Abstract: Some common sea surface temperature (SST) observation methods and four datasets which are widely used in present climate change analysis are introduced and compared in this article. There are many kinds of errors including in blended SST data which is based on different observation methods. As to these errors, there already exists a systematic error analysis method and highly qualified, more sufficient datasets as ICOADS and HadISST et al are obtained. They are widely used in ocean-atmosphere interaction, climate modeling and climate change analysis. However, the temporal and spatial deficiency of observed data is still the biggest handicap for analyzing multi-scale SST characteristics. Therefore, adding newly observed dataset continuously and improving observation method and data processing method to get global and regional high-quality, and long-term datasets are still the main tasks for SST observation data analysis work.