The decadal coupling of sea surface temperature (SST) and wind in the Taiwan Strait (TWS) during winter (JFM-mean) from 1988 to 2017 was analyzed. The Empirical Orthogonal Function (EOF) method is respectively used to analyze the variability of SST and wind fields, while the Singular Value Decomposition (SVD) method is adopted to analyze the covariance between them. The analysis showed the northeasterly wind from the East Asian Winter Monsoon (EAWM) in the TWS had a decadal decreasing from 1988 to 1998 and then increased from 1998 to 2017. Winter SST in the TWS from 1988 to 2017 increased first, with a maximum rate of 1.5 degree per decade in middle TWS, and then decreased with transition near 1999, with the largest decreasing rate of 1.5 degree per decade located at the western coast of the TWS. The EOF, SVD and sensitive test shows the second model of wind, which is most cross-shore, explaining only 18% of the wind variance although, can cause the 35% SST variance in the TWS. While the first model of wind, which is almost alone-shore with 63% wind variance, only contribute 16% SST variance. Both decreasing mode 1 and increasing model 2 of the wind around the TWS explained the high increasing rate of SST from 1988 to 1999, where the model 2 play more important role. Further researches indicated the response of second model wind to ENSO is regulated by Pacific Decadal Oscillation (PDO), i.e., the second mode wind has negative/positive correction with the ENSO index when the PDO is in positive/negative phase. Both two winters,1976 and 2008, when the cold disasters occurred in the TWS, are La-Nina year in negative PDO phase. The mechanism for these cold disasters is the cold water can be transported into the TWS with the strong alongshore northeasterly wind, and separate to the middle TWS with the offshore wind in the special synoptic process.