A new homogenized temperature data set called the China Homogenized Historical Temperature Dataset (CHHTD-V1.0) has been developed, and it includes daily and monthly mean temperature series from 2419 national stations distributed throughout mainland China for the period from 1951 to the present. The inhomogeneities in individual station series were detected using a penalized maximum t-test (PMT) that accounted for the first-order autocorrelation. Detailed metadata information was applied to validate the changepoints caused by changes in local observation systems. Comparative analyses suggested that the quantile-matching (QM) adjustments that accounted for high-order discontinuities led to more reasonable results than the MEAN adjustments for the daily temperature series. Therefore, the QM method was applied to adjust the discontinuities caused by non-climate changes such as changing of observing site, instrumentation and observation environments. The physical consistency among the daily maximum, mean and minimum temperatures (T_max, T_m and T_min) was also checked for each station. Based on the new homogenized data set, linear trends in the annual and seasonal temperature series from 1960 to 2014 were calculated. In comparison with the original data set, the homogenized data set improves the geographical consistency of the long-term climate trends over the region. The updated nationwide mean warming rate reached 0.22 °C per 10 years for the T_max, 0.27 °C per 10 years for the T_m and 0.38 °C per 10 years for the T_min from 1960 to 2014, which are considerably larger than the previous estimates that were based on the more frequently used networks of a few hundred stations in China.

Inhomogeneities in the daily mean/maximum/ minimum temperature (T_m/T_max/T_min) series from 1960–2008 at 549 National Standard Stations (NSSs) in China were analyzed by using the Multiple Analysis of Series for Homogenization (MASH) software package. Typical biases in the dataset were illustrated via the cases of Beijing (BJ), Wutaishan (WT), Ürümqi (UR) and Henan (HN) stations. The homogenized dataset shows a mean warming trend of 0.261/0.193/0.344ºC/decade for the annual series of T_m/T_max/T_min, slightly smaller than that of the original dataset by 0.006/0.009/0.007ºC/decade. However, considerable differences between the adjusted and original datasets were found at the local scale. The adjusted T_min series shows a significant warming trend almost everywhere for all seasons, while there are a number of stations with an insignificant trend in the original dataset. The adjusted T_m data exhibit significant warming trends annually as well as for the autumn and winter seasons in northern China, and cooling trends only for the summer in the middle reaches of the Yangtze River and parts of central China and for the spring in southwestern China, while the original data show cooling trends at several stations for the annual and seasonal scales in the Qinghai, Shanxi, Hebei, and Xinjiang provinces. The adjusted T_max data exhibit cooling trends for summers at a number of stations in the mid–lower reaches of the Yangtze and Yellow Rivers and for springs and winters at a few stations in south–western China, while the original data show cooling trends at three/four stations for the annual/autumn periods in the Qinghai and Yunnan provinces. In general, the number of stations with a cooling trend was much smaller in the adjusted T_m and T_max dataset than in the original dataset. The cooling trend for summers is mainly due to cooling in August. The results of homogenization using MASH appear to be robust; in particular, different groups of stations with consideration of elevation led to minor effects in the results.

During 1977–1981 the Beijing (BJ) meteorological station was at a suburban location. In 1981 it was moved to a more urban location, but in 1997 it was subsequently moved back to the same suburban location. The daily BJ temperature series, together with those from 18 nearby stations, form a unique database for studying how site‐change and possible urbanisation influences affect climate changes at a local scale. The site‐change‐induced biases were quantified, between 0.43 and 0.95 °C, based on comparisons between multi‐year‐mean seasonal temperature anomalies at BJ and the mean of those from a cluster of nearby stations. The annual mean urban‐suburban difference was 0.81 °C around 1981 and 0.69 °C around 1997, indicating a growing urbanisation effect in the suburban compared to the downtown area. The linear warming trend in the adjusted (for site moves only) BJ temperature series during 1977–2006 was 0.78 °C/decade. Comparing with several rural and less‐urban sites, we suggest that the BJ records include an urbanisation‐related warming bias between 0.20 and 0.54 °C/decade, likely about 0.30 °C/decade, for the recent few decades. The climatic warming at BJ between 1977 and 2006 is likely, therefore, to be about 0.48 °C/decade. Caveats for using these estimates were discussed. Copyright © 2009 Royal Meteorological Society

Key Points: Unprecedented compound heat waves will occur on a regular basis in three urban agglomerations in China since mid‐21st century under RCP8.5 The likelihood of unprecedented compound heat waves in a 1.5°C warmer world will halve compared to 2°C warming scenario Population exposures to future unprecedented compound heat waves will increase by nearly 70, 90, and 60 million by the end of this century