The spatial and temporal variability of sea surface temperatures (SSTs) derived from NOAA/AVHRR infrared imagery obtained between 1991 and 1998 in northeastern Taiwan was investigated. An empirical orthogonal function (EOF) analysis was applied to monthly SST data. Results indicated that the first mode of EOF analysis represented 63% of the total spatial variance. It contained the structure of the Kuroshio frontal pattern, which is the interactive phenomenon of Kuroshio water and continental shelf water of the East China Sea (ECS). In addition, the Kuroshio frontal patterns as well as cold eddies were found to exist throughout the year. They could be classified into three main types of distribution patterns. The first type is the Kuroshio frontal pattern, which is usually found in November to March of the following year. The second type, the cold eddy pattern, is commonly found during summer from June to September, while the third type exists simultaneously with both the Kuroshio frontal and cold eddy patterns, and appears in the interchange period of the northeastern-southwestern monsoon, especially in May, October, and November. Furthermore, a wavelet analysis of the SST image time series was attempted, and the results showed that the dominant period variation of the studied area is an annual cycle (i.e., 12 months). Therefore, the ARIMA(0,1,1)(0,1,1)12 time series was determined to completely imitate the SST forecast model which can be used to predict monthly SST data of one or more future time periods.