How Many Uncertainties Transferred From The Meteorological Data To The Terrestrial Carbon Cycle Modeling Based On CEVSA Model?

The combination of the data and the model is effective in improving thesimulating effect of large scale model, which laid the foundation for more accurateassessment of land ecosystem productivity. However, it is inevitable that uncertaintiesof data sets are affected by the limits of observation technologies and processingmethods in the integration process. And, the uncertainty could significantly affect theresults of the model. In this study, to gain the four sets of meteorological data of thesame number of grid points (96889 grid points), we use the meteorologicalinterpolation software ANUSPLIN4.36 which is more acknowledged by theinternational community to interpolate the meteorological data from PrincetonUniversity, NCAR/NCEP and the day meteorological data from 2374 meteorologicalstations and 659 meteorological stations provided by China MeteorologicalAdministration. To discuss how the uncertainty of temperature and precipitation dataaffects the result of model simulation, we use CEVSA model from the angle ofuncertainty of meteorological data in order to deepen our objective understanding ofthis influence, clear the range of the uncertainty of China’s terrestrial ecosystemscarbon balance and lay the foundation for further enhance the simulation precision ofthe terrestrial ecosystems carbon cycle. Through the study, we get the followingpreliminary conclusion:(1) From 1971 to 2005, the difference among meteorological data is obvious.The trends of four sets of data change are consistent and increasing in the aspect ofannual variability of temperature. The estimated annual temperature from the data ofPrinceton University is the highest among the four data sets and the data ofNCAR/NCEP is in the next place. While the temperature from the data of 2374meteorological stations and 659 meteorological stations provided by China Meteorological Administration are close to the average annual temperature and lowerthan the former two. In terms of interannual variability of precipitation, the trends offour sets of data change are consistent, while annual change has large difference. Theannual precipitation data estimated by the data from Princeton University,NCAR/NCEP and 659 meteorological stations are close to each other. In the aspect ofspace, the difference of temperature and precipitation mainly concentrated in thesoutheast coast, east of China northeast, eastern Yunnan-Guizhou plateau, the Tibetanplateau and the northwest inland region. In data uncertainty, the data quality of eastcoast region is obvious better than that of west inland region. The data uncertainty of659 meteorological stations is less than that of Princeton University and NCAR/NCEP.The RMSEs of precipitation are 96.98, 179.75 and 195.51, while the RMSEs oftemperature are 1.86, 1.92 and 2.17 respectively.(2) From 1971 to 2005, the annual NPP amount changing trends of China inlandestimated from the four data sets are almost consistent, all of them are increasingalthough not obvious. In space, the estimated NPP data from 659 meteorologicalstations, NCAR/NCEP and Princeton University are consistent basically, though thesethree data sets are all overestimated the NPP in some region, especially in northwestand Tibetan region, east of northeast, northern north China, Sichuan basin, theYangtze river and the southeast coastal area. In data uncertainty, the estimateduncertainty of NPP from 659 meteorological stations is less than that of PrincetonUniversity and NCAR/NCEP. The RMSEs of NPP are 46.93, 99.14 and 99.76respectively.The difference between temperature and precipitation will affect the result ofmodel simulation indirectly. It is necessary to improve the precision of NPP in Chinamainland by lowering the error of the meteorological data from Princeton Universityand NCAR/NCEP.