基于时空加权KNN算法的1988-2015年渤海海冰空间分布重建  

Reconstruction of spatial distribution of sea ice in Bohai Sea from 1988 to 2015 based onspace-time weighted KNN algorithm

作　　者：孙静琪 李晨睿 许映军[1,2,3,5] 颜钰 邓磊 SUN Jingqi;LI Chenrui;XU Yingjun;YAN Yu;DENG Lei(Beijing Normal University,School of National Security and Emergency Management,Beijing 100875,China;International Research Laboratory of Catastrophe Simulation and Systemic Risk Governance,Beijing Normal University at Zhuhai,Zhuhai 519087,China;Institute of Disaster Reduction and Emergency Management,Ministry of Emergency Management-Ministry of Education,Beijing 100875,China;Capital Normal University,College of Resources,Environment and Tourism,Beijing 100037,China;Beijing Normal University,State Key Laboratory of Earth Surface Processes and Resource Ecology,Beijing 100875,China;China University of Geosciences,Beijing 100083,China)

机构地区：[1]北京师范大学国家安全与应急管理学院,北京100875 [2]北京师范大学教育部巨灾模拟与系统性风险应对国际合作联合实验室,广东珠海519087 [3]应急管理部−教育部减灾与应急管理研究院,北京100875 [4]首都师范大学资源环境与旅游学院,北京100037 [5]北京师范大学地表过程与资源生态国家重点实验室,北京100875 [6]中国地质大学,北京100083

出　　处：《海洋环境科学》2024年第3期438-447,共10页Marine Environmental Science

基　　金：国家自然科学基金项目(41977406);冰上丝绸之路东北亚航段海冰时空变化及其对通航性影响研究(42206246)。

摘　　要：利用AVHRR和MODIS遥感解译数据,结合与渤海海冰面积相关程度高的日平均温度、3 d-1.8℃积温、累积冻冰度日和累积融冰度日等气象因子数据,基于时空加权KNN算法构建了空间分辨率为1 km海冰空间补全模型,重建了1988-2015年渤海海冰空间分布连续日数据集。渤海海冰空间分布补全均方误差为0.03,分类正确率均为87%以上,28年平均正确率为91.87%,均方误差与海冰遥感影像数据缺失率呈中度正相关。结果表明,该模型均方误差较小,且分类正确率高,可以用于渤海海冰空间分布数据补全,空间分辨率高且补全速度快,在海洋环境安全管理领域,尤其对有冰海域海冰灾害风险管理方面有重要的价值。This study used AVHRR and MODIS remote sensing interpretation data,combined with meteorological factors such as daily average temperature,accumulated temperature of 3 d-1.8℃,accumulated frozen ice days and accumulated melted ice days that are highly correlated with Bohai sea ice area,constructed a spatial completion model with spatial resolution of 1km based on space-time weighted KNN algorithm,and reconstructed the continuous daily data set of Bohai sea ice spatial distribution from 1988 to 2015.The complete mean square error of the spatial distribution of sea ice in Bohai Sea is 0.03 km2,the classification accuracy is above 87%, and the average accuracy in 28 years is 91.87%. The mean square error is positivelycorrelated with the missing rate of sea ice remote sensing image data. As shown by results, the model has smallmean square error and high classification accuracy, and can be used to complete the spatial distribution data ofsea ice in Bohai Sea, with high spatial resolution and fast completion speed. It is of important practical value inthe field of marine environmental safety management, especially in the risk management of sea ice disasters inicy waters.

关 键 词：渤海海冰 加权KNN算法 海冰空间分布 海冰数据重建 

分 类 号：P731.15[天文地球—海洋科学] X87[环境科学与工程—环境工程]