Elsevier

Applied Ocean Research

Volume 47, August 2014, Pages 344-351
Applied Ocean Research

Prediction of sea water levels using wind information and soft computing techniques

https://doi.org/10.1016/j.apor.2014.07.003Get rights and content

Highlights

  • Sea water levels at four tidal stations along the USA coastline are predicted.

  • Estimation models and 1 day forecast models are developed.

  • Shear wind velocity components of the present time and up to past 12 h are used as inputs.

  • Genetic Programming and Artificial Neural Network techniques are employed.

  • Water level plots, scatter plots and the error measures indicate satisfactory predictions.

Abstract

Large variations of sea water levels are a matter of concern for the offshore and coastal locations having shallow water depths. Safety of maritime activities, and properties, as well as human lives at such locations can be ensured by using the accurately predicted water levels. Harmonic analysis is traditionally employed for tide predictions, but often the values of predicted tides and observed (measured) water levels are not identical. The difference between them is called sea level anomaly. This can be attributed to non-inclusion of meteorological parameters as an input for tide prediction. Therefore other prediction techniques become necessary. The earlier studies on sea level predictions indicate better efficiency of alternate techniques such as Artificial Neural Network (ANN) and Genetic Programming (GP), and that most researchers have used sea level time series as model inputs. Present work predicts sea levels indirectly by predicting sea level anomalies (SLAs) using hourly local wind shear velocity components of the present time and up to the previous 12 h as inputs at four stations near the USA coastline with the techniques of GP and ANN. The error measures and graphs indicate that predictions are satisfactory.

Introduction

Knowledge of sea water levels and their variations is required for planning, operation, and maintenance works such as construction of jetties and harbors in the coastal and offshore locations, grounding of ships, navigation of the vessels with deep draft, installation of platforms, loading and unloading in the high tide zones. Tidal fluctuations causing large variations in the waterfront distances in low lying areas, small islands, and the lands with gentle slope are critical for the safety of properties and human lives. Working and safety of ocean based nonconventional energy power plants also depends on such fluctuations. A significant increase in the hydro-metrological events such as rise of the sea water levels, frequency of occurrence of cyclones and their severity is experienced in the recent past all over the globe. This highlights reliable and accurate prediction of sea water levels as one of the major challenges for the researchers.

Variations of the sea levels are produced by combination of complex processes involving forces of attraction of the Moon and the Sun on the Earth, bathymetric characteristics as well as meteorological parameters like the atmospheric pressure, air temperature, water temperature, ocean currents, wind, etc. [1]. The observed (or measured) sea water levels consist of tidal and non-tidal portions. The tidal or astronomical portions are created by the astronomical alignment of the Sun–Earth–Moon system while the non-tidal or non-astronomical portions are caused by the combined effect of other parameters. Accuracy of traditional harmonic models depends on the number of harmonics considered, which in turn depends on the length of data used for modeling. Deo and Chaudhari [2] mentioned that in spite of considering as many as 69 harmonic constituents for tide prediction, the harmonic analysis becomes ineffective if non-periodic meteorological events such as hurricanes and cold fronts predominate. Makarynska and Makarynskyy [3] stated that the conventional harmonic analysis for sea level predictions may suffer from 30% residual errors if the hydro-meteorological effects involved in the process are neglected.

Therefore alternate modeling techniques and strategies are tried out by ocean researchers for improving sea level predictions. In the classical approach based on harmonic analysis, tides are considered as a resultant of different harmonics and their frequencies derived from astronomical observations. Other techniques such as analysis of tides as propagation of long waves, regression, and the numerical model in combination with tidal dynamic equations are also tried in sea level research. Chang and Lin [4] reported that numerical method can catch the tidal characteristics successfully even if the hydrodynamic characteristics of tides are not known but generation of the numerical grid requires a lot of time for predicting the tides.

Traditionally the emphasis of knowledge discovery has been on some theory that demands appropriate data obtained through observations or experiments. In this process, a theory based model is obtained in mathematical form; which can be termed as ‘theory-driven approach’. It makes an extensive use of the associated mathematical methods and the physical process or processes involved in the phenomena. On the other hand, the ‘data-driven approach’ utilizes input–output data sets to derive a model that fits the data optimally. It aims at providing the tools to facilitate conversion of data into a convenient form to convey a better understanding of the processes hidden in these data and a mathematical statement of the relations between various parameters in the data can be obtained [5]. Babovic and Abbott [6] explained three criteria for an evolutionary process to occur as given by Maynard-Smith [7] and three case studies of evolution of equations from the hydraulic data [8]. Hybrid techniques combining deterministic numerical model and chaos theory is tried out by Sannasiraj et al. [9], Wang et al. [10], etc. for water level predictions. Modified strategies such as data assimilation, application of error corrections through neural networks, etc. along with Singapore Regional (numerical) Model are demonstrated by Kurniawan et al. [11], Sun et al. [12], Karri et al. [13] in the Singapore Regional Waters for improving water level predictions.

Soft computing techniques such as the Artificial Neural Network (ANN) and Genetic Programming (GP) fall in the domain of data-driven techniques. Some advantages of the data-driven techniques are: (1) understanding of the underlying physical processes is not a pre-requisite, (2) modeling data need not be huge and exogenous, (3) re-calibration of the developed models can be done in comparatively less time and with less efforts, (4) a reasonable generalization is extracted from the data in spite of the complex physical processes involved in the phenomena, and (5) the developed models are more robust since they are more data tolerant.

Present work follows the suite of Cox et al. [14] and Londhe [15] in that the soft technique is employed for predicting the sea level anomaly and water level is obtained subsequently by adding the predicted anomaly to the harmonic tidal level. A series of wind shear velocity components of 1–3 years is supplied to the model as inputs and the models are developed using the techniques of GP and ANN at four stations near the USA coastline. The methodology is explained in Section 3. The innovation in the present work is the use of hourly wind shear velocity component series – the major causal parameter – as model input for predicting non-astronomical components of water levels (sea level anomaly).

The paper is organized as follows: literature review is summarized in Section 2. The study area and data are described in Section 3. The methodology adopted and model development process is detailed out in Section 4. Results are discussed in Section 5 and Section 6 gives the conclusions and scope for further research.

Section snippets

Literature review

Ample literature is available on sea level predictions using traditional as well as alternate approaches. Literature review pertaining to scope of the present work is summarized in this section.

Data-driven techniques include soft computing techniques, which are found to be more versatile for the phenomena such as ocean waves, water level variations, rainfall, changes in the meteorological parameters, etc. Such phenomena involve nonlinear and complex relationships between several variables.

An

Study area and modeling data

In the present work, sea level anomaly prediction models are developed at four tidal stations near the USA coastline. Locations of these stations are shown in Fig. 1.

Station codes, names, their locations and anemometer elevations at these stations (required for computing the wind shear velocity components) are given in Table 1.

Hourly observed sea levels and harmonically predicted tidal levels as well as meteorological data (wind speed and wind directions) at these stations for 4–5 years were

Modeling procedure

Since the difference between the observed sea water level and corresponding harmonically predicted tidal level is sea level anomaly (SLA), the harmonic tidal level is the astronomical portion while anomaly is the non-astronomical portion of the sea water level.

Alternate techniques of GP and ANN are employed in this work to predict the hourly sea water levels indirectly by first predicting the sea level anomalies (SLAs) at four tidal stations near the US coastline, maintained by National Oceanic

Results and discussion

Estimation and 1 day forecast of sea levels at four stations along the USA coastline were made with the help of the soft computing techniques of Genetic Programming and Artificial Neural Networks using the present and the past wind shear velocity components as inputs. Just like other ocean researchers the accuracy yielded by GP models is better than the ANN models. Qualitative assessment of Models is made using (i) water level plots, and (ii) scatter plots between the observed and predicted sea

Conclusions and scope for future research

Sea water levels are predicted satisfactorily at four stations along the USA coastline with the soft computing techniques of GP and ANN using present and past wind shear velocity components. The accuracy of indirect water level estimation as well as 1 day water level forecasts using wind shear velocity components and soft computing tools is satisfactory. It has been seen that estimations are better than the forecasts. Thus the prediction accuracy decreases as the time horizon increases.

References (35)

  • S.P. Nitsure et al.

    Wave forecasts using wind information and genetic programming

    Ocean Eng

    (2012)
  • M.A. Ghorbani et al.

    Sea water level forecasting using genetic programming and comparing the performance with artificial neural networks

    J Comp Geosci

    (2010)
  • M.C. Deo et al.

    Tide prediction using neural networks

    J Comp-Aided Civil Infrastruct Eng

    (1998)
  • V. Babovic et al.

    Genetic programming as a model induction engine

    J Hydroinform

    (2000)
  • V. Babovic et al.

    The evolution of equations from hydraulic data. Part I: theory

    J Hydraul Res

    (1997)
  • J. Maynard-Smith

    The theory of evolution

    (1975)
  • V. Babovic et al.

    The evolution of equations from hydraulic data. Part II: applications

    J Hydraul Res

    (1997)
  • Cited by (27)

    • Sea level simulation with signal decomposition and machine learning

      2021, Ocean Engineering
      Citation Excerpt :

      Due to the dual effects of climate change and human activities, global hydrometeorological events have increased significantly in recent years, and sea-level rise has intensified, indicating that reliable and accurate sea-level prediction is a significant challenge for researchers (Nitsure et al., 2014).

    • Blended wind fields for wave modeling of tropical cyclones in the South China Sea and East China Sea

      2018, Applied Ocean Research
      Citation Excerpt :

      Long-term reanalysis wind data obtained from data assimilation models have been widely used as wind data products, such as the National Centers for Environmental Prediction (NCEP) [24,25], the European Centre for Medium-Range Weather Forecasts (ECMWF) [26,27], the National Aeronautics and Space Administration (NASA) [28,29], and the Japanese Reanalysis (JRA) [30,31]. Due to their accessibility and accuracy, they are widely used in many related studies, such as analysis of wind and wave characteristics [32–34], wave energy assessments [35–38], prediction of sea water levels [39], hindcasting and prediction of waves [40–43], maritime activity safety assessments [44,45]. In the other aspects, various parametric models have been proposed (e.g. Fujita [46], Jelesnianski [47,48], Wadachi [49], Jelesnianski and Taylor [50], Holland [51], Wijnands et al. [52], Olfateh et al. [53]) for generating realistic air pressure and surface wind distributions under TCs.

    • Investigating the Role of Atmospheric Variables on Sea Level Variations in the Eastern Central Red Sea Using an Artificial Neural Network Approach

      2020, Oceanologia
      Citation Excerpt :

      Cox et al. (2002) was able to improve the sea level predictions of a simple linear model by using a neural network approach with wind components, barometric pressure, and the previous sea level anomaly used as inputs. Using wind shear velocity components as inputs, Nitsure et al. (2014) showed that both Genetic Programming (GP) and Artificial Neural Network (ANN) soft-computing techniques performed well in predicting the sea level anomaly. Using an ANN approach, Moghadam (2016) found that wind velocity, sea surface temperature, and atmospheric pressure are the most significant meteorological factors affecting the regional sea level anomaly.

    View all citing articles on Scopus
    View full text