Most methods of homogenization of climate data are applied to time series of a single variable, such as daily maximum temperature (Tmax) or daily minimum temperature (Tmin). Consequently, the physical relationship among different variables, e.g., Tmax > Tmin, may be distorted after homogenization of climate series of individual variables. The authors develop a solution to improve consistency among diurnal temperature records, while using the Multiple Analysis of Series for Homogenization (MASH) method to homogenize the observation series of daily mean temperature (Tm), Tmin, and Tmax at 545 stations in China for the period 1960–2011, called CHTM 2.0. In the previous version of this homogenized dataset based on MASH (CHTM 1.0) for the period 1960–2008, there are a few records (0.039% of the total) that are physically inconsistent. For developing CHTM 2.0, the authors apply additional adjustments for each day with inconsistent temperature records, in order to hold Tmax > Tm > Tmin. Although the additional adjustments are barely influential for estimating long-term climate trends in China as a whole (because very few records are additionally adjusted), the newly introduced solution improves the physical consistency throughout the dataset. It is also helpful for developing more reasonable homogenized climate datasets with regard to physical consistency among multiple variables. Based on CHTM 2.0, the annual Tmax/Tm/Tmin series averaged over China for the period 1960–2011 show significant warming trends of about 0.19/0.25/0.34°C/10 yr, respectively. Large warming trends of up to 0.425/0.596/0.704°C/10 yr occur in northeastern and northwestern China.