Assimilation Of Multiresolution LAI In Bamboo Forest And Its Application In Carbon Cycle Simulation Based On Multi-source Remote Sensing Data

The leaf area index（LAI）is one of the important physiological parameters of terrestrial vegetation.The spatiotemporal LAI data is always considered as one of the parameters and indicators for simulating carbon cycle,and it is therefore significant for quantitative analysis of carbon cycle process and evolution of vegetation ecosystem.However,the carbon cycle and its dynamic change of forest ecosystem depend on the interaction between the short-term responses of individual organisms and long-term changes of regional ecosystem at different scales.To a certain extent,the lack of multiresolution LAI data affects the multiresolution simulation for spatiotemporal evolution and comprehensive understanding for control mechanisms of carbon cycle of forest ecosystem.Therefore,highly accurate multiscale LAI data plays a big part in quantifying and clarifying the evolution of carbon cycle at different scales in regional and global forest ecosystems.As an important part of forest ecosystem,bamboo forest has strong carbon sequestration capacity and is of great significance for maintaining regional ecological environment and global carbon balance.For these reasons,this study used MODIS LAI data（MOD15A2）and MODIS reflectance data（MOD09Q1）of Moso bamboo forest（MBF）in Shanchuan Town（Anji County,Zhejiang Province）and Lei bamboo forest（LBF）in Taihuyuan Town（Lin’an,Hangzhou,Zhejiang Province）,and constructed the LAI multiresolution assimilation system,which adopted an Hierarchical Bayesian Network algorithm（HBN）coupled with a LAI dynamic model and the PROSAIL model,to assimilate highly accurate LAI time series data with multiresolution（1000,500,and 250 m）at MBF and LBF flux measurement sites.And then,taking the MBF in Shanchuan Town as an example,this study used the assimilation system to obtain and research the multiresolution spatiotemporal LAI data of the MBF.Finally,assimilated multiresolution LAI data was used as the input data to drive the BEPS（Boreal Ecosystem Productivity Simulator）model and research the multiresolution simulation of NEE（Net ecosystem exchange）with time series and spatiotemporal distribution.Several findings from this study are as follows:1.The highly accurate multiresolution LAI time series data for MBF and LBF was obtained at resolution of 1000,500 and 250 m by using the LAI multiresolution assimilation system containing HBN algorithm,LAI dynamic model and the PROSAIL model.（1）The assimilated multiresolution LAI time series data corresponded with the actual growth trend of the MBF and LBF.It showed a trend of slow growth in spring before reaching its annual maximum value in summer,and decrease gradually in autumn with the annual minimum value occurring in winter.（2）The determination coefficient R² between assimilated LAI and observed LAI at 1000 m resolution for MBF and LBF were 0.91 and 0.96,respectively,the RMSE were 0.27 and 0.87,respectively,and the aBIAS were 0.22 and 0.82,respectively.At 500 m resolution,the average R² were 0.90 and 0.92,respectively,the average RMSE were 0.42 and1.11,respectively,and the average aBIAS were 0.34 and 1.05 respectively.At the 250 m resolution,the average R² were 0.86 and 0.89,respectively,the average RMSE were 0.35 and 1.05,respectively,and the average aBIAS were 0.29 and 0.94,respectively.（3）Compared with MODIS LAI,the R² between assimilated LAI at 1000 m resolution and observed LAI increased by 2.7 and 8.6 times,respectively,and the RMSE reduced by 87.8%and 59.7%,respectively.2.Based on the LAI multiresolution assimilation system,the results of multiresolution spatiotemporal LAI data for MBF indicated that:（1）The spatiotemporal LAI assimilation results at the resolution of 1000,500 and 250 m were also consistent with the trend of actual variation,that was,annual maximum and minimum values occurred in summer and winter,respectively.（2）The spatiotemporal distribution of LAI data at three resolutions was consistent with the abundance of MBF,and the multiresolution LAI data showed the trend with higher value in the southwest and lower value in the northeast.3.The multiresolution simulation of time-series and spatiotemporal NEE data for MBF in Shanchuan Town was obtained by driving BEPS model using the assimilated multiresolution LAI data.For the simulation of NEE time series data,（1）the simulated multiscale NEE data was in line with the actual change trend,showing a trend of slow growth in spring before reaching its annual maximum value in summer,and decrease gradually in autumn with the annual minimum value occurring in winter.（2）The simulated NEE data at resolution of 1000,500 and 250 m had high precision and correlated significantly with the observed NEE.The determination coefficient R² between simulated NEE and observed NEE reached 0.64,0.64,0.66,respectively,and RMSE and aBIAS were 1.05,1.03,1.05 g C m^-2 day^-1 and 0.91.0.83,0.86 g C m^-2 day^-1,respectively.（3）Compared with simulated NEE by MODIS LAI product,simulated NEE by assimilated multiresolution LAI effectively improved accuracy that the R² increased by 60%,60%,65%,respectively,RMSE and aBIAS decreased by 39%,40%,39%,and 33%,39%,37%,respectively.（4）The simulated NEE at different resolutions had a strong correlation with the R² of 0.88,0.98 and 0.86,respectively,which had laid an important foundation for revealing the interaction and long-term response mechanism of carbon cycle in bamboo forest ecosystem at different scales.For the simulation of spatiotemporal NEE data,（1）at the resolutions of 1000,500 and 250 m,the seasonal variation of simulated NEE data was also in line with the actual change trend.（2）The multiresolution NEE for MBF in Shanchuan Town showed a trend with higher value in the southwest and lower value in the northeast.（3）Multiresolution spatiotemporal NEE data reflected the regular pattern of spatiotemporal evolution with different resolutions,and laid an important foundation for studying the regular patterns of multiscale spatiotemporal evolution of carbon cycle of MBF.