文献类型：期刊文献

英文题名：Filtering and gap-filling strategies in eddy covariance CO2 flux series of mountainous forests: Comparison based on measurements from Qingyuan-Ker Towers

作者：Li, Tongtong[1,2,3] Gao, Tian[1,3,4,5] Teng, Dexiong[1,3,4,5] Chen, Zhi[6] Yang, Bai[7] Wang, Xingchang[4,8] Zhao, Changming[9] Zhang, Jinsong[10,11] Huang, Hui[10,11] Guan, Chao[9] Wu, Jiabing[1] Yu, Fengyuan[1,3,4,5] Zhang, Jinxin[1,3,4,5] Sun, Yirong[1,3,4,5] Li, Shuangtian[1,3,4,5] Zhou, Xinhua[3,4,7] Zhu, Jiaojun[1,3,4,5]

第一作者：Li, Tongtong

通信作者：Gao, T[1];Gao, T[2];Gao, T[3];Gao, T[4]

机构：[1]Chinese Acad Sci, CAS Key Lab Forest Ecol & Silviculture, Inst Appl Ecol, Shenyang 110016, Peoples R China;[2]Shandong Normal Univ, Coll Geog & Environm, Jinan 250358, Peoples R China;[3]Natl Observat & Res Stn, Qingyuan Forest CERN, Shenyang 110016, Liaoning, Peoples R China;[4]Chinese Acad Sci, Inst Appl Ecol, CAS CSI Joint Lab Res & Dev Monitoring Forest Fl, Shenyang 110016, Peoples R China;[5]Key Lab Terr Ecosyst Carbon Neutral, Shenyang 110016, Liaoning, Peoples R China;[6]Chinese Acad Sci, Key Lab Ecosyst Network Observat & Modeling, Inst Geog Sci & Nat Resources Res, Beijing 100101, Peoples R China;[7]Campbell Sci Inc, Logan, UT 84321 USA;[8]Northeast Forestry Univ, Sch Ecol, Harbin 150040, Peoples R China;[9]Lanzhou Univ, Coll Ecol, State Key Lab Herbage Improvement & Grassland Agro, Lanzhou 730000, Peoples R China;[10]Chinese Acad Forestry, Res Inst Forestry, Beijing 100091, Peoples R China;[11]Henan Xiaolangdi Forest Ecosyst Observat & Res Stn, Jiyuan 454650, Peoples R China

年份：2025

外文期刊名：SCIENCE CHINA-EARTH SCIENCES

收录：;EI(收录号：20255219772728);Scopus(收录号：2-s2.0-105025349682);WOS:【SCI-EXPANDED(收录号:WOS:001641976200001)】；

基金：This work was supported by the National Natural Science Foundation of China (Grant No. 32192435), the National Key Research and Development Program of China (Grant No. 2023YFD2200405), the Application and Demonstration Project of Network Security and Informatization Technology, Chinese Academy of Sciences (Grant No. CAS-WX2022SF-0101), and the Key Research and Development Program of Liaoning Province (Grant No. 2023JH2/101800043).

语种：英文

外文关键词：Complex terrain; Forest ecosystem; Eddy covariance; Friction velocity threshold; Gap-filling

摘要：Although the eddy covariance (EC) technique has been widely adopted to measure the CO2 and water vapor exchanges between ecosystems and the atmosphere, challenges remain for its applications in mountainous regions. Ideally, the EC technique requires homogeneous canopies and flat underlying surfaces. Over complex terrain, heterogeneity of source/sink areas across wind directions complicates data analysis, particularly quality control (filtering) and gap-filling of nocturnal CO2 flux data. In this study, we evaluated the performance of nocturnal CO2 flux filtering in mountainous forests (comparing cases with or without wind direction sector partitioning) and assessed different gap-filling methods for the analyzed data that were measured from the Qingyuan-Ker Towers in Northeast China from January 2020 to December 2022. Results showed that surface roughness varied across wind direction sectors, leading to reduced accuracy in nocturnal CO2 fluxes filtering if the conventional friction velocity (u*) threshold method (a single threshold for all wind sectors) was applied. Alternative to the conventional method, identifying the u* threshold (uc*) based on wind sectors could significantly improve the accuracy of nighttime CO2 flux filtering. The gaps in the CO2 flux series of forest ecosystems over complex terrain are filled using four methods: mean diurnal course (MDC), look-up table (LUT), marginal distribution sampling (MDS), and random forest (RF). Biases in the filled nocturnal net ecosystem carbon exchange (NEE) were substantially reduced by 16%, in comparison to the adopted conventional method, following a treatment of data filtering based on wind-sector-specific uc* values. In general, the accuracy of all four gap-filling methods decreased with increasing gap length; however, the RF method consistently outperforms the others, yielding more robust and reliable estimates of annual NEE. Through extensive analyses in this study, we recommend identifying uc* values for canopies in different directions surrounding a flux tower, with a moving window and a careful selection of gap-filling methods when processing EC data from a mountainous forest. These findings advance methodologies in handling flux data from forest ecosystems over complex terrains.