Improved partial trend method to detect rainfall trends in Hainan Island

Abstract

Partial trend method (PTM) is an innovative and efficient trend analysis method, especially for sub-trends in different magnitudes. However, it depends on graphs and identifies trends subjectively, which limit the application for a large mass of data and cannot reveal significant variations from natural variability and randomness. To remedy these shortcomings, this paper applies a trend index derived from the PTM plot, and develops a nonparametric bootstrap approach to identify statistically significant trends. The improved PTM is used to detect trends of annual rainfall, monthly rainfall, monsoon rainfall, annual number of rainy days, and average intensity of daily rainfall in Hainan Island from 1950s to 2014. The Mann–Kendall test is also employed for a comparison purpose. The PTM reveals that, except at the Sanya station, annual rainfall and monsoon rainfall generally show an insignificant change in various magnitudes. Rainy days and daily rainfall intensity show a prevailing decreasing and increasing trend, respectively. The opposite trends explain the no trend in annual rainfall. The trends revealed by the PTM and the Mann–Kendall test show high consistency. This demonstrates that the PTM index combined with the nonparametric bootstrap method is a reliable technique. The improved PTM makes it flexible to detect overall and partial trends by graphical analysis, or detect significant trends for a mass of data by an index. Not limited to rainfall variables, this method is hoped to analyze trends in other fields.