Estimating Fractional Cover Of Photosynthetic/Non-photosynthetic Vegetation In The Xilingol Typical Grassland Region With Remote Sensing Data

Photosynthetic Vegetation?PV?and Non-photosynthetic Vegetation?NPV?play an important role in grassland ecosystem,because they affect the ecosystem carbon storage,CO2 exchange capacity,the productivity of the vegetation and the surface energy balance,so it is an important index to measure the surface vegetation cover conditions.The quantitative estimation of fractional cover of photosynthetic vegetation(f _PV),non-photosynthetic vegetation(f_NPV),and bare soil(f_BS)is critical for grassland ecosystem carbon storage,vegetation productivity,soil erosion and wildfire monitoring.In this paper,the data source of the measured spectral and sample coverage data is presented.Firstly,this study analyzes the spectral characteristics of NPV,PV,and BS endmembers,and evaluate the correlation between different spectral indices and f_NPV by simulating mixed scenes and real mixed scenes in the field.On this basis,explore the NDVI-DFI feature space correspond to the fundamental assumption of ternary linear mixed model.Secondly,using the MOD09A1 surface reflectance 8-day composite product as the data source,the NDVI-DFI model was used to estimate the f_PV,f_NPV and f_BS in the typical grassland area of Xilingol,and the qualitative and quantitative accuracy evaluation was performed.Finally,the time series f_PV,f_NPV and f_BS for the Xilingol typical grassland in 2017 were estimated to analyze seasonal changes.The main conclusions are as follows:?1?In the presence of three components of NPV,PV and BS,DFI index was significantly correlated with f_NPV,and DFI index could effectively estimate non-photosynthetic vegetation coverage in steppe.Through regression analysis of several spectral indices,it was found that DFI and CAI indices were significantly correlated with f_NPV under simulated NPV-PV-BS mixed scenes,The Coefficient of Determination?R²?was 0.953 and 0.852,and the Root Mean Square Error?RMSE?was 0.09 and0.05?n=66,p<0.001?,respectively.However,the correlation between NDI and NDSVI index and f_NPV is very low.Additionally,compared to simulate the mixing,The effectiveness of f _NPV estimated by DFI and CAI index under field mixed scenes was reduced to some extent,R²was 0.745 and 0.712,and RMSE was 0.117 and 0.125,respectively.The R²and RMSE of the NDI index were 0.053 and 0.227,respectively;while the NDSVI index was the worst,R²and RMSE were 0.007 and 0.234,respectively.?2?The NDVI-DFI feature space constructed using the MODIS simulation spectrum and the measured spectral reflectance data all appear as distinct triangles,and conform to the basic assumption of the ternary linear mixed model.It is theoretically feasible to construct an NDVI-DFI ternary linear mixed model by selecting the NDVI index to characterize the f_PV and DFI index to characterize the f_NPV.?3?The NDVI-DFI ternary linear mixed model can be used to estimate the f_PV,f _NPVPV and f_BS of typical grasslands.The estimation accuracy of f_PV,f_NPV and f_BS using the image endmember method is higher than that of the measured endmember method.Among them,for f_NPV,the RMSE estimated by the image endmember method is 0.1210?R²=0.70?,and the RMSE estimated by the measured endmember method is 0.1501?R²=0.71?;For f_PV,the RMSE estimated by image endmember method is 0.0482?R²=0.75?,and the measured endmember method appeared to be significantly overestimated with an estimated RMSE of 0.0736?R²=0.71?;For f_BS,the RMSE estimated by the image endmember method is 0.1299?R²=0.73?,which is higher than the estimated RMSE of the measured endmember method by 0.1813?R²=0.74?.?4?The NDVI-DFI ternary linear mixed model has reasonableness and applicability in the typical grassland area.The time series f_PV,f _NPV and f_BS estimated by this model are basically consistent with the characteristics of grassland vegetation.The study extends the application range of the ternary linear mixed model and fully explores the potential of multispectral remote sensing estimation of f_PV,f _NPV and f_BS,providing a basis for future monitoring of f_PV,f_NPV and f_BS in grasslands with large-scale,high-precision and long-term sequences.