It has been pointed out that the disclosure of the detailed land price data is significant in improving transparency in the real estate market. Although the officially published land prices had played a major role in price making in Japan, and transaction prices will do the same in the future, they do not provide enough information for all market participants. Thus, it is necessary to estimate the land prices of interested land lots using the observed land price data, which is distributed spatially and temporally. In this study, we focus on the recently proposed spatio-temporal stationary covariance function, which models the correlation through space and time. We empirically analyze the applicability of the spatio-temporal covariance function with universal kriging to land price estimation by using the officially published residential land prices in Tokyo 23 wards, and visually illustrate the land price distribution and its change in Tokyo in 30 years.