Due to the temperature sensitivity of many plant and microbial processes, climate warming generally stimulates terrestrial ecosystem respiration (ER), the largest land-to-air CO2 flux annually. However, climate change is also steadily enhancing drought risk in most regions on the Earth, and given sensitivity of plant and microbial metabolism to soil moisture, this consequently makes uncertain the degree to which and dynamics of how, where, and whether climate change will stimulate ER at the global scale. Here, we provide a data-driven estimate of global ER product from 1989 to 2018 using a modified CO2 flux partitioning model based on eddy covariance, a Random Forest model, meteorological and remote-sensing observations. Our results showed that global ER increased at a rate of 0.110 ± 0.097 Pg C yr− 2 in 1989–1998 but then decreased at a rate of -0.090 ± 0.018 Pg C yr− 2 in 1998–2018. This declining trend in the global terrestrial ER was primarily driven by increasing moisture limitation, especially in a majority of tropical and temperate regions. However, current global land models do not adequately capture this apparent decreased trend in ER over the past two decades, likely because they overestimate impacts of rising temperature on global ER while underestimating the associated soil moisture effect. Our findings pose new scientific challenges and opportunities for model benchmarking, hypothesis generation and testing, and ecological forecasting.
Article
Increasing moisture limitation predominates recent decline trend in ecosystem respiration
https://doi.org/10.21203/rs.3.rs-3350160/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted
You are reading this latest preprint version
ecosystem respiration
soil moisture
decreasing trend
eddy covariance
Terrestrial ecosystem respiration (ER) represents a major CO2 flux between the atmosphere and biosphere1, and is the sum of autotrophic respiration from primary producers and heterotrophic respiration from consumers and detritivores2. Several decades of climate change had stimulated ER and some components3,4, returning biosphere carbon to CO2 in the atmosphere and thus amplifying global warming. Controls on ER include a complex, intertwining array of biotic and abiotic factors, and ER is typically approximated using models at continental and global scales owing to observational and measurement challenges. Uncertainties associated with estimates of the magnitude and temporal dynamics of such ER stimulation at the global scale remains one of the largest unknowns for the terrestrial C cycle and climate feedbacks4.
Based on multi-model comparisons and estimations of diverse approaches, a majority of researchers reported that terrestrial ER had shown a positive trend on average over several recent decades1,5, although a decreasing trend was also detected during the shorter “warming hiatus” period (i.e., a slowdown of global surface warming from 1998–2012)6. Resolving any potential discrepancy between these studies remains undone. Additionally, although the relative contributions of different global change agents on long-term changes in ER remains uncertain, recent reports suggest that climate has played a much more important role in controlling ER and its stability than community structure or biodiversity7–9.
Climate is known to control ER by direct (e.g., temperature) and indirect (e.g. soil moisture or substrate quality) effects on metabolism of ER components and drivers10,11. Meanwhile, the sensitivity of ER to temperature as contingent on moisture conditions is emerging as a key regulator of ER temporal dynamics9. Yet debates remain on whether the dominant driver of ER is temperature, moisture variability, or their interaction at long temporal scales6,11,12. At regional/global scale, the mean annual temperature anomaly has been often considered as the main factor to explain interannual ER temporal changes1,5,11. On the contrary, other recent studies underlined the dominance of water availability for ER, even once temperature was accounted for as a co-predictor13. Thus, disentangling the climate sensitivity of ER is imperative for advancing our knowledge of global C cycle processes and their potential feedbacks on future climate predictions14.
Currently, ER at continental and global scales is typically approximated using both “top-down”6 and “bottom-up” approaches15–17. The “top-down” atmospheric inversion method employs an atmospheric transport model coupled with atmospheric CO2 observations to quantify ER from terrestrial ecosystems; such estimates often have low spatial resolution and large uncertainty due to the low number of atmospheric CO2 observation sites6. The “bottom-up” approaches use direct eddy-covariance observations to upscale and constrain terrestrial ER with empirical statistical models, data-driven models, or process-based models15,17,18. Often these deploy data-driven models with machine learning algorithms to estimate ER at regional and global scales by upscaling of point data, based on prior empirically driven relationships between ER and drivers11,16,19. For example, FLUXNET-based CO2 fluxes from two flux partitioning methods were upscaled to global scale by training three machine learning algorithms to estimate terrestrial ER12, with a range from 99.5 to 117.5 Pg C yr− 1. However, ER estimates contain extensive uncertainties in constraining regional-scale to global-scale C fluxes, including those related to the upscaling of sparse or less reliable observations to a global scale (input data), choice of machine learning algorithms, and flux partitioning methods as well as forcing data. Thus, the spatial and temporal changes in terrestrial ER as well as its driving factors are still not well understood.
Here, the goals of this study are to (1) examine the temporal change trend of terrestrial ER during 1989–2018 and (2) investigate the ‘temperature versus water’ debate involving to what extent temperature and moisture availability control ER across spatial and temporal scales. Towards this end, we first designed an approach to generate a monthly global ER dataset at 0.5° spatial resolution during 1989–2018. It integrates in situ ER estimates from a modified CO2 flux partitioning model (referred to as DT-RH, see Methods) at the FLUXNET observing sites with geospatial information from meteorological and remote sensing data in a Random Forest model. The approach is data-driven and thus largely dependent of input data quality. In previous work, DT-RH method has been described and verified in three different ecosystems20, and we found that the diurnal and seasonal patterns of ER from DT-RH matched more closely with isotopic data than those from two other widely-used methods21,22. DT-RH modeling integrates the diurnal variation in temperature sensitivity of ER and the effect of moisture conditions on ER. We also diagnosed the robustness of DT-RH approach on ER estimations across global flux sites, which appears to be an improvement over previous widely used methods (e.g., nighttime and daytime methods21,22; Fig. S9). Finally, we used our model estimates of ER in combination with other model outputs (i.e., FLUXCOM23 and the Coupled Model Intercomparison Project models24–29 (CMIP6)) to assess the temporal change trend of terrestrial ER over the past three decades and investigated how climate sensitivity of ER has changed across the globe.
Our results showed that global ER increased from1989-1998 but then decreased from 1998–2018. This declining trend in the global terrestrial ER was primarily driven by increasing moisture limitation. This study provides evidence for temporal behavior and global spatial distribution of ER over the past 30 years, and identifies that changes in soil moisture were the likely cause of the recent declining trend in global ER.
Deriving global ecosystem respiration product from 1989 to 2018. We estimated a mean annual global ER value from 1989 to 2018 of 122.77 ± 0.73 Pg C yr− 1 (Fig. 1), with the highest ER values per hectare located in the tropics, and the lowest ER values in high-latitude regions of the northern hemisphere and Australia. The estimate of DT-RH falls within the likely and very broad range of 103.83-150.37 Pg C yr− 1, which was estimated by FLUXCOM and the Coupled Model Intercomparison Project models (CMIP6) (Table S1). The validity of our ER product is supported by internal cross-validation at FLUXNET sites (see Methods).
Temporal trends of global ecosystem respiration. The temporal changes in global terrestrial ER derived from DT-RH showed an apparent shift from an increasing trend before 1998 (slope = 0.110 ± 0.097 Pg C yr− 2, p = 0.29) to a decreasing one after 1998 (slope = -0.090 ± 0.018 Pg C yr− 2, p < 0.001; Fig. 1). The turning point of ER that occurred in 1998 was confirmed by piecewise linear regression (p < 0.001; Fig. 1b-c). The global trend toward decreasing ER in 1998–2018 was driven in part by widely observed decreases in the tropical and midlatitude temperate regions (Fig. 2). Among the 12 months, ER showed a significant temporally decreasing trend from May to August but not in other months (Figs. S3b-c).
The declined trend of global ER from 1998–2018 was opposite to those derived from FLUXCOM23 and CMIP6 models24–29, which reported a significant increasing trend over their studied periods (1989–2018 and 1989–2014, respectively; Figs. 1b-c and S3a and S4). Specifically, the results from FLUXCOM and CMIP6 displayed that the annual ER increased by a slope of 0.02 ± 0.01 Pg C yr− 2 from 1989–2018 (p = 0.007), and 0.39 ± 0.01 Pg C yr− 2 from 1989 to 2014, respectively (Fig. S3a).
Additionally, these three approaches also gave different spatiotemporal patterns. At the global scale, our results showed that 42% of global land experienced a decreasing trend in ER during 1989–2018, and the regions with a significant decreasing trend (14% by pixel) were mainly located at the tropical and midlatitude temperate regions. Specifically, the tropics accounted for 54% of the total decrease in the global ER, with 35% of those due to change in the temperate region. In contrast, FLUXCOM data indicated that approximately 26% of the global land experienced a decreasing trend over the period 1989–2018, and the regions with a significantly decreasing trend were mainly located in the Amazon Basin and central Africa in pixels covering approximately 7% of the land (Fig. S4). Meanwhile, CMIP6 simulations showed 6% of the global land experienced a decreasing trend in ER where was only in central Australia and Argentina from 1989 to 2014 (Fig. 1b-c and S4).
Influence of climate factors on the temporal changes of ER. From 1989 to 2018, global soil moisture significantly decreased (Fig. S5), especially in tropical and temperate regions (Fig. S6), which was generally spatially consistent with ER changes (Figs. 2 and S4). Variable importance analysis using a Random Forest model showed that the most important factors controlling the ER changes were soil moisture and MAT, with MAT dominant during 1989 to 1998 and soil moisture during 1998–2018 (Figs. 3a-c). The moving subset window analysis illustrated that the impacts of soil moisture on ER increasingly became more significant than that of MAT after 1998 (Fig. 3d). Partial correlation analysis at the pixel scale also displayed that the detrended ER was more positively correlated with the detrended soil moisture than MAT in a majority of tropical and temperate regions after 1998, causing a dominant effect of soil moisture on ER during 1998–2018, whereas rising MAT primarily increased ER in boreal regions (Figs. 4 and S7). In contrast, the ER changes derived from FLUXCOM and CMIP6 models did not have a significant positive correlation with soil moisture at most regions during 1989–2018, especially for CMIP6 simulations (Fig. S7).
Quantifying global C fluxes in terrestrial ecosystems using bottom-up approaches requires that we translate spatially sparse measurements into consistent, gridded flux estimates30. However, large uncertainties exist in current estimations of global ER by diverse approaches, mainly resulting from the upscaling of local ER by flux partitioning methods at site level to a global scale and from the choice of machine learning algorithms31,32. In this study, we provided new data-driven estimates for global land ER by reducing the uncertainty of the input data and integrating daytime versus nighttime temperature and relative humidity sensitivities of ER, and then examined spatiotemporal patterns of global ER as well as their control by temperature and moisture.
Our bottom-up analyses revealed that global terrestrial ER increased during 1989–1998, but significantly declined from 1998 to 2018 at -0.090 ± 0.018 Pg C yr− 2 due largely to limited soil moisture. This result distinctly differs from other bottom-up ER estimates (e.g., FLUXCOM and CMIP6) and previous findings that exhibited an increasing trend of global terrestrial respiration primarily driven by rising temperature1,11,33. Generally, temperature is assumed to be the dominant factor in regulating ER. However, experimental and seasonal warming studies suggest that indirect effects through warming-induced or naturally varying water stress may be stronger than the direct effects of rising temperature, and the magnitude of the respiration response to warming decreases linearly with the degree of soil drying in water-limited ecosystems34–36. If ER is also sensitive to carbon inputs from photosynthesis, a similar sensitivity of photosynthesis to both warming and its indirect effects on soil moisture37 could be partially at work, constraining new carbon acquisition and allocation, and reducing both carbon supply and respiratory demand for plants and microbes38. The study of ref.6 supported our findings with a top-down global estimate using atmospheric and satellite data during 1998–2012 (a warming hiatus). However, global terrestrial ER still decreased (Fig. 1b) when land surface temperature continually increased from 2013 with a higher anomaly than the warming hiatus4, suggesting the strong effects of soil moisture on ER.
Across the globe, our results showed that, for the first time of our knowledge, the decreasing trend of terrestrial ER from 1998 to 2018 mainly resulted from large decreases in the temperate and tropical regions (Fig. 2). Strong correlation between ER trends and changes in soil moisture was found over large parts of South America, Central Africa, Australia, Central Asia, Southwest Russia, Mongolia, and Eastern China (Figs. 2, 4 and S6), some of which are in dryland regions39. Our findings underscored the importance of soil moisture for ER at low and mid-latitudes regions, which can involve multiple mechanisms. First, lower soil moisture can directly and nearly instantaneously suppress plant root and microbial activities, and thus ecosystem respiration, especially in arid and semiarid regions40,41. Second, and complementarily, reduced average soil moisture can also alter species composition and growth of the plant and soil communities, and carbon allocation from plants to soils over time, resulting in decreased supply of carbon substrate for roots and microorganisms and thereby limiting respiratory metabolism per unit ground area (and thus slowing decomposition of plant litter and soil organic matter) even when measured at a comparable soil moisture43–46. In contrast, increasing ER trends driven by warming in boreal region have been found, probably owing to more positive (than in earlier decades) temperature impacts on photosynthesis, growth and respiration47,48 due to alleviation of severe low temperature constraints combined with modest moisture shortfalls (or even alleviation of anoxic conditions in high latitude wet forests), resulting in larger carbon stocks operating at slightly higher metabolic rates per unit carbon in these high-latitude regions2,49,50.
The ER changes we observed in these three biome zones (tropical, temperate, boreal) were also somewhat consistent with a recent report of global soil respiration observations4. Such a comparison is useful as soil respiration, consisting of autotrophic (root) respiration and heterotrophic respiration, accounts for 60 ~ 90% of ER41,42. Lei et al (2021) found a negative temporal in soil respiration in lower latitudes but a positive one in higher latitudes during 2000–2016, and a global increase in soil respiration to 1999 followed by a stable plateau4.
We also examined whether the simulations of FLUXCOM and Earth system models (CMIP6) can adequately capture the observed responses of ER to soil moisture changes over the study periods. We found that the simulated ER trends of FLUXCOM and CMIP6 ensemble models did not match the temporal and/or spatio-temporal ER trends documented above, especially CMIP6 simulations of ER (Figs. 1, S3 and S4), indicating that the significant increasing ER of Earth system models is likely due to fluxes being incorrectly parameterized, or because the responses of respiration to moisture and temperature are incorrectly represented. The Q10 equation remains the most common representation in temperature sensitivity of respiration across all biomes in Earth system models51, and a constant value of Q10 was often used in models. As it is well known that temperature-dependent Q10, thermal respiratory acclimation, and a non-exponential instantaneous temperature response of respiration are features of plants and soils52–54 and matter to modeled ER estimates55,56, the continued reliance on a fixed Q10 should questioned. Additionally, the most common representation of the moisture sensitivity of soil respiration in CMIP6 is a monotonically increasing function, i.e., that the respond of respiration to moisture is in the same way regardless of location or climate51,57. However, several studies have reported that the responses of respiration to daytime and nighttime temperature may be different58–60. Additionally, a previous study found that increases in ER under experimental warming were mostly detected in wet ecosystems, whereas warming-induced decreases in ER were mainly reported in dry ecosystems61. Thus, a monotonic function of soil respiration to temperature may not be warranted.
To address this potential shortcoming, in our analysis, we adapted a modified empirical model that integrated the different daytime/nighttime sensitivities of ER to temperature and relative humidity at site scale and upscaled those to global ER estimates. Our estimate of ER, like FLUXCOM values, did not directly account for other major carbon loss pathways from biosphere, such as fire and deforestation emissions, because these carbon loss pathways are not sampled by eddy-covariance measurements from FLUXNET23. Nonetheless, these CO2 sources are relatively moderate and have been demonstrated to exhibit no prominent trends at global scale6, and thus may not be making major contributions to decadal variability in our ER estimates.
In conclusions, on the basis of our findings, there is more of an apparent effect of moisture than of temperature on the most recent two decades trend in global terrestrial ER. The strong spatial consistency of the patterns in our ER estimates and soil moisture trends suggests that decreasing soil moisture supply in tropical and temperate regions is the main mechanism contributing to the reversal of the increasing ER trend after 1998. Hence, state-of-the-art Earth ecosystem models24–29 may presently overestimate global ER because they account for direct sensitivity of respiration to temperature but not for moisture effects, including those driven by temperature change. This suggests that process-based models should identify and account for more ways, times, and places in which temperature and moisture influence ER, which will be necessary to improve the estimations of global C fluxes. The two decades of decline in ER we observed may also help contribute to the ongoing growth in the land carbon sink, along with the other potential drivers of both C inputs and effluxes62. Our findings therefore provide new insight to help reconcile previous global C cycle studies and directional guidance for current projections of terrestrial carbon fluxes to climate change.
Eddy-covariance observations. We used eddy-covariance observations of carbon fluxes between ecosystems and the atmosphere from FLUXNET 2015 dataset63 (Tier 1; https://fluxnet.org/data/download-data/). The data used half-hourly and hourly carbon fluxes and their concurrent meteorological observations from 197 sites around the world (Fig. S8), including net ecosystem exchange (NEE_VUT_USTAR50), ecosystem respiration (RECO_DT_VUT_USTAR50 estimated from a daytime partitioning method21 (DT) and RECO_NT_VUT_USTAR50 estimated from a nighttime partitioning method22 (NT), air temperature (TA_F), solar radiation (SW_IN_F), and vapour pressure deficit (VPD_F) and relative humidity (RH). For the missing RH value, we applied random forest (RF) method to fill the gap using TA_F and VPD_F as drivers20. We also used a modified method20 to estimate ecosystem respiration from NEE_VUT_USTAR50, that is RECO_DT-RH.
Modified partitioning method (DT-RH). The DT-RH partitioning method has been successfully applied and demonstrated at three sites20. Here we briefly recapitulate our method. DT-RH modeling strategy is based on the following assumptions: (1) temperature sensitivity of ER was different between daytime and nighttime; (2) the effect of moister conditions on ER might vary from negative in wet ecosystems to positive in dry ones. This method was modified from the study of ref.21 by adding a relative humidity (RH) term as follows:
where α (µmol C J− 1) is the canopy-scale quantum yield, which represents the initial slope of the light-response curve; β (µmol C m− 2 s− 1) is the maximum CO2 uptake rate of the canopy at light saturation, which in turn is regulated by VPD. The threshold of VPD (VPD0) is set to 10 hPa. k is a coefficient to quantify the response of the maximum CO2 uptake to VPD, and Q is the global radiation (W m− 2). Rref (µmol C m− 2 s− 1) is the base respiration rate at the reference temperature (Tref = 15℃). E0 (℃) is the temperature sensitivity of ER and Tair (℃) is the air temperature. \(\stackrel{-}{RH}\) is the mean value of RH at a certain data window and ε is a fitted site-specific parameter, which indicates the positive and negative relationship between \(\stackrel{-}{RH}\) and the standardized respiration rate64 (i.e., the ratio of daytime ER (DER) and Rref, day, DER/Rref, day, referred to as site characteristics) across different sites. The standardized respiration rate (DER/Rref, day) was first calculated from the DT-E0 method20. The parameter \(\epsilon\)was set to -1 if a negative correlation existed between DER/Rref, day and \(\stackrel{-}{RH}\); otherwise, it was set to 1. The other parameters (α, β0, k, daytime Rref (Rref, day) and E0 (E0, day)) were estimated by fitting the entire model to the daytime data, whereas nighttime Rref (Rref, night) and E0 (E0, night) were estimated using nighttime data. We then applied Rref, day and E0, day to estimate daytime ER, and used Rref, night and E0, night to estimate nighttime ER.
Global forcing data. To compute the global flux products at monthly resolution via upscaling, we collected the global meteorological data and Normalized Difference Vegetation Index (NDVI). We extracted gridded monthly diurnal temperature range (DTR), daily mean temperature (Tair), monthly average daily minimum (Tmin) and maximum temperature (Tmax), frost day frequency (Frs), precipitation (Pre), vapour pressure (Vap), wet day frequency (Wet) and potential evapotranspiration (PET) time series data at 0.5° spatial resolution which were obtained from Climate Research Unit (CRU TS v. 4.04, https://crudata.uea.ac.uk/cru/data/hrg/). We also used monthly surface solar radiation (Rg), air relative humidity (RH) and surface volumetric soil moisture (0-7cm, SWC) data from the fifth-generation ECMWF atmospheric reanalysis of the global climate (ERA5) at 0.1°, 0.25° and 0.25° spatial resolution separately. In parallel, we collected global standardized precipitation-ET index (SPEI) data with 0.5° spatial resolution and a monthly time resolution as a measure of drought intensity (https://spei.csic.es/database.html). We obtained the biweekly Global Inventory Modeling and Mapping Studies third-generation (GIMMS-3g) NDVI dataset (https://ecocast.arc.nasa.gov/data/pub/gimms/3g.v1/) with a spatial resolution of 1/12° (~ 8km). Here, we composited the biweekly GIMMS-NDVI3g data to monthly temporal resolution by selecting the higher of the two composites in the same month and then aggregated to 0.5° × 0.5° to match the resolution of meteorological data. To analyze the effects of elevated atmospheric CO2 concentrations on long-term changes of ER, we used global atmospheric CO2 concentration averages from the National Oceanic and Atmospheric Administration (NOAA, https://gml.noaa.gov/ccgg/trends/gl_data.html) from 1989 to 2018.
Global ecosystem respiration estimation. A machine learning method (i.e., random forest65) was applied to estimate global terrestrial ecosystem respiration (ER) from 1989 to 2018. Firstly, we estimated ER based on half-hourly (or hourly) eddy covariance data from 197 globally distributed flux towers using DT-RH method20, which modified daytime flux partitioning method to estimate ER by integrating the different daytime and nighttime temperature sensitivities of respiration along with relative humidity. Secondly, we convert half-hourly (or hourly) scale ER data to a monthly scale at each site, and yield a global dataset composed of monthly estimates of local respiration and forcing data (i.e., DTR, Tair, Tmin, Tmax, Frs, Vap, PET, RH, Wet, Rg, SPEI, SWC) from each tower. Thirdly. we used these global dataset to train the random forest model (RF) and then created maps of gridded monthly ER at 0.5° resolution covering the whole period 1989–2018, namely DT-RH products. In addition, in order to consider the impact of land cover change on ER estimates, we also added NDVI into the global training dataset in steps two and three. The other steps remained consistent, and this ER product named DT-RH_NDVI. Since the results were highly consistent between DT-RH and DT-RH_NDVI (Table S1 and Fig. S4), we only used ER dataset from DT-RH for the following analyses. RF model was created by using h2o package (https://github.com/h2oai/h2o-3) in R. To evaluate the performance of our approach, we firstly evaluated the accuracy of training data by comparing with available measured monthly night-time ecosystem respiration at site scale (Fig. S9) using R2 (coefficient of determination). Mean monthly estimates across global sites showed considerably higher accuracy in DT-RH method than DT and NT methods (RMSEDT−RH = 4.37 ± 0.24, RMSEDT = 11.83 ± 0.67, and RMSENT = 6.0 ± 0.46 g C m− 2 month− 1, respectively; Fig. S1). Then, the validity of our data-driven ER product was supported by internal cross-validation at FLUXNET sites (R2 = 0.715, RMSE = 26.322 g C m− 2 month− 1; Table S1), which was better performance than the standard methods18 (i.e., R2DT = 0.64, RMSEDT = 34.2 g C m− 2 month− 1; R2NT = 0.64, RMSENT = 34.5 g C m− 2 month− 1).
For this study, we also used ER data from FLUXCOM (version RS + METEO)18,66 and the Coupled Model Intercomparison Project Phase 6 (CMIP6) models24–29. The FLUXCOM ER product is based on machine learning methods that upscale FLUXNET67 observations of carbon fluxes, which estimate from two flux partitioning approaches21,22. We collected FLUXCOM-ER product generated by random forest (RF) method at 0.5° spatial resolution and monthly time scale over the period 1989–2018. The six of model simulations from CMIP6 were considered in our analysis, including BCC-CSM2-MR, BCC-ESM1, CESM2, CMCC-CM2-SR5, CMCC-ESM2, MPI-ESM1-2-LR. From these model outputs, we extracted years 1989 to 2014 for comparison with our inferred estimates of ER.
Statistical analyses. All statistical analyses were carried out in R version 4.0.568. We used the piecewise linear regression method to determine the turning point of ER during 1989–2018. The temporal trends of ER were calculated by combining the results of a Theil-Sen median trend analysis and a Mann-Kendall test with R package “trend”. Before conducting the variable importance and partial correlation analysis, we first detrended all time series data to avoid auto-correlation and to make the time series stationary. We used random forest models69 to identify those statistically significant factors contributing to ER, including MAT, SWC, SPEI, MAP, Rg. Variable importance was determined by the increase in node purity (IncNodePurity) which was measured as the decrease in the residual sum of squares if the variable was excluded70,71. Greater values of IncNodePurity indicated a greater importance of variables. Random forest analyses were implemented in “rfPermute” package, and variable importance was tested for significance by 1000 permutations. The direct response of ER to temperature and soil moisture was further examined using the partial correlation analysis. To compare with other studies on global and regional TER estimates, we defined three regions10,11 based on annual air temperature (MAT): tropical (MAT > 17 ℃), temperate (2 ℃ ≤ MAT ≤ 17 ℃), and boreal (MAT < 2 ℃). ALL these analyses were performed at the pixel, regional, and global scales.
- Li, W. et al. Recent changes in global photosynthesis and terrestrial ecosystem respiration constrained from multiple observations. Geophys. Res. Lett. 45, 1058–1068 (2018).
- Niu, B. et al. Warming homogenizes apparent temperature sensitivity of ecosystem respiration. Sci. Adv. 7, eabc7358 (2021).
- Bond-Lamberty, B. et al. Globally rising soil heterotrophic respiration over recent decades. Nature 560, 80–83 (2018).
- Lei, J. et al. Temporal changes in global soil respiration since 1987. Nat. Commun. 12, 403 (2021).
- Keenan, T. F. et al. Recent pause in the growth rate of atmospheric CO2 due to enhanced terrestrial carbon uptake. Nat. Commun. 7, 13428 (2016).
- Ballantyne, A. et al. Accelerating net terrestrial carbon uptake during the warming hiatus due to reduced respiration. Nat. Clim. Change 7, 148–152 (2017).
- Fernández‐Martínez, M. et al. The role of climate, foliar stoichiometry and plant diversity on ecosystem carbon balance. Glob. Change Biol. 26, 7067–7078 (2020).
- Niu, S. et al. Interannual variability of ecosystem carbon exchange: from observation to prediction. Global Ecol. Biogeogr 26, 1225–1237 (2017).
- Piao, S. et al. Interannual variation of terrestrial carbon cycle: issues and perspectives. Glob. Change Biol. 26, 300–318 (2020).
- Bond-Lamberty, B. & Thomson, A. Temperature-associated increases in the global soil respiration record. Nature 464, 579–582 (2010).
- Huang, N. et al. Spatial and temporal variations in global soil respiration and their relationships with climate and land cover. Sci. Adv. 6, eabb8508 (2020).
- Jung, M. et al. Compensatory water effects link yearly global land CO2 sink changes to temperature. Nature 541, 516–520 (2017).
- Wang, K. et al. Regional and seasonal partitioning of water and temperature controls on global land carbon uptake variability. Nat. Commun. 13, 3469 (2022).
- Friedlingstein, P. et al. Uncertainties in CMIP5 Climate Projections due to carbon cycle feedbacks. Journal of Climate 27, 511–526 (2014).
- Ai, J. et al. MODIS-based estimates of global terrestrial ecosystem respiration. J. Geophys. Res. Biogeosci. 123, 326–352 (2018).
- Jung, M. et al. Global patterns of land-atmosphere fluxes of carbon dioxide, latent heat, and sensible heat derived from eddy covariance, satellite, and meteorological observations. J. Geophys. Res. 116, G00J07 (2011).
- Lei, L. et al. Water response of ecosystem respiration regulates future projection of net ecosystem productivity in a semiarid grassland. Agric. For. Meteorol. 252, 175–191 (2018).
- Tramontana, G. et al. Predicting carbon dioxide and energy fluxes across global FLUXNET sites withregression algorithms. Biogeosciences 13, 4291–4313 (2016).
- Beer, C. et al. Terrestrial gross carbon dioxide uptake: global distribution and covariation with climate. Science 329, 834–838 (2010).
- Li, N. et al. Improving estimations of ecosystem respiration with asymmetric daytime and nighttime temperature sensitivity and relative humidity. Agric. For. Meteorol. 312, 108709 (2022).
- Lasslop, G. et al. Separation of net ecosystem exchange into assimilation and respiration using a light response curve approach: critical issues and global evaluation. Glob. Change Biol. 16, 187–208 (2010).
- Reichstein, M. et al. On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm. Glob. Change Biol. 11, 1424–1439 (2005).
- Jung, M. et al. Scaling carbon fluxes from eddy covariance sites to globe: synthesis and evaluation of the FLUXCOM approach. Biogeosciences 17, 1343–1365 (2020).
- Danabasoglu, G. NCAR CESM2 model output prepared for CMIP6 CMIP. Earth System Grid Federation https://doi.org/10.22033/ESGF/CMIP6.7627 (2019).
- Lovato, T. & Peano, D. CMCC CMCC-CM2-SR5 model output prepared for CMIP6 CMIP historical. Earth System Grid Federation https://doi.org/10.22033/ESGF/CMIP6.3825 (2020).
- Lovato, T., Peano, D. & Butenschön, M. CMCC CMCC-ESM2 model output prepared for CMIP6 CMIP historical. Earth System Grid Federation https://doi.org/10.22033/ESGF/CMIP6.13195 (2021).
- Wieners, K.-H. et al. MPI-M MPI-ESM1.2-LR model output prepared for CMIP6 CMIP historical. Earth System Grid Federation https://doi.org/10.22033/ESGF/CMIP6.6595 (2019).
- Wu, T. et al. BCC BCC-CSM2MR model output prepared for CMIP6 CMIP historical. Earth System Grid Federation https://doi.org/10.22033/ESGF/CMIP6.2948 (2018).
- Zhang, J. et al. BCC BCC-ESM1 model output prepared for CMIP6 CMIP historical. Earth System Grid Federation https://doi.org/10.22033/ESGF/CMIP6.2949 (2018).
- Xiao, J. et al. Upscaling carbon fluxes from towers to the regional scale: Influence of parameter variability and land cover representation on regional flux estimates. J. Geophys. Res. 116, G03027 (2011).
- Keenan, T. F. et al. Widespread inhibition of daytime ecosystem respiration. Nat. Ecol. Evol. 3, 407–415 (2019).
- Wehr, R. et al. Seasonality of temperate forest photosynthesis and daytime respiration. Nature 534, 680–683 (2016).
- Hashimoto, S. et al. Global spatiotemporal distribution of soil respiration modeled using a global database. Biogeosciences 12, 4121–4132 (2015).
- Carey, J. C. et al. Temperature response of soil respiration largely unaltered with experimental warming. Proc. Natl. Acad. Sci. USA 113, 13797–13802 (2016).
- Sierra, C. A. et al. Sensitivity of decomposition rates of soil organic matter with respect to simultaneous changes in temperature and moisture. J. Adv. Model. Earth Syst. 7, 335–356 (2015).
- Xu, W. et al. A meta-analysis of the response of soil moisture to experimental warming. Environ. Res. Lett. 8, 044027 (2013).
- Reich, P. B. et al. Effects of climate warming on photosynthesis in boreal tree species depend on soil moisture. Nature 562, 263–267 (2018).
- Fernandez, C. W. et al. Ectomycorrhizal fungal response to warming is linked to poor host performance at the boreal‐temperate ecotone. Glob. Change Biol. 23, 1598–1609 (2017).
- Hulme, M. Recent climatic change in the world’s drylands. Geophys. Res. Lett. 23, 61–64 (1996).
- Gomez-Casanovas, N. et al. Net ecosystem exchange modifies the relationship between the autotrophic and heterotrophic components of soil respiration with abiotic factors in prairie grasslands. Glob. Change Biol. 18, 2532–2545 (2012).
- Ryan, E. M. et al. Antecedent moisture and temperature conditions modulate the response of ecosystem respiration to elevated CO2 and warming. Glob. Change Biol. 21, 2588–2602 (2015).
- Tedeschi, V. et al. Soil respiration in a Mediterranean oak forest at different developmental stages after coppicing. Glob. Change Biol. 12, 110–121 (2006).
- Davidson, E. A. & Janssens, I. A. Temperature sensitivity of soil carbon decomposition and feedbacks to climate change. Nature 440, 165–173 (2006).
- Humphrey, V. et al. Soil moisture–atmosphere feedback dominates land carbon uptake variability. Nature 592, 65–69 (2021).
- Liu, W., Zhang, Z. & Wan, S. Predominant role of water in regulating soil and microbial respiration and their responses to climate change in a semiarid grassland. Glob. Change Biol. 15, 184–195 (2009).
- Xu, L., Baldocchi, D. D. & Tang, J. How soil moisture, rain pulses, and growth alter the response of ecosystem respiration to temperature. Global Biogeochem. Cycles 18, GB4002 (2004).
- Atkin, O. et al. Global variability in leaf respiration in relation to climate, plant functional types and leaf traits. New Phytol. 206, 614–636 (2015).
- Reich, P. B. & Oleksyn, J. Climate warming will reduce growth and survival of Scots pine except in the far north: Scots pine growth and survival following climate transfer. Ecology Letters 11, 588–597 (2008).
- Fernández-Martínez, M. et al. Global trends in carbon sinks and their relationships with CO2 and temperature. Nat. Clim. Change 9, 73–79 (2019).
- Koven, C. D., Hugelius, G., Lawrence, D. M. & Wieder, W. R. Higher climatological temperature sensitivity of soil carbon in cold than warm climates. Nat. Clim. Change 7, 817–822 (2017).
- Varney, R. M. et al. Evaluation of soil carbon simulation in CMIP6 Earth system models. Biogeosciences 19, 4671–4704 (2022).
- Tjoelker, M. G., Oleksyn, J. & Reich, P. B. Modelling respiration of vegetation: evidence for a general temperature-dependent Q10. Glob. Change Biol. 7, 223–230 (2001).
- Heskel, M. A. et al. Convergence in the temperature response of leaf respiration across biomes and plant functional types. Proc. Natl. Acad. Sci. USA 113, 3832–3837 (2016).
- Reich, P. B. et al. Boreal and temperate trees show strong acclimation of respiration to warming. Nature 531, 633–636 (2016).
- Huntingford, C. et al. Implications of improved representations of plant respiration in a changing climate. Nat. Commun. 8, 1602 (2017).
- Butler, E. E. et al. Updated respiration routines alter spatio-temporal patterns of carbon cycling in a global land surface model. Environ. Res. Lett. 16, 104015 (2021).
- Hawkes, C. V., Waring, B. G., Rocca, J. D. & Kivlin, S. N. Historical climate controls soil respiration responses to current soil moisture. Proc. Natl. Acad. Sci. USA 114, 6322–6327 (2017).
- Gessler, A. et al. Night and day – Circadian regulation of night-time dark respiration and light-enhanced dark respiration in plant leaves and canopies. Environmental and Experimental Botany 137, 14–25 (2017).
- Järveoja, J., Nilsson, M. B., Crill, P. M. & Peichl, M. Bimodal diel pattern in peatland ecosystem respiration rebuts uniform temperature response. Nat. Commun. 11, 4255 (2020).
- Juszczak, R., Acosta, M. & Olejnik, J. Comparison of daytime and nighttime ecosystem respiration measured by the closed chamber technique on a temperate mire in Poland. Polish Journal of Environmental Studies 21, 643–658 (2012).
- Quan, Q. et al. Water scaling of ecosystem carbon cycle feedback to climate warming. Sci. Adv. 5, eaav1131 (2019).
- Ruehr, S. et al. Evidence and attribution of the enhanced land carbon sink. Nat Rev Earth Environ 4, 518–534 (2023).
- Pastorello, G. The FLUXNET2015 dataset and the ONEFlux processing pipeline for eddy covariance data. 27 (2020).
- Jägermeyr, J. et al. A high-resolution approach to estimating ecosystem respiration at continental scales using operational satellite data. Glob. Change Biol. 20, 1191–1210 (2014).
- Altmann, A., Toloşi, L., Sander, O. & Lengauer, T. Permutation importance: a corrected feature importance measure. Bioinformatics 26, 1340–1347 (2010).
- Jung, M. et al. The FLUXCOM ensemble of global land-atmosphere energy fluxes. Sci. Data 6, 74 (2019).
- Baldocchi, D. et al. FLUXNET: A new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities. Bulletin of the American Meteorological Society 82, 20 (2001).
- R Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna https://www.R-project.org/ (2021).
- Breiman, L. Random Forest. Machine Learning 45, 5–32 (2001).
- GFBI consortium et al. Climatic controls of decomposition drive the global biogeography of forest-tree symbioses. Nature 569, 404–408 (2019).
- Zhong, Y. et al. Arbuscular mycorrhizal trees influence the latitudinal beta-diversity gradient of tree communities in forests worldwide. Nat. Commun. 12, 3137 (2021).
There is NO Competing Interest.
- SupplementaryInformation.docx
Supplementary Information