Research On The Suitability Of GPP Models Based On MODIS And European Flux Data

Based on MODIS and European carbon flux measurements,GPP models,including Vegetation Photosynthesis Model(VPM),Eddy Covariance Light Use Efficiency model(EC-LUE),Temperature and Greenness(TG)model,Greenness and Radiation(GR)model,Vegetation Index(VI)model and MOD 17 are used to estimate the gross primary productivity(GPP)of eight flux towers across European region at first.Then,methods of the linear correlation and quantitative analysis are used to evaluate and compare the simulation accuracy and suitability of the above six models at different time scales(eg,eight day,seasons and annual).Finally,the sensitivity of model parameters are also discussed and the influences of model parameters including meteorological data,empirical parameters,vegetation index,the maximum light use efficiency(εmax)and the light conversion coefficient(m),land cover classification accuracy and spatial scale on the simulation results are further analyzed in this study.Our results show that:(1)Vegetation indexes,including NDVI,EVI and LSWI were obtained by using the surface reflectance data(MOD09A1)and the vegetation index data was further smoothed by Savitzky-Golay(S-G)method.The results showed that the vegetation index curves can better show the seasonal variation of the actual seasonal variations of vegetation indexes,which means that the Savitzky-Golay(S-G)method can improve the quality of remote sensing data and realize the goal of cloud reconstruction of the vegetation index time series data.(2)Analysis of model performance at different time scales(8 days,seasons and annual)showed that:the TG model performed best at 8 days and annual scales,and GR and EC-LUE followed,while the estimation accuracy of VPM,VI and MOD17A2 models were not as good as the formers.Models have exhibited the highest estimation precision for summer GPP,spring followed,while,when compared with spring,the precision of autumn is slightly lower and the winter is the lowest at seasonal scale.Analysis of model suitability at different biome types showed that:for specific vegetation types,the suitability of GPP model is quite different,for example,GR is suitable for GPP estimation of vegetation types such as mixed forest(MF),evergreen needleleaf forest(ENF),deciduous broadleaf forest(DBF)and grassland(GRA).At the same time,TG is also suitable for mixed forest(MF)and grassland(GRA).EC-LUE and MOD17A2 are more suitable for evergreen broadleaf forest(EBF).For cropland(CRO),EC-LUE and VPM models are more applicable.(3)Analysis of factors that may impact model accuracy indicated that:1)meteorological data,including PAR,temperature and water pressure will affect the the performance of GPP models.For cropland(CRO),water pressure is the main restrictive factor,while the other natural vegetation types are more sensitive to temperature;2)as for photosynthetic temperatures,the Topt has the greatest impact on GPP estimation,Tmin followed,while,Tmax has minimal impact on GPP estimation;3)compared with the original vegetation index(EVI/NDVI)curves,the S-G reconstructed vegetation index data as model input can significantly improve the model estimation accuracy;4)the maximum light use efficiency(εmax)and the light conversion coefficient(m)obtained by 70%of the measured GPP can also significantly improve the accuracy of GPP models;5)using the mean value of 9 pixels within 3×3 matrix as model input can improve the accuracy of GPP models.