Machine Learning Optimized Method For Spatial Estimation Of Forest Biomass Of Tianshan Spruce

In natural succession,forests will form complex internal structures due to climate,topography,soil,human disturbance,and interspecific changes,resulting in spatial heterogeneity at the regional scale.As the most basic characteristic parameter of forest,forest biomass is often used to study forest productivity and spatial structure.With the expansion and application of geostatistics in the field of natural ecology,it can help solve technical problems in forest spatial distribution research,such as:using semi-variogram to explore the spatial heterogeneity and spatial correlation of forest biomass;using geostatistics to analyze forest biological Spatial interpolation and mapping are carried out to obtain the regional spatial distribution.Therefore,the application of geostatistical analysis to the study of forest biomass spatial estimation,combined with the environment to explain the spatial distribution pattern of forest biomass,will gradually become a forest research hotspot.This study took spruce in the western section of the Tianshan Mountains in Xinjiang as the research object,preprocessed the 2012 Tianshan spruce forest resource inventory data,and screened 308 plots.For the diameter at breast height of spruce in each plot,use the allometric growth model to calculate the unit biomass of the plot;draw the biomass semivariogram of the plot and fit the theoretical model parameters to explore the spatial heterogeneity of spruce biomass in the study area.Spatial correlation;extract remote sensing,terrain,soil,and meteorological data and analyze the biomass of the sample plot to obtain highly significant auxiliary variables,participate in the construction of spatial interpolation models,and improve model accuracy;use the optimal spatial interpolation model to draw clouds in the study area The spatial distribution map of Chinese fir biomass,and the spatial pattern analysis was carried out.It mainly includes the following research process:(1)Biomass and multi-source data preprocessing of sample plots.The remote sensing,terrain,soil,and meteorological auxiliary data were preprocessed and fused with the Tianshan spruce biomass plot data to extract 34 auxiliary variables.The correlation,significance and collinearity analysis were carried out between the variables and the biomass of the plot,and 10 auxiliary variables with significance and no multicollinearity were finally screened out to participate in the construction of the spatial model and improve the model accuracy.(2)Fitting and analysis of biomass semivariogram of plots.After the square root change of the biomass of the plot,it conforms to the normal distribution and can be analyzed for variation;the linear model,exponential model,spherical model,and Gaussian model are used to fit the semi-variogram of the biomass of the plot,and the exponential model is fitted in the model accuracy evaluation.The effect is the best,the residual value:6.55,the coefficient of determination:0.52,the block-to-base ratio:0.651,and the biomass of the plot has a moderate spatial correlation;using the least squares method to optimize the parameters of the variogram fitted by the exponential model,we get Range:30.15km,nugget value:17.02,abutment value:25.58,least squares fitting residual:6.32.(3)Spatial interpolation and analysis of spruce biomass in the western Tianshan Mountains based on multi-model.Using the spatial autocorrelation model,the spatial heterocorrelation model and the combined model to spatially interpolate the biomass of spruce in the western Tianshan section and draw the interpolation surface;in the model accuracy evaluation,the ordinary Kriging combined model of random forest combined with residuals is the best,and the coefficient of determination is:0.642,root mean square error:40.18t/hm~2,relative root mean square error:0.36;the optimal interpolation surface was clipped using the spruce growth subclass information to obtain the spatial distribution map of biomass in the growing area of spruce;Moran index analysis showed that Tianshan spruce biomass is positively correlated in space;high/low cluster analysis shows that the spruce biomass in Turks,Gongliu,and Xinyuan areas shows obvious high-value clusters;for spruce sample points combined with environmental analysis,It can be seen that due to the influence of topography,meteorology and soil,the biomass of Tianshan spruce forms different spatial distribution patterns in the region.