Study On Vegetation Biomass Estimation Methods Of Terrestrial Ecosystem In China

Terrestrial ecosystem provides an important material basis for human survival.Accurate assessment of terrestrial biomass is a very important issue.At present,there are two main problems in the accurate assessment of terrestrial biomass: 1.There are differences between the results of various evaluation methods.2.The layout of biomass sample points needs to be optimized.Forest biomass is an important indicator of carbon sequestration capacity and ecological service function of forest ecosystem.Accurate estimation of forest biomass can provide decision support for the government to formulate development strategies.Therefore,we can take the assessment of forest biomass in China as an example to explore the differences of assessment results of various methods.Therefore,we can take the assessment of forest biomass in China as an example to explore the differences of assessment results of various methods.In order to quantify the impact of different forest biomass estimation methods on the uncertainty of evaluation results,this paper uses 6474 standardized and systematic field survey data of forest biomass in China to quantitatively evaluate the differences between the estimation results of 10 methods including 2 forest type methods(M1),6 land statistical interpolation methods(M2)and 2 remote sensing models(M3),and further discusses the estimation results of 10 methods The difference of fruit in different ecological regions.The results showed that: for the whole forest region of China,the biomass reserves estimated by the two forest types were15.68%,the biomass reserves estimated by the six interpolation methods were 9.43%,the biomass reserves estimated by the two remote sensing methods were 42.70%,and the biomass reserves estimated by the three methods were 15.81%.As far as the overall estimation value of various methods is concerned,the difference between most methods is not great;but as far as the ecological region is concerned,there are still considerable variations in the estimation of various methods in different regions.In order to explore the most suitable estimation method,we do error analysis and frequency domain analysis on the evaluation results of various methods.The Taylor diagram of error analysis shows that the estimated biomass of forest type method vibrates near the average biomass,which can not well reflect the spatial variation of biomass.Most interpolation methods have low predicted values in high biomass areas,while remote sensing methods are similar to forest type method.It seems that the interpolation method is more suitable for the estimation of forest biomass,and the random forest method and the inverse distance weight method are better than the support vector machine and other traditional interpolation methods.The results of frequency domain analysis show that the estimation results of machine learning interpolation methods contain more abundant trends,which can effectively simulate the trend part of forest biomass distribution and the autocorrelation part of forest biomass distribution.Therefore,machine learning interpolation has the ability to approach the real distribution of forest biomass.In conclusion,random forest is the most suitable estimation method.In order to combine the advantages of the temporal resolution of remote sensing data,we combined the remote sensing data with the random forest method to establish a new forest biomass estimation model.The validation results show that the new prediction model has the same prediction value as the original interpolation,but its extreme value is worth controlling,which can better reflect the local variation of biomass distribution.The estimated forest biomass of the new model is 19.26 ±3.04 Pg.The research on methodology is endless.We hope that our research can provide reference for the future research on the methodology of Forest Biomass Assessment.Scientific terrestrial biomass estimation scheme should include two parts: scientific spatial expansion scheme and sample plot setting scheme.Considering that the most convenient spatial expansion method is interpolation method.Based on the data of 11500 Chinese terrestrial biomass samples from the "carbon project" and the literature,we used the inverse distance weight method,Kriging method,empirical Bayes Kriging method and radial basis function method to estimate the terrestrial biomass of China according to the interpolation schemes of forest,grassland,desert and wetland ecosystems and no ecosystem.Comparing different spatial expansion schemes,it is found that the estimated terrestrial biomass is different,and the inverse distance weight method,empirical Bayesian Kriging method and radial basis function method are the best spatial expansion schemes.We further use the method of sampling by ecological area and circular interpolation to explore the optimal sample plot setting scheme corresponding to the optimal spatial expansion scheme.The optimal sampling scheme is:random sampling according to a certain proportion in different ecological areas,and the total number of sampling points should not be less than2900.To obtain 75% estimation accuracy,2950 sampling points are needed;to obtain 90% estimation accuracy,4900 sampling points are needed;to obtain 95% estimation accuracy,6100 sampling points are needed.The 95% accuracy of the land biomass estimation scheme included 4103 forest samples,1403 grassland samples,428 desert samples and 66 wetland samples.Based on the optimal spatial expansion scheme and the optimal plot setting scheme,we recommend using the radial basis function(RBF)method combined with the optimal plot setting scheme to estimate the terrestrial biomass in China.The estimated land biomass,forest biomass,grassland biomass,desert biomass and wetland biomass in China are 32.57 pg,22.23 pg,6.37 pg,3.01 pg and0.96 pg respectively.Taking the terrestrial biomass as the research object,this paper discusses the biomass estimation scheme based on the combination of sample plot setting and spatial expansion method,which can also provide reference for carbon storage estimation,resource census,ecosystem service function value evaluation,etc.