Study On Biomass Inversion Method Of Reclaimed Vegetation In Mining Area Based On Worldview-3 And Sentinel-1 SAR Data

Remote sensing quantitative inversion of reclaimed vegetation biomass in mining area is the basis of dynamic monitoring and evaluation of mining ecological environment.This paper took the reclaimed vegetation in Inner Mongolia grassland open-pit coal mine as the research object,and based on Worldview-3 and Sentinel-1 SAR data,combined the advantages of remote sensing optics and radar data to explore the inversion method of reclaimed vegetation biomass.Principal component-wavelet transform algorithm was selected for data fusion,This paper revealed the correlation between parameters such as band reflectivity,vegetation index,backscatter coefficient and texture feature and biomass,established a multivariate biomass inversion model,and analyzed the spatial uncertainty of different biomass models.The following conclusions are drawn:（1）The image fusion of the WV-3 data and the VV polarization of Sentinel-1 SAR data was conducted by the wavelet principal component analysis（W-PCA）,the sharpness as well as the spectral and structural texture features of the fused image are visually better than those of a single optical or SAR image.The higher entropy value reflects that more details are present in the fused image compared to the optical WV-3 image.Also,the higher average gradient indicates that the sharpness and texture information of the fused image are better than those of the SAR image.The 8th band of the fused image generates higher similarity coefficients and lower spectral distortion values in comparison to the WV-3 multi-spectra image,which shows more texture features and less spectral changes.Based on the WV-3 data,Sentinel-1 SAR data and the feature set extracted by the fusion data,the random forest classification method is used to extract the vegetation information,the results show that the fusion data has the highest classification accuracy（total classification accuracy:80.3%,KAPPA coefficient:0.77）,further confirming the potential of Sentinel-1 SAR and WV-3 fusion data in vegetation research.（2）Through correlation analysis,EVI,NDVI,VH polarization scattering coefficient,VH mean texture and the 8th band after fusion have a significant positive correlation with biomass.Compared with a single variable,the NDVI of WV-3 and the VH mean texture of Sentinel-1 joint variables modeling accuracy(R²=0.8340,RMSE=16.4646 g·m^-2,Ac=81.52%)is the highest;the 8th band after fusion verification accuracy(R²=0.7983,RMSE=22.8283 g·m^-2,Ac=74.64%)is the highest.（3）Based on the biomass grade distribution and residual uncertainty analysis of different models,the biomass estimated value greater than 150 g·m^-2 was mainly distributed in the reclamation area in 2013 and 2015 with high uncertainty.The Sentinel-1 data variables are more likely to overestimate and saturate.The joint variables can achieve complementary advantages.The fusion variable significantly improved the saturation and significantly reduced its uncertainty when the biomass is greater than 100 g·m-1,and the uncertainty decreased by 2.42 g·m^-2 to 9.68 g·m^-2 on average.