Comparative Analysis On The Spatial Effects Of Individual Tree Biomass In Typical Subtropical Forests

Currently,the global warming is the serious issues faced by all countries in the world.As the main body of land ecosystem,the forest plays a very important role in slowing down the elevated atmospheric CO₂concentration.Trees are an important element of forest ecosystem,therefore accurate estimates of tree biomass are critical for quantifying forest biomass,carbon stocks and fluxes as well as for the forest management and for assessing climate change impacts on forest ecosystems.However,spatial effect is a common phenomenon in forestry growth and yield data（i.e.correlation among these data）,and this would violate the regression assumptions led to individual tree biomass model（ordinary least squares model）no longer applies.Therefore,it is of great significance to explore the existence or non-existence of spatial effect in biomass data for selecting appropriate biomass modeling theories.In addition,research on spatial effect of forest biomass is of great significance to accurately grasp its spatial distribution law.Based on this,this paper is based on the measured data of different components（i.e.wood,bark,stem,branch,foliage,crown and above-ground）biomass from clear cutting plots（100m×30m）of three typical subtropical forests（i.e.Pinus kesiya natural forest,Pinus kesiya plantation and Eucalyptus spp.plantation）in yunnan province,R software as a tool to:（1）analyse change rule of spatial effect of biomass from different components in three typical subtropical forests by using spatial effect analysis method.（2）On the basis of spatial effect analysis,consider the spatial location information of individual tree,using the spatial regression models and mixed effects model modeling technology,build global spatial regression models（SLM,SEM and SDM）,local spatial regression model（GWR）and mixed effect model（NMEM）of biomass from different components,and the ability was analysed to process data with spatial effects for spatial regression models and mixed effects model.The results showed that:（1）spatial effect among biomass data from different components of Pinus kesiya natural forest,Pinus kesiya plantation and Eucalyptus spp.plantation,but the change rule of spatial effect did not exhibits regular change.Taken alone,different components biomass in three typical subtropical forests showed a certain degree of spatial effect,from the perspective of spatial distribution pattern,the biomass of different components of Pinus kesiya natural forest presents discrete distribution trend basically,Pinus kesiya plantation is similar.Conversely,Eucalyptus spp.plantation presents aggregating distribution trend basically.From the perspective of clustering pattern,Pinus kesiya natural forest and plantation presents positive spatial autocorrelation basically,but Eucalyptus spp.plantation presents positive and negative spatial autocorrelation basically.From the perspective of spatial heterogeneity,different components biomass in three typical subtropical forests presents the tendency of increase with distance scale,stabilizing at 10m.In the same stand,different components biomass with similar spatial effects change rule,e.g.wood,stem and above-ground biomass,but from the perspective of the sequence and degree of the phenomena presents by the changes of the biomass data with the distance scale,the change rule of the spatial effects of each components biomass are different.By comparison,the change rule of spatial effect of same components biomass in three typical subtropical forests also has similarities and differences.From the perspective of spatial distribution pattern,Pinus kesiya natural forest and plantation are similar,and there are significant differences with Eucalyptus spp..From the perspective of clustering pattern,Eucalyptus spp.,Pinus kesiya natural forest and plantation are relatively consistent,e.g.wood,stem and above-ground biomass.From the perspective of spatial heterogeneity,the variation trend of spatial heterogeneity of biomass from the same components between stands was similar,basically increasing with distance scale,and stabilizing at 10m,but the heterogeneity intensity is different.（2）In this paper,global moran’s index and intra-block variance were used to describe the spatial effects of model residuals at different scales and results show that:the residual of OLS model had the problems of spatial effect.Global spatial regression models（SLM,SEM and SDM）can reduce both spatial autocorrelation and spatial heterogeneity in model residuals.GWR model can effectively reduce the spatial heterogeneity in residuals,and also shows the ability to reduce the spatial autocorrelation in residuals.NMEM model has poor ability to deal with the spatial autocorrelation in residuals and can hardly reduce the spatial heterogeneity in residuals.In general,GWR model has the best capabilities for deal with spatial effect in model residuals,followed by the global spatial regression model,and NMEM is the worst.Furthermore,global spatial regression models,local spatial regression model and mixed effect model can basically improve the fitting performance,and prediction performance is similar to OLS.