年层自动识别方法在青藏高原冰芯定年中的应用  

Application of automatic annual layer identification method in ice core dating on the Qinghai-Xizang Plateau

作　　者：罗语晴 姚檀栋[1,3] 徐柏青 杨丹丹[1] 王理 邬光剑 李真[4] 王宁练 曲冬梅 武小波[4] LUO Yuqing;YAO Tandong;XU Baiqing;YANG Dandan;WANG Li;WU Guangjian;LI Zhen;WANG Ninglian;QU Dongmei;WU Xiaobo(Key Laboratory of Tibetan Plateau Earth System,Environment and Resources,Institute of Tibetan Plateau Research,Chinese Academy of Sciences,Beijing 100101,China;University of Chinese Academy of Sciences,Beijing 100049,China;College of Earth and Environmental Sciences,Lanzhou University,Lanzhou 730000,China;Northwest Institute of Eco-Environment and Resources,Chinese Academy of Sciences,Lanzhou 730000,China;Shaanxi Key Laboratory of Earth Surface System and Environmental Carrying Capacity,Xi’an 710127,China)

机构地区：[1]中国科学院青藏高原研究所青藏高原地球系统与资源环境重点实验室,北京100101 [2]中国科学院大学,北京100049 [3]兰州大学资源环境学院,甘肃兰州730000 [4]中国科学院西北生态环境资源研究院,甘肃兰州730000 [5]陕西省地表系统与环境承载力重点实验室,陕西西安710127

出　　处：《冰川冻土》2024年第3期731-741,共11页Journal of Glaciology and Geocryology

基　　金：第二次青藏高原综合科学考察研究项目(2019QZKK0208)资助。

摘　　要：冰芯是重建过去气候环境变化的良好载体,高时间分辨率冰芯信息的提取依赖于冰芯准确定年。本文结合人工定年与基于隐马尔科夫模型的年层自动识别算法,在青藏高原的作求普冰芯、达索普冰芯和阿尼玛卿冰芯上进行了测试、验证与分析。在约束参考层、同步其他冰芯记录及审核复盘时,该方法在一定程度上减少了人工逐层识别的时间,并能通过算法自动推断年层所在位置及提供误差范围。首先,由人工提供一套基础定年模板及约束点,随后基于算法学习模板,计算冰芯代理变量的准周期性的循环,找到目标段落并探究其年层存在的可能性,通过前向-后向算法寻找所有可能的年层点位,并通过最大似然法从中选出最优解。在年层自动识别算法辅助下的作求普、达索普冰芯上部δ^(18)O定年结果与已知年表的累计误差小于2 a(3%),具有较好的一致性。进一步将此方法应用于阿尼玛卿冰芯上部的定年,得出该算法与人工定年在e_(1)[2.07,4.47]、e_(2)[8.31,9.71]部分存在1~2 a的差异。再经污化层校正,冰芯56.48 m处的定年结果为74 a,误差范围±3 a。本文设计的方法结合了人工判定年层的优势与算法的标准化分层处理能力,降低了在年代校正过程中的人工参与,可为未来青藏高原地区高分辨率冰芯数年层方法定年的误差量化研究提供科学参考。Ice cores are an effective proxy for past climate changes,and high-resolution reconstruction necessi-tates precise dating methods.This study combines manual dating with an automated layer recognition algorithm(StratiCounter)based on a Hidden Markov Model(HMM)framework to test,validate,and analyse the perfor-mance of the algorithm in assisting the dating of Zuoqiupu,Dasuopu,and Anemaqen ice cores.The method re-duces the time required for manual layer-by-layer identification during binding reference layers,synchronizing with other ice core records,and reviewing processes by automatically inferring layer positions and providing er-ror ranges.Initially,a collection of fundamental dating templates and layer constraint points are manually pro-vided.Next,the algorithm familiarises itself with the template for calculating the quasi periodic cycle of the ice core proxy variable.It then identifies the target paragraph and examines the presence of layer constraints.Per-form a comprehensive search for potential points while considering layer constraints,and employ the maximum likelihood method to determine the most suitable solution from the available options.The findings show that for the upper sections of the Zuoqiupu and Dasuopu ice cores,the algorithm-assisted dating results have an accumu-lated error of less than 2 years(3%)when compared to known chronologies,indicating a high level of consisten-cy.Even under simulated sample loss conditions and the absence of tie points,the algorithm remains operation-al,resulting in a cumulative error of less than five years.Further analysis of the upper section of the Anemaqen ice core reveals 1 to 2 year differences between algorithmic and manual dating results at specific sections e_(1)[2.07,4.47]and e_(2)[8.31,9.71].After using pollution layer data,the ice core at 56.48 metres was dated to 74 years with an error range of±3 years.Integrating reference layer findings likeβ-activity and ^(137)Cs withδ^(18)O counting layers helped determine the age sequence of the upper 0.05~56.48

关 键 词：冰芯定年 青藏高原 年层自动识别算法(StratiCounter) 阿尼玛卿冰芯 

分 类 号：P343.6[天文地球—水文科学] P532[天文地球—地球物理学]