[Research on Remote Sensing Inversion of Suaeda Salsa’s Biomass Based on TSAVI for OLI Band Simulation]

Suaeda salsa(S.salsa) is a typical vegetation of coastal wetland in the north of Liaodong Bay. The S. salsa biomass assessment plays an important role in understanding the ecosystem productivity of coastal wetland and the formation of ecosystem structure and function. Usually the S.salsa coverage is inhomogeneous. The low S.salsa coverage can be found at a natural condition, the soil background has a strong influence on S.salsa spectral data. The Transformed Soil Adjusted Vegetation Index (TSAVI) used as independent variable was derived by the Landsat 8 OLI simulation data. The S.salsa biomass inversion models were built based on the regression analysis of TSAVI and ground measured biomass in this study. The correlation between TSAVI (600~687, 820~880 nm) and biomass was significant, the correlation coefficient was about 0.9, up to 0.92. The results of linear and quadratic models were better than those of logarithmic, exponential and power models, the determination coefficient r2 of linear and quadratic models were 0.83. Combined with F value and operation efficiency, the linear model was the best option for mature S.salsa biomass inversion. The linear model was applied to invert the S.salsa biomass by using the Landsat 8 OLI data in the study area and it was further validated using in-situ data. The correlation coefficient between the in-situ value and retrieved value was 0.962, the relative error was 0.106. For higher S.salsa coverage, the relative error was lower. The relative error of the low-cover S.salsa biomass inversion was around 0.18. The results showed that the established model has good accuracy for different coverage. In addition, with the introduction of ±5% error of soil line parameters a and b, the average relative errors were relatively stable, and the correlation coefficients were reduced, but all the correlative coefficients were above 0.9. The results showed that the established model is stable.