| Forest biomass is closely related to the processes of material cycle and energy flow of terrestrial ecosystems,it is an important parameter for assessing forest carbon budgets.As an important indicator reflecting forest quality and ecological benefits,forest biomass plays a crucial role in guiding the management of regional forest resources and environmental protection.In this study,taking the upper reaches of the Heihe River as an example,we based on airborne Light Detection And Ranging(LiDAR)data,Ziyuan-3(ZY-3)optical data,terrain data and measured forest aboveground biomass,and used five prediction methods including Random Forest(RF),Support Vector Machines(SVM),Back Propagation Neural Network(BPNN),K-nearest Neighbor(KNN)and Generalized Linear Mixed Model(GLMM)to estimate Forest Aboveground Biomass(AGB).By comparison of predicted results used various data sources and different modelling methods,we select the optimal model to estimate and map the forest Aboveground Biomass.Based on the predicted result of Forest AGB,we combine the measured soil particle composition in different soil horizons,terrain variables and vegetation indexs,and use the RF method to explore the effect of predicted result of soil composition in different soil horizons with incorporation of biomass.The main research conclusions were summarized as follows:(1)In the study area,Forest AGB were significantly correlation with Normalized difference Vegetation Index(NDVI)and texture information extracted from optical data at 0.05 and 0.01 levels,respectively.In the correlation analysis between Forest AGB and predicted variables derived from the Airborne LiDAR data,the mean of laser hight in sample plot has a strong positive correlation with Forest AGB,and the Pearson correlation coefficient reached 0.854 at 0.01 level.Followed by the maximum of laser height and canopy cover,there was a significant positive correlation at the 0.01 levelt and the Pearson correlation coefficient was 0.808 and 0.793,respectively.In the relationship between terrain variables and Forest AGB,there was no significant difference in distribution of Forest AGB at different elevation gradients.The Forest AGB on the flat slope was higest in different slope gradients,and it was reaching 104.767 t/ha,and Forest AGB distribution on the flat slope has a significantly different with the steep slope at 0.05 level.The Forest AGB on the north aspect was higest than others aspect gradients,and it was significantly different with the others.Forest AGB has a negetaive correlation with Terrain Wetness Index(TWI)and Multi-resolution Valley bottom flatness(MRVBF)at the 0.01 level,and the Pearson correlation coefficients were-0.451 and-0.245,respectively.(2)In the study,wer built five predicted varibles datasets:optical data airborne LiDAR data,optical data and terrain data,airborne LiDAR data and terrain data,and integrated airborne LiDAR and optical data;and used five RF,SVM,BPNN,KNN and GLMM five modeling methods to estimate.the ForestˇAGB in small catchmet located on the upper reaches of the Heihe River.Models based on the RF algorithm emerged best among the five algorithms irrespective of datasets used.The Random Forest AGB model,using only LiDAR data(R2=0.899,RMSE=14.0 t/ha)as input data,was more effective than one using optical data(R2=0.835,RMSE=22.724 t/ha).Compared to LiDAR or optical data alone,the AGB model(R2=0.913,RMSE=13.352 t/ha)which used the Rf algorithm and integrated LiDAR and optical data was found optimal.Incorporation of terrain variables with optical data resulted in only slight improvements in accuracy.(3)We analyzed the spatial distribution characteristics of soil particle composition and its correlation with biomass,vegetation index and terrain factors in the study area.The results indicated that the formation and evolution of soil particle composition was affected by factors of terrain and vegetation.In the.analysis of soil particle composition and vegetation factors,biomass and NDVI was significantly positively correlated with soil sand content at 0.01 and 0.05 level in 0-10 cm soil horizon,respectively.In the 10-20 cm soil horizon,biomass and soil silt content were significantly negatively correlated at 0.01 level,and significantly positively correlated with soil sand content.In the 20-30 cm soil horizon,biomass has no correlated relationship with soil particle composition.In the analysis of the correlation between terrain variables and soil particle composition,the soil clay content was gradually increased with the decrease of elevation,and the distribution of soil particle composition was affected by the slope factor.The soil clay,soil silt and soil sand contents were significant differences in different slopes.The distribution of soil clay content and silt content were less affected by the aspect,while the distribution of soil sand content has significant differences in different apect gradients.In general,the distribution of soil particle composition has a certain regularity in the study area,and formation and migration of it was affected by the factors such as vegetation and terrain.(4)Based on the terrain data,terrain data and vegetation indexs,and terrain data and biomass,we used the RF method to predict the distribution of soil particle composition in different soil horizons.The result showed that the predicted result of the soil silt content which based on the terrain data was optimal in different soil horizons.The predicted result of the soil clay content which based on the terrain data and vegetation indexs was optimal in different horizons,the soil sand content was followed and the soil silt content was the worst.The predicted result of the soil silt content which based on the terrain data and biomass was optimal in different horizons,the soil silt content was followed and the soil sand content was the worst.In the 0-10 cm soil horizon,the addition of biomass can improve the prediction accuracy of soil clay content and soil silt content.In the 10-20 cm soil horizon,the addition of vegetation indexs can improve the prediction accuracy of soil clay content,while have a little effect on predicted result of soil silt content and soil sand content.In the 20-30 cm soil horizon,the addition of vegetation indexs and biomass have a little effect on predicted result of soil particle composition. |
PDF Full Text Request