An Improved Estimation and Gap-Filling Technique for Sea Surface Wind Speeds Using NARX Neural Networks

This work presents the use of a nonlinear autoregressive neural network to obtain an improved estimate of sea surface winds, taking Placentia Bay, Newfoundland and Labrador, Canada, as a study case. The network inputs and delays were chosen through cross correlation with the target variable. The proposed method was compared with five other wind speed estimation techniques, outperforming them in correlation, precision, accuracy, and bias levels. As an extension, the temporal gap filling of missing wind speed data during a storm has been considered. Data containing a measurement gap from a 40-yr windstorm that hit the same location has been used. The proposed method filled the gaps in the dataset with a high degree of correlation with measurements obtained by surrounding stations. The method presented in this work showed promising results that could be extended to estimate wind speeds in other locations and filling gaps in other datasets.

[1]  Mark D. Powell,et al.  The Transition of the Hurricane Frederic Boundary-Layer Wind Field from the Open Gulf of Mexico to Landfall , 1982 .

[2]  Rohit Jain,et al.  A Coupled Numerical and Artificial Neural Network Model for Improving Location Specific Wave Forecast , 2016 .

[3]  Hao Zhou,et al.  Wind Speed Inversion in High Frequency Radar Based on Neural Network , 2016 .

[4]  Shang-Ping Xie,et al.  Wave- and Anemometer-Based Sea Surface Wind (WASWind) for Climate Change Analysis* , 2010 .

[5]  C. Vincent,et al.  Estimation of winds over the Great Lakes , 1976 .

[6]  Christopher S. Ruf,et al.  Bayesian Wind Speed Estimation Conditioned on Significant Wave Height for GNSS-R Ocean Observations , 2017 .

[7]  Sergios Theodoridis,et al.  Surface wind speed extraction from HF sky wave radar Doppler spectra , 1982 .

[8]  Mohammad Reza Nikoo,et al.  Wave Height Prediction Using Artificial Immune Recognition Systems (AIRS) and Some Other Data Mining Techniques , 2017, Iranian Journal of Science and Technology, Transactions of Civil Engineering.

[9]  Ed Hawkins,et al.  An empirical model for probabilistic decadal prediction: global attribution and regional hindcasts , 2017, Climate Dynamics.

[10]  M. Wei,et al.  Evaluation of a Support Vector Machine–Based Single-Doppler Wind Retrieval Algorithm , 2017 .

[11]  Marcello Passaro,et al.  Validation of Significant Wave Height From Improved Satellite Altimetry in the German Bight , 2015, IEEE Transactions on Geoscience and Remote Sensing.

[12]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[13]  Eduardo D. Sontag,et al.  Feedback Stabilization Using Two-Hidden-Layer Nets , 1991, 1991 American Control Conference.

[14]  D Hoffman,et al.  THE THEORY OF THE RAYLEIGH DISTRIBUTION AND SOME OF ITS APPLICATIONS , 1975 .

[15]  Stephan D. Howden,et al.  Hurricane Katrina Winds Measured by a Buoy-Mounted Sonic Anemometer , 2008 .

[16]  C. L. Bretschneider.  The generation and decay of wind waves in deep water , 1952 .

[17]  George Cybenko,et al.  Approximation by superpositions of a sigmoidal function , 1989, Math. Control. Signals Syst..

[18]  Weimin Huang,et al.  HF radar wave and wind measurement over the Eastern China Sea , 2002, IEEE Trans. Geosci. Remote. Sens..

[19]  Nan Li,et al.  A support vector machine‐based VVP wind retrieval method , 2015 .

[20]  Hava T. Siegelmann,et al.  Computational capabilities of recurrent NARX neural networks , 1997, IEEE Trans. Syst. Man Cybern. Part B.

[21]  P. Dexter.  Tests on Some Programmed Numerical Wave Forecast Models , 1974 .

[22]  M. C. Deo,et al.  Real-time wave forecasting using genetic programming , 2008 .