Spatial Pattern Of Net CO₂ Exchange In Terrestrial Ecosystems And Optimization Of Its Model Parameters

Net ecosystem CO₂ exchange(NEE)of terrestrial ecosystems,resulting from a delicate balance between GPP and RE,represents the capacity to absorb atmospheric CO₂.Under global climate change caused by elevated atmospheric CO₂ concentration,it is critical to shed light on the magnitude,spatial pattern,and temporal changes of NEE in terrestrial ecosystems.However,the spatial variability in NEE and its influencing factors are largely unknown,resulting in large uncertainty in model simulation and prediction.Currently,most of the terrestrial carbon cycle models are based on plant functional types(PFTs)to simulate and predict ecosystem carbon sink,which largely masks the difference of physiological and ecological parameters among diverse ecosystems,even within the same plant functional type.The PFT-based modeling method is questionable to accurately simulate terrestrial carbon cycle and to understand the underlying mechanism of spatial variability in NEE.In order to probe the mechanism of spatial variability in NEE of terrestrial ecosystems and to simulate and predict it more accurately,this study focused on the combined influence of climatic variables,soil properties and vegetation characters(plant functional traits and biome types)on NEE.First,we explored the spatial variability in NEE and the relative importance of climatic variables,soil properties and plant traits based on the analysis of long-term flux sites(?4 years)at the worldwide.Second,the spatial variability of carbon cycle parameters was estimated and analyzed based on the dataset of NEE,global climatic variables,soil properties and plant traits.Third,the terrestrial carbon sink was estimated based on the optimized parameters.Finally,we predicted the terrestrial carbon sink and its spatial pattern under different global warming scenarios(1-8?).The main results were as follows:(1)Based on the global flux observation dataset of 147 long-term sites(?4 years),the pattern of spatial variability in NEE was analyzed by the generalized additive model and principal component regression.The results showed that the spatial variability in NEE was mainly caused by the different effects of climatic variables,soil properties and plant traits on gross primary productivity(GPP)and ecosystem respiration(RE).Specially,NEE in forests was mainly influenced by climatic variables,such as mean annual precipitation and mean annual potential evapotranspiration,explaining 23.8%of the spatial variability in NEE.The spatial variability in NEE in grasslands was mainly affected by soil properties(41.4%),such as available water capacity and soil clay content.These factors,which are important for spatial variability in NEE but have little effect on GPP and RE,are not well represented in current terrestrial biosphere models,which need to be incorporated to more accurately predict the spatial pattern of carbon cycling across forests and grasslands globally.(2)Based on NEE of half-hour and one hour time scale(?4 years)from global flux sites,four key parameters of each site were obtained by inversing carbon cycle process with flux based ecosystem model(FBEM)using data assimilation and conditional inversion.Four key parameters were the maximum rate of carboxylation at 25?(V_m²⁵),the ratio of the maximum rate of electron transport and the maximum rate of carboxylation(r_JmVm),basal ecosystem respiration at 0?(R_eco⁰),and temperature sensitivity of ecosystem respiration(Q₁₀).Then,Automatic machine learning(AutoML)was applied to predict spatial distribution of four key parameters(R²=0.86-0.99).Our results showed that the four parameters had different pattern of spatial distribution,and the spatial pattern of the key parameters of photosynthesis process was significantly different between forests and grasslands.r_JmVm in forests was higher in tropics,which was mainly controlled by light,leaf area index and plant height.However,r_JmVm)in grasslands was higher in high latitudes,which was mainly related to temperature.V_m²⁵ in forests increased with latitude,which was mainly caused by leaf functional traits,while that in grasslands showed contrast pattern dominated by soil pH.The key parameters of ecosystem respiration were regulated by climatic variables and plant traits.R_eco⁰ in forests and grasslands was higher in low latitudes,while Q₁₀ was higher in high latitudes.This study used climatic variables,soil properties and plant traits together to explore the spatial variability in carbon cycle parameters and their drivers,and to extrapolate the site-based parameters into the global terrestrial ecosystems,which strengthen model development for more accurate simulation and prediction of terrestrial carbon sink.(3)Using the global spatial heterogeneous maps of the four optimized parameters,annual global terrestrial carbon sink(2000-2014)estimated by FBEM was 5.35 PgC,which was higher than the estimation of plant functional type-based models(8 MsTMIP).Compared with the results from traditional models,our spatial distribution of NEE was closer to that of observation data from flux sites,especially in tropical regions where great uncertainty among models remains.The FBEM on the base of these optimized key parameters better captured the spatial variation of RE in evergreen broadleaf forests,reflecting the spatial heterogeneity in NEE across tropics.These results from the spatial heterogeneity of terrestrial eco-physiological processes indicate that our method could improve the performance of models to simulate and predict the spatial variability in NEE.(4)Based on the above optimized parameters,FEBM predicted the capacity of terrestrial carbon sink under global warming scenarios(1-8?).The results showed that global warming had reduced the capacity of terrestrial ecosystems to uptake CO₂,especially in high latitudes.In most terrestrial ecosystems(82.7% of land area),the negative effects of global warming on terrestrial carbon sink can be mitigated by incorporating thermal assimilation of photosynthesis and respiration into models.However,as global temperature continues to rise,69% of terrestrial ecosystems would gradually reduce their thermal acclimation,and thus their ability to maintain the capacity of carbon sink.In the tropics where evergreen broadleaved forests are widely distributed,the effect of thermal acclimation on maintaining the capacity of carbon sink decreased more obviously than other areas,while it increased in high latitudes.Therefore,incorporating the thermal acclimation into photosynthesis and respiration process might improve the overestimated impact of global warming on terrestrial carbon sink,and reduce uncertainty of model prediction.In summary,the mechanism underlying the influence of climatic variables,soil properties and plant traits on spatial variability in NEE was different between forests and grasslands.The difference of spatial pattern in NEE might result from spatial difference of model parameters,especially for the divergent spatial pattern of key photosynthetic parameters.Compared with traditional plant functional type-based models,our model employing spatial heterogeneous parameters greatly reduced the uncertainty of carbon sink estimation in tropical regions,and better reflected the complex spatial pattern of terrestrial carbon sink.The prediction of terrestrial carbon sink under global warming in the future using the optimized parameters is largely influenced by the thermal acclimation of photosynthesis and respiration,and the thermal acclimation could improve the overestimation of global warming effect.In this study,we disentangled the main driver of spatial variability in NEE and its key eco-physiological process parameters,assessed the spatial distribution of terrestrial NEE at the present and in the future,and highlighted the importance of thermal acclimation of eco-physiological processes to mitigate warming influences.Our study provided important scientific basis for more accurately simulating and predicting terrestrial carbon balance under global change.