A method for homogenization of complex daily mean temperature data: Application at Beijing Observatory (1915–2021) and trend analysis

A homogenized daily mean temperature series from 1915 to 2021 at Beijing Observatory (BO) has been developed. A thorough investigation of observation records and historical metadata was carried out to identify specific non-climatic biases. The inhomogeneities were detected and adjusted with the optimally selected reference stations using a statistical reduction methodology. The results indicated three types of non-climatic biases in temperature records: the first was the relocation of BO from suburban to urban, which caused positive shifts in temperature records, especially for wintertime; the second arose from different methods for calculating daily mean temperature, which caused different sign biases varying with months; and the last was caused by the transition from manual to automatic measurements with site-specific biases. The corresponding adjustments due to three type biases ranged from −1.13 to 0.63, −0.29 to 0.23 and −0.13 to 0.00°C, respectively. The new homogenized annual mean temperature series showed a warming trend of 0.199°C/decade during 1915–2021. The trend was more pronounced in winter than in the other three seasons. The warming trend in the new series is greater than those in the previous homogenized series (0.136–0.177°C/decade), primarily due to the more effective adjustment associated with the site relocation, particularly for more urban locations.