Estimation Of Grassland GPP In Northern China And Mongolia Based On Satellite SIF Data

Gross primary productivity（GPP）is a key parameter of ecosystem carbon cycle,which refers to the total amount of organic carbon fixed per unit area and per unit time during photosynthesis.Grassland is the largest terrestrial ecosystem.GPP estimation of grassland ecosystem is one of the important topics in global carbon cycle research.GPP acquisition on the ground is mainly based on vorticity tower flux observation,and combined with statistical model or ecological process model to extrapolate GPP distribution on regional scale.With the development of remote sensing technology,inversion methods based on different remote sensing information sources have become the main way of large-scale GPP estimation,among which the most widely used is the light energy utilization model.The efficiency and accuracy of the GPP estimation model are limited because the ecological process model and the solar energy utilization model need more parameters and the calculation is more complex.Solar induced chlorophyll II fluorescence（SIF）is a spectral signal（650-800 nm）emitted by plants in the process of photosynthesis under the condition of sunlight.Compared with vegetation index and other parameters,SIF can more directly reflect the relevant information of vegetation photosynthesis,which brings a new way for large-scale GPP estimation.However,the current satellite SIF data either have the shortcomings of coarse resolution or the limitations of spatial discontinuity.So it is difficult to apply to the estimation of large-scale continuous GPP.Oco-2 SIF data has high spatial resolution,but it is spatial discrete data.In view of the above problems,this paper focuses on the continuous prediction method of discrete oco-2 SIF data to generate the high-precision continuous SIF dataset of northern China and Mongolia grassland ecosystem;based on the surface flux observation data of different grassland types,the relationship model between GPP and SIF of different grassland types is established,and combined with the continuous SIF data set to realize the large-scale estimation and analysis of GPP.The main conclusions are as follows:（1）through Cubist regression tree algorithm,combined with MODIS reflectance data,meteorological data and land use types,a continuous SIF data set with 0.05°resolution every 8 days is established,and the prediction accuracy is R~2=0.66,RMSE=0.097.Among them,the prediction accuracy of SIF for crop is the highest,with R~2=0.71 and RMSE=0.116,followed by the prediction for forest and grassland,with R~2 and RMSE of 0.65/0.098 and 0.61/0.099,respectively.（2）Based on the SIF data and the measured GPP data of ground vorticity stations,the gpp-sif estimation models of different grassland types were constructed,and the accuracy of the evaluation model was verified by ten fold cross validation.Among them,the fitting accuracy of GPP of alpine grassland is the highest,R~2=0.9644,RMSE=1.17,followed by temperate typical grassland,temperate meadow grassland and desert grassland,R~2 and RMSE are 0.7003/1.39,0.6459/1.64 and 0.6141/1.56 respectively.（3）By analyzing the spatial distribution of GPP and SIF annual mean of different grassland types,it was found that although the SIF annual mean of alpine grassland was slightly lower than that of temperate meadow grassland in 2019,the GPP of alpine grassland was the highest,and the total yield in 2019 was between 2.5-3.5 kg·C·m^-2·y^-1,and its average GPP was twice that of temperate meadow grassland.The total yield of temperate meadow grassland and temperate typical grassland in 2019 is 0.9-1.3kg·C·m^-2·y^-1.The annual GPP of desert steppe was lower than 3 g·C·m^-2·d^-1,and the annual total yield was less than 1 kg·C·m^-2·y^-1.