Abstract:The historical monthly mean surface relative humidity at 13 observation stations in Tianjin are homogenized by two methods establishing reference series of NCEP/DOE AMIP Ⅱ Reanalysis (R 2) specific humidity data and observed relative humidity data. The results indicate that nine stations have significant breakpoints, accounting for 69% of Tianjins total stations. The main reason for the abrupt change is the automation referring to metadata information, and then is the relocation and instrument change. For the correction amplitude, the proportion of negative Quantile Matching (QM) correction is nearly 96.3%, ranging in -5.0%～-1.5% which accounts for more than 80% of the total correction. Here the changes of variances and trends of monthly mean surface relative humidity series before and after corrections are compared, implying that the artificialities due to some inhomogeneity factors in the series are largely corrected, and have reduced the abnormal dry tendency in relative humidity for a long time. Moreover, error analysis reveals that there is much higher consistency between the two datasets here and the data by Zhu Yani et al. (2015) by setting the latter as the reference. However, due to some subjective factors, the error values of MAE and SE between the two datasets are more than 3.0% at some individual stations, resulting in only above 84.6% and 76.9% stations with error values of MAE and SE from 0 to 2.0%, respectively. Accordingly, the technology of data processing used in this paper is more superior to those before to a certain extent.