Forest Biomass Estimation In Nanchuan District Of Chongqing City Using A Combination Of Sentinel-1 And Sentinel-2 Data

Ecological forest is the main body of terrestrial ecosystem,which contains huge carbon sequestration energy,plays an important role in the global carbon cycle,and has a good regulation effect on heat,water vapor and climate.With the development of forest biomass estimation research,the combined use of optical and radar data has great development space and potential for forest parameter extraction and biomass estimation.This paper mainly focuses on the estimation of forest biomass,taking Nanchuan District of Chongqing City as the research area,combining optical data Sentinel-2 and radar data Sentinel-1,initially extracting 77 eigenvalues,after Pearson correlation analysis,a total of selected 15 factors are involved in the modeling,using multiple linear regression,BP neural network,and support vector machine three biomass estimation modeling methods.The estimation accuracy based on single source data and multi-source data is compared and analyzed,and finally according to the optimal inversion The model constructed the spatial distribution map of forest biomass in Nanchuan District.The results and conclusions of this research are as follows:(1)The texture factors in the optical data Sentinel-2 and the radar data Sentinel-1have good sensitivity to biomass estimation,and have great potential in biomass estimation research.The forest biomass inversion model based on multiple linear regression,BP neural network,and support vector machine based on Sentinel-1 data has R~2 of 0.295,0.5449,0.499,and RMSE of 49.78 t/ha,49.43 t/ha,and 49.87 t,respectively./ha;Based on the biomass inversion model of Sentinel-2 data,the R~2 of the model is 0.495,0.6495,0.592,and the RMSE is 39.97 t/ha,42.88 t/ha,and 51.4t/ha,respectively.Overall,the biomass inversion model based on Sentinel-2 data is more accurate than the biomass inversion model based on Sentinel-1 data.(2)The forest biomass model based on the cooperative inversion of Sentinel-1data and Sentinel-2 data,the R~2 of multiple linear regression,BP neural network,and support vector machine model inversion are 0.548,0.8209,0.7347,and the RMSE is respectively 37.42 t/ha,32.39 t/ha,41.65 t/ha.The research results show that the determination coefficients of the biomass inversion model based on collaborative inversion are all higher than the inversion model based on a single data source,and the BP neural network based on machine learning supports The vector machine modeling and inversion accuracy is slightly higher than the multiple linear regression model.(3)The BP neural network model of Sentinel-1 and Sentinel-2 was used to produce the forest biomass distribution map in Nanchuan District.Through statistical analysis,the total forest biomass in Nanchuan District is 1.3×10~7 t,and the biomass density in the forest area is mostly between 100-150 t/ha,accounting for 19.55%of the total area of the study area and 48.2%of the total biomass.The spatial distribution of biomass is analyzed in combination with topographical factors,and its distribution characteristics are as follows:1)As the altitude increases,the biomass density first rises and then decreases,and the biomass density decreases significantly above 1500meters.2)With the change of slope,the greater the slope,the higher the density of forest biomass.The biomass ratio of the study area is slope>gentle slope>steep slope>steep slope>flat slope>dangerous slope.Biomass is mainly concentrated in areas with gentle slopes,slopes and steep slopes,accounting for 84.12%of the total biomass.3)The proportion of the slope area in the study area is relatively uniform,and the biomass density is semi-sun slope>shade slope>sun slope>semi-shady slope.The highest biomass proportion is semi-sun slope,which accounts for 30.83%,and the total amount is 403.04×10~4 t;the lowest biomass proportion is sunny slope,which accounts for 18.46%,and the total amount is 241.36×10~4 t.