Impacts of nationally determined contributions on 2030 global greenhouse gas emissions: uncertainty analysis and distribution of emissions

We first consider a dataset of emissions for reference years used in the NDCs. For consistency with work from the Intergovernmental Panel on Climate Change (IPCC), we rely on published emission inventories [16, table S2]. Historical emissions data come from three distinct databases: the Emissions Database for Global Atmospheric Research (EDGAR) for emissions of CO₂ except from LULUCF, CH₄, N₂O, HFC, PFC, and SF₆ [17], and the BLUE (bookkeeping of land use emissions) model [18 and personal communication] for CO₂ emissions from LULUCF. For the United States, Canada and Russia, whose NDCs are expressed in 'net-net emissions', we also consider their CO₂ LULUCF balance (emissions + sinks) reported to the UNFCCC. Since the UNFCCC accounting for LULUCF balance differs from that used in Earth system models [19], we only consider half of these supplemental carbon sinks as anthropogenic (table S4). Uncertainties in historical emissions, including in Chinese emissions [20, 21], are not accounted for. Rather, in the light of the recent study by Liu et al considering specific emission factors for Chinese coal types [22], we use average values between Chinese emissions data from EDGAR and those of Liu et al This results in Chinese CO₂ emissions about 5% smaller than EDGAR over the 1990–2010 period (table S3). We discuss the impact of this correction on our results in the supplementary material (section 10 and figure S6). Combining these datasets, we estimate 2010 GHG emissions to be 50.6 Gt CO₂eq yr⁻¹ accounting for an anthropogenic carbon sink due to LULUCF. We are not able to derive a strict uncertainty range for this quantity as neither EDGAR nor BLUE provides an uncertainty analysis for their data. Our estimate is within the range of other estimates: 50.9 GtCO₂eq yr⁻¹ for EDGAR [23]; 47.6 Gt CO₂eq yr⁻¹ for the Climate Action Tracker [24] and 49.3 Gt CO₂eq yr⁻¹ for PRIMAP [25]. While reassuring, this does not mean much as most studies rely on identical or similar sources for emissions. Uncertainties in historical emissions at the country level are also difficult to assess and not considered in this study.

We consider 20 GDP scenarios, illustrating the five Shared Socioeconomic Pathways (SSP) [10] for four quantifications (the marker quantification by OECD [13], and three alternative quantifications by CEPII [12], IIASA [14], and PIK [15]). SSP scenarios aim at facilitating the integrated analysis of future climate impacts, vulnerabilities, adaptation, and mitigation [10] and represent a framework to examine possible future states of the world. Because the GDP scenarios diverge after 2005, we correct them on the basis of the historical 2005–2015 data from the World Bank World Development Indicators database (release from 15 September 2017). We propagate the difference between the SSP and actual GDP growth rates until 2030 with a transition from a purely historical GDP driven trajectory to a purely SSP driven trajectory (see supplementary material section 4). This generally results in smaller GDP levels in 2030 than assumed in the SSP scenarios because actual growth rates are lower in the historical data over 2005–2015 than assumed in the SSP scenarios. A noticeable exception is the CEPII scenarios, which show a late rise in GDP growth for some countries, including China (see supplementary material section 5). Figures S1(a)–(b) illustrate the 20 economic GDP scenarios before and after correction, respectively, for China. Note that we only use the GDP quantifications of the SSP, not the emissions quantifications (currently not available at the country level). However, we compare our projected emissions with those projected by IAMs for the SSP scenarios for their five usual meta-regions.

Countries used different formats to express their mitigation effort in their NDC [26]. A first set of countries put forward an absolute target, whereby GHG emissions levels for the target year are provided or can be readily inferred from the data provided (reduction with respect to a reference year or a business as usual (BAU) scenario which is explicited in the NDC, emission value in target year). A second set of countries express their target relative to a BAU scenario that is not explicited in the NDC or in terms of a reduction in the carbon intensity of GDP without providing a forecast of the GDP. A third set of countries only provide sectoral indications that do not per se allow to infer an emission reduction target (table S1). Furthermore, China has indicated a CO₂ emission peaking date target before 2030. China and India also provide additionnal targets in terms of share of non-fossil fuels, of the total primary energy supply for China and of the power generation capacity for India.

We use the latest updated versions of NDCs, as of November 2017. We treat one by one all 103 Parties (among which the European Union represents 28 countries) for which the NDC contains enough information to compute GHG emissions levels for the target year. We lump the remaining 64 countries in several groups for which additional assumptions are required. We present results for the 23 largest emitting countries (representing 80% of 2010 global GHG emissions), but the results for all countries are presented in a file and code provided as supplementary material. The countries other than the 23 largest emitters are separated in four groups (table S1): Other Annex 1, Other Emerging, Other Oil exporting countries and Rest of World (representing 0.7, 1.7, 0.4 and 14.7% of 2010 global GHG emissions, respectively). A detailed description of our treatment of NDCs is provided in the supplementary material (section 1).

Two of the main elements of the Chinese NDC (i.e. a reduction in the carbon intensity and a peak emission before 2030) refer specifically to CO₂ emissions other than from LULUCF. We take this into account by first estimating CO₂ emissions for 2030 (excluding LULUCF), which are then converted to CO₂eq using a CO₂eq to CO₂ emissions ratio (R). Our derivation of this ratio is described in the supplementary material (section 7).

The Chinese NDC also includes a target to peak CO₂ emissions before 2030. This requires the reduction rate of the carbon intensity in 2030 to exceed the growth rate in the same year. In order to account for this in our analysis, we first interpolate the GDP growth rate in 2030. We then convert the overall carbon intensity reduction target into a time profile and, if necessary, iteratively increase it until the annual rate of carbon intensity reduction in 2030 exceeds the GDP growth rate in the same year. We further describe and justify our approach in the supplementary material (section 6), including the reasons and limitations of not considering the non-fossil fuel targets for China and India (section 8). Estimated emissions for China are compared to current trends on figure S3.

Our treatment of LULUCF emissions is indicated in table S4 and described in the supplementary material (section 3). We also account for the LULUCF elements of NDCs when provided.

The first part of our study focuses on an uncertainty analysis of the 2030 global GHG emission level. To determine the uncertainty range surrounding the 2030 emission level, a Monte Carlo method (N = 50 000 runs) is used for each of the 20 GDP scenarios. Uncertainty for each GDP scenario is generated by varying the following parameters: ambition level of the NDC (including the ranges mentioned in some NDCs); our set of assumption for countries with NDCs providing either only sectoral or incomplete information; emissions level for international aviation and shipping; LULUCF emissions reduction for India and the Rest of World; the choice of the intensity reduction profile (parameter λ) for China, and the CO₂eq/CO₂ ratio for China (parameter μ). Note that the above-mentioned ambition levels are assumed to be the same for all countries. In other words, they are perfectly correlated in our Monte Carlo simulation. This follows the idea of a virtuous circle between countries as part of the Paris Agreement. Whenever there is a range, whether it is indicated in the NDC, refers to unconditional and conditional statements in the NDC or assumed from our expert judgement, we consider variations within the range and refer in a generic way to the 'ambition level' of the NDCs. Note that when only a conditional target is provided, we assume the unconditional target to be no effort compared to BAU emissions, thus generating a range. Unless stated otherwise we assume a uniform probability distribution within the range. This implies in particular that on average half of the additional effort of going from low to high ambition level would be carried out. Finally, a 1 000 000 member ensemble is made by aggregating the 20 ensembles corresponding to the 20 GDP scenarios considered with equal weight. The uncertainty attributable to a particular driver is estimated as the ratio in the variance of the 2030 emissions caused by that driver to the total variance. Covariances between the different drivers are gathered in a residue termed 'Interactions of all drivers'. In the second part of this study, where we analyze the evolution in emissions distribution among countries, we do not consider sources of uncertainty other than that of the GDP scenarios, but use average values for all relevant uncertain parameters. Unless otherwise stated, all our confidence intervals represent 5%−95% uncertainty ranges.

The GHG emissions levels in 2030 are available at the country level in the supplementary material. We find that NDCs imply 2030 global GHG emissions of 61.7 (56.8–66.5) Gt CO₂eq yr⁻¹ (figure 1(a)). This is higher than in previous studies (see table 1 for a comparison) with the lower bound of our uncertainty range corresponding to the upper bound of most published estimates. Our uncertainty range is also larger than most of those published so far, which is essentially because we consider a range of GDP scenarios rather than a single scenario. Figure S5 displays the probability distributions of 2030 global emissions grouped by SSP narrative and by GDP data source. For the SSP2 OECD scenario, which is often used as a marker scenario, we find 2030 GHG emissions to be 61.8 (58.4–65.0) GtCO₂eq yr⁻¹. The uncertainty range is then comparable to previous estimates, but it is shifted to higher values, meaning that our mean estimate is still higher than previous estimates. It is unfortunately not possible to gain a full understanding as to where the differences in the median value come from because no single study publishes their emission estimates for all countries.

Zoom In Zoom Out Reset image size

Nevertheless one likely reason for this difference in emissions is that, unlike previous studies, we do not consider the impact of current policies which are not (yet) reflected into a high-level NDC target. In particular, a large part of the difference appears to come from the Chinese CO₂ emissions. We find a 5%–95% uncertainty range of 11.8–15.7 Gt CO₂ yr⁻¹ (without LULUCF) in 2030, which accounts for uncertainties in the GDP scenario and the profile of the carbon intensity reduction (see figures S2 and S7). For the SSP2 OECD scenario, the range in Chinese CO₂ emissions is 13.3 to 15.0 Gt CO₂ yr⁻¹ in 2030. The peak target does or does not constrain the 2030 Chinese emissions depending on the choice of parameters in our Monte Carlo. As expected the peak target is constraining more often when the decrease in the carbon intensity is assumed to be more exponential and less linear. Overall, accounting for the peak target reduces the lower bound of the 2030 Chinese CO₂ emissions by ~0.5 Gt CO₂ and the upper bound by ~1.2 Gt CO₂, while the average is reduced by ~0.6 Gt CO₂. For all scenarios and GHG, we find 2030 emissions of 16.4 (14.5–18.2) Gt CO₂eq yr⁻¹. One reason for our larger emissions for China is the occurrence of high, long-lasting GDP growth in many of the scenarios considered (see figure S1(c) for 2030 growth rate descriptions and figure S2 for their impact on the peak constraint). For comparison, analyses based on Chinese five-year plans up to 2030 assume GDP growth decreasing from 5.5% yr⁻¹ for 2021–2025 to 4% yr⁻¹ for 2025–2030 and 3.5% yr⁻¹ for 2030–2035, which corresponds to the lower bound of the SSPs scenarios, and find a peak of Chinese emissions between 2025 and 2030 at around 12.7 GtCO₂ yr⁻¹ [27, 28]. The other reason for our larger emissions is that in a number of previous estimates, the carbon intensity reduction target and/or the peaking before 2030 target are overachieved (see section 8 of the supplementary material). As Chinese emissions appear to have increased again over the past year following a stagnation for a couple of years [29], it will be important to monitor future trends to confirm whether the Chinese NDC will indeed be overachieved.

An equal part of the difference between our estimate and previous estimates comes from Indian emissions. The GHG emissions for India are projected to be 7.6 (5.8–9.1) Gt CO₂eq yr⁻¹ in 2030. It should be noted that the GDP growth projected by India in its NDC is in the higher end of the range of GDP growth of the SSP scenarios. Therefore, here the main reason for our larger estimates comes from the fact the Indian carbon intensity target is overachieved in previous studies.

The differences about Chinese and Indian estimates represent together about 70% of the difference in global 2030 emissions. The rest of the difference between our estimate and previous estimates may be due to the treatment of other countries, the level of disaggregation for small countries, the choice of global warming potentials to compute carbon dioxide equivalent emissions, the treatment of emissions related to land use, and the treatment of international aviation and maritime shipping.

Our results in terms of 2030 emissions can also be compared to the IAM quantifications of emissions trajectories for the combinations of the five SSP socioeconomic pathways and various radiative forcing pathways. Section 12 from the supplementary material presents such a comparison at the level of five world regions (see figure S9). We find that our quantification of emissions levels in 2030 corresponding to NDC for Asia is at the higher end of IAM quantifications, whereas they are at the lower end for OECD countries and in the middle of the range for other regions.

The uncertainty range for 2030 global emissions results from multiple drivers (figure 1(b)). With 44% of the total variance explained, the NDC ambition level is the most important source of uncertainty. Our set of assumptions to interpret NDCs further explains 5% of the total variance, Future economic growth is also a sizeable source of uncertainty, across SSP scenarios (7%) and especially across GDP data sources within each SSP scenario (20%) (figure S7). The relative importance of uncertainty related to economic growth is not surprising given the fact that we consider a large range of GDP values for 2030, while a number of key emerging countries have expressed their NDCs with carbon-intensity targets. However the fact that the different GDP interpretations of the SSP are leading to larger differences in global emissions than the different SSP is more unexpected, highlighting the heterogeneity of GDP projections even when following similar narratives. The choice of carbon intensity reduction profile for China explains 7.6% of the total variance. Other identified sources of uncertainty, related to LULUCF or international aviation and shipping, are small. This does not mean the corresponding emissions are not important though, since the share of international aviation and shipping in global emissions increase from 2.3% in 2010 to 3.0 to 3.7% in 2030. The total uncertainty due to LULUCF is more important than shown in the figure, as our estimate only bears on LULUCF emissions from the few countries that have provided separate targets for LULUCF in their NDC; for numerous countries this uncertainty is contained in the NDCs themselves [30]. The term 'interaction of all drivers' arises from covariances between all drivers. In terms of countries, uncertainties on 2030 emissions are largely driven by uncertainties on Chinese, Indian and Rest of World countries' emissions (figure S7).

Notwithstanding the uncertainty surrounding 2030 global emissions, NDCs announce a significant change in the distribution of emissions among countries. This evolution reflects the combination of demographic and economic dynamics with mitigation efforts across the world. As shown in figure 2(a), emissions from the United States, the European Union and LEA countries are all lower in 2030 than in 2010, whereas emissions from China, India, countries from the 'World Other' group, international transport and to a lesser extent LENA countries are larger in 2030 than in 2010, whichever GDP scenario is considered.

Zoom In Zoom Out Reset image size

Emerging and developing countries will represent in 2030 a larger share of global emissions than in 2010 (figure 2(b)). In particular, the combined shares of China and India could reach up to 37%–41% in 2030, against 28% in 2010. The share of the United States, the European Union and LEA countries would decrease, from 38% in 2010, to a range from 22%−25% in 2030, depending on the GDP scenario considered.

We also analyse international inequalities in per capita GHG emissions. We find a noticeable reduction in such inequalities (see supplementary material section 13, figures S10–12).

Assessing whether current NDCs put countries on track to the long-term goal of 'keeping warming well below 2 °C and do best efforts to maintain it under 1.5 °C' [1] is not a straightforward task. Indeed, it implies building scenarios until the end of the 21st century [31] and judging on how fast emissions can be reduced after 2030. A less questionable way of assessing aggregated NDCs is to compare the implied 2030 global emissions to the worldwide GHG emissions reduction target of 40%−70% in 2050 compared to 2010 reported in the IPCC Fifth Assessment Report for scenarios aiming at limiting warming to 2 °C with a >66% probability [32, 33]. These scenarios correspond to a range of permissible emissions of 15 to 30 Gt CO₂eq yr⁻¹ in 2050. As shown in figure 3, our interpretation of NDCs results in an average global emissions growth of 1.0% ± 0.7% per year between 2010 and 2030. However, this hides large regional disparities, with e.g. an average emission reduction of 5.1% per year in Brazil, and 5.1% per year emission increase in India. Connecting the NDCs' 2030 global emission level to the 2 °C compatible range in 2050 requires an average annual worldwide reduction rate in emissions of 4.9% ± 2.3%. Note that uncertainty for 2030 emissions is mainly due to economic GDP scenarios and NDCs interpretation (figure 1(b)), whereas the 2050 range results from uncertainty on achievable emissions trajectories and technology developments during the second half of the 21st century.

Zoom In Zoom Out Reset image size