Multimodal Simulation And Dynamic Analysis Of Forest Carbon Fluxes

The quantitative relationship between carbon dioxide（CO₂） emissions and sinks according to the global carbon project（GCP） in 2015 showed that the carbon stocks and fluxes are dynamic and hard to quantify among atmosphere, land ecosystem and ocean. Land ecosystem has acted as a substantial CO₂ sink and sequestered about 40% of CO₂ emissions. Of the terrestrial ecosystems, forest land is the main contributor to CO₂ emissions and removals.Therefore, quantifying forest carbon fluxes is important both to researches of carbon cycle,uncertainty analysis and the response of forest to climatic variation. Currently, observation and simulation were the main techniques for quantifying carbon fluxes. The Eddy Covariance（EC）technique is considered the standard tool, measuring carbon, water and energy fluxes between ecosystem and atmosphere directly. This technique is considered reliable based on the site scale.An alternative approach was developed for large scale simulating and forecasting carbon fluxes through models, which includes remote sensing-based model and process-based model. The two types of models adapt to different scales and study objects.However, the EC technique only provides carbon flux measurements over small footprint areas, and is expensive and spatially limited. Additionally, it is difficult to calibrate the complicated and non-linear process-based model, and the errors existed in the input data and model structure. So in this study, the incorporation of the remote sensing-based model and process-based model was conducted to calibrate Biome-BGC model and satisfy the requirement of temporally and spatially continuious products using multi-source data,including meteorological data, EC measurements, remote sensing data, field measurements.Considering the errors during the whole simulation, data assimilation was introduced in the estimations of carbon fluxes. The regional forest carbon fluxes were simulated using data assimilation based on regional GLASS LAI. Finally, the responses of forest NPP to climatic variation were analyzed. The specific conclusions are as follows:（1） The remote sensing-based model（MODIS MOD₁7 GPP（MOD₁7）） provides a simple yet accurate and quantitative measure of the spatial pattern of forest carbon fluxes on regional scales. The original MODIS GPP products agreed well with EC measurements at Guantan and Changbai Mountains forest flux site, but they were significantly underestimated.Therefore, the original MOD₁7 model was optimized using observed meteorological data,recalibrated maximum light use efficiency, and GLASS f PAR. GPP estimations from the optimized MOD₁7 model yielded high agreement with EC measurements（Gutan: R2 = 0.91,RMSE = 5.05 g C/m2/8d; Changbai Mountains: R2 = 0.93,RMSE = 8.74 g C/m2/8d）.（2） As a complicated and nonlinear process-based model, Biome-BGC contains multiple parameters and it is difficult to calibrate the parameters individually. So the sensitivity analysis was conducted using Extended Fourier Amplitude Sensitivity Test（EFAST） to acquire the first sensitivity index and total index of the parameters. Then key parameters were determined and calibrated through the integration of the optimized MOD₁7 and Biome-BGC model. When the GPPs obtained from Biome-BGC model and those from the optimized MOD₁7 reached the best agreement, the parameters was determined. The carbon fluxes obtained from the calibrated Biome-BGC model agreed well compared with EC measurements at the two forest sites（Guantan: GPP,R2 = 0.79,RMSE = 1.15 g C/m2/d,NEE, R2 = 0.69,RMSE = 1.09 g C/m2/d; Changbai Mountain: GPP, R2=0.87, RMSE=1.58 g C/m2/d, NEE, R2=0.62,RMSE=0.34 g C/m2/d）.（3） Data assimilation was designed to create the best analysis of the state of the system and find the optimal model performance, most consistent with observations. In this study,Ensemble Kalman Filter（En KF） was used to improve the performances of forest carbon fluxes at Guantan and Changbai Mountains forest site. The ensemble window is 8-day. After assimilation, the performances of GPP and NEE improved significantly（Guantan: GPP,R2 =0.84,RMSE = 1.01 g C/m2/d,NEE, R2 = 0.74,RMSE = 0.99 g C/m2/d; Changbai Mountains:GPP,R2=0.92, RMSE=1.26 g C/m2/d,NEE,R2=0.67,RMSE=3.04 g C/m2/d）.（4） The forest carbon fluxes were simulated over the Qilian Mountains and Greater Khingan using the calibrated and assimilated Biome-BGC model during 2000 and 2012. Thenthe trend of regional NPP was analyzed and 98% of Qilian Mountains experienced a downward trend, including 44.5% forest area with significant downward. The forest NPP over Greater Khingan showed upward trend during 2000 and 2012, with the significant upward area accounting for 44.5%. The partial correlation between the interannual average NPPs over Qilian Mountains and climatic variation showed that the annual average temperature,precipitation and shortwave downward radiation generally had significant positive impacts on NPP, but the vapor pressure deficit had a negative impact on NPP. The partial correlation between the interannual average NPPs over Greater Khingan and cliamitic variation resulted that annual precipitation and shortwave downward radiation generally had significant positive impacts on NPP, but anuual average temperature and the vapor pressure deficit had a negative impact on NPP.A series of issues occurred in past studies. It is difficult to extend forest carbon fluxes through temporal and spatial scale simultaneously, and calibrate the parameters with local environment. It is innovative in this study to integrate the optimized MOD₁7 and Biome-BGC model globally, realizing the temporally and spatially continuious simulations of carbon fluxes and dynamic analysis. The regional GLASS LAI products were assimilated into the calibrated Biome-BGC to correct the errors timely, integrating the advantages of the observations and simulations. The results in this study could provide technical support for sustainable management of forest resource and scientific reference for solving the ecological issues.