Elevated carbon dioxide (eCO2) acts as a fertiliser for photosynthesis, driving an increase in gross primary production (GPP). However, it is unclear how effectively increased GPP propagates along the 'carbon (C) cascade' to increase net primary production (NPP) and vegetation C stocks (Cveg) in different plant compartments. Vegetation models were criticised for being overly sensitive to photosynthesis (source-driven), neglecting sink-driven processes which may attenuate (or amplify) changes in NPP and vegetation C stocks. Here, we introduce an analytical framework to diagnose linearity (L) as ratios of relative changes in linked fluxes and pools. We then apply this framework to 16 models of the TRENDY v11 ensemble and to observation-based estimates of CO2 sensitivities. We found widely varying global patterns in L across models. Six models showed a majority of grid cells with larger relative changes in NPP than in GPP (LNPP:GPP > 1 for > 60% of gridcells), indicating increased vegetation carbon use efficiency under eCO2. Only three models had LNPP:GPP < 1 for > 60% of gridcells. Four models showed a majority of gridcells with larger relative changes in estimated steady-state Cveg than in NPP, while five models showed the opposite-in both cases with a large spread of LCveg*:NPP across grid cells within models. Observations-based analysis reveals median LCveg*:NPP < 1 overall but evidence is insufficient for a conclusion. Three models showed a larger relative increase in root C than in Cveg, (LCroot:Cveg > 1) while five models showed the opposite. Most field evidence shows LCroot:Cveg > 1. Widely differing distributions of L among models and links in the C cascade reveal a strong influence of nonlinear behaviour in individual models. However, due to the spread in L, across the whole models ensemble, L deviations from 1 were roughly balanced, leading to an overall linear behaviour of terrestrial C cycle representations in the multi-model-mean.

Accurate assessment of anthropogenic carbon dioxide (CO2) emissions and their redistribution among the atmosphere, ocean, and terrestrial biosphere in a changing climate is critical to better understand the global carbon cycle, support the development of climate policies, and project future climate change. Here we describe and synthesise datasets and methodologies to quantify the five major components of the global carbon budget and their uncertainties. Fossil CO2 emissions (E-FOS) are based on energy and cement production data. Emissions from land-use change (E-LUC) are estimated by bookkeeping models based on land-use data. The global atmospheric CO2 growth rate (G(ATM)) is computed from changes in concentration measured at surface stations. The global net uptake of CO2 by the ocean (S-OCEAN) is estimated with global ocean biogeochemistry models and observation-based fCO(2)-products. The global net uptake of CO2 by the land (S-LAND) is estimated with dynamic global vegetation models. Additional lines of evidence are provided by atmospheric inversions, atmospheric oxygen measurements, ocean interior observation-based estimates, and Earth System Models. This year, we introduced corrections on the E-LUC, S-OCEAN and S-LAND estimates. The sum of all sources and sinks results in the carbon budget imbalance (B-IM), a measure of imperfect data and incomplete understanding of the contemporary carbon cycle. All uncertainties are reported as +/- 1 sigma. For the year 2024, E-FOS increased by 1.1 % relative to 2023, with fossil emissions at 10.3 +/- 0.5 GtC yr(-1) (including the cement carbonation sink, 0.2 GtC yr(-1)), E-LUC was 1.3 +/- 0.7 GtC yr(-1), for total anthropogenic CO2 emissions of 11.6 +/- 0.9 GtC yr(-1) (42.4 +/- 3.2 GtCO(2) yr(-1)). Also, for 2024, G(ATM) was 7.9 +/- 0.2 GtC yr(-1) (3.73 +/- 0.1 ppm yr(-1)), 2.2 GtC above the 2023 growth rate. S-OCEAN was 3.4 +/- 0.4 GtC yr(-1) and S-LAND was 1.9 +/- 1.1 GtC yr(-1), leaving a large negative B-IM (-1.7 GtC yr(-1)), suggesting that the total sink or G(ATM) is strongly overestimated in 2024. The global atmospheric CO2 concentration averaged over 2024 reached 422.8 +/- 0.1 ppm. Preliminary data for 2025 suggest an increase in E-FOS relative to 2024 of +1.0 % (0.2 % to 1.7 %) globally, and atmospheric CO2 concentration increasing by 2.1 ppm reaching 425.6 ppm, 53 % above the pre-industrial level (around 278 ppm in 1750). Overall, the mean and trend in the components of the global carbon budget are consistently estimated over the period 1959-2024, with a near-zero overall budget imbalance, although discrepancies of up to around 1 GtC yr(-1) persist for the representation of annual to decadal variability in CO2 fluxes. Comparison of estimates from multiple approaches and observations shows: (1) a persistent large uncertainty in the estimate of land-use change emissions, (2) a low agreement between the different methods on the magnitude of the land CO2 flux in the northern extra-tropics, and (3) a discrepancy between the different methods on the mean ocean sink.

Abstract. Drylands cover ~41 % of the Earth’s land surface and contribute more than one third of the global net primary productivity. Several studies have demonstrated that drylands play a crucial role in global carbon cycle interannual variability. However, drylands are vulnerable to the impacts of climate change. To predict changes in dryland productivity under climate change we depend on dynamic global vegetation models (DGVMs). Compared to more mesic ecosystems, DGVM carbon cycle dynamics have not been widely evaluated against data. Existing studies are mostly focused at site scale; rarely have these models been assessed or benchmarked against dryland carbon flux products at regional to global scales. Global gross primary productivity (GPP) products have poor performance in dryland regions. Only recently upscaled in situ flux products have been developed specifically for drylands. Here, we evaluated GPP inter-annual variability (IAV) simulated by 15 DGVMs from the TRENDY v11 model intercomparison project against theDryFlux GPP, which is newly developed upscaled GPP product that considers dryland- specific ecohydrological responses. Comparing model simulated GPP IAV to DryFlux, we identified two groups of models: a one group of models with generally lower GPP IAV than DryFlux (e.g., lower standard deviation in annual GPP than DryFlux and slope values of the linear regression between each model and the DryFlux product that are less than 1.0) and a second group of models with generally higher GPP IAV than DryFlux. We examined if including a representation of dynamic vegetation (i.e., changes in the spatial distribution of plant functional type (PFT) fractional cover) or fire in the models can explain the inter-model spread and model performance in comparison to DryFlux. Models that do not include a representation of fire and/or dynamic changes in plant functional type distribution over time generally have lower annual GPP variability compared to DryFlux (1st group of models), except for the eastern and southeastern region of the study area with high rainfall variability. We also found that models with dynamic vegetation exhibit high variability in grass fractional cover (that was higher than two independent reference fractional cover datasets), which was strongly correlated with high GPP IAV. Only some models that included fire simulated burnt area annual variability that correlated well with GPP IAV. Other models that included fire simulated low burnt area variability and therefore we did not find any strong relation between burnt area and GPP IAV. Finally, we examined the relationship between the dominant PFT and GPP IAV. We did not find a strong correlation between the spatial mean of the slope of the linear regression between each model and DryFlux annual GPP and their spatial mean woody, grass, or C3 grass fractional cover (although many models with generally low GPP IAV had higher woody plant cover). However, we did find a high correlation between the slope of the linear regression between each model and DryFlux annual GPP and spatial mean C4 grass cover. Therefore, our findings suggest that DGVMs inability to accurately represent the spatial distribution of herbaceous (specifically C4 grass) cover as well as processes controlling dynamically changing vegetation distributions over time (including fire) contribute to poor model performance in capturing annual variability in dryland productivity. Our findings can provide a roadmap for DGVM teams seeking to improve vegetation representation in sparsely vegetated dynamic dryland ecosystems.

The assessment of forest-based climate change mitigation strategies relies on computationally intensive scenario analyses, particularly when dynamic vegetation models are coupled with socioeconomic models in multi-model frameworks. In this study, we developed surrogate models for the LPJ-GUESS dynamic global vegetation model to accelerate the prediction of carbon stocks and fluxes, enabling quicker scenario optimization within a multi-model coupling framework. We trained two machine learning methods: random forest and neural network. We assessed and compared the emulators using performance metrics and Shapley-based explanations. Our emulation approach accurately captured global and biome-specific forest carbon dynamics, closely replicating the outputs of LPJ-GUESS for both historical (1850–2014) and future (2015–2100) periods under various climate scenarios. Among the two trained emulators, the neural network extrapolated better at the end of the century for carbon stocks and fluxes and provided more physically consistent predictions, as verified by Shapley values. Overall, the emulators reduced the simulation execution time by 95 %, bridging the gap between complex process-based models and the need for scalable and fast simulations. This offers a valuable tool for scenario analysis in the context of climate change mitigation, forest management, and policy development.

Forests play a crucial role in Europe's strategy for achieving carbon neutrality. Carbon turnover time - the time that carbon spends in the ecosystem - is a fundamental component in determining forest potential to mitigate climate change. However, there is a significant knowledge gap regarding how current and future forest management practices will affect carbon turnover time. This study aims to compare the effects of various forest management strategies on carbon turnover time in European forests. To achieve this, we used the dynamic global vegetation model LPJ-GUESS to simulate carbon pools and fluxes under stylised forest management scenarios mainly based on changing species composition. We calculated carbon turnover times under two conditions: first, with constant climate and CO2 concentration to assess the isolated impact of forest management; second, under a climate change scenario (SSP3-RCP7.0) to evaluate the combined effects of forest management and climate change. Our results indicate that unmanaged forests and the transition to broadleaved deciduous forests have a similar ecosystem carbon turnover time, which is the longest among all the management options across all the European climatic zones. Climate change decreases ecosystem carbon turnover time in any forest management, in a similar way, especially in cold climates. This study is the first step to include forest management when modelling carbon turnover time and indicates how the shift towards broadleaved forests, which is seen as an important climate-change adaptation strategy in many European regions, can also provide co-benefits for climate-change mitigation.

Animal herbivory can have large and diverse impacts on vegetation and hence on the state and function of ecosystems. Despite this, quantitative understanding of vegetation responses to consumption of green leaf tissue by herbivores is currently lacking. The large-scale impacts of changes in herbivore abundance on ecosystem function have yet to be investigated. Process-based modelling can help to quantify how animals affect important processes, such as ecosystem carbon cycling. To do so, we linked the dynamic global vegetation model LPJ-GUESS with Madingley, a model of multi-trophic functional diversity. This implementation allows us to simulate feedbacks between the availability of green vegetation biomass, herbivory and the whole trophic chain in response to monthly consumption of leaf biomass. In the coupled model system, we see an overall reduction in ecosystem productivity (NPP −5.2 %), leaf area index (−9.0 %) and carbon mass (−9.7 %), compared to the stand-alone version of LPJ-GUESS, with the highest impact on carbon mass in the boreal ecosystems (−42 %). We observe ecosystem composition to shift from boreal coniferous forests (without animals) to boreal mixed forests (with animals), as well as a general increase in herbaceous vegetation. Indirect effects like an increased light transfer facilitating growth of lower canopy layers are also captured by the model system. Overall, the results of this study underpin the important role of animals in ecosystem functioning and highlight the important contribution of process-based modelling towards a better understanding of complex food web interconnections.

Nitrogen (N) transformation processes by soil microbes account for significant nitrous oxide (N2O) emissions from natural ecosystems and cropland. However, understanding and quantifying global soil N2O emissions and their responses to changing environmental conditions remain challenging. Here, we implemented a soil nitrification–denitrification module into the dynamic vegetation model LPJ-GUESS to estimate N2O emissions from global lands. The performance of this new development is examined using observed N2O fluxes from natural-soil and cropland field trials and independent global-scale estimates. LPJ-GUESS broadly reproduces the cumulative N2O emissions under different climate conditions and N fertilizer applications that are observed in the field experiments, with some deviations in emission seasonality. Globally, simulated soil N2O emissions from terrestrial ecosystems increase from 5.6±0.2 Tg N yr−1 in the 1960s to 9.9±0.3 Tg N yr−1 in the 2010s, with croplands contributing about two-thirds of the total increase. East Asia and South Asia show the fastest growth rates in N2O emissions over the study period due to the expansion of fertilized croplands. On a global scale, N fertilization (including synthetic fertilizer and manure use), atmospheric N deposition, and climate change contribute 58 %, 46 %, and 24 %, respectively, to the simulated soil N2O emissions in the 2010s. Rising CO2 levels in the atmosphere reduce the simulated emissions by 32 % through increased plant N uptake, whereas land use changes have varied spatial effects on emissions depending on N management intensity after land cover conversion. Our estimates only account for the direct soil N2O emissions, excluding those from fertilized pastures. This study highlights the importance of environmental factors in influencing global soil N2O emissions, particularly for assessing greenhouse gas mitigation potential in agricultural ecosystems.

The terrestrial biosphere plays a major role in the global carbon cycle, and there is a recognized need for regularly updated estimates of land-atmosphere exchange at regional and global scales. An international ensemble of Dynamic Global Vegetation Models (DGVMs), known as the "Trends and drivers of the regional scale terrestrial sources and sinks of carbon dioxide" (TRENDY) project, quantifies land biophysical exchange processes and biogeochemistry cycles in support of the annual Global Carbon Budget assessments and the REgional Carbon Cycle Assessment and Processes, phase 2 project. DGVMs use a common protocol and set of driving data sets. A set of factorial simulations allows attribution of spatio-temporal changes in land surface processes to three primary global change drivers: changes in atmospheric CO2, climate change and variability, and Land Use and Land Cover Changes (LULCC). Here, we describe the TRENDY project, benchmark DGVM performance using remote-sensing and other observational data, and present results for the contemporary period. Simulation results show a large global carbon sink in natural vegetation over 2012-2021, attributed to the CO2 fertilization effect (3.8 +/- 0.8 PgC/yr) and climate (-0.58 +/- 0.54 PgC/yr). Forests and semi-arid ecosystems contribute approximately equally to the mean and trend in the natural land sink, and semi-arid ecosystems continue to dominate interannual variability. The natural sink is offset by net emissions from LULCC (-1.6 +/- 0.5 PgC/yr), with a net land sink of 1.7 +/- 0.6 PgC/yr. Despite the largest gross fluxes being in the tropics, the largest net land-atmosphere exchange is simulated in the extratropical regions. Around one third of human-induced CO2 emissions are absorbed by land ecosystems and thus act to mitigate climate change. It is essential to understand the processes, ecosystems and regions responsible for this natural carbon sink, to inform on the efficiency of the sinks into the future. These sinks are susceptible to year-to-year variation in response to climate variations and extremes. At the same time deforestation and other forms of land management are changing the land surface, which overall adds significantly to the human-induced CO2 emissions. There is a need to regularly update our estimate of land carbon dynamics to aid global stock takes for the Paris agreement to avoid dangerous climate change. Here we present an international initiative that on an annual basis assesses "Trends and drivers of the regional scale terrestrial sources and sinks of carbon dioxide" (TRENDY) using computer models of the land carbon cycle. We quantify the land sink during the contemporary period (2012-2021), and attribute to processes, mainly the large opposing effects of CO2 fertilization enhancing plant productivity and land-use change. Forests and semi-arid ecosystems are largely responsible for the mean and trend in the land sink, with the latter most important for its year-to-year variation. We quantify and attribute land carbon dynamics to underlying processes at regional scales, contributing bottom-up estimates to RECCAP-2 Models simulate a contemporary net land sink of 1.7 +/- 0.6 PgC/yr, with large opposing effects of CO2 fertilization and land-use change Despite the largest gross fluxes being in the tropics, the largest net land-atmosphere exchange is simulated in the extratropical regions

Accurate representation of the turbulent exchange of carbon, water, and heat between the land surface and the atmosphere is critical for modelling global energy, water, and carbon cycles in both future climate projections and weather forecasts. Evaluation of models' ability to do this is performed in a wide range of simulation environments, often without explicit consideration of the degree of observational constraint or uncertainty and typically without quantification of benchmark performance expectations. We describe a Model Intercomparison Project (MIP) that attempts to resolve these shortcomings, comparing the surface turbulent heat flux predictions of around 20 different land models provided with in situ meteorological forcing evaluated with measured surface fluxes using quality-controlled data from 170 eddy-covariance-based flux tower sites. Predictions from seven out-of-sample empirical models are used to quantify the information available to land models in their forcing data and so the potential for land model performance improvement. Sites with unusual behaviour, complicated processes, poor data quality, or uncommon flux magnitude are more difficult to predict for both mechanistic and empirical models, providing a means of fairer assessment of land model performance. When examining observational uncertainty, model performance does not appear to improve in low-turbulence periods or with energy-balance-corrected flux tower data, and indeed some results raise questions about whether the energy balance correction process itself is appropriate. In all cases the results are broadly consistent, with simple out-of-sample empirical models, including linear regression, comfortably outperforming mechanistic land models. In all but two cases, latent heat flux and net ecosystem exchange of CO2 are better predicted by land models than sensible heat flux, despite it seeming to have fewer physical controlling processes. Land models that are implemented in Earth system models also appear to perform notably better than stand-alone ecosystem (including demographic) models, at least in terms of the fluxes examined here. The approach we outline enables isolation of the locations and conditions under which model developers can know that a land model can improve, allowing information pathways and discrete parameterisations in models to be identified and targeted for future model development.

Increasing crop productivity while keeping detrimental side-effects on the environment low is a major challenge for global agriculture. Cover crops (CCs), mostly grown during the fallow period and incorporated in soils, are expected to improve soil fertility and crop yields while reducing chemical fertilizer use, with climate change mitigation co-benefits. However, quantifying these ecosystem services across global agricultural lands remains uncertain. In this study we investigate how the use of herbaceous CCs with and without biological nitrogen (N) fixation affects yields and cropland carbon and nitrogen balances using the dynamic global vegetation model LPJ-GUESS. Model performance is evaluated against observations from field trials worldwide as well as other published model-based estimates. LPJ-GUESS generally captures the observed enhanced soil carbon, reduced N leaching, and yield changes caused by CCs. We found that the combination of N-fixing CCs with no-tillage management could potentially increase soil carbon storage by 7% (+0.32 Pg C yr-1 in global croplands) while reducing N leaching by 41% (-7.3 Tg N yr-1) compared with bare fallows after 36 years of simulation. This integrated practice is accompanied by a 2% increase in total crop production (+37 million tonnes yr-1 including wheat, maize, rice, and soybean) in the last decade of the simulation. Legume cover cropping is found to contribute more to increasing the subsequent crop yields than non-legumes. The effects of CCs on crop productivity are highly dependent on the main food crop types, chemical fertilizer use, and management duration, with smallest yield changes found in soybean systems and highly fertilized agricultural soils. Our results demonstrate the possibility of conservation agriculture when targeting long-term environmental sustainability without compromising crop production in global croplands.

Accurate assessment of anthropogenic carbon dioxide (CO2) emissions and their redistribution among the atmosphere, ocean, and terrestrial biosphere in a changing climate is critical to better understand the global carbon cycle, support the development of climate policies, and project future climate change. Here we describe and synthesize data sets and methodology to quantify the five major components of the global carbon budget and their uncertainties. Fossil CO2 emissions (EFOS) are based on energy statistics and cement production data, while emissions from land-use change (ELUC), mainly deforestation, are based on land-use and land-use change data and bookkeeping models. Atmospheric CO2 concentration is measured directly, and its growth rate (GATM) is computed from the annual changes in concentration. The ocean CO2 sink (SOCEAN) is estimated with global ocean biogeochemistry models and observation-based fCO2 products. The terrestrial CO2 sink (SLAND) is estimated with dynamic global vegetation models. Additional lines of evidence on land and ocean sinks are provided by atmospheric inversions, atmospheric oxygen measurements, and Earth system models. The resulting carbon budget imbalance (BIM), the difference between the estimated total emissions and the estimated changes in the atmosphere, ocean, and terrestrial biosphere, is a measure of imperfect data and incomplete understanding of the contemporary carbon cycle. All uncertainties are reported as ±1σ. For the year 2022, EFOS increased by 0.9 % relative to 2021, with fossil emissions at 9.9±0.5 Gt C yr−1 (10.2±0.5 Gt C yr−1 when the cement carbonation sink is not included), and ELUC was 1.2±0.7 Gt C yr−1, for a total anthropogenic CO2 emission (including the cement carbonation sink) of 11.1±0.8 Gt C yr−1 (40.7±3.2 Gt CO2 yr−1). Also, for 2022, GATM was 4.6±0.2 Gt C yr−1 (2.18±0.1 ppm yr−1; ppm denotes parts per million), SOCEAN was 2.8±0.4 Gt C yr−1, and SLAND was 3.8±0.8 Gt C yr−1, with a BIM of −0.1 Gt C yr−1 (i.e. total estimated sources marginally too low or sinks marginally too high). The global atmospheric CO2 concentration averaged over 2022 reached 417.1±0.1 ppm. Preliminary data for 2023 suggest an increase in EFOS relative to 2022 of +1.1 % (0.0 % to 2.1 %) globally and atmospheric CO2 concentration reaching 419.3 ppm, 51 % above the pre-industrial level (around 278 ppm in 1750). Overall, the mean of and trend in the components of the global carbon budget are consistently estimated over the period 1959–2022, with a near-zero overall budget imbalance, although discrepancies of up to around 1 Gt C yr−1 persist for the representation of annual to semi-decadal variability in CO2 fluxes. Comparison of estimates from multiple approaches and observations shows the following: (1) a persistent large uncertainty in the estimate of land-use changes emissions, (2) a low agreement between the different methods on the magnitude of the land CO2 flux in the northern extra-tropics, and (3) a discrepancy between the different methods on the strength of the ocean sink over the last decade. This living-data update documents changes in methods and data sets applied to this most recent global carbon budget as well as evolving community understanding of the global carbon cycle. The data presented in this work are available at https://doi.org/10.18160/GCP-2023 (Friedlingstein et al., 2023).

Cover crops (CCs) can improve soil nutrient retention and crop production while providing climate change mitigation co‐benefits. However, quantifying these ecosystem services across global agricultural lands remains inadequate. Here, we assess how the use of herbaceous CCs with and without biological nitrogen (N) fixation affects agricultural soil carbon stocks, N leaching, and crop yields, using the dynamic global vegetation model LPJ‐GUESS. The model performance is evaluated with observations from worldwide field trials and modeled output further compared against previously published large‐scale estimates. LPJ‐GUESS broadly captures the enhanced soil carbon, reduced N leaching, and yield changes that are observed in the field. Globally, we found that combining N‐fixing CCs with no‐tillage technique could potentially increase soil carbon levels by 7% (+0.32 Pg C yr −1 in global croplands) while reducing N leaching loss by 41% (−7.3 Tg N yr −1 ) compared with fallow controls after 36 years of simulation since 2015. This integrated practice is accompanied by a 2% of increase in total crop production (+37 million tonnes yr −1 including wheat, maize, rice, and soybean) in the last decade of the simulation. The identified effects of CCs on crop productivity vary widely among main crop types and N fertilizer applications, with small yield changes found in soybean systems and highly fertilized agricultural soils. Our results demonstrate the possibility of conservation agriculture when targeting long‐term environmental sustainability without compromising crop production in global croplands.

Climate change, growing populations and economic shocks are adding pressure on the global agricultural system’s ability to feed the world. In addition to curbing the emissions from fossil fuel use, land-based actions are seen as essential in the effort to mitigate climate change, but these tend to reduce areas available for food production, thereby further increasing this pressure. The actors of the food system have the capacity to respond and adapt to changes in climate, and thereby reduce the negative consequences, while potentially creating additional challenges, including further greenhouse gas emissions. The food system actors may respond autonomously based on economic drivers and other factors to adapt to climate change, whereas policy measures are usually needed for mitigation actions to be implemented. Much research and policy focus has been given to land-based climate change mitigation, but far less emphasis has to date been given to the understanding of adaptation, or the interaction between adaptation and mitigation in the land use and food system. Here, we present an approach to better understand and plan these interactions through modelling. Climate change adaptation and mitigation strategies and the impacts on the global food system and socio-economic development can be simulated over long-term predictions, thanks to the new combination of multiple models into the Land System Modular Model (LandSyMM). LandSyMM takes into account the impacts in changes in climate (i.e. temperature, precipitation, atmospheric greenhouse gas concentrations) and land management on crop yields with its implications for land allocation, food security and trade. This new coupled model integrates, over fine spatial scale, the interactions between commodities consumption, land use management, vegetation and climate into a worldwide dynamic economic system. This study offers an outline description of the LandSyMM as well as the perspectives of uses for climate adaptation assessment.

Drylands cover ca. 40% of the land surface and are hypothesised to play a major role in the global carbon cycle, controlling both long-term trends and interannual variation. These insights originate from land surface models (LSMs) that have not been extensively calibrated and evaluated for water-limited ecosystems. We need to learn more about dryland carbon dynamics, particularly as the transitory response and rapid turnover rates of semi-arid systems may limit their function as a carbon sink over multi-decadal scales. We quantified aboveground biomass carbon (AGC; inferred from SMOS L-band vegetation optical depth) and gross primary productivity (GPP; from PML-v2 inferred from MODIS observations) and tested their spatial and temporal correspondence with estimates from the TRENDY ensemble of LSMs. We found strong correspondence in GPP between LSMs and PML-v2 both in spatial patterns (Pearson's r = 0.9 for TRENDY-mean) and in inter-annual variability, but not in trends. Conversely, for AGC we found lesser correspondence in space (Pearson's r = 0.75 for TRENDY-mean, strong biases for individual models) and in the magnitude of inter-annual variability compared to satellite retrievals. These disagreements likely arise from limited representation of ecosystem responses to plant water availability, fire, and photodegradation that drive dryland carbon dynamics. We assessed inter-model agreement and drivers of long-term change in carbon stocks over centennial timescales. This analysis suggested that the simulated trend of increasing carbon stocks in drylands is in soils and primarily driven by increased productivity due to CO2 enrichment. However, there is limited empirical evidence of this 50-year sink in dryland soils. Our findings highlight important uncertainties in simulations of dryland ecosystems by current LSMs, suggesting a need for continued model refinements and for greater caution when interpreting LSM estimates with regards to current and future carbon dynamics in drylands and by extension the global carbon cycle.

Land biosphere processes are of central importance to the climate system. Specifically, ecosystems interact with the atmosphere through a variety of feedback loops that modulate energy, water, and CO2 fluxes between the land surface and the atmosphere across a wide range of temporal and spatial scales. Human land use and land cover modification add a further level of complexity to land–atmosphere interactions. Dynamic global vegetation models (DGVMs) attempt to capture land ecosystem processes and are increasingly incorporated into Earth system models (ESMs), which makes it possible to study the coupled dynamics of the land biosphere and the climate. In this work we describe a number of modifications to the LPJ-GUESS DGVM, aimed at enabling direct integration into an ESM. These include energy balance closure, the introduction of a sub-daily time step, a new radiative transfer scheme, and improved soil physics. The implemented modifications allow the model (LPJ-GUESS/LSM) to simulate the diurnal exchange of energy, water, and CO2 between the land ecosystem and the atmosphere and thus provide surface boundary conditions to an atmospheric model over land. A site-based evaluation against FLUXNET2015 data shows reasonable agreement between observed and modelled sensible and latent heat fluxes. Differences in predicted ecosystem function between standard LPJ-GUESS and LPJ-GUESS/LSM vary across land cover types. We find that the emerging ecosystem composition and carbon fluxes are sensitive to both the choice of stomatal conductance model and the response of plant water uptake to soil moisture. The new implementation described in this work lays the foundation for using the well-established LPJ-GUESS DGVM as an alternative land surface model (LSM) in coupled land–biosphere–climate studies, where an accurate representation of ecosystem processes is essential.