The statistical relationship of cementsurface temperature and temperature observed simultaneously by the automatic temperature recorder in screen at Jiaxing Station in the summers of 2006 and 2007 (June to September) is analyzed. The results indicate that the average and extreme maximum summer cementsurface temperatures change simultaneously with the temperature in screen at the scales of day, month and quarter, but there is obvious difference between them, which is mainly manifested in the maximum temperature. Comparing with the temperature in screen, the cementsurface temperature has such features as earlier appearing of extreme maximum temperature, shorter high temperature episodes, greater difference, minimum temperature in morning close to temperature in screen or slightly higher. The cementsurface maximum temperature in summer is related closely to sunshine duration, precipitation, wind direction and wind speed, relative humidity, etc. By means of the second Kalman filter and Optimal subset regression, the quantitative prediction outputs of the cementsurface maximum temperature are given and revised artificially according to different weather situations (sunny, cloudy, precipitation). Those equations proved satisfactory in accuracy in forecasting operation and can help improve the specialized meteorological services.