• 研究简报 •

### GF-3全极化SAR数据极化分解估算人工林冠层生物量

1. 东北林业大学林学院 森林生态系统可持续经营教育部重点实验室 哈尔滨 150040
• 收稿日期:2018-06-25 出版日期:2020-09-25 发布日期:2020-10-15
• 通讯作者: 范文义
• 基金资助:
"十三五"国家重点研发课题(2017YFB0502700)

### Polarimetric Decomposition Parameters for Artificial Forest Canopy Biomass Estimation Using GF-3 Fully Polarimetric SAR Data

Jingyu Wei,Wenyi Fan*,Ying Yu,Xuegang Mao

1. Key Laboratory of Sustainable Forest Ecosystem Management of Ministry of Education School of Forestry, Northeast Forestry University Harbin 150040
• Received:2018-06-25 Online:2020-09-25 Published:2020-10-15
• Contact: Wenyi Fan

Abstract:

Objective: The object of this study was to explore the potential of GF-3 fully polarized SAR data in the estimation of artificial forest canopy biomass and to make an attempt to propose an accurate estimation method for forest canopy biomass. Method: In this study,based on the data from 22 sample plots of Pinus tabulaeformis and Larix principis-rupprechtii artificial forest in Wangyedian Forest Farm,Chifeng,Inner Mongolia and from GF-3 fully polarimetric SAR,Freeman three-component decomposition,Freeman two-component decomposition and Yamaguchi three-component decomposition were used to obtain polarization decomposition components. The volume-to-ground scattering ratio parameters (R1,R2 and R3) corresponding to each polarization decomposition method were constructed respectively. The multiple stepwise regression method was used to establish regression models for forest canopy biomass and SAR extraction parameters,and the model was evaluated using the leave-one-out cross test. Result: There were significant negative correlations between the components obtained by different polarization decomposition methods and the canopy biomass. The correlation between the volume scattering component of the Freeman three-component decomposition and the canopy biomass (r=-0.68) was higher than that between the double-bounce scattering component and the canopy biomass (r=-0.6),and also higher than that between the surface scattering component and the canopy biomass (r=-0.424). Similarly,the correlation between the volume scattering component of the Freeman two-component decomposition and the canopy biomass (r=-0.718) was higher than that between the ground scattering component and the canopy biomass (r=-0.62). The correlation between the double-bounce scattering component of Yamaguchi three-component decomposition and the canopy biomass was the highest (r=-0.743). Compared with Freeman three-component decomposition,the decomposition components of Freeman two-component decomposition and Yamaguchi three-component decomposition had better correlations with the canopy biomass. The optimal parameters obtained by stepwise regression were the volume-to-ground scattering ratio parameters R2 and R3 corresponding to Freeman two-component decomposition and Yamaguchi three-component decomposition,respectively. Then the canopy biomass estimation model was established (R2=0.658,RMSE=4.943 t·hm-2). The cross-validation results of the model showed that the prediction error of the model was relatively low (ME=-0.665 t·hm-2,MAE=4.845 t·hm-2,MRE=3.33%,AMRE=23.233%,P=91.5%). The model was tested by means of confidence ellipse F test,and the estimated values were consistent with the measured values. The simulation results were good,and the predicted values were generally distributed near the 1:1 line. No saturation point appears in the model. Conclusion: The decomposition components obtained by the three polarization decomposition methods of Freeman three-component decomposition,Freeman two-component decomposition,and Yamaguchi three-component decomposition were significantly correlated with the canopy biomass. The rotation transformation of polarization coherence and the optimization of volume scattering model could effectively improve polarization decomposition effects of forest area. The volume-to-ground scattering ratio parameters were more sensitive to the canopy biomass than each single polarization decomposition component. Collaborative use of multiple polarization decomposition components could estimate forest canopy biomass more accurately. Polarization decomposition was feasible for estimating forest canopy biomass,and there was no obvious saturation problem.