Development of the century-long homogenized daily temperature data set in China

The meteorological observation data of the early period (1840-1950) in China were  manually corrected the input and clerical errors, and then according to the length or coverage of time, the main series was determined. The observation time system of unknown sites is determined by the difference method introduced. After these operations, the data of all sites are unified into the same format. Then, through the ridge regressions established by data from modern reference stations, the missing maximum temperature (T-max) and minimum temperature (T-min) are interpolated, and then combined with modern data to form the heterogeneous extreme temperature data set of 1840-2020 in China. RHtest software was used to adjust the homogenate problem in  data set. Finally, the century-long homogenized daily temperature data set including 45 key city stations in China was obtained. There are 20 stations with observation record more than one hundred year. The length of temperature observation series of 17 stations is between 80 and 100 years. The series length of the remaining 7 sites is between 68 and 80 years. The data integrity of 5 stations is 50 - 60%, 9 stations is 60 - 70%, 8 stations is 70 - 75%, 5 stations is 75 - 80% and 18 stations is 85 - 100%. Finally, the angular distance weighting (ADW) method is used to interpolate the dataset into grid products, and the grid accuracy is 2.5 ° * 2.5 °. Climatic research unit dataset (CRU data set) was compared to verify the new dataset. The correlation between the four longest sequences in the newly developed data set  and the corresponding sites in the CRU data set is very good, and the correlation coefficient is very high, and the correlation between the remaining sites of the two data sets is also very good so the quality of CUG dataset is still trustworthy.