The Analysis Of The Variabilities And Forecasting Viability Of The Main Sea Surface Factors In The South China Sea

The topic of the climate variability is one of the central subjects in the recentearth sciences, which has a great impact on the human's living and development.The South China Sea (SCS) is the largest marginal sea of China, located upwind thesummer monsoon, and has dominant interannual variability as a marginal sea of thetropical Pacific Ocean. With the increasing satellite observations and product data,we have the chance to directly view the main oceanic parameters of the SCS and toexamine the climate variability and to study the predictability of the SCS.The present paper focuses on the detailed analysis of the climate variability onthe basis of the recent satellite data products including surface wind (SW), seasurface temperature (SST) and sea surface height anomalies (SSHA). We also testthe predictability of the SCS SST anomalies using canonical correlation analysis(CCA) with tropical Pacific Nino index and Indian Ocean Dipole mode index aspredictors. The nonlinear characteristics of the above parameters in SCS are alsodiscussed through nonlinear empirical orthogonal function (NLEOF) analysis. Themain scope of work includes :1. To obtain the SCS absolute dynamic topography (ADT) relative to the geoid, ajoint empirical orthogonal function (EOF) analysis is performed using the resultsof altimeter measurements and numerical model results. The altimeter data usedin the present study is the climatological monthly SSH anomalies, while themodel data is the climatological monthly mean SSH from a variable-grid globalocean circulation model. The monthly genostropic currents with 1/3°×1/3°spatial resolution are then calculated from monthly ADT of the SCS. From thegeostrophic current fields it can be seen that the interior SCS surface currents arecharacterized by three types of currents, that is, the west boundary currents,offshore currents, and eddies.2. Trends of the surface winds, sea surface height and sea surface temperature ofthe SCS in 1993-2003 are analyzed using monthly products from satelliteobservations. Time series are smoothed with a 12-month running mean filter.The east and north components of the SW, SSH, and SST have linear trends of0.53 ± 0.35 ms-1 decade-1, -0.04 ± 0.17 ms-1 decade-1, 6.7 ± 2.7 cm decade-1, and0.50 ± 0.26 K decade-1, respectively. The sea level rising rate and sea surfacewarming rate are significantly higher than the corresponding global rates.3. An EOF analysis is performed to evaluate the interannual variability. Resultsshow that the first EOF of the SW is characterized by a basin-wide anticyclonicpattern. The corresponding time coefficient function (TCF) correlates with theNino3.4 index at the 99% confidence level, with a lag of 3 months. The first EOFof the SSH is characterized by a low sea level along the eastern boundary. Thecorresponding TCF correlates with the Nino3.4 at the 99% level, with a lag of 2months. The first EOF of the SST is characterized by a basin-wide warming withthe highest anomalies in the north deep basin. The corresponding TCF correlateswith the Nino3.4 at the 95% level, with a lag of 8 months. Based on the EOFanalysis, the ENSO-associated correlation patterns of the SW, SSH and SST arealso presented.4. Based on the EOF analysis of the OISST data with period of 1982-2003 afterremoving the climatological mean and trends of SST in SCS, the correspondingTCF correlates best to Dipole Mode Index, Nino1+2, Nino3.4, Nino3 and Nino4indices with a time lag of 10-month, 3-month, 6-month, 5-month and 6-month,respectively. A statistical linear prediction for SSTA in the SCS is tested basedon CCA model using the above indices as predictors. The forecast skill isevaluated over an independent test period of more than 11 yr (1993/94–Oct.2004)by comparing the model performance with a simple prediction strategy involvingthe persistence of sea surface temperature anomalies over 1-12 months' lead time.Prediction based on CCA has a significant improvement especially with anincreasing lead time (longer than 3 months) over the persisted prediction. TheCCA model performs steadily and the correlation coefficients between theobserved and predicted SSTA in the SCS are about 0.6 on the average, whichexceed the thresholds at the 99% level of confidence, for all the lead times, andthe root mean square errors are about 0.2 standard deviation. In addition, the SCSwarm event in 1998 can be successfully predicted for the 1-12 months' lead time.The seasonal differences of the prediction performance for 1-12 months leadtime are also examined.5. The nonlinear EOF (NLEOF) analysis, which is based on the feed forward neuralnetwork method, of the satellite data including SW, SST and SSH anomalies inthe recent decade is performed. The first mode of NLEOFs explain morevariances as compared with the first mode of linear EOFs and can capture moredetailed spatial characteristics. Among them, the SW and SSHA show muchstronger nonlinear characters, and both of the nonlinear-to-linear error ratiosbelow 0.8 occupy the most part of the SCS. The first mode of NLEOFs of SWand SSHA explain 67.26% versus 54.75% and 60.24% versus 50.43%, explainedby the first linear EOF mode respectively. In contrast, the variance explained bythe first nonlinear mode of SSTA is not improved significantly since the varianceexplained by linear EOF is already up to 79%. The nonlinear character of SSTAis not dominant, since nonlinear-to-linear error ratios below 0.8 only appear inthe northern and southern coastal regions. The analysis further indicates that thenonlinear characters of SSTA in the recent decade are stronger than the previousdecade, since the nonlinear-to-linear error ratios are even higher during1982-2003.