Aligning climate scenarios to emissions inventories shifts global benchmarks

Abstract

Taking stock of global progress towards achieving the Paris Agreement requires consistently measuring aggregate national actions and pledges against modelled mitigation pathways¹. However, national greenhouse gas inventories (NGHGIs) and scientific assessments of anthropogenic emissions follow different accounting conventions for land-based carbon fluxes resulting in a large difference in the present emission estimates^2,3, a gap that will evolve over time. Using state-of-the-art methodologies⁴ and a land carbon-cycle emulator⁵, we align the Intergovernmental Panel on Climate Change (IPCC)-assessed mitigation pathways with the NGHGIs to make a comparison. We find that the key global mitigation benchmarks become harder to achieve when calculated using the NGHGI conventions, requiring both earlier net-zero CO₂ timing and lower cumulative emissions. Furthermore, weakening natural carbon removal processes such as carbon fertilization can mask anthropogenic land-based removal efforts, with the result that land-based carbon fluxes in NGHGIs may ultimately become sources of emissions by 2100. Our results are important for the Global Stocktake⁶, suggesting that nations will need to increase the collective ambition of their climate targets to remain consistent with the global temperature goals.

Fig. 1. Difference in present estimates of LULUCF carbon fluxes under NGHGI and model-based accounting conventions.

Schematic showing the difference in accounting conventions between NGHGIs (green) and scientific models (bookkeeping models in red and vegetation models in blue). Models such as IAMs are based on bookkeeping approaches and consider direct fluxes due to land use (for example, wood harvest) and land-cover changes. Additional indirect fluxes due to evolving environmental conditions can be estimated by processed-based vegetation models. NGHGIs consider a wider managed land area and are generally based on physical observations, and thus include both direct and indirect fluxes. We use the term ‘unmanaged’ to describe land not considered managed by NGHGIs to be consistent with previous literature, but recognize that this includes land that has been managed by indigenous and traditional communities for centuries to millienia^,. In this study, we estimate the alignment factor to translate between both conventions (the indirect flux considered in NGHGIs but not in models, blue). This factor will change over time based on future land-use decisions and overall mitigation efforts because of, for example, changing atmospheric CO₂ levels.

Fig. 2. Land-use emissions in re-analysed IPCC pathways with model-based and NGHGI-based accounting conventions.

a,b, Land-use emissions pathways before and after alignment to match NGHGIs for 1.5 °C (a) and 2,0 °C (b) pathways. Historical estimates^, are shown with carbon-cycle uncertainty (1σ), and the median of scenario pathways are shown with the scenario interquartile range in shaded plumes. Pathways consistent with model-based convention are shown in red, whereas the NGHGI convention is shown in green. c, Comparing the two conventions results in a difference between re-analysed and NGHGI-adjusted pathways—that is, an alignment factor, which evolves as a function of the strength of land-based climate mitigation.

Fig. 3. Changes in global mitigation benchmarks across assessed scenarios.

a–c, Scenario-wise distributions of the estimated change in the net-zero CO₂ year (a), 2020–2030 CO₂ emission reductions (b) and cumulative emissions until net-zero CO₂ (c) between the re-analysed model-based and the NGHGI LULUCF accounting conventions are shown for 1.5 °C (blue, IPCC category C1), 1.5 °C-OS (green, IPCC category C2) and 2.0 °C (purple, IPCC category C3) scenarios. A positive value indicates that the benchmark comes later (for net-zero years) or is higher (for cumulative emissions) in the model-based framework compared with the NGHGI-based framework, whereas a negative value indicates that the benchmark is higher in the NGHGI-based framework (for emission reductions). Across all benchmarks, NGHGI-based accounting tends to result in more stringent outcomes (earlier net-zero years, higher emission reductions and lower cumulative emissions to net-zero CO₂ emission). A comparison with the original AR6 benchmarks is shown in Extended Data Fig. 1. a.u., arbitrary units.

Fig. 4. The future role of indirect fluxes in national climate targets.

In a future with strong mitigation action in line with the goals of the Paris Agreement (bottom row), stabilizing or even decreasing atmospheric CO₂ will result in a weakening of the indirect sink (blue arrows), whereas a future with weak mitigation action will see the indirect sink increase (as long as CO₂ fertilization dominates over climate feedbacks, top row). The direct component of LULUCF fluxes (red arrows) is due entirely to land-use management decisions (columns). Future estimates of net LULUCF emissions (green arrows) will differ between conventions depending on how much overall mitigation occurs and how much land-based mitigation occurs, which can have unexpected consequences on national climate target achievement.

Fig. 5. CDR characteristics in mitigation and current-policy pathways.

a, Net land-use carbon removal levels from direct fluxes (green bars) are compared with non-land CDR (brown bars) and total levels (summing land-use and CDR, grey bars) with whiskers denoting the interquartile range of each estimate across 1.5 °C and 2.0 °C scenarios. Here, non-land CDR comprises technologies included in the IAM pathways assessed in AR6 other than those due solely to land-use change, such as bio-energy with carbon capture and storage, direct air capture of CO₂ with storage and enhanced mineral weathering. b, The share of land-based CDR reduces over time across both 1.5 °C and 2.0 °C pathways with the median (solid line) and interquartile range (shaded area) shown for the population of scenarios assessed. The direct component of land-based removal flux, which constitutes land-based CDR, and the indirect component of the removal flux evolve differently across pathways. c, In the near term, until 2030, the 1.5 °C pathways see a strong enhancement of additional removals (pink bar), whereas the 2.0 °C pathways see a similar addition of total removals as current-policy pathways. d, By mid-century, additional removals in current-policy pathways out-pace both the 1.5 °C and 2.0 °C pathways, because of the continued enhancement of indirect removals compared with an overall weakening of this flux in mitigation pathways. Scenario uncertainty in c,d is estimated by the interquartile range of scenario-based estimates, whereas the carbon-cycle uncertainty is estimated by the interquartile range of the median ensemble of climate runs (Methods).

Extended Data Fig. 1. Scenario-wise mitigation benchmark shift.

The change between estimates of mitigation benchmarks for 1.5 C (blue, IPCC category C1), 1.5C-OS (green, IPCC category C2), and 2 C (purple, IPCC category C3) scenarios. Original values from the AR6 database (which follows IAM reporting conventions) are shown as circles whereas values derived from reanalyzed scenarios in this study (in line with NGHGI reporting conventions) are shown as triangles. The estimates of the year of global net-zero CO₂ (panel a), emissions reductions between 2020 and 2030 (panel b), and cumulative CO₂ emissions (panel c) are shown. Each pair of dots and triangles represents results from a single scenario, with scenarios ordered along the y-axis based on the values in the original AR6 dataset.

Extended Data Fig. 2. NGHGI-adjusted global GHG pathways compared with NDCs and current policies.

The interquartile range shown and median highlighted is plotted together with current estimates of 2030 aggregated national climate target levels and current policy estimates from den Elzen et al. (2022).

Extended Data Fig. 3. The 2030 emissions gap between current policies and pledges.

1.5 C and 2 C as assessed in this study and by den Elzen (2022) is compared against levels of current policies, conditional NDCs, and unconditional NDCs as reported in den Elzen (2022). Median estimates of all values are used to compute the respective emission gaps.

Extended Data Fig. 4. The change in the estimated 2030 emission gap between due to alignment to NGHGI conventions.

The total magnitude, left, relative value, right. Each bar represents the median value with the interquartile range of the estimate across scenarios. These changes occur differently across different regions between pathways following model-based conventions and adjusted pathways following NGHGI-based conventions. A positive value means that the gap is larger when considering both (i.e. when aligned to NGHGIs), and a negative value means the gap is smaller. Regions labels conform to IPCC 5-region labels for Asia, Latin America, Middle East and Africa, the OECD and EU, and the Reformed Economies, respectively (Extended Data Table 4).

Extended Data Fig. 5. Gross carbon removal levels.

Gross carbon removal levels from LULUCF (reanalyzed with OSCAR) by direct effects (green) and indirect effects (purple) across 1.5 C and 2 C pathways. Interquartile ranges of each estimate are shown by error bars.

Extended Data Fig. 6. Cumulative carbon sequestered on land starting from 2020.

Gross cumulative carbon removal levels starting from 2020 from LULUCF (reanalyzed with OSCAR) by direct effects (green) and indirect effects (purple) across 1.5 C and 2 C pathways. Removals in both categories increase by midcentury, but at different levels. Both pathway categories see similar total cumulative removal levels by the end of the century with varying strength of indirect removals.

Extended Data Fig. 7. K-S test of scenario distributions.

Kolmogorov-Smirnov (K-S) test results for key mitigation indicators for the full set of C1-C3 scenarios, those scenarios having all land-cover variables defined at the R5 region level, and those not having all land-cover variables defined at the R5 region level. The null hypothesis of the K-S test is that two dataset values are from the same distribution. For all indicators derived from scenarios including land-cover variables data at the R5 region level, we can not reject the null hypothesis (p > 0.05). Some indicators of the scenario set without land-cover data (not used in this analysis) do reject the null hypothesis.

Extended Data Fig. 8. Mitigation metrics from scenario subsets.

Key mitigation metrics where scenarios without R5 region coverage (in red) cannot replicate the full database outcome. The blue bar presents the outcome for the full database, scenarios with global values of land-cover variables and R5 values are shown in yellow, and scenarios with land-cover variables at the R5 region are shown in green. The red bar shows how the distribution changes when considering the population of scenarios without full variable coverage (‘No R5 all’).