文献类型：期刊文献

英文题名：Forest aboveground biomass estimation using polarization coherence tomography and PolSAR segmentation

作者：Li, Wenmei[1,2] Chen, Erxue[1] Li, Zengyuan[1] Ke, Yinghai[3] Zhan, Wenfeng[4]

第一作者：Li, Wenmei

通信作者：Chen, EX[1]

机构：[1]Chinese Acad Forestry, Inst Forest Resources Informat Tech, Beijing, Peoples R China;[2]Nanjing Univ Posts & Telecommun, Coll Geog & Biol, Nanjing, Jiangsu, Peoples R China;[3]Capital Normal Univ, Coll Resource Environm & Tourism, Beijing, Peoples R China;[4]Nanjing Univ, Int Inst Earth Syst Sci, Nanjing 210008, Jiangsu, Peoples R China

年份：2015

卷号：36

期号：2

起止页码：530-550

外文期刊名：INTERNATIONAL JOURNAL OF REMOTE SENSING

收录：;EI(收录号：20150600497521);Scopus(收录号：2-s2.0-84961288785);WOS:【SCI-EXPANDED(收录号:WOS:000348710300007)】；

基金：This work was partially supported by National Key Basic Research Development Programme (973 Programme) sub-project [Grant 2013CB733404]; High Technology Research Project [Contract number 2011AA120402]; National Natural Science Foundation of China [Grant 41401480].

语种：英文

外文关键词：Biomass - Budget control - Climate change - Forestry - Image segmentation - Mean square error - Polarimeters - Polarization - Synthetic aperture radar - Tomography

摘要：Forest aboveground biomass (AGB) is essential for monitoring the carbon cycle budget and climate change. This study proposes a method for the estimation of forest AGB based on polarization coherence tomography (PCT) and polarimetric synthetic aperture radar (PolSAR) segmentation. The data used are the single-baseline polarimetric SAR interferometry data acquired by the German Aerospace Center's E-SAR sensor at the L-band over the city of Traunstein and its vicinity in 2003. First, vertical structure tomographic profiles (relative reflectivity distribution in vertical direction) were produced by PCT for each pixel and then averaged within each field-surveyed forest stand to obtain the mean tomographic profile. Next, the mean vertical tomographic profiles were parameterized by 10 shape parameters. Several models for biomass estimation were designed based on the relationships between the definition parameters and the in situ measurements using backward step-wise regression and the M-fold (10-fold) cross-validation method. The best model was chosen by evaluation criteria such as R, root mean square error (RMSE), R-2, etc. Forest polygons (objects) were produced by image segmentation using forest heights, A(1) (produced by coherence optimization representing double-bounce scattering mechanism), A(2) (produced by coherence optimization representing the volume scattering mechanism), and the volume scattering mechanism class (produced by Freeman-Durden decomposition and Whishart classification). Finally, the selected model was used to estimate the AGB of each forest polygon and was validated using ground-measured biomass with R-2 of 0.883 and RMSE of 39.98 tons ha(-1). The results show that the proposed method works well for the estimation of forest AGB. No saturation phenomena have been observed even for the forest stands with AGB larger than 500 tons ha(-1).