Simulation Of NPP In Farm Lands And Sensitivity Analysis Of Parameters Based On Remote Sensing And Process-based Model

Net Primary Production(NPP)is an important variable for the terrestrial ecosystem carbon cycle,reflecting the productivity and carbon sequestration of terrestrial vegetation.With the development of remote sensing,the simulation of NPP based on remote sensing and process-based model has become a hot spot in the global study of carbon cycle in vegetation,and one of the commonly used models called the Boreal Ecosystem Productivity Simulator(BEPS)model.The BEPS model originated from the northern Canadian forest ecosystem and can be coupled with remote sensing data to simulate the regional carbon-water cycle process.It has been successfully used to simulate NPP in different types of global and regional ecosystems.The keys to NPP simulation accuracy are the model's localization and sensitivity of model parameters applied in different ecosystem.In order to get a better application of the BEPS model in farmland ecosystems,the paddy field ecosystem in Dongtai County,Jiangsu Province was.toke as the case study,import the measured data,retired data and the meteorological data obtained by automatic weather stations into the BEPS model to simulate the NPP of rice.And the accuracy of the BEPS model was evaluated by the actual observation data of the eddy flux system.At the same time,the Extended Fourier Amplitude Sensitivity Test(EFAST)method was used to carry out global sensitivity analysis of parameters in BEPS model.The research conclusions mainly include the following aspects:(1)With the research of rice measured data and retired data,we found that the clumping index(CI)has obvious differences in different growth stages,and it generally shows a trend of first decreasing and then increasing.The relative error of the LAI was effectively improved after correction by the CI,especially in the medium term,the improvement of the accuracy was most obvious,and the relative error was increased from 23%to 11%.In the LAI remote sensing inversion algorithm,the LAI retrieved model based on the random forest algorithm and the GF-1 WFV image has a good fitting effect.The determinative coefficient of the predicted and measured values is 0.88,the RMSE is 1.03.The mean error of the LAI measured value and retrieved value is 0.32.The relative error is 15%.The retrieval of LAI using remote sensing data can better reflect the changing trend of LAI in rice growing season.(2)By analyzing the sensitivity of parameters in the BEPS model using the EFAST method,we found that the global sensitivity of the 11 input parameters to NPP simulation is ranked as follows:CO2,LAI,Rad,Vcmax,soilwater,awe,TRAC,Tmin,Tmax,rh,Pre.Combined with the actual changes in the parameters of rice growth,the result showed that:rice's own mechanism parameters(LAI,Vcmax,TRAC)and solar radiation(Rad)have an important impact on its carbon accumulation.And,CO2,Rad,LAI,and Vcmax have a positive correlation effect on the simulation results,while TRAC shows a negative correlation effect.(3)By simulating the NPP value of research area using BEPS model,it can be seen that the curve of NPP values in the rice growing season showed a significant increase in the initial stage,a slight fluctuation in the medium term,and a slow decline in the later stage.The simulated NPP values and the measured NPP values showed high consistency in the growing season.The total NPP accumulation of the model simulation is 533gC/m2,and the value of suit measurement is 550gC/m2,the relative error is only 3.2%.However,NPP was underestimated at the initial stage due to the low LAI value,and then the NPP value was overestimated when the key period moved to reproductive growth.During the whole growing season,NPP showed a high correlation with LAI,especially when the LAI was low at the initial stage(the first 30 days),and the effect of total solar radiation on simulated NPP was lower in the initial stage and evident in the middle and later stages.