Model Optimization And Spatiotemporal Pattern Analysis Of Terrestrial Ecosystem Respiration

Terrestrial ecosystem respiration(ER)is a major flux in the global carbon cycle.Even small changes in ER will lead to great fluctuations in atmospheric CO₂concentration,and then global climate change.Therefore,accurate estimate of ER is crucial to understanding the feedback between global carbon cycle and climate change.However,current estimation methods of ER ignored the diurnal difference of key physiological parameters and the effects of water on respiration,causing a great uncertainty in the spatiotemporal pattern and the influencing factors.In this study,we aimed to explore the spatial and temporal trends of global ER and its drivers over the recent three decades.Specifically,we first integrated the widely-used CO₂ flux partitioning approach and observations of eddy-covariance flux and stable isotope to develop a reliable method for estimating ER at the ecosystem scale.The new method was applied to global eddy-covariance flux network(FLUXNET)to quantify temperature sensitivity of ER(E₀)and basal respiration rate at reference temperature(R_ref),and explore their spatial-temporal variations.Meanwhile,a robust data-driven machine learning model was developed to generate a global ER product,and investigate the spatial distribution pattern and long-term change trend of global ER and its driving factors over the recent three decades.The main findings were as follows:(1)Model optimization of ecosystem respiration.Based on the isotopic data of three flux sites with different ecosystem types(US-Ha1,CN-Qia and US-Tw3),we developed a reliable ER estimation method(DT-RH)by integrating the diurnal differences in E₀ and effects of relative humidity(RH)on ER.The results showed that the diurnal and seasonal variations of ER derived from DT-RH agreed with isotopic data well compared with previous widely used methods(e.g.,nighttime and daytime methods).The daily mean ER would be overestimated by 6%?18%,if the diurnal difference in E₀ was ignored.Considering effects of relative humidity additionally improved the accuracy of estimated ER,especially in water-limited seasons or regions.Overall,these results suggest that an integrated consideration of the diurnal differences in E₀ and effects of relative humidity on respiration is essential for improving ER estimation.(2)Spatiotemporal variation in key physiological parameters of ER.We used DT-RH method to estimate temperature sensitivity of ER(E₀)and basal respiration rate at reference temperature(R_ref)during daytime and nighttime at 196 long-term flux observation sites(including nine ecosystem types).The results showed that the diurnal difference of temperature sensitivity(?E₀)was prevalent,with pronounced seasonalities.Specifically,E₀ was smaller during the daytime than that at night.This diurnal difference was smaller in the growing than non-growing seasons.However,the E₀ was slightly lower at night than during the daytime at the growing season of forest ecosystems.The?E₀ was mainly affected by nighttime temperature(R~2=0.35,p<0.001).Meanwhile,the R_ref also differed during the day and night among different ecosystem types,and water stress significantly reduced daytime R_ref by 23.29%on average.The spatial distribution of effects of water stress on daytime R_ref was mainly affected by annual mean temperature,while there was no significant seasonal pattern.These results highlight the prevalence of diurnal difference in key physiological parameters of ER and their spatiotemporal differences in response to temperature and water changes,which have important scientific implications for accurately assessing the spatiotemporal pattern of ER.(3)Spatiotemporal pattern of terrestrial ER and its drivers.To explore the long-term change trends of global ER and its response to climate change,we used DT-RH and mchine learning methods with climate raster data and remote sensing data to generate monthly-scale global ER dataset during 1989-2018.The results showed that the global annual mean ER was 122.77±0.73 Pg C yr^-1,with a significant decreasing trend over the recent three decades(-0.04±0.01 Pg C yr^-2,p=0.016).Most declines of ER occured in tropical and temperate regions.Globally,the changes of soil moisture and temperature have a significantly greater effect on the long-term change in ER than the elevated atmospheric CO₂ concentration,with a dominant role for soil moisture changes.Specifically,reduced soil moisture primarily decreased ER in majority of mid-low latitude regions,whereas rising mean annual temperature primarily increased ER in high latitude regions.Interesting,the declined trend of global ER was opposite to those derived from FLUXCOM and CMIP6 models,showed an increasing trend over the recent three decades due to the failure of adequately detecting the inhibition effect of soil water deficit accompanied by global warming on ER.Our results suggested that current land surface models may overestimate the positive feedback between global warming and ER because of ignoring the diurnal differences in the temperature sensitivity of ER and basal respiration rates,and overlooking the effects of water changes on respiration.In summary,our study revealed the prevalence of diurnal variation in temperature sensitivity of ER and its importance for ER estimation,and clarified the importance of water content on ER.At the global scale,the decreasing trend of global ER over the recent three decades was mainly attributed to the reduction in soil moisture.Current land surface models may not be able to accurately assess the long-term trend of ER due to ignoring the diurnal differences in key physiological parameters of ER and the effects of water changes on respiration.These results offer directional guidance for improving the accuracy of terrestrial carbon flux projections to under climate change,and provide a solid scientific basis for more accurately assessing terrestrial net carbon sink in China and over the world.