This paper describes a method to estimate the spatio-temporal distribution of urban temperature and its application for evaluation of the urban thermal environment in the Tokyo metropolitan area. First, formulae to estimate the distribution of monthly and hourly mean temperatures were developed through regression analysis, using geographic conditions at meteorological stations, and then mesh temperature data were generated. Second, degree day maps based on mesh temperature data were made and energy consumption for heating and cooling of residential buildings was estimated for each mesh. It was found that energy consumption per family for heating is lower in coastal areas compared to inland. In case of cooling, the energy consumption per family is higher in the central part of Tokyo compared to other areas, because of climatic conditions.