Comparative study of multiple measures on temporal irreversibility of daily air temperature anomaly variations over China

Abstract

As a fundamental property of nonlinear time series, temporal irreversibility (TI) is a hot topic in various fields. Among these studies, many methods and measures have been developed. However, only a few studies contribute to the performance comparison of these methods and measures. Taking the daily mean air temperature from observations over China as an example, the performance of five well developed measures was compared in this study. Three of them are directly related to air temperature increments with A defined from sign ratio, G from magnitude ratio and E from skewness, and the other two measures are L2 from directed horizontal visibility graph (DHVG for short) and L1 from consecutive increasing and decreasing steps (CIDS for short). The results show that five measures reach almost consistent conclusions to TI of daily mean air temperature from observations over China, but with different statistical significance and sensitivity to low-frequency trend and measured noise. They also show that the higher TI occurs over southern China, and lower TI over northeast China and southwest China close to Basin of Sichuan and Yungui-Tibetan Plateau. Further studies indicate that this dominated TI behavior is closely related to the number difference between extreme decreasing temperature variations and extreme increasing temperature variations. When conditional threshold exceeds certain critical values, this marked TI can be well quantified by the extreme decreasing but not increasing temperature variations. At last, combining all aspects considered, measure from CIDS is the best among these five TI measures due to its robustness to low-frequency trend and measured noise.