Annual and Interannual Variability of Scatterometer Ocean Surface Wind over the South China Sea

ZHANG Guosheng, XU Qing,, GONG Zheng, CHENG Yongcun, WANG Lei, and JI Qiyan

1) Key Laboratory of Coastal Disasters and Defence, Ministry of Education, Hohai University, Nanjing 210098, P. R. China

2) College of Harbor, Coastal and Offshore Engineering, Hohai University, Nanjing 210098, P. R. China

3) State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Hohai University, Nanjing 210098, P. R. China

4) National Space Institute, Technical University of Denmark, Copenhagen 2100, Denmark

Annual and Interannual Variability of Scatterometer Ocean Surface Wind over the South China Sea

ZHANG Guosheng1),2), XU Qing1),2),*, GONG Zheng3), CHENG Yongcun4), WANG Lei1),2), and JI Qiyan1),2)

1) Key Laboratory of Coastal Disasters and Defence, Ministry of Education, Hohai University, Nanjing 210098, P. R. China

2) College of Harbor, Coastal and Offshore Engineering, Hohai University, Nanjing 210098, P. R. China

3) State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Hohai University, Nanjing 210098, P. R. China

4) National Space Institute, Technical University of Denmark, Copenhagen 2100, Denmark

To investigate the annual and interannual variability of ocean surface wind over the South China Sea (SCS), the vector empirical orthogonal function (VEOF) method and the Hilbert-Huang transform (HHT) method were employed to analyze a set of combined satellite scatterometer wind data during the period from December 1992 to October 2009. The merged wind data were generated from European Remote Sensing Satellite (ERS)-1/2 Scatterometer, NASA Scatterometer (NSCAT) and NASA’s Quick Scatterometer (QuikSCAT) wind products. The first VEOF mode corresponds to a winter-summer mode which accounts for 87.3% of the total variance and represents the East Asian monsoon features. The second mode of VEOF corresponds to a spring-autumn oscillation which accounts for 8.3% of the total variance. To analyze the interannual variability, the annual signal was removed from the wind data set and the VEOFs of the residuals were calculated. The temporal mode of the first interannual VEOF is correlated with the Southern Oscillation Index (SOI) with a four-month lag. The second temporal interannual VEOF mode is correlated with the SOI with no time lag. The time series of the two interannual VEOFs were decomposed using the HHT method and the results also show a correlation between the interannual variability and El Niño-Southern Oscillation (ENSO) events.

ocean surface wind; annual and interannual variability; scatterometer; South China Sea

1 Introduction

The South China Sea (SCS), with a total area of about 3.5×106 km2, is the largest semi-enclosed sea in the western tropical Pacific Ocean. It is under the influence of monsoon winds and synoptic systems such as fronts and tropical cyclones (Chu et al., 1999a, b, 2000). The East Asian summer monsoon brings precipitation from the ocean to the East Asian continent. The monsoon rainfall is beneficial to agriculture in normal years, but may cause flood disasters in anomalous years (Domrös, 1988).

Studies of the wind system and its relationship with ENSO over the tropical and northwestern Pacific have been conducted for a long time (e.g., Pan et al., 2002; Ho et al., 2008). The SCS climate, which is part of the East Asian monsoon system, also appears to be connected with ENSO in the Pacific (Qu et al., 2004). The connection between the SCS climate and the Pacific ENSO has become a very active area of research.

In this study, the wind system over the SCS was analyzed using seventeen years of monthly satellite wind vector observations from the combined ERS-1/2, NSCAT and QuikSCAT scatterometers. The VEOF method was used to investigate the annual and interannual variability of wind fields, and the HHT, a non-linear and non- stationary time series processing method, was applied to study the detail of interannual oscillations of ocean surface winds.

Section 2 describes the data and methodology. Section 3 presents the interpretation of VEOF decomposition results. Section 4 provides the interannual VEOF modes and the HHT results. The conclusions are drawn in Section 5.

2 Data and Methodology

2.1 Scatterometer Winds

It has long been known that the ocean surface wind field can be observed from space by satellite scatterometers (Xu et al., 2010, 2011), the wind products of which have been widely used for weather and climate studies (Chelton and Freilich, 2005). In this study, the base data sets consist of monthly wind vectors from ERS-1/2, NSCAT and Quik-SCAT scatterometers over the SCS (0°–23°N, 99°–121°E) during December 1992 to October 2009. Table 1 shows the wind obseration periods for each scatterometer. ERS-1/2 winds are processed in the Centre ERS d’Archivage et de Traitement (CERSAT) of the French Research Institute for Exploitation of the Sea (IFR EMER). NSCAT and QuikSCAT winds are from Physical Oceanography Distributed Active Archive Center (PO. DAAC).

Table 1 Satellite scatterometer winds used in this study

Bentamy et al. (2002) and Pan et al. (2002) showed the consistency between different scatterometer winds. To further compare the ERS-2, NSCAT and QuikSCAT wind vectors in this study, the ERS-2 winds were compared with NSCAT and QuikSCAT winds, respectively, during the overlapping observation periods (Fig.1). The comparison statistics shown in Table 2 indicate a good agreement among the different winds with correlation coefficients greater than 0.97. The root mean square errors (RMSEs) between ERS-2 and NSCAT wind speed are smaller than 0.8 m s−1, and those between ERS-2 and QuikSCAT wind speeds are smaller than 1.1 m s−1, which are much smaller than those between scatterometer and buoy winds. The NSCAT winds fit the ERS-2 winds particularly well.

The different scatterometer winds were also compared at three randomly selected locations in the study area. The results shown in Fig.2 also indicate a good agreement among these winds.

The ERS-1/2 winds were originally measured in a spatial grid of 1°×1°. These data were interpolated to the NSCAT and QuikSCAT wind grids (0.5°×0.5°) using the bilinear interpolation method. During the overlapping periods, wind data from different sources were averaged. The merged monthly winds with a spatial grid of 0.5°× 0.5° provide the basis for this study.

Fig.1 Comparisons of zonal (left panel) and meridional (right panel) wind components between ERS-2 and NSCAT (a, b), and ERS-2 and QuikSCAT (c, d) wind products.

Table 2 Comparison statistics of Fig.1

Fig.2 Time series of zonal (left panel) and meridional (right panel) wind components of ERS-2 (solid line), NSCAT (*) and QuikSCAT (+) at three locations in the South China Sea.

2.2 VEOF Method

The EOF analysis decomposes time series of spatial measurements into several important orthogonal modes in space and time to capture the main variability in atmospheric and oceanographic observations (Kelly, 1998). In this study, a method suggested by Hardy and Walton (1978) was used to capture the annual and interannual variability of surface wind over the SCS by applying the EOF decomposition to the satellite wind data set.

First, the zonal (u) and meridional (v) component matrices with m grid points and n months were merged to form a new 2m×n matrix S as

The singular value decomposition (SVD) was then applied to the matrix S:

where Tnis the n-th temporal EOF; Cnis the weighted coefficient for the n-th EOF；and Snis the n-th spatial EOF. In Eq. (2), Sncan be decomposed into spatial vector shown in the following as Vn. The decomposition is changed into

2.3 Hilbert-Huang Transform

The Hilbert-Huang transform (HHT) is a method for analyzing nonlinear and non-stationary processes (Huang et al., 1998), which combines the empirical mode decomposition (EMD) method and the associated Hilbert spectral analysis. It is the first local and adaptive method in the frequency-time analysis, which extracts more intrinsic characteristics of a data set than other methods, such as the spectrogram, wavelet analysis, the Wigner-Vile distribution (Pan et al., 2002). The essence of the EMD is to identify the intrinsic oscillatory modes by their characteristic time scales in data, and then decompose the data accordingly. The EMD is given as

where X(t) is the original data, ciis the i-th empirical mode of an intrinsic mode function (IMF), and rnis a residue that can either show a mean trend or be a constant. Any function is an IMF if (a) the numbers of extrema and zero-crossings in a dataset are either equal or differ at most by one, and (b) the mean value of the envelopes defined by local maxima and local minima is zero at any point. For an IMF time series, ci, the Hilbert transform, Yi(t), can be defined as

With this definition, ci(t) and Yi(t) form a complex conjugate pair, so an analytical signal, Zi(t), can be obtained as

and its phase function θi(t) is

The instantaneous frequency is defined as

The time-dependent spectrum can be derived once the instantaneous frequency of a time series has been generated. In this study, HHT is applied to analyze intrinsic oscillations of interannual temporal VEOF modes.

3 Annual Variability

By removing the temporal mean from the merged scatterometer winds and suppressing the high-frequency variability by a three-point moving average, the decomposition was performed. The VEOF decomposition results show that the first VEOF (VEOF-1) and second VEOF (VEOF-2) of the SCS surface wind field account for 87.3% and 8.3% of the total variance, respectively, which correspond to different annual oscillations.

3.1 VEOF-1

The winter-summer mode and the East Asian monsoon feature are two important characteristics of the first VEOF of the SCS surface wind field. As shown in Fig.3a, the spatial mode of VEOF-1 is dominated by the northeasterly wind over the SCS, and a cross-equator airflow appears over the equatorial region between 103°E and 109°E. Fig.3b shows that the annual oscillation is a typical temporal feature of VEOF-1. In an annual cycle, the temporal mode is close to zero in spring and reaches a negative maximum in summer. Because of this negative maximum, the wind direction over the SCS reverses to the northeast. Following the summer, the temporal mode rises and reaches a peak value in winter. Furthermore, an interannual variability can be identified from the winter-summer mode, the further studies of which will be given in Section 4.2. In terms of annual oscillation, it is meaningful to demonstrate that the VEOF-1 of SCS is dominated by the East Asian monsoon which has the largest amplitude near the southeast of the Indo-China Peninsula. These results are very similar to those found by Pan et al. (2002) based on the analysis of eight years of scatterometer winds over the northwestern Pacific.

Fig.3 (a) The first spatial mode of VEOF of the SCS surface wind field. The arrow represents wind direction and the gray scale denotes the normalized wind speed. (b) The first temporal mode.

3.2 VEOF-2

The spatial and temporal VEOF-2 modes represent another annual oscillation, the spring-autumn mode. Fig.4a shows a large-scale cyclone with strong winds over the northern SCS. Out of phase with the temporal mode of VEOF-1, the time series of VEOF-2 in Fig.4b reach the annual maximum in autumn and the minimum in spring. Hence the VEOF-2 variability over SCS is part of an anticyclone in spring and part of a cyclone in autumn. The anticyclonic wind is associated with the Subtropical High over the northwestern Pacific. The interannual cycle also appears in the VEOF-2 temporal mode and is stronger than that in VEOF-1. This interannual signal will be analyzed in Section 4.1.

Fig.4 Same as Fig.3 except for the second VEOF.

4 Interannual variability

The interannual cycle of the third temporal VEOF mode, accounting for about 1.17% of the total variance, is much more significant than the interannual characteristics in VEOF-1 and VEOF-2 (Fig.5). To investigate the interannual signal, the climatological monthly mean was first removed from the data set. Then the data set was smoothed and the semi-annual signal eliminated using a 5-month moving average. The VEOFs of this residual over SCS are named the interannual VEOFs. The first interannual VEOF (VEOF-1) and second interannual VEOF (VEOF-2) account for 42.0% and 21.6% of the total interannual variance, respectively.

Fig.5 The third temporal VEOF mode of the SCS surface wind field.

4.1 Interannual VEOF-1

The analysis of the annual VEOF-2 and interannual VEOF-1 reveals that the spatial modes of the two decompositions are similar and the temporal undulation of the interannual VEOF-1 corresponds to the interannual variability of the spring-autumn mode. In addition, the correlation between the interannual VEOF-1 temporal mode and ENSO events with a 4-month lag is strong (the correlation coefficient is about 0.62). Fig.6b shows the ENSO signal, represented by the Southern Oscillation Index (SOI) and shifted backward by four months, and the time series of the interannual VEOF-1. Corresponding to an ENSO event, the spring occurrence of the anticyclonic wind anomaly over the northern SCS has a four month lag and the autumn cyclonic wind anomaly (Fig.6a) occurs four months after a La Niña begins. The anticyclonic and cyclonic wind anomalies are also related to the western Pacific Subtropical High (WPSH) variations.

The temporal mode of the interannual VEOF-1 wasdecomposed into eight IMFs and a residual using the EMD method. Fig.7a shows the fifth IMF (IMF-5), the main interannual oscillation signal, which is related to the ENSO cycle with a four-month lag. The Hilbert spectrum of IMF-5 in Fig.7b ranges between 0.7 and 0.2 cycle per year, a period of 1.4 to 5 years, which is within the range of ENSO periods.

Fig.6 Same as Fig.3 except for the first interannual VEOF. The dashed line in (b) designates the shifted SOI.

Fig.7 (a) The IMF-5 of temporal mode of interannual VEOF-1 and the dashed line designates the shifted SOI. (b) Hilbert spectrum of IMF-5.

4.2 Interannual VEOF-2

The spatial and temporal modes of the interannual VEOF-2 are shown in Figs.8a and 8b, respectively. Compared with the first annual temporal mode, the temporal undulation of the interannual VEOF-2 demonstrates the interannual variability in the summer-winter mode. The SOI is also shown in Fig.8b and the correlation coefficient based on the two curves is 0.53. This indicates that no time lag exists between the interannual VEOF-2 and ENSO events. The spatial mode of the interannual VEOF-2 is characterized by the prominent high wind anomaly over the northern SCS. A southwesterly wind anomaly appears in El Niño years and a northeasterly wind anomaly in La Niña years. This wind variability reflects the anomaly of the Hadley cell due to the divergence and convergence of the Walker cells over the Indonesian archipelago region. The Hadley cell over the SCS weakensin El Niño years, and strengthens in La Niña years (Pan et al., 2001, 2002; Wang, 2002). The interannual VEOF-2 also shows that during El Niño years the southwesterly wind anomaly over the northern SCS can weaken the prevailing wind in winter and strengthen the prevailing wind in summer. The situation is reversed during La Niña years.

Fig.8 Same as Fig.3 except for the second interannual VEOF and the SOI (dashed line) in (b).

Fig.9 shows the four important IMFs (IMF-4–7) of the temporal mode of the interannual VEOF-2, which represent the main interannual variability of the temporal mode. The Hilbert spectra of IMF-4–7 are shown in Fig.10. From Figs.9 and 10 it can be seen that the spectrum of IMF4 ranges from 0.4 to 0.8 cycle per year, representing a period from 1.2 to 2.5 years. The spectrum of IMF5 shows a period from 2 to 5 years. Both periods of IMFs are within the range of the ENSO period. The main frequencies of the IMF-6–7 spectra are 0.2 and 0.15 cycle per year, which represents a 5- and 7-year period of oscillation, respectively.

Fig.9 The IMF-4–7 of temporal mode of interannual VEOF-2 and SOI.

Fig.10 Hilbert spectra of IMF-4–7 of temporal mode of interannual VEOF-2.

5 Conclusions

In this study, the VEOF and HHT methods were applied to analyze the annual and interannual variability of the ocean surface wind over the South China Sea based on seventeen years of merged satellite scatterometer wind vector data. The VEOFs of the merged data set show that the two leading VEOF modes are dominated by annual cycles. The first VEOF mode accounts for 87.3% of the total variance, representing the East Asian monsoon features, and has the largest amplitude near the southeast Indo-China Peninsula. The second mode shows a springautumn oscillation with a large-scale anticyclone in spring and a cyclone in autumn.

By removing the annual signal from the merged scatterometer wind, the interannual variability was investigated using both the VEOF and HHT methods. The temporal mode of the first interannual VEOF represents the spring-autumn mode and is related to ENSO events with a four-month lag. The anticyclonic wind anomaly occurs four months after an El Niño starts and the cyclonic wind appears four months after a La Niña. This pattern may be related to the variations of the western Pacific Subtropical High.

Compared with the interannual VEOF-1, the temporal mode of the interannual VEOF-2 is the summer-winter mode and responds to ENSO without a time lag. The spatial mode is characterized by the high wind anomaly over the northern South China Sea, which corresponds to the anomaly of the Hadley cell.

Acknowledgements

This work was supported by the National Natural Science Foundation of China through G41006108, the Open Research Fund of the Shandong Provincial Key Laboratory of Marine Ecology and Environment & Disaster Prevention and Mitigation through G2011001, the Laboratory of Data Analysis and Application, State Oceanic Administration through LDAA-2013-02, and the State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering through G2009586812.

Bentamy, A., Quilfen, Y., and Flament, P., 2002. Scatterometer wind fields: A new release over the decade 1991–2001. Canadian Journal of Remote Sensing, 28: 431-449.

Chelton, D. B., and Freilich, M. H., 2005. Scatterometer-based assessment of 10-m wind analyses from the ECMWF and NCEP numerical weather prediction models. Monthly WeatherReview, 133: 409-429.

Chu, P. C., Lu, S. H., and Liu, W. T., 1999a. Uncertainty of the South China Sea prediction using NSCAT and NCEP winds during tropical storm Ernie 1996. Journal of Geophysical Research, 104 (11): 273-289.

Chu, P. C., Edmons, N. L., and Fan, C. W., 1999b. Dynamical mechanisms for the South China Sea seasonal circulation and thermohaline variabilities. Journal of Physical Oceanography, 29: 2971-2989.

Chu, P. C., Veneziano, J. M., and Fan, C. W., 2000. Response of the South China Sea to tropical cyclone Ernie 1996. Journal of Geophysical Research, 105: 13991-14009.

Domrös, M., 1988. The Climate of China. Springer-Verlag, New York, 362pp.

Hardy, D. M., and Walton, J. J., 1978. Principal components analysis of vector wind measurements. Journal of Applied Meteorology, 17: 1153-1162.

Ho, C., Zheng, Z., Lin, C., Kuo, N., and Tsao, C., 2008. Wind field variations in the tropical Pacific Ocean from satellite scatterometer data. Annals of GIS, 14 (2): 120-123.

Huang, N. E., Shen, Z., Long, S. T., Wu, M. L., Shih, H. H., Zheng, Q., Yen, N. C., Tung, C. C., and Liu, H. H., 1998. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proceedings of the Royal Society of London Series A, 454: 903-995.

Kelly, K. A., 1988. Comment on ‘Empirical orthogonal function analysis of advanced very high resolution radiometer surface temperature patterns in Santa Barbara Channel’ by G. S. E. Lagerloef and R. L. Bernstein. Journal of Geophysical Research, 93: 15753-15754.

Pan, J., Yan, X.-H., Zheng, Q., and Liu, W. T., 2001. Scatterometers observe the variation of the Walker Circulation. EOS, Transactions American Geophysical Union, 82: 536-537.

Pan, J., Yan, X.-H., Zheng, Q., Liu, W. T., and Klemas, V. V., 2002. Interpretation of scatterometer ocean surface wind vector EOFs over the northwestern Pacific. Remote Sensing of Environment, 84: 53-68.

Qu, T., Kim, Y. Y., Yaremchuk, M., Tozuka, T., Ishida, A., and Yamagata, T., 2004. Can Luzon Strait transport play a role in conveying the impact of ENSO to the South China Sea? Journal of Climate, 17: 3644-3657.

Wang, C., 2002. Atmospheric circulation cells associated with the El Niño Southern Oscillation. Journal of Climate, 15: 399-419.

Xu, Q., Lin, H., Li, X., Zuo, J., Zheng, Q., Pichel, W., and Liu, Y., 2010. Assessment of an analytical model for sea surface wind speed retrieval from Spaceborne SAR. International Journal of Remote Sensing, 31 (4): 993-1008.

Xu, Q., Zhang, G., Cheng, Y., and Ji, Q., 2011. Wind resource estimation using QuikSCAT ocean surface winds. In: Proceedings of the 21th International Offshore and Polar Engineering Conference, International Society of Offshore and Polar Engineers (ISOPE), 506-510.

(Edited by Xie Jun)

(Received March 31, 2012; revised July 3, 2012; accepted July 16, 2013)

© Ocean University of China, Science Press and Springer-Verlag Berlin Heidelberg 2014

* Corresponding author. Tel: 0086-25-83787340

E-mail: xuqing0215@aliyun.com