Comparison of three Mann-Kendall methods based on the China's meteorological data

Trend test is an essential topic in climate change research. Three trend test methods, Mann-Kendall, Pre-Whitening Mann-Kendall and Trend-Free Pre-Whitening Mann-Kendall, are employed to analyze trends and autocorrelations of annual precipitation, pan evaporation and average air temperature spatially and temporally across China. The meteorological variables are observed at 317 stations during the period 1956-2005. Specifically, the autocorrelation of annual precipitation is insignificant, resulting in the similar trend analysis results for the three methods. While both annual pan evaporation and average temperature exhibit significant autocorrelations, which result in distinct trend test results using the three methods. Thus, the autocorrelation must be properly considered when conducting a trend test for both annual pan evaporation and average air temperature. In general, meteorological data from northern basins has a more significant autocorrelation than the southern one. In addition, the effects of trend on the estimated autocorrelation coefficient and correlation on the Mann-Kendall statistics are also investigated theoretically. Results indicate that the positive correlation magnifies the series trend significance and the trend existed in the series may contaminate the lag-1 autocorrelation coefficient in turn.