Search

Terrestrial ecosystem acts as important carbon dioxide (CO2) sinks and nitrous oxide (N2O) sources. Ecosystem green-house-gas fluxes could further lead to a climate feedback, which are highly correlated with the C-N coupling. However, magnitudes of such composited feedbacks as well as contributions by individual ecological processes remained certain uncertainties. Here, we applied a terrestrial biosphere model QUINCY with fully C-N-coupling to comprehensively examine responses of CO2 and N2O fluxes to future climate changes and quantify contributions by individual processes. Our results showed that the trade-offs in CO2 and N2O still led a negative feedback (-3386.9 Tg CO2 yr-1) averaged over 2050-2100 under the SSP 5-8.5 scenario relative to SSP 1-2.6 scenario, however, which varies from -1761.7 Tg CO2 yr-1 to -5012.1 CO2 yr-1 with a high or low climate sensitivity to CO2 increases. Further process analysis showed that the CO2 fertilization effects on ecosystem climate feedbacks were equivalent for high and low climate sensitivity, but the higher climate sensitivity led to less carbon sequestration on tropical plants as well as higher N2O emissions. The climate feedbacks attributed to individual soil processes, including decomposition, nitrification, denitrification and nitrogen biological fixation, were also comprehensively quantified. This finding suggests that reducing the uncertainties in climate sensitivity estimates could be of great significance to better predict future terrestrial ecosystem green-house-gas fluxes as well as corresponding climate feedbacks.

Fire is an ubiquitous process within the Earth system that has significant impacts in terrestrial ecosystems. Process-based fire models quantify fire disturbance effects in stand-alone dynamic global vegetation models (DGVMs) and within coupled Earth system models (ESMs), and their advances have incorporated both descriptions of natural processes and anthropogenic drivers. However, we still observe a limited skill in modeling and predicting fire at global scale, mostly due to the stochastic nature of fire, but also due to the limits in empirical parameterizations in these process-based models. As an alternative, statistical approaches have shown the advantages of machine learning in providing robust diagnostics of fire damages, though with limited value for process-based modeling frameworks. Here, we develop a deep-learning-based fire model (DL-fire) to estimate gridded burned area fraction at global scale and couple it within JSBACH4, the land surface model used in the ICON ESM. We compare the resulting hybrid model integrating DL-fire into JSBACH4 (JDL-fire) against the standard fire model within JSBACH4 and the stand-alone DL-fire results. The stand-alone DL-fire model forced with observations shows high performance in simulating global burnt fraction, showing a monthly correlation (Rm) with the Global Fire Emissions Database (GFED4) of 0.78 and of 0.8 at global scale during the training (2004-10) and validation periods (2011-15), respectively. The performance remains nearly the same when evaluating the hybrid modeling approach JDL-fire (Rm=0.76 and 0.86 in training and evaluation periods, respectively). This outperforms the currently used standard fire model in JSBACH4 (Rm=-0.16 and 0.22 in training and evaluation periods, respectively) by far. We further evaluate the modeling results across specific fire regions and apply layer-wise relevance propagation (LRP) to quantify importance of each predictor. Overall, land properties, such as fuel amount and water contents in soil layers, stand out as the major factors determining burnt fraction in DL-fire, paralleled by meteorological conditions, over tropical and high latitude regions. Our study demonstrates the potential of hybrid modeling in advancing the predictability of Earth system processes by integrating statistical learning approaches in physics-based dynamical systems.

Continental North America has been found to be a carbon (C) sink over recent decades by multiple studies employing a variety of estimation approaches. However, several key questions and uncertainties remain with these assessments. Here we used results from an ensemble of 19 state-of-the-art dynamic global vegetation models from the TRENDYv9 project to improve these estimates and study the drivers of its interannual variability. Our results show that North America has been a C sink with a magnitude of 0.37 ± 0.38 (mean and one standard deviation) PgC year$^{-1}$ for the period 2000–2019 (0.31 and 0.44 PgC year$^{-1}$ in each decade); split into 0.18 ± 0.12 PgC year$^{-1}$ in Canada (0.15 and 0.20), 0.16 ± 0.17 in the United States (0.14 and 0.17), 0.02 ± 0.05 PgC year$^{-1}$ in Mexico (0.02 and 0.02) and 0.01 ± 0.02 in Central America and the Caribbean (0.01 and 0.01). About 57% of the new C assimilated by terrestrial ecosystems is allocated into vegetation, 30% into soils, and 13% into litter. Losses of C due to fire account for 41% of the interannual variability of the mean net biome productivity for all North America in the model ensemble. Finally, we show that drought years (e.g., 2002) have the potential to shift the region to a small net C source in the simulations (−0.02 ± 0.46 PgC year$^{-1}$). Our results highlight the importance of identifying the major drivers of the interannual variability of the continental-scale land C cycle along with the spatial distribution of local sink-source dynamics.

Terrestrial ecosystems respond to changes in environmental conditions, mainly via key climatic controls of precipitation and temperature on vegetation activities and decomposition processes (Taylor et al., 2017). Yet, the relationship between climate and the overall spatiotemporal dynamics and uncertainties of the global carbon cycle, i.e., gross primary productivity (GPP), effective ecosystem carbon turnover times (τ), and consequently the total ecosystem carbon stock (Ctotal), are unclear (Anav et al., 2013; Friedlingstein et al., 2014; Friend et al., 2014; Jones et al., 2013). Using a global observation-based synthesis, we first show that the apparent partial spatial climate sensitivities of GPP and τ are associated with relative availability of precipitation and temperature, and are therefore modulated by aridity. The apparent sensitivity of GPP to temperature increases from arid to humid climatic regions. In contrast, its sensitivity to precipitation is invariant throughout different climatic regions. Simultaneously, the τ-precipitation response is strongly non-linear resulting in ~2 times longer τ in arid regions compared to humid regions for a given temperature. Compared with these observed patterns, the offline carbon cycle simulations of seven European Earth System Models (ESMs), that participated in CMIP6, perform relatively better for climate sensitivities of GPP than those of τ. This leads to a large spread and bias in Ctotal in both warm and cold semi-arid and arid regions where only a few models capture the observed τ-precipitation relationship. The emergence of the hydrological controls, modulated by aridity, on global carbon cycle implies that the changes in precipitation may moderate the temperature-driven climate feedback of the global carbon cycle under climate change.

With the focus of climate policy shifting from pledges to implementation, there is an increasing need to track progress on climate change mitigation at country level, especially for the land-use sector. Despite new tools and models offering unprecedented monitoring opportunities, striking differences remain in estimations of anthropogenic land-use CO2 fluxes between the national greenhouse gas inventories (NGHGIs) used to assess compliance with the Paris Agreement, and the Global Carbon Budget and IPCC assessment reports, both based on global bookkeeping models (BMs). Recent evidence showed that these differences are mainly due to inconsistent definitions of anthropogenic forest CO2 fluxes. In particular, the part of the land sink that is caused by the indirect effects of human-induced environmental change (e.g., fertilization effect on vegetation growth due to increase atmospheric CO2 concentration, climate change) on managed lands is treated as non-anthropogenic by BMs, while in most cases is considered anthropogenic in NGHGIs. In addition, countries use a broader definition of managed land than BMs. Building on previous studies, we implement an approach that adds the CO2 sink due to environmental change from countries’ managed forest area (estimated by Dynamic Global Vegetation Models, DGVMs) to the original land-use flux from BMs. This sum is expected to be conceptually more comparable to NGHGIs. Our analysis uses updated and more comprehensive data from NGHGIs than previous studies and provides model results at a greater level of disaggregation in terms of land categories (i.e., forest land, deforestation, organic soils, other land uses) and countries. Our results confirm a large difference in land use CO2 fluxes between the ensemble mean of the BMs, estimating a source of 4.3 GtCO2 yr-1 globally for the period 2001–2020, and NGHGIs, which estimate a sink of -1.7 GtCO2 yr-1. Most of this 6.0 GtCO2 yr-1 gap is found on forest land (3.8 GtCO2 yr-1), with differences also for deforestation (1.1 GtCO2 yr-1), other land uses (1.0 GtCO2 yr-1), and to a lesser extent for organic soils (0.1 GtCO2 yr-1). By adding the DGVM ensemble mean sink arising from environmental change in managed forests (-5.1 GtCO2 yr-1) to BMs estimates, the gap between BMs and NGHGIs becomes significantly smaller both globally (residual gap: 0.9 GtCO2 yr-1) and in most regions and countries. The remaining differences mostly reflect smaller net emissions from deforestation and agricultural land in the NGHGIs of developing countries than in the BMs. By reconciling most of the differences between NGHGIs and global models (BMs and DGVMs), offering a blueprint for operationalizing future comparisons, and identifying areas to be further investigated, this study represents an important step forward for increasing transparency and confidence in land-use CO2 flux estimates at the country level. This is crucial to support land-based mitigation investments and assess the countries’ collective progress under the Paris Agreement’s Global Stocktake.

The trajectories of soil carbon (C) in the changing climate are of utmost importance, as soil carbon is a substantial carbon storage with a large potential to impact the atmospheric carbon dioxide (CO2) burden. Atmospheric CO2 observations integrate all processes affecting C exchange between the surface and the atmosphere. Therefore they provide a benchmark for carbon cycle models. We evaluated two distinct soil carbon models (CBALANCE and YASSO) that were implemented to a global land surface model (JSBACH) against atmospheric CO2 observations. We transported the biospheric carbon fluxes obtained by JSBACH using the atmospheric transport model TM5 to obtain atmospheric CO2. We then compared these results with surface observations from Global Atmosphere Watch (GAW) stations as well as with column XCO2 retrievals from the GOSAT satellite. The seasonal cycles of atmospheric CO2 estimated by the two different soil models differed. The estimates from the CBALANCE soil model were more in line with the surface observations at low latitudes (0 N–45 N) with only 1 % bias in the seasonal cycle amplitude (SCA), whereas YASSO was underestimating the SCA in this region by 32 %. YASSO gave more realistic seasonal cycle amplitudes of CO2 at northern boreal sites (north of 45 N) with underestimation of 15 % compared to 30 % overestimation by CBALANCE. Generally, the estimates from CBALANCE were more successful in capturing the seasonal patterns and seasonal cycle amplitudes of atmospheric CO2 even though it overestimated soil carbon stocks by 225 % (compared to underestimation of 36 % by YASSO) and its predictions of the global distribution of soil carbon stocks was unrealistic. The reasons for these differences in the results are related to the different environmental drivers and their functional dependencies of these two soil carbon models. In the tropical region the YASSO model showed earlier increase in season of the heterotophic respiration since it is driven by precipitation instead of soil moisture as CBALANCE. In the temperate and boreal region the role of temperature is more dominant. There the heterotophic respiration from the YASSO model had larger annual variability, driven by air temperature, compared to the CBALANCE which is driven by soil temperature. The results underline the importance of using sub-yearly data in the development of soil carbon models when they are used in shorter than annual time scales.

Field measurements of aboveground net primary productivity (ANPP) in temperate grasslands suggest that both positive and negative asymmetric responses to changes in precipitation (P) may occur. Under normal range of precipitation variability, wet years typically result in ANPP gains being larger than ANPP declines in dry years (positive asymmetry), whereas increases in ANPP are lower in magnitude in extreme wet years compared to reductions during extreme drought (negative asymmetry). Whether the current generation of ecosystem models with a coupled carbon–water system in grasslands are capable of simulating these asymmetric ANPP responses is an unresolved question. In this study, we evaluated the simulated responses of temperate grassland primary productivity to scenarios of altered precipitation with 14 ecosystem models at three sites: Shortgrass steppe (SGS), Konza Prairie (KNZ) and Stubai Valley meadow (STU), spanning a rainfall gradient from dry to moist. We found that (1) the spatial slopes derived from modeled primary productivity and precipitation across sites were steeper than the temporal slopes obtained from inter-annual variations, which was consistent with empirical data; (2) the asymmetry of the responses of modeled primary productivity under normal inter-annual precipitation variability differed among models, and the mean of the model ensemble suggested a negative asymmetry across the three sites, which was contrary to empirical evidence based on filed observations; (3) the mean sensitivity of modeled productivity to rainfall suggested greater negative response with reduced precipitation than positive response to an increased precipitation under extreme conditions at the three sites; and (4) gross primary productivity (GPP), net primary productivity (NPP), aboveground NPP (ANPP) and belowground NPP (BNPP) all showed concave-down nonlinear responses to altered precipitation in all the models, but with different curvatures and mean values. Our results indicated that most models overestimate the negative drought effects and/or underestimate the positive effects of increased precipitation on primary productivity under normal climate conditions, highlighting the need for improving eco-hydrological processes in those models in the future.

Background Large uncertainty in modeling land carbon (C) uptake heavily impedes the accurate prediction of the global C budget. Identifying the uncertainty sources among models is crucial for model improvement yet has been difficult due to multiple feedbacks within Earth System Models (ESMs). Here we present a Matrix-based Ensemble Model Inter-comparison Platform (MEMIP) under a unified model traceability framework to evaluate multiple soil organic carbon (SOC) models. Using the MEMIP, we analyzed how the vertically resolved soil biogeochemistry structure influences SOC prediction in two soil organic matter (SOM) models. By comparing the model outputs from the C-only and CN modes, the SOC differences contributed by individual processes and N feedback between vegetation and soil were explicitly disentangled. Results Results showed that the multi-layer models with a vertically resolved structure predicted significantly higher SOC than the single layer models over the historical simulation (1900–2000). The SOC difference between the multi-layer models was remarkably higher than between the single-layer models. Traceability analysis indicated that over 80% of the SOC increase in the multi-layer models was contributed by the incorporation of depth-related processes, while SOC differences were similarly contributed by the processes and N feedback between models with the same soil depth representation. Conclusions The output suggested that feedback is a non-negligible contributor to the inter-model difference of SOC prediction, especially between models with similar process representation. Further analysis with TRENDY v7 and more extensive MEMIP outputs illustrated the potential important role of multi-layer structure to enlarge the current ensemble spread and the necessity of more detail model decomposition to fully disentangle inter-model differences. We stressed the importance of analyzing ensemble outputs from the fundamental model structures, and holding a holistic view in understanding the ensemble uncertainty.