from 1953 to 2013 between Kolachel and Kayalpattanam. This approach effectively identified the areas of coastline transgression and regression in the study area. Supervised classifiers such as Maximum Likelihood (Rokni et al., 2015), SVM & ANN & EL (Çelik and Gazioğlu, 2022), RF (Bayram et al., 2017), Minimum-Distance-to-Means, and Mahalanobis distance (Sekovski et al., 2014) also have been employed to classify and detect the coastline position based on the satellite images. As depicted in Figure 9, the average median distance of all shorelines, observed in relation to the reference, suggests that the shorelines produced by the ISODATA and Mahalanobis methods demonstrate the best alignment, with a discrepancy of 2.2 m, thereby being closer to the reference than other methods. Conversely, the Parallelepiped and Maximum Likelihood methods resulted in shorelines with the highest average median distance from the reference shoreline, measuring 5.1 m and 5.6 m respectively.

FIGURE 9

4 Summary and conclusion

This study provides a comprehensive review of machine learning applications to model the marine and coastal environments, with comprehensive coverage from data preprocessing to the application of different models. The review indicates that appropriately implemented and optimized ML methods can significantly contribute to marine and coastal sustainability through developing accurate and robust models for prediction of wave height, oceanographic parameters, and sediment transport, image processing, optimization of coastal and marine structures design.

4.1 Recommendations for future research endeavours

References

• 1

  AbolfathiS.Yeganeh-BakhtiaryA.Hamze-ZiabariS. M.BorzooeiS. (2016). Wave runup prediction using M5′ model tree algorithm. Ocean. Eng.112, 76–81. 10.1016/J.OCEANENG.2015.12.016

  • CrossRef
  • Google Scholar

• 2

  AdamE.MutangaO.OdindiJ.Abdel-RahmanE. M. (2014). Land-use/cover classification in a heterogeneous coastal landscape using RapidEye imagery: evaluating the performance of random forest and support vector machines classifiers. Int. J. Remote Sens.35, 3440–3458. 10.1080/01431161.2014.903435

  • CrossRef
  • Google Scholar

• 3

  AfsarianF.SaberA.PourzangbarA.OlabiA. G.KhanmohammadiM. A. (2018). Analysis of recycled aggregates effect on energy conservation using M5″ model tree algorithm. Energy156, 264–277. 10.1016/j.energy.2018.05.099

  • CrossRef
  • Google Scholar

• 4

  AgarwalP.ManuelL. (2008). Extreme loads for an offshore wind turbine using statistical extrapolation from limited field data. Wind Energy11, 673–684. 10.1002/we.301

  • CrossRef
  • Google Scholar

• 5

  AgrafiotisP.SkarlatosD.GeorgopoulosA.KarantzalosK. (2019). DepthLearn: learning to correct the refraction on point clouds derived from aerial imagery for accurate dense shallow water bathymetry based on SVMs-fusion with LiDAR point clouds. Remote Sens.11, 2225. 10.3390/rs11192225

  • CrossRef
  • Google Scholar

• 6

  AkbarifardS.RadmaneshF. (2018). Predicting sea wave height using Symbiotic Organisms Search (SOS) algorithm. Ocean. Eng.167, 348–356. 10.1016/J.OCEANENG.2018.04.092

  • CrossRef
  • Google Scholar

• 7

  AliM.PrasadR. (2019). Significant wave height forecasting via an extreme learning machine model integrated with improved complete ensemble empirical mode decomposition. Renew. Sustain. Energy Rev.104, 281–295. 10.1016/J.RSER.2019.01.014

  • CrossRef
  • Google Scholar

• 8

  AliM.PrasadR.XiangY.JameiM.YaseenZ. M. (2023). Ensemble robust local mean decomposition integrated with random forest for short-term significant wave height forecasting. Renew. Energy205, 731–746. 10.1016/J.RENENE.2023.01.108

  • CrossRef
  • Google Scholar

• 9

  AltmanN. S. (1992). An introduction to kernel and nearest-neighbor nonparametric regression. Am. Statistician46, 175–185. 10.1080/00031305.1992.10475879

  • CrossRef
  • Google Scholar

• 10

  AnderliniE.HusainS.ParkerG. G.AbusaraM.ThomasG. (2020). Towards real-time reinforcement learning control of a wave energy converter. J. Mar. Sci. Eng.8, 845. 10.3390/jmse8110845

  • CrossRef
  • Google Scholar

• 11

  ArslanO.AkyürekÖ.KayaŞ.ŞekerD. Z. (2020). Dimension reduction methods applied to coastline extraction on hyperspectral imagery. Geocarto Int.35, 376–390. 10.1080/10106049.2018.1520920

  • CrossRef
  • Google Scholar

• 12

  Ba-AlawiA. H.VilelaP.Loy-BenitezJ.HeoS. K.YooC. K. (2021). Intelligent sensor validation for sustainable influent quality monitoring in wastewater treatment plants using stacked denoising autoencoders. J. Water Process Eng.43, 102206. 10.1016/j.jwpe.2021.102206

  • CrossRef
  • Google Scholar

• 13

  BabooS. S.TajudinK. (2013). “Clustering centroid finding algorithm (CCFA) using spatial temporal data mining concept,” in 2013 International Conference on Pattern Recognition, Informatics and Mobile Engineering 2013, Salem, India, February 21–22, 2013 (IEEE (Institute of Electrical and Electronics Engineers), 30–36. 10.1109/ICPRIME.2013.6496443

  • CrossRef
  • Google Scholar

• 14

  BaiY.CaiW. J.HeX.ZhaiW.PanD.DaiM.et al (2015). A mechanistic semi-analytical method for remotely sensing Sea Surface pCO2 in river-dominated coastal oceans: a case study from the east China sea. J. Geophys. Res. Oceans120, 2331–2349. 10.1002/2014JC010632

  • CrossRef
  • Google Scholar

• 15

  BakhtyarR.GhaheriA.Yeganeh-BakhtiaryA.BaldockT. E. (2008). Longshore sediment transport estimation using a fuzzy inference system. Appl. Ocean Res.30, 273–286. 10.1016/J.APOR.2008.12.001

  • CrossRef
  • Google Scholar

• 16

  BalogunA. L.AdebisiN. (2021). Sea level prediction using ARIMA, SVR and LSTM neural network: assessing the impact of ensemble ocean-atmospheric processes on models’ accuracy. Geomatics, Nat. Hazards Risk12, 653–674. 10.1080/19475705.2021.1887372

  • CrossRef
  • Google Scholar

• 17

  BateniS. M.BorgheiS. M.JengD. S. (2007). Neural network and neuro-fuzzy assessments for scour depth around bridge piers. Eng. Appl. Artif. Intell.20, 401–414. 10.1016/j.engappai.2006.06.012

  • CrossRef
  • Google Scholar

• 18

  BayramB.ErdemF.AkpinarB.InceA. K.BozkurtS.Catal ReisH.et al (2017). The efficiency of random forest method for shoreline extraction from landsat-8 and gokturk-2 imageries. ISPRS Ann. Photogrammetry, Remote Sens. Spatial Inf. Sci., 141–145. 10.5194/isprs-annals-IV-4-W4-141-2017

  • CrossRef
  • Google Scholar

• 19

  BelgiuM.DrăguţL. (2016). Random forest in remote sensing: a review of applications and future directions. ISPRS J. Photogrammetry Remote Sens.114, 24–31. 10.1016/J.ISPRSJPRS.2016.01.011

  • CrossRef
  • Google Scholar

• 20

  BerbićJ.OcvirkE.CarevićD.LončarG. (2017). Application of neural networks and support vector machine for significant wave height prediction. Oceanologia59, 331–349. 10.1016/J.OCEANO.2017.03.007

  • CrossRef
  • Google Scholar

• 21

  BeuzenT.GoldsteinE. B.SplinterK. D. (2019). Ensemble models from machine learning: an example of wave runup and coastal dune erosion. Nat. Hazards Earth Syst. Sci.19, 2295–2309. 10.5194/nhess-19-2295-2019

  • CrossRef
  • Google Scholar

• 22

  BishopC. M.NasrabadiN. M. (2006). Pattern recognition and machine learning. Springer.

  • Google Scholar

• 23

  BooijN.HolthuijsenL. H.RisR. C. (1997). The swan wave model for shallow water. Coast. Eng. 1996. 10.1061/9780784402429.053

  • CrossRef
  • Google Scholar

• 24

  BowesB. D.TavakoliA.WangC.HeydarianA.BehlM.BelingP. A.et al (2021). Flood mitigation in coastal urban catchments using real-time stormwater infrastructure control and reinforcement learning. J. Hydroinformatics23, 529–547. 10.2166/HYDRO.2020.080

  • CrossRef
  • Google Scholar

• 25

  BowesB. D.WangC.ErcanM. B.CulverT. B.BelingP. A.GoodallJ. L. (2022). Reinforcement learning-based real-time control of coastal urban stormwater systems to mitigate flooding and improve water quality. Environ. Sci. Water Res. Technol.8, 2065–2086. 10.1039/d1ew00582k

  • CrossRef
  • Google Scholar

• 26

  BreimanL. (2001). Random forests. Mach. Learn.45, 5–32. 10.1023/A:1010933404324

  • CrossRef
  • Google Scholar

• 27

  BrownJ. M.YellandM. J.PullenT.SilvaE.MartinA.GoldI.et al (2021). Novel use of social media to assess and improve coastal flood forecasts and hazard alerts. Sci. Rep.11, 13727. 10.1038/s41598-021-93077-z

  • CrossRef
  • Google Scholar

• 28

  ÇelikO. İ.GazioğluC. (2022). Coast type based accuracy assessment for coastline extraction from satellite image with machine learning classifiers. Egypt. J. Remote Sens. Space Sci.25, 289–299. 10.1016/J.EJRS.2022.01.010

  • CrossRef
  • Google Scholar

• 29

  ChandolaV.BanerjeeA.KumarV. (2009). Anomaly detection: a survey. ACM Comput. Surv.41, 1–58. 10.1145/1541880.1541882

  • CrossRef
  • Google Scholar

• 30

  ChenJ.AbbadyS.DuggimpudiM. B. (2016). Spatiotemporal outlier detection: did buoys tell where the hurricanes were?Pap. Appl. Geogr.2, 298–314. 10.1080/23754931.2016.1149874

  • CrossRef
  • Google Scholar

• 31

  ChenW.MaW. (2010). “Applications based on genetic neural network model of Lianyungang marine water quality optimization techniques and algorithms Technology”. 2010 International Conference of Information Science and Management Engineering 20101, 526–529. 10.1109/ISME.2010.253

  • CrossRef
  • Google Scholar

• 32

  ChoH. Y.OhJ. H.KimK. O.ShimJ. S. (2013). Outlier detection and missing data filling methods for coastal water temperature data. J. Coast. Res.165, 1898–1903. 10.2112/si65-321.1

  • CrossRef
  • Google Scholar

• 33

  ChoiH. M.KimM. K.YangH. (2023). Deep-learning model for sea surface temperature prediction near the Korean Peninsula. Deep Sea Res. Part II Top. Stud. Oceanogr.208, 105262. 10.1016/J.DSR2.2023.105262

  • CrossRef
  • Google Scholar

• 34

  CiortanS.RusuE. (2018). Prediction of the wave power in the Black Sea based on wind speed using artificial neural networks. E3S Web Conf.51, 01006. 10.1051/e3scconf/20185101006

  • CrossRef
  • Google Scholar

• 35

  Copernicus (2023). The copernicus marine service. Available at: https://marine.copernicus.eu/about.

  • Google Scholar

• 36

  Cornejo-BuenoL.Nieto-BorgeJ. C.García-DíazP.RodríguezG.Salcedo-SanzS. (2016). Significant wave height and energy flux prediction for marine energy applications: a grouping genetic algorithm – extreme learning machine approach. Renew. Energy97, 380–389. 10.1016/J.RENENE.2016.05.094

  • CrossRef
  • Google Scholar

• 37

  CortesC.VapnikV. (1995). Support-vector networks. Mach. Learn.20, 273–297. 10.1023/A:1022627411411

  • CrossRef
  • Google Scholar

• 38

  CuadraL.Salcedo-SanzS.Nieto-BorgeJ. C.AlexandreE.RodríguezG. (2016). Computational intelligence in wave energy: comprehensive review and case study. Renew. Sustain. Energy Rev.58, 1223–1246. 10.1016/j.rser.2015.12.253

  • CrossRef
  • Google Scholar

• 39

  DarandaA.DzemydaG. (2020). Navigation decision support: discover of vessel traffic anomaly according to the historic marine data. Int. J. Comput. Commun. CONTROL15. 10.15837/IJCCC.2020.3.3864

  • CrossRef
  • Google Scholar

• 40

  DavidsonM. A.BirdP. A. D.BullockG. N.HuntleyD. A. (1996). A new non-dimensional number for the analysis of wave reflection from rubble mound breakwaters. Coast. Eng.28, 93–120. 10.1016/0378-3839(96)00012-9

  • CrossRef
  • Google Scholar

• 41

  DemetriouD.MichailidesC.PapanastasiouG.OnoufriouT. (2021). Coastal zone significant wave height prediction by supervised machine learning classification algorithms. Ocean. Eng.221, 108592. 10.1016/j.oceaneng.2021.108592

  • CrossRef
  • Google Scholar

• 42

  den BiemanJ. P.WilmsJ. M.van den BoogaardH. F. P.van GentM. R. A. (2020). Prediction of mean wave overtopping discharge using gradient boosting decision trees. Water12. 10.3390/W12061703

  • CrossRef
  • Google Scholar

• 43

  DengY.YangJ.ZhaoW.LiX.XiaoL. (2016). Freak wave forces on a vertical cylinder. Coast. Eng.114, 9–18. 10.1016/j.coastaleng.2016.03.007

  • CrossRef
  • Google Scholar

• 44

  DeroliyaP.GhoshM.MohantyM. P.GhoshS.RaoK. D.KarmakarS. (2022). A novel flood risk mapping approach with machine learning considering geomorphic and socio-economic vulnerability dimensions. Sci. Total Environ.851, 158002. 10.1016/j.scitotenv.2022.158002

  • CrossRef
  • Google Scholar

• 45

  DezvarehR.ShafaghatM. (2020). Predicting the sediment rate of Nakhilo Port using artificial intelligence. Int. J. Coast. offshore Eng.4 (2), 41–49. 10.22034/IJCOE.2020.149345

  • CrossRef
  • Google Scholar

• 46

  DiZ.ChangM.GuoP.LiY.ChangY. (2019). Using real-time data and unsupervised machine learning techniques to study large-scale spatio-temporal characteristics of wastewater discharges and their influence on surface water quality in the Yangtze River Basin. WaterSwitzerl.11, 1268. 10.3390/w11061268

  • CrossRef
  • Google Scholar

• 47

  DiesingM.StephensD. (2015). A multi-model ensemble approach to seabed mapping. J. Sea Res.100, 62–69. 10.1016/j.seares.2014.10.013

  • CrossRef
  • Google Scholar

• 48

  DoganG.FordM.JamesS. (2021). “Predicting ocean-wave conditions using buoy data supplied to a hybrid RNN-LSTM neural network and machine learning models,” in Proceedings of the 2021 IEEE International Conference on Machine Learning and Applied Network Technologies, Soyapango, El Salvador, December 16–17, 2021 (IEEE (Institute of Electrical and Electronics Engineers)). ICMLANT 2021. 10.1109/ICMLANT53170.2021.9690528

  • CrossRef
  • Google Scholar

• 49

  DonnellyJ.AbolfathiS.PearsonJ.ChatrabgounO.DaneshkhahA. (2022). Gaussian process emulation of spatio-temporal outputs of a 2D inland flood model. Water Res.225, 119100. 10.1016/j.watres.2022.119100

  • CrossRef
  • Google Scholar

• 50

  DuongN. T.TranK. Q.LuuL. X.TranL. H. (2023). Prediction of breaking wave height by using artificial neural network-based approach. Ocean. Model.182, 102177. 10.1016/J.OCEMOD.2023.102177

  • CrossRef
  • Google Scholar

• 51

  El-HaddadB. A.YoussefA. M.PourghasemiH. R.PradhanB.El-ShaterA. H.El-KhashabM. H. (2021). Flood susceptibility prediction using four machine learning techniques and comparison of their performance at Wadi Qena Basin, Egypt. Nat. Hazards105, 83–114. 10.1007/s11069-020-04296-y

  • CrossRef
  • Google Scholar

• 52

  El-RahmanS. A. (2016). “Hyperspectral imaging classification using ISODATA algorithm: big data challenge”. in Proceedings - 2015 5th International Conference on e-Learning, Manama, Bahrain, October 18–20, 2015 (IEEE (Institute of Electrical and Electronics Engineers)). 10.1109/ECONF.2015.39

  • CrossRef
  • Google Scholar

• 53

  ElsayedS.IbrahimH.HusseinH.ElsherbinyO.ElmetwalliA. H.MoghanmF. S.et al (2021). Assessment of water quality in lake qaroun using ground-based remote sensing data and artificial neural networks. Water13, 3094. 10.3390/w13213094

  • CrossRef
  • Google Scholar

• 54

  EmmanouilS.AguilarS. G.NaneG. F.SchoutenJ. J. (2020). Statistical models for improving significant wave height predictions in offshore operations. Ocean. Eng.206, 107249. 10.1016/j.oceaneng.2020.107249

  • CrossRef
  • Google Scholar

• 55

  EnnoualiZ.FannassiY.LahssiniG.BenmohammadiA.MasriaA. (2023). Mapping coastal vulnerability using machine learning algorithms: a case study at north coastline of sebou estuary, Morocco. Regional Stud. Mar. Sci.60, 102829. 10.1016/J.RSMA.2023.102829

  • CrossRef
  • Google Scholar

• 56

  EsterM.KriegelH.-P.SanderJ.XuX. (1996). “A density-based algorithm for discovering clusters in large spatial databases with Noise,” in Proceedings of the 2nd International Conference on Knowledge Discovery and Data Mining.

  • Google Scholar

• 57

  EwuzieU.AkuN. O.NwankpaS. U. (2021). An appraisal of data collection, analysis, and reporting adopted for water quality assessment: a case of Nigeria water quality research. Heliyon7, e07950. 10.1016/J.HELIYON.2021.E07950

  • CrossRef
  • Google Scholar

• 58

  EyringV.BonyS.MeehlG. A.SeniorC. A.StevensB.StoufferR. J.et al (2016). Overview of the coupled model intercomparison project phase 6 (CMIP6) experimental design and organization. Geosci. Model. Dev.9, 1937–1958. 10.5194/gmd-9-1937-2016

  • CrossRef
  • Google Scholar

• 59

  FanS.XiaoN.DongS. (2020). A novel model to predict significant wave height based on long short-term memory network. Ocean. Eng.205, 107298. 10.1016/J.OCEANENG.2020.107298

  • CrossRef
  • Google Scholar

• 60

  FernándezJ. C.Salcedo-SanzS.GutiérrezP. A.AlexandreE.Hervás-MartínezC. (2015). Significant wave height and energy flux range forecast with machine learning classifiers. Eng. Appl. Artif. Intell.43, 44–53. 10.1016/J.ENGAPPAI.2015.03.012

  • CrossRef
  • Google Scholar

• 61

  FormentinS. M.ZanuttighB.Van Der MeerJ. W. (2017). A neural network tool for predicting wave reflection, overtopping and transmission. Coast. Eng. J.59, 1750006-1–1750006-31. 10.1142/S0578563417500061

  • CrossRef
  • Google Scholar

• 62

  FreemanB.TangY.HuangY.VanZwietenJ. (2021). Rotor blade imbalance fault detection for variable-speed marine current turbines via generator power signal analysis. Ocean. Eng.223, 108666. 10.1016/j.oceaneng.2021.108666

  • CrossRef
  • Google Scholar

• 63

  GandomiM.Dolatshahi PiroozM.VarjavandI.NikooM. R. (2020). Permeable breakwaters performance modeling: a comparative study of machine learning techniques. Remote Sens.12, 1856. 10.3390/rs12111856

  • CrossRef
  • Google Scholar

• 64

  GauciA.DeidunA.AbelaJ.Zarb AdamiK. (2016). Machine Learning for benthic sand and maerl classification and coverage estimation in coastal areas around the Maltese Islands. J. Appl. Res. Technol.14, 338–344. 10.1016/j.jart.2016.08.003

  • CrossRef
  • Google Scholar

• 65

  GawehnM.van DongerenA.de VriesS.SwinkelsC.HoekstraR.AarninkhofS.et al (2020). The application of a radar-based depth inversion method to monitor near-shore nourishments on an open sandy coast and an ebb-tidal delta. Coast. Eng.159, 103716. 10.1016/J.COASTALENG.2020.103716

  • CrossRef
  • Google Scholar

• 66

  GoldsteinE. B.CocoG.PlantN. G. (2019). A review of machine learning applications to coastal sediment transport and morphodynamics. Earth-Science Rev.194, 97–108. 10.1016/j.earscirev.2019.04.022

  • CrossRef
  • Google Scholar

• 67

  GomezC. (2022). “Point-cloud technology for coastal and floodplain geomorphology,” in Point cloud technologies for geomorphologists from data acquisition to processing (Springer), 53–81. 10.1007/978-3-031-10975-1

  • CrossRef
  • Google Scholar

• 68

  GünaydinK. (2008). The estimation of monthly mean significant wave heights by using artificial neural network and regression methods. Ocean. Eng.35, 1406–1415. 10.1016/J.OCEANENG.2008.07.008

  • CrossRef
  • Google Scholar

• 69

  HagenaarsG.de VriesS.LuijendijkA. P.de BoerW. P.ReniersA. J. H. M. (2018). On the accuracy of automated shoreline detection derived from satellite imagery: a case study of the sand motor mega-scale nourishment. Coast. Eng.133, 113–125. 10.1016/J.COASTALENG.2017.12.011

  • CrossRef
  • Google Scholar

• 70

  HallJ. W.MeadowcroftI. C.LeeE. M.Van GelderP. H. A. J. M. (2002). Stochastic simulation of episodic soft coastal cliff recession. Coast. Eng.46, 159–174. 10.1016/S0378-3839(02)00089-3

  • CrossRef
  • Google Scholar

• 71

  HashemiM. R.GhadampourZ.NeillS. P. (2010). Using an artificial neural network to model seasonal changes in beach profiles. Ocean. Eng.37, 1345–1356. 10.1016/J.OCEANENG.2010.07.004

  • CrossRef
  • Google Scholar

• 72

  HessamiM.GachonP.OuardaT. B. M. J.St-HilaireA. (2008). Automated regression-based statistical downscaling tool. Environ. Model. Softw.23, 813–834. 10.1016/J.ENVSOFT.2007.10.004

  • CrossRef
  • Google Scholar

• 73

  HeumannB. W. (2011). An object-based classification of mangroves using a hybrid decision tree-support vector machine approach. Remote Sens.3, 2440–2460. 10.3390/rs3112440

  • CrossRef
  • Google Scholar

• 74

  HodgeV. J.AustinJ. (2004). A survey of outlier detection methodologies. Artif. Intell. Rev.22, 85–126. 10.1023/b:aire.0000045502.10941.a9

  • CrossRef
  • Google Scholar

• 75

  HongJ.-H.ChiewY.-M.ChengN.-S. (2013). Scour caused by a propeller jet. J. Hydraul. Eng.139, 1003–1012. 10.1061/(asce)hy.1943-7900.0000746

  • CrossRef
  • Google Scholar

• 76

  HoonhoutB. M.RadermacherM.BaartF.van der MaatenL. J. P. (2015). An automated method for semantic classification of regions in coastal images. Coast. Eng.105, 1–12. 10.1016/j.coastaleng.2015.07.010

  • CrossRef
  • Google Scholar

• 77

  HosseinzadehS.Etemad-ShahidiA.KooshehA. (2021). Prediction of mean wave overtopping at simple sloped breakwaters using kernel-based methods. J. Hydroinformatics23, 1030–1049. 10.2166/hydro.2021.046

  • CrossRef
  • Google Scholar

• 78

  HotellingH. (1933). Analysis of a complex of statistical variables into principal components. J. Educ. Psychol.24, 417–441. 10.1037/h0071325

  • CrossRef
  • Google Scholar

• 79

  HuaX. G.NiY. Q.KoJ. M.WongK. Y. (2007). Modeling of temperature–frequency correlation using combined principal component analysis and support vector regression technique. J. Comput. Civ. Eng.21, 122–135. 10.1061/(asce)0887-3801(2007)21:2(122)

  • CrossRef
  • Google Scholar

• 80

  HuangD.ZhaoD.WeiL.WangZ.DuY. (2015). Modeling and analysis in marine big data: advances and challenges. Math. Problems Eng.2015, 1–13. 10.1155/2015/384742

  • CrossRef
  • Google Scholar

• 81

  IrishJ. L.WhiteT. E. (1998). Coastal engineering applications of high-resolution lidar bathymetry. Coast. Eng.35, 47–71. 10.1016/S0378-3839(98)00022-2

  • CrossRef
  • Google Scholar

• 82

  IzadiM.SultanM.KadiriR. E.GhannadiA.AbdelmohsenK. (2021). A remote sensing and machine learning-based approach to forecast the onset of harmful algal bloom. Remote Sens.13, 3863. 10.3390/rs13193863

  • CrossRef
  • Google Scholar

• 83

  JainP.DeoM. C. (2008). Artificial intelligence tools to forecast ocean waves in real time. Open Ocean Eng. J.1, 13–20. 10.2174/1874835x00801010013

  • CrossRef
  • Google Scholar

• 84

  JamesS. C.ZhangY.O’DonnchaF. (2018). A machine learning framework to forecast wave conditions. Coast. Eng.137, 1–10. 10.1016/j.coastaleng.2018.03.004

  • CrossRef
  • Google Scholar

• 85

  JayaratneM. P. R.PremaratneB.AdewaleA.MikamiT.MatsubaS.ShibayamaT.et al (2016). Failure mechanisms and local scour at coastal structures induced by Tsunami. Coast. Eng. J.58, 1640017-1–1640017-38. 10.1142/S0578563416400179

  • CrossRef
  • Google Scholar

• 86

  JiangT.GradusJ. L.RoselliniA. J. (2020). Supervised machine learning: a brief primer. Behav. Ther.51, 675–687. 10.1016/J.BETH.2020.05.002

  • CrossRef
  • Google Scholar

• 87

  JirakittayakornA.KormongkolkulT.VateekulP.JitkajornwanichK.LawawirojwongS. (2017). “Temporal kNN for short-Term ocean current prediction based on HF radar observations,” in Proceedings of the 2017 14th International Joint Conference on Computer Science and Software Engineering, NakhonSiThammarat, Thailand, July 12–14, 2017 (IEEE (Institute of Electrical and Electronics Engineers)). 10.1109/JCSSE.2017.8025921

  • CrossRef
  • Google Scholar

• 88

  JoyceK. E.FickasK. C.KalamandeenM. (2023). The unique value proposition for using drones to map coastal ecosystems. Camb. Prisms Coast. Futur.1, e6. 10.1017/cft.2022.7

  • CrossRef
  • Google Scholar

• 89

  Kabiri-SamaniA. R.Aghaee-TarazjaniJ.BorgheiS. M.JengD. S. (2011). Application of neural networks and fuzzy logic models to long-shore sediment transport. Appl. Soft Comput.11, 2880–2887. 10.1016/J.ASOC.2010.11.021

  • CrossRef
  • Google Scholar

• 90

  KagemotoH. (2020). Forecasting a water-surface wave train with artificial intelligence- A case study. Ocean. Eng.207, 107380. 10.1016/J.OCEANENG.2020.107380

  • CrossRef
  • Google Scholar

• 91

  KalkanK.BayramB.MaktavD.SunarF. (2013). “Comparison of support vector machine and object based classification methods for coastline detection,” in International archives of the photogrammetry, remote sensing and spatial information sciences - ISPRS archives. 10.5194/isprsarchives-XL-7-W2-125-2013

  • CrossRef
  • Google Scholar

• 92

  KaloopM. R.KumarD.ZarzouraF.RoyB.HuJ. W. (2020). A wavelet - particle swarm optimization - extreme learning machine hybrid modeling for significant wave height prediction. Ocean. Eng.213, 107777. 10.1016/J.OCEANENG.2020.107777

  • CrossRef
  • Google Scholar

• 93

  KaplanD.Muñoz-CarpenaR.RitterA. (2010). Untangling complex shallow groundwater dynamics in the floodplain wetlands of a southeastern U.S. coastal river. Water Resour. Res.46. 10.1029/2009WR009038

  • CrossRef
  • Google Scholar

• 94

  KarimpourA.ChenQ.TwilleyR. R. (2016). A field study of how wind waves and currents may contribute to the deterioration of saltmarsh fringe. Estuaries Coasts39, 935–950. 10.1007/s12237-015-0047-z

  • CrossRef
  • Google Scholar

• 95

  KartalS. (2023). Assessment of the spatiotemporal prediction capabilities of machine learning algorithms on Sea Surface temperature data: a comprehensive study. Eng. Appl. Artif. Intell.118, 105675. 10.1016/J.ENGAPPAI.2022.105675

  • CrossRef
  • Google Scholar

• 96

  KelleherJ. D.NameeB. M.D’ArcyA. (2015). Fundamentals of machine learning for predictive data analytics: algorithms, worked examples, and case studies.

  • Google Scholar

• 97

  KimH. D.AokiS. I. (2021). Artificial intelligence application on sediment transport. J. Mar. Sci. Eng.9, 600. 10.3390/jmse9060600

  • CrossRef
  • Google Scholar

• 98

  KimJ.KimJ. (2020). Estimation of water surface flow velocity in coastal video imagery by visual tracking with deep learning. J. Coast. Res.95, 522. 10.2112/SI95-101.1

  • CrossRef
  • Google Scholar

• 99

  KimJ.KimJ.KimT.HuhD.CairesS. (2020). Wave-tracking in the surf zone using coastal video imagery with deep neural networks. Atmos. (Basel)11, 304. 10.3390/atmos11030304

  • CrossRef
  • Google Scholar

• 100

  KimT.KwonY.LeeJ.LeeE.KwonS. (2022). Wave attenuation prediction of artificial coral reef using machine-learning integrated with hydraulic experiment. Ocean. Eng.248, 110324. 10.1016/J.OCEANENG.2021.110324

  • CrossRef
  • Google Scholar

• 101

  KitsikoudisV.SidiropoulosE.HrissanthouV. (2015). Assessment of sediment transport approaches for sand-bed rivers by means of machine learning. Hydrological Sci. J.60, 1566–1586. 10.1080/02626667.2014.909599

  • CrossRef
  • Google Scholar

• 102

  KnightP. J.BirdC. O.SinclairA.PlaterA. J. (2020). A low-cost GNSS buoy platform for measuring coastal sea levels. Ocean. Eng.203, 107198. 10.1016/J.OCEANENG.2020.107198

  • CrossRef
  • Google Scholar

• 103

  KongX.SunY.SuR.ShiX. (2017). Real-time eutrophication status evaluation of coastal waters using support vector machine with grid search algorithm. Mar. Pollut. Bull.119, 307–319. 10.1016/J.MARPOLBUL.2017.04.022

  • CrossRef
  • Google Scholar

• 104

  KramerO. (2013). Dimensionality reduction with unsupervised nearest neighbors. Intell. Syst. Ref. Libr.51. 10.1007/978-3-642-38652-7

  • CrossRef
  • Google Scholar

• 105

  KroonA.LarsonM.MöllerI.YokokiH.RozynskiG.CoxJ.et al (2008). Statistical analysis of coastal morphological data sets over seasonal to decadal time scales. Coast. Eng.55, 581–600. 10.1016/j.coastaleng.2007.11.006

  • CrossRef
  • Google Scholar

• 106

  KumarN. K.SavithaR.MamunA. A. (2017). Regional ocean wave height prediction using sequential learning neural networks. Ocean. Eng.129, 605–612. 10.1016/J.OCEANENG.2016.10.033

  • CrossRef
  • Google Scholar

• 107

  KumarN. K.SavithaR.MamunA. A. (2018). Ocean wave height prediction using ensemble of Extreme Learning Machine. Ocean. Eng.277, 605–612. 10.1016/J.NEUCOM.2017.03.092

  • CrossRef
  • Google Scholar

• 108

  KuntojiG.RaoM.RaoS. (2020). Prediction of wave transmission over submerged reef of tandem breakwater using PSO-SVM and PSO-ANN techniques. ISH J. Hydraulic Eng.26, 283–290. 10.1080/09715010.2018.1482796

  • CrossRef
  • Google Scholar

• 109

  KuoY. M.LiuW.ZhaoE.LiR.Muñoz-CarpenaR. (2019). Water quality variability in the middle and down streams of Han River under the influence of the Middle Route of South-North Water diversion project, China. J. Hydrology569, 218–229. 10.1016/j.jhydrol.2018.12.001

  • CrossRef
  • Google Scholar

• 110

  LatifS. D.ChongK. L.AhmedA. N.HuangY. F.SherifM.El-ShafieA. (2023). Sediment load prediction in johor river: deep learning versus machine learning models. Appl. Water Sci.13, 79–13. 10.1007/s13201-023-01874-w

  • CrossRef
  • Google Scholar

• 111

  LazuardiW.ArdiyantoR.MarfaiM. A.MutaqinB. W.KusumaD. W. (2021). Coastal reef and seagrass monitoring for coastal ecosystem management. Int. J. Sustain. Dev. Plan.16, 557–568. 10.18280/IJSDP.160317

  • CrossRef
  • Google Scholar

• 112

  LeeT.-L. (2004). Back-propagation neural network for long-term tidal predictions. Ocean. Eng.31, 225–238. 10.1016/S0029-8018(03)00115-X

  • CrossRef
  • Google Scholar

• 113

  LiS. P.WangH. L. (2011). “Control stratory in coastal area using Markov chain and Random Forest,” in 2011 IEEE 18th International Conference on Industrial Engineering and Engineering Management, IE and EM 2011, Changchun, China, September 3–5, 2011 (IEEE (Institute of Electrical and Electronics Engineers)). 10.1109/ICIEEM.2011.6035480

  • CrossRef
  • Google Scholar

• 114

  LiuB.YangB.Masoud-AnsariS.WangH.GaheganM. (2021). Coastal image classification and pattern recognition: Tairua beach, New Zealand. Sensors21, 7352. 10.3390/s21217352

  • CrossRef
  • Google Scholar

• 115

  LouR.LvZ.DangS.SuT.LiX. (2021). “Application of machine learning in ocean data,” in Multimedia systems. 10.1007/s00530-020-00733-x

  • CrossRef
  • Google Scholar

• 116

  MacayealD. R.AbbotD. S.SergienkoO. V. (2011). Iceberg-capsize tsunamigenesis. Ann. Glaciol.52, 51–56. 10.3189/172756411797252103

  • CrossRef
  • Google Scholar

• 117

  MahjoobiJ.Adeli MosabbebE. (2009). Prediction of significant wave height using regressive support vector machines. Ocean. Eng.36, 339–347. 10.1016/J.OCEANENG.2009.01.001

  • CrossRef
  • Google Scholar

• 118

  MahjoobiJ.Etemad-ShahidiA. (2008). An alternative approach for the prediction of significant wave heights based on classification and regression trees. Appl. Ocean Res.30, 172–177. 10.1016/J.APOR.2008.11.001

  • CrossRef
  • Google Scholar

• 119

  MahmoodiK.GhassemiH. (2018). Outlier detection in ocean wave measurements by using unsupervised data mining methods. Pol. Marit. Res.25, 44–50. 10.2478/pomr-2018-0005

  • CrossRef
  • Google Scholar

• 120

  MakarynskyyO.Pires-SilvaA. A.MakarynskaD.Ventura-SoaresC. (2005). Artificial neural networks in wave predictions at the west coast of Portugal. Comput. Geosciences31, 415–424. 10.1016/J.CAGEO.2004.10.005

  • CrossRef
  • Google Scholar

• 121

  MartinsG. M.ThompsonR. C.NetoA. I.HawkinsS. J.JenkinsS. R. (2010). Enhancing stocks of the exploited limpet Patella candei d’Orbigny via modifications in coastal engineering. Biol. Conserv.143, 203–211. 10.1016/j.biocon.2009.10.004

  • CrossRef
  • Google Scholar

• 122

  MasmoudiO.JaouaM.JaouaA.YacoutS. (2021). Data preparation in machine learning for condition-based maintenance. J. Comput. Sci.17, 525–538. 10.3844/JCSSP.2021.525.538

  • CrossRef
  • Google Scholar

• 123

  MengF.SongT.XuD.XieP.LiY. (2021). Forecasting tropical cyclones wave height using bidirectional gated recurrent unit. Ocean. Eng.234, 108795. 10.1016/J.OCEANENG.2021.108795

  • CrossRef
  • Google Scholar

• 124

  MészárosL.van der MeulenF.JongbloedG.El SerafyG. (2022). Coastal environmental and atmospheric data reduction in the Southern North Sea supporting ecological impact studies. Front. Mar. Sci.9, 1–23. 10.3389/fmars.2022.920616

  • CrossRef
  • Google Scholar

• 125

  MillerJ. K.DeanR. G. (2007). Shoreline variability via empirical orthogonal function analysis: part II relationship to nearshore conditions. Coast. Eng.54, 133–150. 10.1016/j.coastaleng.2006.08.014

  • CrossRef
  • Google Scholar

• 126

  MoncadaA. M.MelesseA. M.VithanageJ.PriceR. M. (2021). Long-term assessment of surface water quality in a highly managed estuary basin. Int. J. Environ. Res. Public Health18, 9417. 10.3390/ijerph18179417

  • CrossRef
  • Google Scholar

• 127

  MoodyD. I.BrumbyS. P.RowlandJ. C.AltmannG. L. (2014). Land cover classification in multispectral imagery using clustering of sparse approximations over learned feature dictionaries. J. Appl. Remote Sens.8, 084793. 10.1117/1.jrs.8.084793

  • CrossRef
  • Google Scholar

• 128

  NajafiM. R.MoradkhaniH.WherryS. A. (2011). Statistical downscaling of precipitation using machine learning with optimal predictor selection. J. Hydrol. Eng.16, 650–664. 10.1061/(asce)he.1943-5584.0000355

  • CrossRef
  • Google Scholar

• 129

  NakamuraT.KuramitsuY.MizutaniN. (2008). Tsunami scour around a square structure. Coast. Eng. J.50, 209–246. 10.1142/S057856340800179X

  • CrossRef
  • Google Scholar

• 130

  NeshatM.AbbasnejadE.ShiQ.AlexanderB.WagnerM. (2019). “Adaptive neuro-surrogate-based optimisation method for wave energy converters placement optimisation,” in Lecture notes in computer science (including subseries lecture notes in artificial intelligence and lecture notes in bioinformatics. 10.1007/978-3-030-36711-4_30

  • CrossRef
  • Google Scholar

• 131

  NeumannB.OttK.KenchingtonR. (2017). Strong sustainability in coastal areas: a conceptual interpretation of SDG 14. Sustain. Sci.12, 1019–1035. 10.1007/s11625-017-0472-y

  • CrossRef
  • Google Scholar

• 132

  NikooM. R.KerachianR.AlizadehM. R. (2018). A fuzzy KNN-based model for significant wave height prediction in large lakes. Oceanologia60, 153–168. 10.1016/J.OCEANO.2017.09.003

  • CrossRef
  • Google Scholar

• 133

  OehmckeS.ZielinskiO.KramerO. (2015). “Event detection in marine time series data,” in Lecture notes in computer science (including subseries lecture notes in artificial intelligence and lecture notes in bioinformatics. 10.1007/978-3-319-24489-1_24

  • CrossRef
  • Google Scholar

• 134

  PalM.MatherP. M. (2003). An assessment of the effectiveness of decision tree methods for land cover classification. Remote Sens. Environ.86, 554–565. 10.1016/S0034-4257(03)00132-9

  • CrossRef
  • Google Scholar

• 135

  ParadinasL. M.JamesN. A.QuinnB.DaleA.NarayanaswamyB. E. (2021). A new collection tool-kit to sample microplastics from the marine environment (sediment, seawater, and biota) using citizen science. Front. Mar. Sci.8. 10.3389/fmars.2021.657709

  • CrossRef
  • Google Scholar

• 136

  ParkJ.OhJ. (2022). Analysis of collected data and establishment of an abnormal data detection algorithm using principal component analysis and K-nearest neighbors for predictive maintenance of ship propulsion engine. Processes10, 2392. 10.3390/pr10112392

  • CrossRef
  • Google Scholar

• 137

  ParkS. J.LeeD. K. (2020). Prediction of coastal flooding risk under climate change impacts in South Korea using machine learning algorithms. Environ. Res. Lett.15, 094052. 10.1088/1748-9326/aba5b3

  • CrossRef
  • Google Scholar

• 138

  PearsonK. (1901)., 2. London, Edinburgh, 559–572. 10.1080/14786440109462720LIII. On lines and planes of closest fit to systems of points in spaceLond. Edinb. Dublin Philosophical Mag. J. Sci.

  • CrossRef
  • Google Scholar

• 139

  PeñaE.FerrerasJ.Sanchez-TemblequeF. (2011). Experimental study on wave transmission coefficient, mooring lines and module connector forces with different designs of floating breakwaters. Ocean. Eng.38, 1150–1160. 10.1016/j.oceaneng.2011.05.005

  • CrossRef
  • Google Scholar

• 140

  PereiraG. C.EbeckenN. F. F. (2009). Knowledge discovering for coastal waters classification. Expert Syst. Appl.36, 8604–8609. 10.1016/J.ESWA.2008.10.009

  • CrossRef
  • Google Scholar

• 141

  PimentelM. A. F.CliftonD. A.CliftonL.TarassenkoL. (2014). A review of novelty detection. Signal Process.99, 215–249. 10.1016/j.sigpro.2013.12.026

  • CrossRef
  • Google Scholar

• 142

  PlaatA.KostersW.PreussM. (2023). High-accuracy model-based reinforcement learning, a survey. Artif. Intell. Rev.56, 9541–9573. 10.1007/s10462-022-10335-w

  • CrossRef
  • Google Scholar

• 143

  PourghasemiH. R.GayenA.ParkS.LeeC. W.LeeS. (2018). Assessment of landslide-prone areas and their zonation using logistic regression, LogitBoost, and naïvebayes machine-learning algorithms. Sustainability10, 3697. 10.3390/su10103697

  • CrossRef
  • Google Scholar

• 144

  PourzangbarA.BrocchiniM. (2022). A new process-based, wave-resolving, 2DH circulation model for the evolution of natural sand bars: the role of nearbed dynamics and suspended sediment transport. Coast. Eng.177, 104192. 10.1016/J.COASTALENG.2022.104192

  • CrossRef
  • Google Scholar

• 145

  PourzangbarA.BrocchiniM.SaberA.MahjoobiJ.MirzaaghasiM.BarzegarM. (2017a). Prediction of scour depth at breakwaters due to non-breaking waves using machine learning approaches. Appl. Ocean Res.63, 120–128. 10.1016/j.apor.2017.01.012

  • CrossRef
  • Google Scholar

• 146

  PourzangbarA. (2012). Determination of the most effective parameters on scour depth at seawalls using genetic programming (GP). 10th Int. Conf. coasts, ports Mar. Struct. (ICOPMASS 2012).

  • Google Scholar

• 147

  PourzangbarA.LosadaM. A.SaberA.AhariL. R.LarroudéP.VaeziM.et al (2017b). Prediction of non-breaking wave induced scour depth at the trunk section of breakwaters using Genetic Programming and Artificial Neural Networks. Coast. Eng.121, 107–118. 10.1016/j.coastaleng.2016.12.008

  • CrossRef
  • Google Scholar

• 148

  PourzangbarA.SaberA.Yeganeh-BakhtiaryA.AhariL. R. (2017c). Predicting scour depth at seawalls using GP and ANNs. J. Hydroinformatics19, 349–363. 10.2166/hydro.2017.125

  • CrossRef
  • Google Scholar

• 149

  PrataJ. C.da CostaJ. P.DuarteA. C.Rocha-SantosT. (2019). Methods for sampling and detection of microplastics in water and sediment: a critical review. TrAC Trends Anal. Chem.110, 150–159. 10.1016/J.TRAC.2018.10.029

  • CrossRef
  • Google Scholar

• 150

  ProvostE. J.ButcherP. A.ColemanM. A.KelaherB. P. (2020). Assessing the viability of small aerial drones to quantify recreational Fishers. Fish. Manag. Ecol.27, 615–621. 10.1111/fme.12452

  • CrossRef
  • Google Scholar

• 151

  QiaoX.ChuT.TissotP.AliI.AhmedM. (2023). Vertical land motion monitored with satellite radar altimetry and tide gauge along the Texas coastline, USA, between 1993 and 2020. Int. J. Appl. Earth Observation Geoinformation117, 103222. 10.1016/J.JAG.2023.103222

  • CrossRef
  • Google Scholar

• 152

  RamaswamyS.RastogiR.ShimK. (2000). Efficient algorithms for mining outliers from large data sets. SIGMOD Rec. (ACM Spec. Interes. Gr. Manag. Data)29, 427–438. 10.1145/335191.335437

  • CrossRef
  • Google Scholar

• 153

  RanasingheR.LarsonM.SavioliJ. (2010). Shoreline response to a single shore-parallel submerged breakwater. Coast. Eng.57, 1006–1017. 10.1016/j.coastaleng.2010.06.002

  • CrossRef
  • Google Scholar

• 154

  RaoS.MandalS. (2005). Hindcasting of storm waves using neural networks. Ocean. Eng.32, 667–684. 10.1016/J.OCEANENG.2004.09.003

  • CrossRef
  • Google Scholar

• 155

  ReggianniniM.PapiniO.PieriG. (2022). An automated analysis tool for the classification of Sea surface temperature imagery. Pattern Recognit. Image Anal.32, 631–635. 10.1134/S1054661822030336

  • CrossRef
  • Google Scholar

• 156

  RengarajanD.VaidyaG.SarveshA.KalathilD.ShakkottaiS. (2022). “Reinforcement learning with sparse rewards using guidance from offline demonstration,” in ICLR 2022 - 10th International Conference on Learning Representations, Virtual, April 25–29, 2022.

  • Google Scholar

• 157

  RizianizaI.AisjahA. S. (2015). “Prediction of significant wave height in the java sea using artificial neural network,” in Proceeding 2015 International Seminar on Intelligent Technology and Its Applications, ISITIA, Surabaya, Indonesia, May 20–21, 2015 (IEEE (Institute of Electrical and Electronics Engineers)). 10.1109/ISITIA.2015.7219944

  • CrossRef
  • Google Scholar

• 158

  RokniK.AhmadA.SolaimaniK.HaziniS. (2015). A new approach for surface water change detection: integration of pixel level image fusion and image classification techniques. Int. J. Appl. Earth Observation Geoinformation34, 226–234. 10.1016/J.JAG.2014.08.014

  • CrossRef
  • Google Scholar

• 159

  Ruiz de Alegría-ArzaburuA.Pedrozo-AcuñaA.Horrillo-CaraballoJ. M.MasselinkG.ReeveD. E. (2010). Determination of wave-shoreline dynamics on a macrotidal gravel beach using Canonical Correlation Analysis. Coast. Eng.57, 290–303. 10.1016/j.coastaleng.2009.10.014

  • CrossRef
  • Google Scholar

• 160

  SajjadM.LinN.ChanJ. C. L. (2020). Spatial heterogeneities of current and future hurricane flood risk along the U.S. Atlantic and Gulf coasts. Sci. Total Environ.713, 136704. 10.1016/j.scitotenv.2020.136704

  • CrossRef
  • Google Scholar

• 161

  SakaaB.ElbeltagiA.BoudibiS.ChaffaïH.IslamA. R. M. T.KulimushiL. C.et al (2022). Water quality index modeling using random forest and improved SMO algorithm for support vector machine in Saf-Saf river basin. Environ. Sci. Pollut. Res.29, 48491–48508. 10.1007/s11356-022-18644-x

  • CrossRef
  • Google Scholar

• 162

  Santamaria CervantesM.Díaz-CarrascoP.MoraguesM. V.ClaveroM.LosadaángelM. (2022). “Uncertainties of the actual engineering formulas for coastal protection slopes. The dimensional analysis and experimental method,” in Proceedings of the 39th IAHR World Congress From Snow to Sea, Granada, Spain, June 19–24, 2022 (IAHR (International Association for Hydro-Environment Engineering and Research)). 10.3850/iahr-39wc252171192022900

  • CrossRef
  • Google Scholar

• 163

  SarkarD.ContalE.VayatisN.DiasF. (2016). Prediction and optimization of wave energy converter arrays using a machine learning approach. Renew. Energy97, 504–517. 10.1016/j.renene.2016.05.083

  • CrossRef
  • Google Scholar

• 164

  SarkarS.GundechaV.GhorbanpourS.ShmakovA.BabuA. R.PichardA.et al (2022). “Skip training for multi-agent reinforcement learning controller for industrial wave energy converters,” in 2022 IEEE 18th International Conference on Automation Science and Engineering (CASE), Mexico City, Mexico, August 20–24, 2022 (IEEE (Institute of Electrical and Electronics Engineers)), 212–219. 10.1109/CASE49997.2022.9926561

  • CrossRef
  • Google Scholar

• 165

  ScottT.MasselinkG.RussellP. (2011). Morphodynamic characteristics and classification of beaches in England and Wales. Mar. Geol.286, 1–20. 10.1016/j.margeo.2011.04.004

  • CrossRef
  • Google Scholar

• 166

  SekovskiI.StecchiF.ManciniF.Del RioL. (2014). Image classification methods applied to shoreline extraction on very high-resolution multispectral imagery. Int. J. Remote Sens.35, 3556–3578. 10.1080/01431161.2014.907939

  • CrossRef
  • Google Scholar

• 167

  ShafaghatM.DezvarehR. (2021). Support vector machine for classification and regression of coastal sediment transport. Arab. J. Geosci.14, 2009. 10.1007/s12517-021-08360-0

  • CrossRef
  • Google Scholar

• 168

  ShamshirbandS.MosaviA.RabczukT.NabipourN.ChauK. w. (2020). Prediction of significant wave height; comparison between nested grid numerical model, and machine learning models of artificial neural networks, extreme learning and support vector machines. Eng. Appl. Comput. Fluid Mech.14, 805–817. 10.1080/19942060.2020.1773932

  • CrossRef
  • Google Scholar

• 169

  ShenbagarajN.ManiN. D.MuthukumarM. (2014). Isodata classification technique to assess the shoreline changes of Kolachel to Kayalpattanam coast. Int. J. Eng. Res. Technol.3. 10.17577/IJERTV3IS040136

  • CrossRef
  • Google Scholar

• 170

  ShiF.KirbyJ. T.HarrisJ. C.GeimanJ. D.GrilliS. T. (2012). A high-order adaptive time-stepping TVD solver for Boussinesq modeling of breaking waves and coastal inundation. Ocean. Model.43-44, 36–51. 10.1016/J.OCEMOD.2011.12.004

  • CrossRef
  • Google Scholar

• 171

  ShuiP. L.XiaX. Y.ZhangY. S. (2020). Sea-land segmentation in maritime surveillance radars via k-nearest neighbor classifier. IEEE Trans. Aerosp. Electron. Syst.56, 3854–3867. 10.1109/TAES.2020.2981267

  • CrossRef
  • Google Scholar

• 172

  ShuvoS. S.YilmazY.BushA.HafenM. (2022). Modeling and simulating adaptation strategies against sea-level rise using multiagent deep reinforcement learning. IEEE Trans. Comput. Soc. Syst.9, 1185–1196. 10.1109/TCSS.2021.3122282

  • CrossRef
  • Google Scholar

• 173

  SierraC.Flor-BlancoG.OrdoñezC.FlorG.GallegoJ. R. (2017). Analyzing coastal environments by means of functional data analysis. Sediment. Geol.357, 99–108. 10.1016/j.sedgeo.2017.06.008

  • CrossRef
  • Google Scholar

• 174

  SmitM. W. J.AarninkhofS. G. J.WijnbergK. M.GonzálezM.KingstonK. S.SouthgateH. N.et al (2007). The role of video imagery in predicting daily to monthly coastal evolution. Coast. Eng.54, 539–553. 10.1016/J.COASTALENG.2007.01.009

  • CrossRef
  • Google Scholar

• 175

  SoloyA.TurkiI.LecoqN.Gutiérrez BarcelóÁ. D.CostaS.LaignelB.et al (2021). A fully automated method for monitoring the intertidal topography using Video Monitoring Systems. Coast. Eng.167, 103894. 10.1016/J.COASTALENG.2021.103894

  • CrossRef
  • Google Scholar

• 176

  SzmytkiewiczM.BiegowskiJ. X.KaczmarekL. M.OkrójT.OstrowskiR. X.PruszakZ.et al (2000). Coastline changes nearby harbour structures: comparative analysis of one-line models versus field data. Coast. Eng.40, 119–139. 10.1016/S0378-3839(00)00008-9

  • CrossRef
  • Google Scholar

• 177

  TanJ.ChenS.LeeC. Y.DongG.HuW.WangJ. (2021). Projected changes of typhoon intensity in a regional climate model: development of a machine learning bias correction scheme. Int. J. Climatol.41, 2749–2764. 10.1002/joc.6987

  • CrossRef
  • Google Scholar

• 178

  TanJ.LiuH.LiM.WangJ. (2018). A prediction scheme of tropical cyclone frequency based on lasso and random forest. Theor. Appl. Climatol.133, 973–983. 10.1007/s00704-017-2233-3

  • CrossRef
  • Google Scholar

• 179

  TayfurG.KarimiY.SinghV. P. (2013). Principle component analysis in conjuction with data driven methods for sediment load prediction. Water Resour. Manag.27, 2541–2554. 10.1007/s11269-013-0302-7

  • CrossRef
  • Google Scholar

• 180

  TimmermansB. W.GommengingerC. P.DodetG.BidlotJ. R. (2020). Global wave height trends and variability from new multimission satellite altimeter products, reanalyses, and wave buoys. Geophys. Res. Lett.47. 10.1029/2019GL086880

  • CrossRef
  • Google Scholar

• 181

  TsiakosC. A. D.ChalkiasC. (2023). Use of machine learning and remote sensing techniques for shoreline monitoring: a review of recent literature. Appl. Sci.13, 3268. 10.3390/app13053268

  • CrossRef
  • Google Scholar

• 182

  TsujimotoG.TamaiM.YamadaF. (2012). LONG-TERM prediction of beach profile and sediment grain size characteristic at low energy beach. Coast. Eng. Proc.1, 14. 10.9753/icce.v33.sediment.14

  • CrossRef
  • Google Scholar

• 183

  TurnerI. L.HarleyM. D.AlmarR.BergsmaE. W. J. (2021). Satellite optical imagery in coastal engineering. Coast. Eng.167, 103919. 10.1016/J.COASTALENG.2021.103919

  • CrossRef
  • Google Scholar

• 184

  UddinM. G.NashS.RahmanA.OlbertA. I. (2023). Performance analysis of the water quality index model for predicting water state using machine learning techniques. Process Saf. Environ. Prot.169, 808–828. 10.1016/J.PSEP.2022.11.073

  • CrossRef
  • Google Scholar

• 185

  UhlF.Græsdal RasmussenT.OppeltN. (2021). Classification ensembles for beach cast and drifting vegetation mapping with sentinel-2 and PlanetScope. Geosciences12, 15. 10.3390/geosciences12010015

  • CrossRef
  • Google Scholar

• 186

  van GentM. R. A.van den BoogaardH. F. P.PozuetaB.MedinaJ. R. (2007). Neural network modelling of wave overtopping at coastal structures. Coast. Eng.54, 586–593. 10.1016/j.coastaleng.2006.12.001

  • CrossRef
  • Google Scholar

• 187

  Van KomenD. F.HowarthK.NeilsenT. B.KnoblesD. P.DahlP. H. (2022). A CNN for range and seabed estimation on normalized and extracted time-series impulses. IEEE J. Ocean. Eng.47, 833–846. 10.1109/JOE.2021.3134719

  • CrossRef
  • Google Scholar

• 188

  VaralakshmiP.VasumathiN.VenkatesanR. (2021). Tropical Cyclone prediction based on multi-model fusion across Indian coastal region. Prog. Oceanogr.193, 102557. 10.1016/j.pocean.2021.102557

  • CrossRef
  • Google Scholar

• 189

  VerwegaM. T.TrahmsC.AntiaA. N.DickhausT.PriggeE.PrinzlerM. H. U.et al (2021). Perspectives on marine data science as a blueprint for emerging data science disciplines. Front. Mar. Sci.8. 10.3389/fmars.2021.678404

  • CrossRef
  • Google Scholar

• 190

  VosK.SplinterK. D.HarleyM. D.SimmonsJ. A.TurnerI. L. (2019). CoastSat: a Google Earth engine-enabled Python toolkit to extract shorelines from publicly available satellite imagery. Environ. Model. Softw.122, 104528. 10.1016/J.ENVSOFT.2019.104528

  • CrossRef
  • Google Scholar

• 191

  WattelezG.DupouyC.JuillotF. (2022). Unsupervised optical classification of the seabed color in shallow oligotrophic waters from sentinel‐2 images: a case study in the voh‐koné‐pouembout lagoon (New Caledonia). Remote Sens.14, 836. 10.3390/rs14040836

  • CrossRef
  • Google Scholar

• 192

  WongY. J.ShimizuY.KamiyaA.ManeechotL.BharambeK. P.FongC. S.et al (2021). Application of artificial intelligence methods for monsoonal river classification in Selangor river basin, Malaysia. Environ. Monit. Assess.193, 438. 10.1007/s10661-021-09202-y

  • CrossRef
  • Google Scholar

• 193

  XieP.ArkinP. A. (1996). Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Clim.9. 10.1175/1520-0442(1996)009<0840:AOGMPU>2.0.CO;2

  • CrossRef
  • Google Scholar

• 194

  XuL.LiQ.YuJ.WangL.XieJ.ShiS. (2020). Spatio-temporal predictions of SST time series in China’s offshore waters using a regional convolution long short-term memory (RC-LSTM) network. Int. J. Remote Sens.41, 3368–3389. 10.1080/01431161.2019.1701724

  • CrossRef
  • Google Scholar

• 195

  XuX.ZhanY.ZhengJ.GengB. (2021). “Classification of coastal altimetric waveforms using machine learning technology,” in 2021 4th International Conference on Information Communication and Signal Processing, ICICSP 2021, Shanghai, China, September 24–26, 2021 (IEEE (Institute of Electrical and Electronics Engineers)). 10.1109/ICICSP54369.2021.9611971

  • CrossRef
  • Google Scholar

• 196

  YaoH.YangY.FuX.MiC. (2018). An adaptive sliding-window strategy for outlier detection in wireless sensor networks for smart port construction. J. Coast. Res.82, 245–253. 10.2112/SI82-036.1

  • CrossRef
  • Google Scholar

• 197

  Yeganeh-BakhtiaryA.EyvazoghliH.ShabakhtyN.KamranzadB.AbolfathiS. (2022). Machine learning as a downscaling approach for prediction of wind characteristics under future climate change scenarios. Complexity2022. 10.1155/2022/8451812

  • CrossRef
  • Google Scholar

• 198

  Yeganeh-bakhtiaryA.GhorbaniM. A.PourzangbarA. (2012). Determination of the most important parameters on scour at coastal determination of the most important parameters on scour at coastal structures. J. Civ. Eng. Urbanism2, 68–71.

  • Google Scholar

• 199

  YuL.SunJ.GuoY.ZhangB.YangG.ChenL.et al (2022). Research on outlier detection in CTD conductivity data based on cubic spline fitting. Front. Mar. Sci.9. 10.3389/fmars.2022.1030980

  • CrossRef
  • Google Scholar

• 200

  ZanuttighB.FormentinS. M.van der MeerJ. W. (2016). Prediction of extreme and tolerable wave overtopping discharges through an advanced neural network. Ocean. Eng.127, 7–22. 10.1016/J.OCEANENG.2016.09.032

  • CrossRef
  • Google Scholar

• 201

  Zelada LeonA.HuvenneV. A. I.BenoistN. M. A.FergusonM.BettB. J.WynnR. B. (2020). Assessing the repeatability of automated seafloor classification algorithms, with application in marine protected area monitoring. Remote Sens.12, 1572. 10.3390/rs12101572

  • CrossRef
  • Google Scholar

• 202

  ZhuangX.LiW.XuY. (2022). Port planning and sustainable development based on prediction modelling of port throughput: a case study of the deep-water dongjiakou port. Sustainability14, 4276. 10.3390/su14074276

  • CrossRef
  • Google Scholar

• 203

  ZimekA.SchubertE.KriegelH. P. (2012). A survey on unsupervised outlier detection in high-dimensional numerical data. Stat. Analysis Data Min.5, 363–387. 10.1002/sam.11161

  • CrossRef
  • Google Scholar

• 204

  ZouS.ZhouX.KhanI.WeaverW. W.RahmanS. (2022). Optimization of the electricity generation of a wave energy converter using deep reinforcement learning. Ocean. Eng.244, 110363. 10.1016/J.OCEANENG.2021.110363

  • CrossRef
  • Google Scholar