Predicted Sea Level Research Articles

Sea Level Prediction Using Machine Learning

Sea level prediction is essential for the design of coastal structures and harbor operations. This study presents a methodology to predict sea level changes using sea level height and meteorological factor observations at a tide gauge in Antalya Harbor, Turkey. To this end, two different scenarios were established to explore the most feasible input combinations for sea level prediction. These scenarios use lagged sea level observations (SC1), and both lagged sea level and meteorological factor observations (SC2) as the input for predictive modeling. Cross-correlation analysis was conducted to determine the optimum input combination for each scenario. Then, several predictive models were developed using linear regressions (MLR) and adaptive neuro-fuzzy inference system (ANFIS) techniques. The performance of the developed models was evaluated in terms of root mean squared error (RMSE), mean absolute error (MAE), scatter index (SI), and Nash Sutcliffe Efficiency (NSE) indices. The results showed that adding meteorological factors as input parameters increases the performance accuracy of the MLR models up to 33% for short-term sea level predictions. Moreover, the results contributed a more precise understanding that ANFIS is superior to MLR for sea level prediction using SC1- and SC2-based input combinations.

Sea-level prediction for early warning information of coastal inundation in Belawan coastal area using Delft3D model

Coastal inundation has a great impact on the environment, such as damage to infrastructure and pollution of land and water. One of the efforts to prevent coastal inundation is to predict the water level. Delft3D is a hydrodynamic model that's able to simulate the water level. Coastal inundation research using the Delft3D model is still rarely done in Indonesia, especially on the east coast of Sumatra. This research is conducted in Belawan coastal area by simulating the water level that caused the coastal inundation using the Delft3D model. The best bathymetry for the prediction of water level and the magnitude of the wind effect was obtained from the simulation. The final step is to predict the water level in Belawan coastal area. The result of this research shows that the Delft3D model can simulate the water level which causes the coastal inundation in the Belawan coastal area. The correlation of the Delft3D model is 0.9, and the RMSE of GEBCO bathymetry is 0.39 meters and the RMSE of NOAA bathymetry is 0.46 meters. The GEBCO bathymetry is better than NOAA bathymetry in describing the water level in the Belawan coastal area. The wind effect on the water level simulations is not significant because the coefficient of determination is 0.47%. Besides, the Delft3D model with GEBCO bathymetry input can predict the water level which causes the coastal inundation with correlation reaches 0.92 and RMSE is 0.39 meters.

Sea level simulation with signal decomposition and machine learning

In recent years, under the background of global warming, influenced by changing climate conditions and human activities, the sea level time series have shown great non-stationarity and randomness, making it difficult to predict sea levels scientifically and accurately. Sea level simulations and predictions under climate change are hot topics in global environment and climate studies. In this paper, three prediction models based on signal decomposition (time-varying filtering based empirical mode decomposition (TVF-EMD), wavelet transform (WT), and complementary ensemble empirical mode decomposition (CEEMD) coupled with the Elman neural network (ENN) are established, and the prediction performances of the monthly mean sea level (MMSL, time series length between 16 and 61 years) at 12 stations and daily mean sea level (DMSL, time series length is 790 days) at six stations are evaluated. The results show that the MMSL has multiple time scales. Among the TVF-EMD-ENN, WT-ENN and CEEMD-ENN models, the TVF-EMD-ENN model is the optimal model for predicting the MMSL and DMSL, in MMSL and DMSL prediction, the Nash-Sutcliffe efficiency coefficient (NSE) range is 0.98–1.0 and 0.974–0.995, respectively, NSE is greater than 0.97 and performing better than the ENN and long short-term memory (LSTM) model. The difference in the prediction results between the TVF-EMD-ENN, WT-ENN, and CEEMD-ENN are mainly caused by the first subcomponent (Sc-1). The performance of the TVF-EMD-ENN model is optimal because the contribution of Sc-1 to the original sequence is small, and the complexity of the Sc-1 model is low. The predicted future MMSL exhibits a period of alternating sea level rise and decline, and the rate of increase is relatively low. The proposed approach has great significance for sea-level prediction and environmental management.

Forecasting of Sea Level Time Series using RNN and LSTM Case Study in Sunda Strait

Sea-level forecasting is essential for coastal development planning and minimizing their signi?cantconsequences in coastal operations, such as naval engineering and navigation. Conventional sealevel predictions, such as tidal harmonic analysis, do not consider the in?uence of non-tidal elementsand require long-term historical sea level data. In this paper, two deep learning approachesare applied to forecast sea level. The ?rst deep learning is Recurrent Neural Network (RNN), andthe second is Long Short Term Memory (LSTM). Sea level data was obtained from IDSL (InexpensiveDevice for Sea Level Measurement) at Sebesi, Sunda Strait, Indonesia. We trained themodel for forecasting 3, 5, 7, 10, and 14 days using three months of hourly data in 2020 from 1stMay to 1st August. We compared forecasting results with RNN and LSTM with the results of theconventional method, namely tidal harmonic analysis. The LSTM’s results showed better performancethan the RNN and the tidal harmonic analysis, with a correlation coef?cient of R2 0.97 andan RMSE value of 0.036 for the 14 days prediction. Moreover, RNN and LSTM can accommodatenon-tidal harmonic data such as sea level anomalies.

Using neural network to improve sea level prediction along the southeastern Brazilian coast

The objective of this study is to apply data-based models to improve sea level prediction along the southeastern Brazilian coast. Focus is placed on the atmospheric-forced low-frequency variability (∼0,3 to 0,03 day −1), which is associated primarily with propagation of Coastal Trapped Waves along the continental shelf, and possibly with local wind forcing. Process-based hydrodynamic models, that solve complex physical equations to simulate a dynamical system, have been traditionally used to predict sea level. These models, however, may present inaccuracies related to their physical formulation, numerical approximations, unresolved small scale processes, model parameterizations, or initial and boundary conditions. A combined approach, using a neural network to update the outputs of a hydrodynamic model (Hycom) in order to correct the errors, allows for the integration of all available knowledge to generate more reliable predictions. Time series of coastal sea level at three stations (Cananéia, Ilha Fiscal and Imbituba) and atmospheric data from the CFSR reanalysis were used to train the data-based model. A novel methodology to preprocess the input data is presented, and it was critical to the reduction of predictions errors. Correlation maps between coastal sea level and the atmospheric forcing were calculated, and the areas of highest correlation were defined. While a purely data-based model is able to match the predictions of the hydrodynamic model for these stations, the combined approach goes a step further, reducing the errors in ∼50% for the frequency band of the atmospherically forced sea level variability.

Evaluation of a global total water level model in the presence of radiational S[formula omitted] tide

The development of a computationally efficient scheme for predicting the global distribution of total water level (TWL) is discussed. The ocean model is barotropic, has a horizontal grid spacing of 1/12°, and is based on the NEMO modeling framework. It is forced by the gravitational potential and hourly atmospheric fields for 2008. Hourly time spacing was required to resolve the S2 tide in global air pressure and wind. The predicted tide in water deeper than 400 m was nudged to TPXO8 “observations” of tidal elevation or current using a scheme called tidal nudging (Kodaira et al., 2019). The benefit of nudging horizontal velocity in the momentum equation, compared to sea level in the continuity equation, is discussed. Tidal nudging is shown to improve tidal predictions of sea level at the coast, particularly at the S2 tidal frequency. The predicted radiational S2 tide in sea level forced solely by the S2 tide in global air pressure reaches amplitudes exceeding 80 cm. Decreasing the time spacing of the air pressure forcing from 1 h to 3 h reduces the S2 amplitude in air pressure by a factor of 0.82, consistent with expectations based on Fourier analysis. This highlights the importance of using hourly atmospheric forcing when predicting the global sea level response to atmospheric forcing. The radiational S2 tide in sea level is subject to strong nonlinear interaction with the gravitational tide, leading to a pronounced attenuation of the radiational S2 tide. The attenuation is explained by an increase in effective bottom friction at the S2 frequency due to the presence of the gravitational tide. Four schemes for predicting TWL are evaluated to quantify the impact of tidal nudging and nonlinear interaction of tide and surge. Using TWLs observed by 304 coastal tide gauges, we show it is necessary to include both tidal nudging and nonlinear interaction. Plans for the further development of an operational flood forecast system for the Canadian coast, based on the above model, are discussed.

Artificial Intelligence Based Prediction of Seawater Level: A Case Study for Bosphorus Strait

Sea level prediction is an important phenomenon for making reliable oceanographic and ship traffic management decisions especially for Bosphorus Strait that has no permanent sea level measurement stations due to high cost. This study presents artificial intelligence (AI) techniques, such as Artificial Neural Networks (ANNs) and Support Vector Machines (SVM) to predict the seawater level in the Bosphorus Strait. In addition, the Multiple Linear Regression model (MLR) is constructed and employed as a benchmark. The dataset employed in developing the models are wind speed, atmospheric pressure, water surface salinity, and temperature data, which were measured between September 2004 and January 2006. The results reveal that all ANN and SVM models outperform MLR and can predict the water levels quite accurately. ANN has a better performance than SVM for predicting sea level in the Bosphorus by coefficient of correlation (R) = 0.76 and root mean square error (RMSE) = 0.059. Moreover, the influence of the Danube River discharge in the prediction is investigated in the present study. The discharge of the Danube River by the lag time of 70 days yields the highest performance on ANN by increasing R to 0.82 and decreasing RMSE to 0.048.

Reducing Uncertainty in 21st Century Sea-Level Predictions and Beyond

Sea-level rise is one of the most critical issues the world faces under global warming. Around 680 million people (10% of the world’s population) live in low-lying coastal regions that are susceptible to flooding through storm surges and from sea-water infiltration of fresh groundwater reserves, degradation of farmland and accelerated coastal erosion, among other impacts. Rising sea level will exacerbate these problems and lead to societal impacts ranging from crop and water-supply failures to breakdowns of city infrastructures. In time, it is likely such changes will necessitate the migration of people with substantial economic cost and social upheaval. Here, we discuss the physical processes influencing 21st Century sea-level rise, the importance of not using 2100 alone as a benchmark, the changes that are already locked in, especially after 2100, and those that can be avoided. We also consider the need for both adaptation and mitigation measures and early warning systems in this challenging global problem. Finally, we discuss how the scientific prediction of sea level rise can improved through international coordination, cooperation and cost sharing.

Sea Level Rise Mitigation by Global Sea Water Desalination Using Renewable-Energy-Powered Plants

This work suggests a solution for preventing/eliminating the predicted Sea Level Rise (SLR) by seawater desalination and storage through a large number of desalination plants distributed worldwide; it also comprises that the desalinated seawater can resolve the global water scarcity by complete coverage for global water demand. Sea level rise can be prevented by desalinating the additional water accumulated into oceans annually for human consumption, while the excess amount of water can be stored in dams and lakes. It is predicted that SLR can be prevented by desalination plants. The chosen desalination plants for the study were Multi-Effect Desalination (MED) and Reverse Osmosis (RO) plants that are powered by renewable energy using wind and solar technologies. It is observed that the two main goals of the study are fulfilled when preventing an SLR between 1.0 m and 1.3 m by 2100 through seawater desalination, as the amount of desalinated water within that range can cover the global water demand while being economically viable.

Prediction and Allocation of EDP Based on Gray Model and Fuzzy Comprehensive Evaluation

Global warming is accelerating the sea level to rise, which increases the risk of major coastal areas being submerged. Then, the residents may become environmentally displaced persons (EDP). A method of prediction and allocation of EDP is proposed in this paper. Firstly, the gray model is used to predict sea level and the amount of EDP. Then, the evaluation criterion for the responsibility ability of the EDP migration target countries is given based on the entropy and fuzzy comprehensive evaluation method. In addition, a two-way selection mechanism for EDP is constructed. Finally, the amounts of EDP for the next 10 years are predicted, and the allocation plan in 2030 is made by applying the proposed method.

Machine learning methods applied to sea level predictions in the upper part of a tidal estuary

Sea levels variations in the upper part of estuary are traditionally approached by relying on refined numerical simulations with high computational cost. As an alternative efficient and rapid solution, we assessed here the performances of two types of machine learning algorithms: (i) multiple regression methods based on linear and polynomial regression functions, and (ii) an artificial neural network, the multilayer perceptron. These algorithms were applied to three-year observations of sea levels maxima during high tides in the city of Landerneau, in the upper part of the Elorn estuary (western Brittany, France). Four input variables were considered in relation to tidal and coastal surge effects on sea level: the French tidal coefficient, the atmospheric pressure, the wind velocity and the river discharge. Whereas a part of these input variables derived from large-scale models with coarse spatial resolutions, the different algorithms showed good performances in this local environment, thus being able to capture sea level temporal variations at semi-diurnal and spring-neap time scales. Predictions improved furthermore the assessment of inundation events based so far on the exploitation of observations or numerical simulations in the downstream part of the estuary. Results obtained exhibited finally the weak influences of wind and river discharges on inundation events.

A deep-learning model for national scale modelling and mapping of sea level rise in Malaysia: the past, present, and future

In this study, we conducted a holistic evaluation of current and future trend in coastal sea level at the 21 stations along Malaysia’s coastline. For sea level prediction, univariate and 3 scenarios of multivariate Long Short Term Memory (LSTM) neural networks were trained with absolute sea level data and ocean-atmospheric variables. The result from the four scenario predictive models revealed that multivariate LSTM neural network trained with combined ocean-atmospheric variables performed best for modelling sea level variation, giving a mean RMSE and R accuracy of 0.060 and 0.861, respectively. The national sea level rise estimated from the average of sea level trend at all stations is 3.72 mm/yr for relative sea level and 3.68 mm/yr for absolute sea level. The 2050 and 2100 projections indicate that sea level will continue to rise but at a very slow rate with no acceleration.

Development of a Training Set of Contemporary Salt-Marsh Foraminifera for Late Holocene Sea- Level Reconstructions in southeastern Australia

We collected contemporary foraminiferal training sets from two salt marshes to enable more precise and accurate proxy historical sea-level reconstructions from southeastern Australia. Combined with an existing training set from Tasmania, this new regional set consists of 112 samples and 16 species of foraminifera, of which 13 are agglutinated. Cluster analyses group the regional training set into a high–elevation cluster, dominated by Trochamminita salsa, a mid–elevation cluster, dominated by Entzia macrescens and Trochammina inflata, and a mid–low elevation cluster dominated by Miliammina fusca and tidal-flat species. We develop transfer functions using local and regional training sets and assess their performance. Our resulting site-specific and chosen regional models are capable of predicting sea level with decimetre-scale precision (95% confidence intervals of 0.12–0.22 m). These results are comparable to other examples from around the world. When developing regional training sets, we advocate that the similarity in the environmental settings (particularly salinity) should be assessed as an alternative way of grouping sites, rather than simply using spatial proximity. We compare our findings with global results and conclude that salt marshes along microtidal coasts yield models with the lowest vertical uncertainties. Studies with the lowest uncertainties are located in the western Pacific and the western Atlantic, whereas those from the eastern Atlantic generally have larger tidal ranges and carry larger vertical uncertainties. Our models expand the existing region available for sea-level reconstruction and can be used to generate new late Holocene sea-level reconstructions across southeastern Australia.

Millennial-scale vertical deformation of the Hachinohe coastal plain (NE Japan)

We report geomorphological and sedimentological analyses in the Hachinohe plain (northernmost Sanriku coast, NE Japan) that yield new insights into vertical deformation around the source region of the 2011 Tohoku-oki earthquake. Landform classification and analyses of seven sediment cores document the transgression and regression of a sandy barrier and back-barrier estuary in a wave-dominated estuary system during the Holocene. We compare the heights of (i) the earliest Holocene marine flooding surface overlying fluvial sediments and (ii) the onset of a Middle Holocene marsh environment with non-tectonic hydro-isostatic sea-level predictions to constrain the millennial-scale vertical deformation rate in the Hachinohe plain to within −0.7 to +0.7 mm/yr (uplift positive). This result is in contrast to observed subsidence along the central to southern Sanriku coast but is consistent with relative uplift along the northern Sanriku coast. Our results support deformation characterized by southward tilting along the Sanriku coast during the last 100 kyr as suggested by previous studies south of Hachinohe. Furthermore, this tectonic pattern is consistent with deformation trends before and during the 2011 Tohoku-oki earthquake. This collective consistency implies that the effects of subduction and megathrust slip on coastal deformation have persisted for at least 100 kyr around the source region of the 2011 earthquake.

Time-frequency dependency of temperature and sea level: a global perspective.

The importance of environmental sustainability to all human aspects has been spiking as the world keeps evolving. Economies around the world are on the move to ensure sustainable economic development and a clean atmosphere through the use of renewable energy sources. Since the end of the nineteenth century, it is recognized that a global mean sea level rise speed about 1.7 ± 0.2 mm/year, but the rate has increased to 3.2 ± 0.4 mm/year. In this study, we investigated the dynamic linkage between average temperature and sea level within the global framework covering 1881 to 2013. The paper employs wavelet analysis to investigate the short-term and long-run causal links between global average temperature anomalies and global sea level. In this respect, our findings indicate that (i) a significant vulnerability in global average temperature and sea level is observed over the selected study period; (ii) global average temperature has considerable power for predicting sea level, particularly in the long-term. The causality test revealed a bi-directional causal relationship between global average temperature and sea level and unidirectional flow from global average temperature to sea level. And the study confirms the "conservation hypothesis," and it has long-run implications for environmental quality. Thus, minimizing global average temperature anomalies is a decisive ingredient for minimizing sea level rise. Based on these outcomes, both developed and developing countries' policymakers should support the Paris agreement (COP21) agreement to control CO2 emissions growth. Policymakers should encourage the usage of environmentally friendly energy sources that will enhance environmental quality and hold the increase in global average temperature below 1.5 °C.

Seasonal Forecasting Skill of Sea‐Level Anomalies in a Multi‐Model Prediction Framework

AbstractCoastal high water level events are increasing in frequency and severity as global sea‐levels rise, and are exposing coastlines to risks of flooding. Yet, operational seasonal forecasts of sea‐level anomalies are not made for most coastal regions. Advancements in forecasting climate variability using coupled ocean‐atmosphere global models provide the opportunity to predict the likelihood of future high water events several months in advance. However, the skill of these models to forecast seasonal sea‐level anomalies has not been fully assessed, especially in a multi‐model framework. Here, we construct a 10‐model ensemble of retrospective forecasts with future lead times of up to 11 months. We compare predicted sea levels from bias‐corrected forecasts with 20 years of observations from satellite‐based altimetry and shore‐based tide gauges. Forecast skill, as measured by anomaly correlation, tends to be highest in the tropical and subtropical open oceans, whereas the skill is lower in the higher latitudes and along some continental coasts. For most locations, multi‐model averaging produces forecast skill that is comparable to or better than the best performing individual model. We find that the most skillful predictions typically come from forecast systems with more accurate initializations of sea level, which is generally achieved by assimilating altimetry data. Having relatively higher horizontal resolution in the ocean is also beneficial, as such models seem to better capture dynamical processes necessary for successful forecasts. The multi‐model assessment suggests that skillful seasonal sea‐level forecasts are possible in many, though not all, parts of the global ocean.

Ocean Circulation Model Applications for the Estuary-Coastal-Open Sea Continuum

Coastal zones are among the most variable environments. As such, they require adaptive water management to ensure the balance of economic and social interests with environmental concerns. High quality marine data of hydrographic conditions e.g., sea level, temperature, salinity, and currents are needed to provide a sound foundation for the decision making process. Operational models with sufficiently high forecasting quality and resolution can be used for a further extension of the marine service toward the coastal-estuary areas. The Limfjord is a large and shallow water body in Northern Jutland, connecting the North Sea in the West and the Kattegat in the East. It is currently not covered by the CMEMS service, despite its importance for sea shipping, aquaculture and mussel fisheries. In this study, we use the operational HIROMB-BOOS Model (HBM) to resolve the full Baltic-Limfjord-North Sea system with a horizontal resolution of 185.2 m in the Limfjord. The study shows several factors that are essential for successfully modeling the coastal-estuary system: (a) high computational efficiency and flexible grids to allow high resolution in the fjord, (b) an improved short wave radiation scheme to model the thermodynamics and the diurnal variability of the temperature in very shallow waters, (c) high resolution atmospheric forcing, (d) adequate river forcing, and (e) accurate bathymetry in the narrow straits. With properly resolving these issues, the system is able to provide high quality sea level forecast for storm surge warning and hydrography forecasts: temperature, salinity and currents with sufficiently good quality for ecosystem-based management. The model is able to simulate the complex spatial and temporal pattern of sea level, salinity and temperature in the Limfjord and to reproduce their diurnal, seasonal and interannual variability and stratification rather well. Its high computational efficiency makes it possible to model the transition from the basin-scales to coastal- and estuary-scales seamlessly. In total, The HBM model has been successfully extended, to include the complex estuary system of the Limfjord, and shows an adequate model performance with regards to sea level, salinity and temperature predictions, suitable for storm surge warning applications and coastal management applications.

Transfer sea level learning in the Bosphorus Strait by wavelet based machine learning methods

Accurate prediction of sea level fluctuations is fundamental to coastal engineering. However, limited monitoring service on a regional scale is the most important constraint in analyzing, identifying, or predicting sea-level fluctuations driven with several nonlinearly integrated deterministic processes. The present study investigates the prediction performance of machine learning (ML) models based on available sea level time series information recorded in different conditions. Discrete Wavelet Transform (DWT) combined with Support Vector Machine (SVM), k-Nearest Neighbor (KNN), and Decision Tree (DT) methods were used to transfer sea level information between 3 stations located in different regions of the Bosphorus Strait. The developed models are tested in predicting sea level lead-time up to 7 days based on the Root Mean Square Errors (RMSE) and Nash-Sutcliffe Efficiency (NSE) indicators. Modeling strategy is determined by taking the sensitivity of a classical regression technique, Multi-linear Regression (MLR) into account, to additional decomposition or standardization processes. The developed models are found to be more successful in the information transfer between spatially close stations than periodically close stations. Considering the relative success of ML methods in defining the sea level fluctuations, SVM and KNN models provide relatively close results while DT model results are far behind the others.

A multi-model architecture based on Long Short-Term Memory neural networks for multi-step sea level forecasting

The intensification of extreme events, storm surges and coastal flooding in a climate change scenario increasingly influences human processes, especially in coastal areas where sea-based activities are concentrated. Predicting sea level near the coasts, with a high accuracy and in a reasonable amount of time, becomes a strategic task. Despite the developments of complex numerical codes for high-resolution ocean modeling, the task of making forecasts in areas at the intersection between land and sea remains challenging. In this respect, the use of machine learning techniques can represent an interesting alternative to be investigated and evaluated by numerical modelers.This article presents the application of the Long-Short Term Memory (LSTM) neural network to the problem of short-term sea level forecasting in the Southern Adriatic Northern Ionian (SANI) domain in the Mediterranean sea. The proposed multi-model architecture based on LSTM networks has been trained to predict mean sea levels three days ahead, for different coastal locations. Predictions were compared with the observation data collected through the tide-gauge devices as well as with the forecasts produced by the Southern Adriatic Northern Ionian Forecasting System (SANIFS) developed at the Euro-Mediterranean Center on Climate Change (CMCC), which provides short-term daily updated forecasts in the Mediterranean basin. Experimental results demonstrate that the multi-model architecture is able to bridge information far in time and to produce predictions with a much higher accuracy than SANIFS forecasts.

Alternative approaches to medium-long term sea level rise mapping in Southern Miami Beach (Florida, USA)

The purpose of this paper is to evaluate three temporal mapping approaches to predict Sea Level Rise (SLR) in the Southern Miami-Dade County, Florida, USA. These three temporal approaches provide an alternative to SLR prediction by the common binary method based on Digital Elevation Models (DEMs). The first temporal approach gathers seven theoretical and semi-empirical scenarios of SLR by the end the century in a single map. The second temporal approach is based on the calculation of the time horizon to inundate each cell of the DEM during ordinary high tides. Finally, the third temporal approach maps the minimum average rate of SLR by which a cell will be inundated by the year 2100. These approaches indicate that, in the second half of the 21st century, a large area on the coast of Miami will be inundated due to SLR. A survey conducted with a group of 73 experts concluded that these approaches were more suitable than other classical approaches for mapping SLR in urban areas.