A new early warning indicator of abrupt climate change based on the changing standard deviation-based range
A new early warning indicator of abrupt climate change based on the changing standard deviation-based range
复制标题
基于变化的归一化动态范围的新的气候突变预警指标
DOI：
10.1002/joc.7000
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Wenping He
中科院分区：
文献类型：
--
作者：
Wuqiong;Xiaoqiang Xie;Ying Mei;Wenping He
当系统被迫跨越临界阈值时，会发生突变。本研究提出一种新的预警指标-变化归一化动态范围。我们使用一个单变量气候模型和三个简单的折叠模型来测试新的预警指标的性能。结果表明，在具有折叠分叉的动力系统接近临界点之前，在α= 0.05的显著性水平下，本指标在所有检验中均表现出统计上显著的增加或减少。将著名的自相关系数与归一化动态范围进行比较，发现自相关系数在大多数测试中表现良好，对于即将到来的突变，其表现出几乎单调的增加或减少趋势。然而，自相关系数在二维蔬菜模型中表现出与噪声引起的突变完全相反的趋势。峰度系数在本研究中使用的大多数测试中都能表现良好，但在特定情况下无法给出预警。此外，自相关和峰度的变化幅度远小于归一化的动态范围之前，四个折叠模型接近其临界点。因此，新指标的突变预警性能优于自相关系数和峰度系数，在一定程度上弥补了峰度系数和自相关系数的不足。
An abrupt change will occur when the system is forced to across a critical threshold. This study presents a novel early warning indicator—the changing normalized dynamic range. We use a one‐variable climate model and three simple folding models to test the performance of the new early warning indicator. The results show that the present indicator exhibits a statistically significant increase or decrease in all of tests at a significance level ofα= .05 before a dynamical system with a folding bifurcation approaches a tipping point. Comparing the well‐known autocorrelation coefficient with the normalized dynamic range, it is found that autocorrelation coefficient performs well in most of tests, which shows an almost monotonous increasing or decreasing trend for an upcoming abrupt change. However, autocorrelation coefficient exhibits a diametrically opposite trend for a noise‐induced abrupt change in a two‐dimensional vegetable model. The kurtosis coefficient can perform well in most of the tests used in this study, however, fails to give an early warning in a particular case. Moreover, the magnitude of change of the autocorrelation and kurtosis is much smaller than that of the normalized dynamic range before the four folding models approach its tipping point. Therefore, the performance of the new indicator for warning an abrupt change in advance is better than that of the autocorrelation coefficient as well as kurtosis coefficient, which can make up for the deficiency of the kurtosis coefficient and autocorrelation coefficient to some extent.