The Research On The Impact Of Data And Model Uncertainty On The Simulation Of Gross Primary Production Of Sample Plots And Global Scale

Accurate estimates of gross primary productivity(GPP)at regional and global scales are important for understanding the carbon cycle and its biochemical processes in terrestrial ecosystems.A large number of studies have revealed that there are many uncertainties in existing GPP models,among which the most critical is the uncertainty of input data and model.Therefore,it is necessary to conduct an in-depth study on its impact of the uncertainty of GPP simulation.First,this study compared 8 GPP models based on remote sensing data,and analyzed the influence of input data on GPP simulation accuracy.Further calibrate the input data,optimize the model parameters and structure,and develop two new GPP models based on the global flux tower data:LUE-EF model--driven by the EF(Evaporative Fraction);LUE-NDWI model--driven by NDWI(Normalized Difference Water Index).Finally,a new global GPP product dataset is generated based on the LUE-NDWI model.The main results of this study are as follows:(1)A comprehensive comparison of eight GPP models has found that there are huge differences between the models.All models R2 mainly focus on between 0.46 and 0.57 at each flux site.Through model comparison,it is found that cloud cover has a great impact on the model.The analysis of the model structure shows that different model structures affect the performance of the model(2)This study developed the new GPP model,which mainly considered the influence of cloud cover index(CI)and carbon dioxide(CO2)on GPP,and optimized the model structure and parameters.The new model shows a strong ability to reproduce GPP,LUE-EF and LUE-NDWI models have the best fitting performance on the site_year scale R2 can reach 0.85 and 0.75,respectively.The new models are evaluated at different temp-spatial scales and have good simulation performance.(3)Remote sensing products data is an indispensable data for calculating regional and global GPP.By comparing the measured data of flux tower,it is found that the remote sensing products data has deviation.Therefore,it is necessary to correct the remote sensing products data when conducting GPP simulation.By comparison on global scale,it was found that the global total GPP generated after data correction was about 125Pg Cyr-1,which was 18%less than the GPP before remote sensing products correction.(4)Considering the convenience and directivity data input in model optimization,we calculated the influence of water on GPP with normalized water index(NDWI)data,and developed the LUE-NDWI model.Therefore,this study used this model to generate new global GPP product from 2000 to 2015,and analyzed the overall growth trend of global GPP,with the growth rate is about 0.32Pg Cyr-1.GPP growth area accounted for about 76.8%of the global land area,and decrease accounted is about 10.2%(p<0.05).The main growth area occurred in the central South American region,and the main decrease area occurred in parts of South Africa,central Australia,and the northern Caspian sea.This study evaluated the uncertainty of model structure and parameters on the GPP model,developed a new model of GPP,reflects the change tendency of global GPP.At the same time,we emphasize the remote sensing product data error in the important role of GPP simulation.This study is helpful to improve the accurate simulation of the global GPP and reduce the uncertainty of GPP simulation,deepen the understanding of carbon balance between the biosphere and atmosphere.