Complete and reliable monthly mean temperature series at 500hPa for 50 stations of the Tibetan Plateau which locate at a more than 3000m height for 1950-2000 are presented by making use of the bilinear interpolation for NCEP/NCAR data.According to the close relation between surface and 500hPa temperatures over the Tibetan Plateau,we design the regression scheme for each month of every station,and make the new regressions supply the lack,and reconstruct the whole surface monthly average temperature of the Tibetan Plateau during the period from 1950 to 2000.The contrast experiment of the regressive effect shows that the regression schemes are complete feasible.The characters of the new reconstructed monthly average air temperature represent the actual ones of temperature over the Tibetan Plateau fairly well.