| Forest biomass estimation not only provides a foundation for the research of forest carbon storage,but also an important basis of global climate change analysis.Remote sensing determines the main direction of the research of biomass estimation because of its advantages of speedy,high efficiency and dynamic performance.Shitai County,the National Key Biosphere Reserves,act as an important carbon base of Anhui province and even in the middle and lower reaches of Yangtze River,is located in southern Anhui Province.The biomass research of Shitai County has important significant to its further research on carbon storage,the evaluation of the function of National Biosphere Reserves and its role in global climate change.Based on the major problem-how to estimate the biomass rapidly and exactly in the process of forest biomass estimation,this study choose Shitai County as the study area,and combining Rapideye image and the ground biomass survey data of different scales of Shitai County to build the biomass reversion model of different forest types.The main conclusions are as follows:(1)Compared with traditional characteristics of remote sensing factors,this article introducing the chlorophyll red edge model(CRM)and chlorophyll Green model(CGM)into model building based on the internal mechanism of forest biomass formation.Results shows that both CRM and CGM has significant correlation with the biomass of the main forest types in Shitai County at the 0.01 level.It illustrates that CRM and CGM can be used for forest biomass estimation.In addition,the texture measurements which have strong correlation with biomass are mainly contained in the band of red or red edge,and only MEAN,VAR and SM filter contributes great to the process of estimation.(2)Comparing between the prediction accuracy of random forest model and precision of multiple linear regression model turns out random forest algorithm is superior to multiple linear regression method.It main caused by the selection of variables of linear regression is very strict,which lead to the variables participate in the modeling tend to be less and the underutilization of remote sensing information.(3)Through the contrast of the precisions of regression results of the biomass reversion model of different forest types,the effect of coniferous and broadleaved mixture forest is significant because of less quantity of modeling samples.The prediction model of broadleaved forest got a higher value of adjust R2 than conifer forest prediction model,but the random forest model shows opposite results.It means that random forest model can be better explained the non-linear relationship between AGB and indices extract from images.Inversion model of conifer forest AGB has an higher accuracy of prediction than broadleaved forest AGB inversion model.It would be concerned with the estimate deviation to deciduous broadleaved tree species which created by the season of the image we selected are different to the inventory season.(4)Using the random forest regression model of different forest types to estimate the layer biomass of the main forest types in Shitai County,calculate the total biomass of it then,get the total aboveground biomass of 3501008.614 ton of the study area in Shitai County,of which 1818037.1 ton of broadleaved forest with the AGB of 92.3820t/ha and a total area of 19679.5600 hectares,accounting for 51.93% of total AGB;883844.3676 ton of conifer forest with the AGB of 60.4387t/ha and a total area of 14623.815 hectares,accounting for 25.25% of total AGB and 799127.1314 ton of broadleaved and conifer mixture forest with the AGB of 63.7372t/ha and a total area of12537.845 hectares,accounting for 22.82% of total AGB.(5)The distribution of AGB of Shitai County mainly located in three levels as 20-40t/ha,40-60t/ha and 60-80t/ha.And the forests with high biomass are distributed in the sunny slope of the southern part of whole study area intensively. |
PDF Full Text Request