Spatial Simulation Precision Analysis For The Distribution Of Multi-scale Forest Carbon

Global warming is not only the key to global ecological environment research but also getshighly attention from every government to the problems of global warming brings seriouslythreatens the global economic development and the living quality of human being. As the principalpart of terrestrial ecosystem, forest plays an irreplaceable role in balancing global carbon andslowing global warming. It is important for carbon sink function to use multi-method to estimatecarbon stock on the foundation of forest inventory data and remote sensing data.This article estimates the forestry carbon density of study area in the methods of OrdinaryKriging Interpolation, simulation (sequence Gauss co-simulation), block simulation andregression simulation. And we use the data from323plots and different remote sensing datumfrom in Xian-Ju, Zhejiang province after summarizing the research achievement and researchmethod at home and abroad. The results are as following:(1)The distribution trends of Ordinary Kriging Interpolation map, simulation map, blocksimulation map and regression simulation are in accordance with the true situation. The areas withrelatively high density carbon mainly locate in the Ping-xi forest farm with northeast-southwesttrend, Miao-Liao forest farm in the south and the regions with sparse villages and small towns inthe east, all of which are affected by landform, hydrologic condition, maintenance andmanagement etc. The areas with lightly density carbon and non-stocked land mainly locate on themiddle region where Yong-Xi river runs over and is notably affected by residence, villages andsmall towns human factors.(2) The results of Ordinary Kriging Interpolation, simulation, block simulation map andregression simulation show the smooth effect from statistical eigenvalues. The maximum valuereduces and the minimum value increases. Ordinary Kriging Interpolation shows better smootheffect than analogue simulation, reflecting the week ability of non-forest areas.Block simulation block transforms the cartographic unit from30m×30m to900m×900m.Andsmooth effect appears, but its statistical value is close to true value. The transformation is notaccurate for the heterogeneous effect of the surface, however it provides a new method for scale transformation.Regression simulation is affected by MODIS image and scale transformation error、regression error etc. The most obvious smooth effect emerges in regression and clusteringphenomena appears too. There is little difference between regression simulation other methods.Point data is transformed into polygon data through analogue simulation and is built by upscalingwith MODIS data. This method can match the ground plot data with multi-resolution remotesensing image, being a new trial for estimating forest carbon by remote sensing.(3)Image resolution is dropping with scaling up and the heterogeneity of surface decreases.Its capability to present original data reduces too. Cross check which uses average error, meanaverage error, root mean square error and normalize root mean square root shows the precisioncomparison of the four methods: analogue simulation>block analogue simulation> OrdinaryKriging Interpolation> regression simulation.