Abstract: On the basis of several datasets of temperature observations for the recent century and the metadata of each station, problems for the warm bias in the 1940s was explored in this work, partly via the reconstruction of the early developed temperature series. The results indicated that relocations of most stations in the early 1950s were the main reason for inhomogeneities and caused relatively warm biases in earlier records. Therefore, the temperature series based on the original data has a distinct warm peak in the 1940s, while the series based on the homogenized data does not exhibit this warm peak. The new temperature series for China has a long-term warming trend of 1.41°C/100 a, which is larger than most of the early estimates.