| Terrestrial ecosystems which fix about 30%artificial CO2 emissions,which are one of the important ways to mitigate global warming.Since the 1980s,the global terrestrial ecosystem has gradually increased from a weak carbon sink close to equilibrium to a larger carbon sink.However,how to find and evaluate the“lost”terrestrial carbon sink and its pattern is still an important scientific issue.In the past 30 years,based on inventory method and model simulation,scientists have done a lot of research on the size of carbon sinks in terrestrial ecosystems,but the spatial and temporal patterns and mechanisms of carbon sinks are still uncertain.Therefore,the development of terrestrial ecosystem carbon cycle model to assess the status,potential and mechanism of ecosystem carbon sequestration from different spatial scales is not only the core issue of ecosystem carbon cycle process research,but also the basis for formulating policies to control greenhouse gases,mitigate and adapt to global warming.It is urgent for ecologists to make breakthroughs in research methods and mechanism.This study takes the uncertainty of terrestrial ecosystem carbon-nitrogen coupling model as the breakthrough point.Base on a matrix approach representation framework,we conducted the feasibility of matrix approach,coupling nitrogen module,the spatial effects of different representations of N processes,combining with the methods of data assimilation,traceability analysis and semi-analytical solution spin-up?SASU?,to explore the source of uncertainty in the simulation and prediction of carbon storage in terrestrial ecosystem carbon-nitrogen models.The main findings were as follows:?1?The matrix approach reorganizes the C and N balance equations in the original model into two matrix equations whitout change any C and N processes and mechanisms,respectively.We applied the matrix approach to the Community Land Model?CLM5?with the biogeochemical carbon-nitrogen coupling cycle.The matrix equations exactly reproduce both C and N dynamics of the vegetation,litter and soil organic material?SOM?in CLM5.We proved that the matrix approach can be introduced in the complex terrestrial ecosystem models to improve their diagnostic ability.?2?In order to investigate the model uncertainties caused by new processes and/or data sets remain largely unclear,we explore uncertainties resulting from additional nitrogen?N?data and processes in a terrestrial ecosystem?TECO?model framework using a data assimilation system.Three assimilation experiments were conducted with TECO-C-C?carbon?C?-only model?,TECO-CN-C?TECO-CN coupled model with only C measurements as assimilating data?,and TECO-CN-CN?TECO-CN model with both C and N measurements?.Our results showed that additional N data had greater effects on ecosystem C storage?+68%and+55%?compared with added N processes?+32%and-45%?at the end of the experimental period?2009?and the long-term prediction?2100?,respectively.The uncertainties mainly resulted from woody biomass?relative information contributions are+50.4%and+36.6%?and slow soil organic matter pool?+30.6%and-37.7%?at the end of the experimental period and the long-term prediction,respectively.During the experimental period,the additional N processes affected C dynamics mainly through process-induced disequilibrium in the initial value of C pools.However,in the long-term prediction period,the N data and processes jointly influenced the simulated C dynamics by adjusting the posterior probability density functions of key parameters.These results suggest that additional measurements of slow processes are pivotal to improving model predictions.Quantifying the uncertainty of the additional N data and processes can help us explore the terrestrial C-N coupling in ecosystem models and highlight critical observational needs for future studies.?3?In order to clarify how the diverse representations of C–N interactions affect C storage dynamics in terrestrial ecosystem models,we explored how the simulated ecosystem C storage capacity in the terrestrial ecosystem?TECO?model varied with three different commonly used schemes of C–N coupling.The three schemes?SM1,SM2,and SM3?have been used in three different coupled C–N models?i.e.,TECO-CN,CLM 4.5,and O-CN,respectively?.They differ mainly in the stoichiometry of C and N in vegetation and soils,plant N uptake strategies,downregulation of photosynthesis,and the pathways of N import.We incorporated the three C–N coupling schemes into the C-only version of the TECO model and evaluated their impacts on the C cycle with a traceability framework.Our results showed that all three of the C–N schemes caused significant reductions in steady-state C storage capacity compared with the C-only version with magnitudes of-23%,-30%,and-54%for SM1,SM2,and SM3,respectively.This reduced C storage capacity was mainly derived from the combined effects of decreases in net primary productivity?NPP;-29%,-15%,and-45%?and changes in mean C residence time?MRT;9%,-17%,and-17%?for SM1,SM2,and SM3,respectively.The differences in NPP are mainly attributed to the different assumptions on plant N uptake?PNU?,plant tissue C:N ratio?PS?,downregulation of photosynthesis?DRP?,and biological N fixation?BNF?.In comparison,the alternative representations of the plant vs.microbe competition strategy?PMC?and PNU,combined with the flexible C:N ratio in vegetation and soils,led to a notable spread in MRT.These results highlight the fact that the diverse assumptions on N processes represented by different C–N coupled models could cause additional uncertainty for land surface models.Understanding their difference can help us improve the capability of models to predict future biogeochemical cycles of terrestrial ecosystems.?4?One of the key bottlenecks that limit land model development and evaluation is computational cost of spin-up to reach steady state of modeled biogeochemical cycles.The matrix equations can be used to achieve semi-analytical spin-up?SASU?periodically to calculate the steady-state pool sizes of carbon and nitrogen.We applied SASU at global scale,a tropic site in Brazil,and a permafrost site in Alaska.Compared with traditional native dynamics?ND?method,SASU saved 99%and 95%of computational time roughly for site-level and global spin-up,respectively.Compared with accelerating the decomposition of soil organic matter?AD?method,SASU saved similar computational time by 69%and 6.7%at both the global and site level,respectively.However,the AD method overshot steady-state total ecosystem C sizes which were estimated from SASU methods,especially in permafrost regions.For the global scale,the estimated steady-state total ecosystem C size was 6047.6 Pg C with the AD method,3.7%higher than that with the SASU method.For the Alaska site,the estimated steady-state total ecosystem C size in AD was 156.1 kg C m-2,which was 33%higher than SASU.Overall,the SASU method improved the computational efficiency and estimation accuracy of steady-state pool sizes,especially for the permafrost regions.?5?In order to study the effects of carbon-nitrogen coupling process on carbon storage at the global scale,we conducted the sensitivity test of three main carbon-nitrogen coupling processes on carbon storage based on the SASU method in CLM5.The results showed that the DRP process had a positive effect on carbon storage above ground,underground and GPP,and its effect on total carbon storage of ecosystem was31.3%.The relative contribution of DRP process to NPP and MRT is 52%and 22.5%,respectively.The PMC process mainly affected underground carbon storage with a negative effect.The impact on total carbon storage of ecosystem is 67.2%,which can be contributed to both the relative contribution to NPP is 24.8%,and the relative contribution to MRT is 77%.The impact of PNU on carbon storage in ecosystems was less than 0.1%,mainly because the global values of PNU impact factors calculated by CLM5 model are more uniform and close to 1.Our research highlighted the importance of matrix representation of carbon and nitrogen dynamics in terrestrial ecosystem C-N coupling models,identified the sources of uncertainty in C storage simulation and prediction by coupling N cycles from ecosystem to global scale,and revealed the influences of the model structure and observation data,as well as the representations of nitrogen processes.Our research also clarified the effects of DRP,biometric relationship,and plant and microbe competition strategies in the model simulation and prediction.It is valuable to point out key research directions for more accurate assessment of terrestrial carbon and nitrogen dynamics. |
PDF Full Text Request