Uncertainty Of Forest Carbon Storage Estimation In Zhejiang Province

Forest is the largest organic carbon pool in terrestrial ecosystem and plays an important role in global carbon balance.At present,the study on regional forest carbon storage estimation is a hot issue,but the field is facing two challenges.One common challenge is to quantify spatial patterns and distribution of forest vegetation carbon and scale up point measurements to larger national,regional,and global scales.In addition,the estimates of forest carbon sinks have uncertainties because of inaccurate data;inadequate methods;and gaps in understanding of the physiological processes and relationships among carbon,plants,and soils.So,it’s important to solve the problem of up-scaling and quantifing the uncertainty which has an important influence on the accurate estimation of forest carbon storage.Based on these facets,Zhejiang province is chosen as the representative of large-scale area in this thesis,by using aboveground biomass data for Cunninghamia lanceolate,MOD13Q1 image,Landsat TM data,National Forest Inventory data in 2009,combined with Sequential Gaussian co-simulation model and error back-propagation neural network(BPNN)to estimate forest carbon storage and its distribution in the study area.Then the error propagation law was used to quantify the uncertainty in the estimation of forest carbon storage,including the uncertainty based on the forest inventary data and the uncertainty based on remote sensing data.The study revealed that for the model with diameter at breast height(DBH),the uncertainty caused by the measurement uncertainty,model uncertainty and sampling uncertainty was about 1.58%,18.51%,9.12% of the total biomass,respectively.And the total uncertainty was 20.70% of the estimation from the model with DBH.As for the model with DBH and tree height(H),the uncertainty of the measurement,model uncertainty and sampling uncertainty was about 3.42%,11.43% and 9.15% of the total biomass,respectively.And the total uncertainty was 15.03% of the estimation.Among the three sources of uncertainty,the model uncertainty has the greatest effect in the estimated results,followed by sampling uncertainty,and the measurement uncertainty is relatively small.In addition,the uncertainty of the residual variability decreased with an increase of the sample size.When the sample size was increased from 30 to 42 and 48,the residual variability uncertainty of the unary model decreased from 15.2% to 12.3% and 11.7%,respectively.As for the binary model,the residual variability uncertainty decreased from 13.3% to 9.4% and 8.3%,respectively.The values of plot tree above-ground carbon density was scaled up from the plots and pixels of 28.5 m 28.5 m to map units of 240 m× 240 m.And a coefficient 0.81 of correlation between the estimates at the larger blocks and the observations at smaller sample plots was obtained which indicated the larger blocks value could reflect the distribution of carbon storage to a certain extent.The estimation results using BPNN showed that the mean carbon density was 14.71 Mg/hm2 which was higher than the average from the sample plots with a relative error of 7.21%.As fo the uncertainty based on remote sensing data,the model uncertainty has the greatest effect in the estimated results.The relative uncertainty was about 56.32%,while the plot uncertainty was only about 7.97%.And the total uncertainty was about 56.88%.From the above results,we can draw the following conclusions:(1)As for uncertainty based on the forest inventary data,the model uncertainty has the greatest effect in the estimated results,followed by sampling uncertainty,and the measurement uncertainty is relatively small.Therefore,it is important to improve the accuracy of single tree biomass model to improve the estimation accuracy of forest carbon storage.(2)As for the uncertainty of forest biomass estimation based on remote sensing,the main uncertainty sources was the uncertainty of remote sensing model,and the plot uncertainty is much less.Therefore,the accuracy of remote sensing model is very important to the estimation accuracy when the forest carbon storage is estimated by remote sensing.To improve the estimation accuracy of forest carbon storage,the accuracy of remote sensing model should be taken as the key point.(3)Finer spatial resolution satellite data could been used as an intermediate step when relating ground reference data with coarser spatial resolution data.