SST (Sea Surface Temperature) is an essential meteorological factor that enables us to discover atmosphere-ocean interaction and understand climate change. Because SST observation using satellites can provide spatially continuous coverage, it is thought of as a useful tool for the analysis of spatial characteristics of SST and the spatial complement of in-situ point observations. This paper examined a quality improvement method for GK2A (Geostationary Korea Multi-Purpose Satellite-2A) daily SST product in terms of the outlier detection using DSAT (Deviation from Spatial Autocorrelation Trend) and the gap-filling by FMM (Fast Marching Method) and MLR (Multiple Linear Regression). The outliers were found mainly around coastal areas, presumably because it is difficult to retrieve SST around complex coastlines and small islands. After the removal of such outliers, GK2A SST showed a somewhat improved coincidence with in-situ observations. MLR, a multivariate method, produced better accuracy than FMM, a univariate method. The MLR model using appropriate explanatory variables yielded a more spatially detailed result for the gap-filling and quite high accuracy with the correlation coefficients between 0.951 and 0.964. Future works will be necessary to enhance the outlier detection and gap-filling process, including ensemble and filtering methods.

Key Words:GK2A, sea surface temperature, outlier detection, gap-filling