We have produced global daily high-resolution analytical sea surface temperature (SST) data for the latter half of the 1990s from four kinds of different satellite SST products, including microwave SST, from 1998. The optimum interpolation (OI) method is applied to SST anomaly from climatological annual signal with the decorrelation scales of the first-guess error of less than 200 km. The effective spatial resolution of the OI SST is a few hundred kilometers due to these scales. The monthly mean difference between this OI SST and in situ SST averaged in 60¿S--60¿N are almost within ¿0.1 K. The root-mean square differences between them are less than 0.6 K in most of open oceans. From 1995 to 1997, when only one or two infrared SST products are available, the difference from the in situ SST tends to be negative because one of the infrared SST used here has a slight cool bias throughout the period. This bias decreases or becomes positive upon being combined with other satellite SST data. Fine spatial structure that cannot be seen in a 1¿-grid analytical SST is clearly represented, and mesoscale spatial variations with a wavelength of 400--500 km are captured in this OI SST. The temporal variability of the OI SST is also examined at a few points using a wavelet technique. The OI SST can nearly capture variations in the period of more than 1 month, although there are still some differences in the variations between the OI SST and buoy SST.