| Terrestrial ecosystems are vital for human survival and sustainable development. Not only the structure and function but also the patterns and processes of the ecosystem are closely related to the change of atmosphere and soil environment. Global climate change and intensive human activities have all-round, multi-level impact on terrestrial ecosystem’s pattern and structure, the ecosystem’s supply services, regulating services, cultural services, and support services in all aspects. Accurate estimates of the carbon and water fluxes on the regional scale are of great significance for understanding the feedback mechanism of terrestrial ecosystems and the atmosphere in the context of global change and climate decision. Large-scale, long-term and continuous carbon, water, energy flux observation datas between biosphere and atmosphere are specially needed in the key processes study of global carbon and water cycle.Eddy covariance method by calculating turbulent fluxes with the physical quantities pulsation and wind speed pulsation is the only way for directly measuring CO2exchange between the communities and the upper atmosphere. It’s the universal and standard method for determinating the carbon, water and energy fluxes. Upscaling the flux site observation data provides a new way for accurately estimating the carbon, water and energy fluxes in regional, continental and global scale. However, the premise of upscaling the flux site data is that the flux measurement network has a higher representation for the ecosystem of the region, in order to guarantee flux observation network detecting the space-time variability of ecosystem carbon and water fluxes in the monitoring area.There are more than500flux sites with the eddy covariance technique all over the world. They belong to different regional flux observation network and locate in different types of climate and ecosystem. Most flux sites were built only based on the scientific issues, address feasibility as well as previous experience. They lack of a scientific basis for the sites’distribution overall the region, and flux observation network itself has not been scientifically designed. Fluxsite observation data can only represent the flux footprint values, and the carbon and water fluxes have relatively large spatial and temporal variability. All these characteristics caused that there is greater uncertainty by using flux observation network to estimate carbon and water fluxes on a regional scale. So the representativeness analysis of flux observation network over the region would determine the accuracy of upscaling the flux observation stations to the the regional-scale. China has vast region, complex ecosystems, from cold temperate to tropical climate zones, and special plant geographical area. All these characteristics provide a good experimental platform for the research of global carbon cycle. So the systematical representativeness analysis of flux observation sites over China would provide a theoretical basis for the flux observation data upscale and it is very important for accurate estimation of China’s regional terrestrial ecosystem carbon and water fluxes.The paper’s mainly research ideas:we selected the key environmental variables affecting carbon flux, compared and analyzed the site and regional’s environmental elements data. By analyzing resprentativeness of the flux site data relative to the regional data, we indirectly analyzed flux observation networks’s representation of regional carban flux. Therefore, we collected data of91normally operating sites in land areas which utilized eddy covariance technique for flux observation. Considering data accuracy, variable representation and data accessibility, we selected the total radiation, average temperature, average vapor pressure, vegetation index(EVI) and vegetation type as the key environment environmental variables that affect the net amount of carton exchange between vegetation and the atmosphere. With the support of high-performance computing cluster, MATLAB, parallel algorithms we analyzed and discussed the regional representation status quo of the existing flux observation from space representation, time representation and site contribution rate. And on the basis of the results of qualitative analysis of the spatial representativeness and K-means clustering analysis algorithm, we predicted the specific location of flux sites that should be added to optimize existing flux observation network and compared with space representation of the original flux observations, then obtained the following conclusions:(1) In the space representation analysis of the flux observation, over the entire area there are60.9%of the pixel values below the national average. Highly representative pixels mainly concentrate in the north of Qinling and east of the Qinghai-Tibet Plateau, including temperate grassland, warm temperate deciduous broadleaf forest, cold temperate coniferous forest, temperate coniferous and broad-leaved mixed forest, and the Tarim Basin in the temperate desert. Poorly representative pixels mainly concentrate in the southern subtropical evergreen broadleaf forest, tropical monsoon rainy forest, temperate desert of Xinjiang and alpine vegetation region. So there is necessary for four specific vegetation area to add the flux sites to improve the regional representation of flux measurement;(2) In the time representation analysis of the flux observation, mean significant test results show that the total radiation and the average temperature have relatively higher level of representation, representative of meadows, forests and farmland ecosystem is above wetlands and bare land ecosystems; variance significant test results show that trends of flux site data variance and regional variance trends are in good agreement; Interannual sequence comparison analysis shows that whatever the environment elements are, the annual variations of grass, farmland, bare land and wetland ecosystem between flux site data and regional data are consistent, but there are still large differences between the absolute variance between the site data and regional data; Combined with results and analysis there are several suggestions such as increasing the fluxsite number in bare land and wetland in order to increase the flux measurement intensity, increasing the grass fluxsite in order to enhance the representatives of the grassland ecosystem’s regional diversity. Overall minimizing the gap between various ecosystems flux site data and regional data in discrete condition and annual variations to achieve higher time representation.(3)In the analysis of the status quo on the site contribution rates, the total contribution rate of the grass flux sites is up to32.12%, followed by the forest fluxsite contribution rate of31.21%; For the forest fluxsites, flux measurement and effectiveness in subtropical evergreen broad-leaved forest was more obvious. For non-forest fluxsites, sites with higher contribution rate has a high level of consistency with the typical vegetation in the area. Grass flux observation sites have a higher contribution rate in the Qinghai-Tibet Plateau and temperate grassland. Farmland flux measurement sites have a higher contribution rate in the subtropical evergreen broadleaf forest and warm temperate deciduous broad-leaved forest. Wetlands flux observation sites in the Qinghai-Tibet Plateau and subtropical evergreen broadleaf forest have a higher contribution rate. Desert flux observation site in temperate desert area has the highest contribution rate.(4) Combined with qualitative analysis of the conclusions and recommendations of the status quo of regional representation about adding flux sites, using K-means clustering analysis method to quantitatively determine the specific location to add flux sites. Then by comparing the representation of predicted flux observation network and the previous representation, there are following conclusions:there is a good consistency between the predicted spatial distribution of flux sites to add by quantitative algorithm and the original qualitative analysis; spatial representation of the predicted flux observation network have a larger increase in the average condition and degree of dispersion. Part of the space representation to improve accounts for27.5%of the total area, particularly in the area of alpine vegetation. The newly added flux sites have a higher site contribution rate. |
PDF Full Text Request