Analysis of trends in air temperature at Chinese stations considering the long-range correlation effect

Abstract

In this study, we consider the long-range correlation (LRC) effect on trend analysis. We use daily maximum and minimum air temperature at 590 Chinese meteorological stations in recent 46 years. Before estimating trends in daily air temperature, we filter out the periodic component by removing the mean record for each calendar date. With considering the LRC effect, we extract the observed trends in both daily maximum and minimum air temperature, test their significance, and compare the results to those obtained by the traditional linear regression which ignores the LRC effect. It is substantiated that by inflating the estimation uncertainty LRC can increase the chance of the observed trends to be insignificant by about 4 times and 10 times for daily maximum and minimum air temperature at Chinese stations, respectively. In addition, the observed trends in daily minimum air temperature are found to be much more significant than those in the daily maximum records. With consideration of the LRC effect, we apply a method proposed by Rybski and Bunde (2009) based on the detrended fluctuation analysis (DFA) to estimate the most-likely trend in temperature at five representative stations (Tianjin, Hohhot, Nanjing, Haikou, and Xining), which correspond to the five typical climate in China, i.e., temperate monsoon climate, temperate continental climate, subtropical monsoon climate, tropical monsoon climate, and plateau mountain climate, respectively. Comparing to the DFA-based method, it is demonstrated that the traditional linear regression may overestimate the most-likely trends at these five stations by more than 100% and up to 36% for daily maximum and minimum air temperature, respectively. Therefore, the LRC effect must be considered in trend analysis of temperature at Chinese stations.