In terms of the canonical variable regression(CVR)of principal component,an interpolation experiment is carried out based on integrated analysis of the change of Jones’ data coverage rate with time,correlation structure and stability between data missing and non data missing fields together with the reconstruction of continuous and homogeneous series of global monthly mean temperature field from 1900 to 1998.The independent sample test demonstrates that the ratio of error variance of data missing to variance of origin data is below 0.40.There are the rough agreements between the original and reconstructed series of global and zonal mean temperature.Linear warming rate of the former gives 0.11℃/(100a)higher than the latter,which may be the result of inhomogeneous distribution of Jones’ series.