| The impact of human disturbance such as forest management activities on carbon stocks has been become the focus of attention with the global warming.The harvesting,as a main human disturbance means that human access to timber resources,will have an important influence on forest carbon stocks.The research about the effect of harvesting disturbance on forest carbon stocks is important to further understand the role of forest ecosystem in carbon cycle and to study the forest carbon dynamics.Forest biomass estimation is the basis for the analysis of carbon cycling and carbon dynamics in terrestrial ecosystems,and has become one important content of the ecology and global change researches.The remote sensing technology could make up for the lack of traditional methods used in forest biomass estimation.And the application of multi-source remote sensing data could improve the accuracy of forest biomass estimation and further achieve large regional forest biomass estimation.In this study,Dailing forestry operation management zone,located in Xiaoxing'anling Mountains,was selected as the study area.The multi-source remote sensing data were processed to estimate the large-scale forest biomass in real time.Then the interference evaluation system of harvesting disturbance on forest carbon stocks was established to study and predict the effect of different harvesting disturbances on forest carbon dynamic.The results could provide methods and technical support for forest resource management.The main results are as follows:(1)The cellular automata(CA)model algorithm and BP neural network model was used to classify the forest types.The overall accuracy from CA was 88.71%and the Kappa coefficient was 0.829.At the same time,the overall accuracy from BP was 86.67%and the Kappa coefficient was 0.798.Both the classification accuracy of two methods were high enough to achieve the purposes of forest type classification.By comparing the results,it was found that the accuracy from CA is slightly higher than that of BP.Therefore,the diagrams of forest type distribution from 3 separate times were obtained with cellular automata model algorithm.(2)The cumulative frequency of normalized difference vegetation index(NDVI)data was calculated and the values at 5%and 95%were chosen as the best threshold value for NDVIcrown and NDVInon-crown.Then the forest canopy density was estimated with dimidiate pixel model and the accuracy was 89.01%.The R2 was 0.859 and the RMSE was 0.039.The mean values of forest canopy density from 3 separate times,derived from the above model,were 0.691,0.668,and 0.457,separately.By leveling the change degree of forest canopy density,the change of forest canopy density was analyzed during the period from 1989 to 2010.Specifically,the general trend was in increasing phase,although forest resource between 1989 and 2000 was decrease.The forest resource during the period from 2000 to 2010 was still under considerable to reduce.Overall,during the period from 1989 to 2010,the destruction of forest coverage was greater than the increase of forest coverage in the study area.(3)Based on the method developed by Xing et al.,the maximum canopy height estimation models of coniferous forest,broad-leaved forest,and coniferous and broad-leaved mixed forest were established using Geoscience Laser Altimeter System(GLAS)data.The accuracy from the model of coniferous and broad-leaved mixed forest was the highest,with an R2 of 0.898 and an RMSE of 1.12 m.The R2 from the model of coniferous forest was 0.859 and the RMSE was 1.63 m.And the accuracy of broad-leaved forest was lowest,with an R2 of 0.761 and an RMSE of 2.01 m.(4)On the basis of getting GLAS maximum forest canopy height,combine with TM multi-spectral data and its converting generated vegetation index,forest canopy density,band reflectances,and considering the effect of terrain factors,using BP neural network algorithm,Constructed adapted to the regional extension of various types of forest canopy height of remote sensing inversion model.The accuracy of coniferous forest was the highest with an R2 of 0.907 and an RMSE of 1.52 m.Then the accuracy of mixed forest followed,with the R2 of 0.883 and the RMSE of 1.79 m.The accuracy of broad-leaved forest was the lowest,with the R2 of 0.849 and the RMSE of 1.96 m.Although there are individual points of overestimation or underestimation,the overall consistency is still good.By analyzing the data of forest canopy height,it was found that the increase of forest canopy height during the period from 1989 to 2000 was more than the decrease.The decrease of forest canopy height during the period from 2000 to 2010 was more than the increase.Overall,the increase of forest canopy height was more than the decrease during the period from 1989 to 2010.It was indicated that the forest height was in negative growth in the past 20 years.(5)The univariate and bivariate forest biomass estimation models of three forest types were established using linear regression,with the independent variables of forest canopy density and forest canopy height and the dependent variable of field measured forest biomass.The results of bivariate estimation model with forest canopy density and forest canopy height were better than the results of univariate estimation model.It was indicated that the fusion of forest spectral information and vertical structure information could improve the accuracy of forest biomass estimation model.Specifically,the accuracy of coniferous forest biomass estimation model was 91.66%and the R2 was 0.841.The accuracy of broad-leaved forest biomass estimation model was 85.75%and the R2 was 0.878.And the accuracy of mixed forest biomass estimation model was 85.47%and the R2 was 0.884.Then the bivariate forest biomass estimation models for three forest types were used to predict the forest biomass throughout the study area.By analyzing the forest biomass data acquired from separate times,it was found that the change of forest biomass during the period from 1989 to 2000 was in decrease trend.And the change of forest biomass during the period from 2000 to 2010 was in increase trend.In general,the change of forest biomass during the period from 1989 to 2010 was in decrease trend.(6)The changes of carbon stocked in forest vegetation were quantitatively analyzed based on the forest canopy density information.The results showed that the harvesting reduced the amount of forest carbon stocks during the period from 1989 to 2000.During the period from 2000 to 2010,the harvesting of over-mature forests reduced the forest carbon stocks.However,the forest tending and updating also compensated for the reduction in forest carbon stocks during the 10 years.In general,the effects of different harvesting disturbances on forest carbon stocks were negative during the 20 years from 1989 to 2010.At the same time,the disturbance evaluation system of forest carbon stocks was established with forest tree species composition,forest canopy closure and forest stand mean tree height.And the influence prediction models of different harvesting disturbances on forest carbon stocks were established by using forest carbon stocks and forest canopy closure data of GLAS footprint.It was found that the prediction accuracy of coniferous forest was 0.879,the prediction accuracy of broad-leaved forest was 0.836;and the prediction accuracy of mixed forest was 0.891.There was a clear linear relationship between forest carbon stocks and harvesting disturbance in each forest type and the change of forest carbon stocks decreased with the increase of change ratio of forest canopy density.The forest would begin to release carbon and become the carbon source when the change ratio of coniferous forest was bigger than 0.351,the change ratio of broad-leaved forest was bigger than 0.381 and the change ratio of mixed forest was bigger than 0.322.It was demonstrated that the anti-interference ability of broad-leaved forest was the best,and the coniferous forest followed and the mixed forest was the lowest.Therefore,the harvesting should be controlled within the above threshold value to increase the forest carbon sequestration capacity. |
PDF Full Text Request