Research On Estimating Sparse Photosynthetic/non-photosynthetic Vegetation Fractional Cove By Fusion Of Sentinel-1 And Sentinel-2

Photosynthetic vegetation and non-photosynthetic vegetation play an important role in the ecosystem of arid and semi-arid regions,affecting the carbon storage of the ecosystem,vegetation productivity and are important indicators for evaluating the status of surface vegetation coverage.Accurate and quantitative inversion of photosynthetic vegetation and non-photosynthetic vegetation coverage plays a vital role in the monitoring of sparse vegetation in arid areas,understanding the carbon cycle process and ecological resource management.At the same time,non-photosynthetic vegetation coverage information is obtained as land desertification and vegetation conversion mechanism research provides important information.This paper mainly conducts research on the estimation method of sparse NPV coverage from two aspects:model and data source.First,analyzed the rationality of the basic assumptions of the tri-endmember linear mixture model in the characteristic space of the photosynthetic vegetation index(Normalized vegetation index(NDVI),red-edge Chlorophyll Index(CI_red-edge),Modified Chlorophyll Absorption Ratio Index(MCARI))and the non-photosynthetic vegetation index(Dead Fuel Index(DFI)).Use Sentinel-2A as the data source,Propose the ratio soil index(RSI)into the tri-endmember linear mixture model with reference to the measured spectrum in the field and linear index model was constructed to estimate the coverage of each component of PV,NPV and BS in a typical sparse vegetation experimental area.Qualitative and quantitative accuracy evaluation based on the measured endmember abundance in the field.Next,Using Sentinel-1 and Sentinel-2 as the data sources,based on the random forest model,we carried out the research on the sensitivity of the VV and VH polarization modes of Sentinel-1 data to PV/NPV for the cooperative inversion of vegetation coverage by microwave and multispectral.Finally,Analyze the PV/NPV estimation accuracy of each model,and select the best estimation model for PV/NPV coverage estimation in Minqin Oasis,Gansu.The main conclusions are as follows:(1)The characteristic spaces of NDVI-DFI,MCARI-DFI,and CI-DFI vegetation index are all represented as triangles,which meet the basic conditions for the construction of the pixel three-division model.Therefore,a three-point pixel model based on NDVI-DFI,MCARI-DFI,and CI-DFI indexes was constructed to realize the estimation of sparse PV/NPV coverage in arid areas.(2)The NDVI-DFI linear vegetation index model integrated with the RSI index can effectively estimate the photosynthetic vegetation coverage and non-photosynthetic vegetation coverage of sparse vegetation in typical arid areas,providing a more reliable theoretical basis for the quantitative estimation of NPV coverage.(3)The random forest model can be effectively applied to the estimation of sparse vegetation coverage in arid areas.The minimum RMSE is 0.01044 and the minimum RMSE%is 32.7%,showing great application potential in the estimation of vegetation coverage.(4)The VV and VH polarization modes of Sentinel-1 data have a good correlation with PV/NPV.The VH polarization mode is more sensitive to NPV,and the VV polarization mode is more sensitive to PV.