基于启发式分割算法的气候突变检测研究

Abrupt climate change detection based on heuristic segmentation algorithm

气候系统的非线性、多层次性和非平稳性对气候突变的检测方法提出了较高的要求.基于t检验将非平稳序列分割为多个不同尺度的自平稳子序列,Bernaola Galvan提出的启发式分割算法(BG算法),对非平稳时间序列的突变检测效果较好.在BG算法的基础上,通过理想时间序列验证BG算法处理非平稳时间序列的有效性,并对近2000a北半球树木年轮距平宽度序列基于不同层次的思想,检测和分析其中包含的各种尺度的气候突变事件,成功地区分不同尺度的突变.定义的新物理量——突变密度的分析表明,自然因素作用的基础上,人为因素影响的加剧可能导致近1000a来突变密集段和稀疏段分布失衡,这可能是全球变化的重要表现之一.

Climate system is nonlinear, non-stationary and hierarchical, which makes even harder to detect and analyze abrupt climate changes. Based on Student＇s t-test, Bernaola Galvan recently proposed a heuristic segmentation algorithm to segment the time series into several subsets with different scales, which is more effective in detecting the abrupt changes of nonlinear time series. In this paper, we try to verify the effectiveness of heuristic segmentation algorithm in dealing with nonlinear time series by an ideal time series.Through detecting and analyzing the information of abrupt climate changes contained in recent 2000a＇s tree annual growth ring, we succeeded in distinguishing abrupt changes with different scales. The research based on the newly defined parameter of abrupt change density shows that human activities might have lead to the recent 1000a＇s unbalanced distribution of serial and spares segments of abrupt climate changes, which may be one of the manifestations of global temperature change.