You might remember that in 2015, after many years of climate negotiations, most of the world’s countries came together at an international climate conference where the Paris Agreement (PA) was born as the successor of the Kyoto Protocol. The PA’s “goal is to limit global warming to well below 2, preferably to 1.5 degrees Celsius, compared to pre-industrial levels” (What is the PA, incl. video, see also Key aspects of the PA). To start with a bit of a bummer: we have already passed a warming level of 1°C (State of the climate: How the world warmed in 2019), and global anthropogenic emissions are still on the rise. And another one: the global pandemic Covid-19 produced a dip in emissions, but only temporarily (Covid-19 paused climate emissions – but they’re rising again).
Why are we talking about temperature and Greenhouse Gas (GHG) emissions as if they were linked? Well, as you most probably know, they are! Warming levels depend on the total atmospheric concentrations of GHGs, which accumulate over time as each year we fuel it with our anthropogenic emissions, and the GHGs stay in the atmosphere following their atmospheric lifetimes. We can translate the global emissions trajectory (sum over national pathways, so each country’s emissions count, in each year), into estimates of the resulting end-of-century warming level (there are models for that, e.g., MAGICC), and check how we can stay in the 1.5-2°C limit (on a certain time horizon, here up to the year 2100). So each country could be allowed a certain prescribed share in the remaining emissions. But defining these shares is complicated in respect to historical responsibilities, industrialisation and development level, different levels of population and Gross Domestic Product (GDP), etc.
Instead of trying to agree on emission sharing rules and trying to get all countries on board of a climate agreement, for the PA the approach was to set a temperature goal and let each country reflect on what seems feasible and fair as a national contribution to this global effort, in light of the countries’ common but differentiated responsibilities and respective capacities, write it down, and submit this document as Intended Nationally Determined Contribution (INDC) to the United Nations Framework Convention on Climate Change (UNFCCC). Already before the conference, many countries submitted their INDCs, which turned into NDCs upon countries’ ratification of the PA (Status of Ratification: 191 Parties out of 197 Parties to the Convention).
NDCs, their content and first assessment challenges
NDCs are the backbone of the agreement, in which Parties communicate envisaged Greenhouse Gas reduction targets (mitigation) or specific measures to reduce emissions, but also plans on how to respond to adverse climate change impacts (adaptation), or support needed (targets can be conditional on international support). To get everyone on board, countries were given a lot of freedom to describe their emission reduction targets. There are very different formats leaving more or less room for interpretation. Here are some examples starting from the clear ones and going to the ones that need more assumptions to allow for a translation into national emissions:
- Absolute target emissions (e.g., “500 megatonnes of CO2 equivalent in 2030”; ABS in Figure 1) is the most straight-forward target, as it directly indicates the emissions level planned in the chosen target year
- Reduction compared to a historical reference year (e.g. “30% reduction in 2030 compared to 1990”; RBY) is a very clear specification. As historical emissions are relatively well known, this target generally is easy to interpret.
- Reduction in relative or absolute terms, compared to a future year in a Business As Usual (BAU) scenario (e.g. “20% reduction in 2025 compared to a BAU scenario without any additional mitigation measures”; RBU/ABU) is already more difficult to interpret as we need an assumption about the BAU emissions accounting for potential future economic development and national population growth.
- Reduction in emission intensities (e.g. “15% reduction in emissions per capita or 40% reduction in emissions per $ GDP”; REI) need assumptions about population or economic growth.
Mitigation target types chosen by countries. For an explanation of the colour-coded target types, please refer to the end of this page. The legend reads as follows: e.g., “ABS (#8, 2.9%)” means that eight countries chose the target type ABS (absolute emissions target) and these countries represent 2.9% of global 2017’ emissions (excluding LULUCF and bunkers emissions, short information at the end of this page). Current state of this figure: only NDCs submitted by 31 December 2020 are analysed, and we are working on an update (nonetheless, the shown information is generally valid). Countries whose “main target type” is not ABS are not marked as ABS here. However, more than 90 countries with targets differing from ABS additionally provide the absolute target value in the NDC, or the numerical information necessary to do so is given (e.g., “we pledge to reduce our emissions by X% in 2030 compared to 1990 levels, which results in 2030 emissions of Y megatonnes CO2 equivalent”: the main target type is a relative reduction compared to base year emissions).
FIG 1 / Image source: A. Günther, Gütschow, and Jeffery (2021)
A mitigation target can be contingent upon, e.g., financial, technological, or capacity building support expected from developed countries, if a country chooses so. In this case the contribution is not unconditional but conditional, which makes their actual implementation more uncertain. Some countries provide a target range (e.g., “50-52% reduction”). For global emissions or temperature pathways this means that one needs to consider the target conditionality & range, leading to four different pathways: unconditional worst, unconditional best, conditional worst, and conditional best (from high to low).
In addition, reduction targets may only refer to individual sectors (e.g. Energy and Industrial Processes and Product Use) or gases (e.g. all Kyoto GHGs or only CO2). In these cases we need to know which sectors and gases are covered, and what happens to the rest of emissions, sometimes complicating even more straight-forward target quantifications. Emissions and removals from Land Use, Land-Use Change and Forestry (LULUCF), f.ex. resulting from activities such as cutting down or planting of forests, can also make target quantifications more difficult and increase the uncertainty of target estimates.
A great outcome of the first round of NDCs would have been that countries’ aggregated efforts are sufficient to meet the 1.5-2°C temperature goal. In this case, all we would need to care about is the implementation and that countries actually act in line with their NDCs (probably this sounds easier than it would have been). But well, it was not the case … and it was not even expected. That is why the PA already includes rules for a ratcheting up, i.e. after a review whether the NDCs are sufficient to meet the target, countries are invited to up-date their targets which are then reviewed every five years. The first round of NDC updates was planned for the year 2020, while the first official global stocktake takes place in 2023 to shed light on what is missing in terms of ambition and how to improve the NDCs. So the struggle continues: countries should submit regular updates of their NDCs and mitigation targets, with increasing levels of ambition over time (it is not allowed to fall behind its previous target), basically until the temperature goal is met. Which is a race against time, and a gamble with how much reductions are feasible in a short period of time. The longer we wait, the shorter the available time and the steeper the necessary emissions cuts to still be able to meet the 1.5-2°C goal. Or for compensation, we need to remove GHGs from the atmosphere, fast and on a large scale, which unfortunately holds another feasibility issue.
So, we say that with the first round of NDCs we will not meet the 1.5-2°C PA temperature goal. But how do we actually know whether there is a lack of ambition? As indicated above, there is a connection between emissions and temperature levels, and if we know the countries’ emissions trajectories in line with their NDCs, we can sum them up and estimate the resulting temperatures. Sounds easy enough. Somehow it’s not that simple, though … or fast to do. I’m German, and luckily my levels of English and French are ok, and my passive Spanish skills are okish as well. But going through all those documents and extracting all the necessary information to quantify the emissions targets takes forever (and the regular NDC updates needed to hopefully get on track mean that “being done” doesn’t equal “being done”). And what is “the necessary information” anyway? That depends on the purpose: here, we try to understand the countries’ planned emissions levels, and don’t really care much about information on adaptation, implementation measures, financial needs, etc.
If you don’t have time you should not start the NDC assessments
From a political perspective you can understand why the NDC formats have not been specified more clearly to not risk the overall goal of getting an agreement. Getting countries on board may have been paid by allowing for softness in specifying the mitigation targets and the missing agreement on formats has to be paid within the review process:
Length. For example, currently the longest NDC with 167 pages is only available in Spanish (NDC Dominican Rep., 2020), which is a lot and as I said: my Spanish reading skills are okish, but fast scanning through a document for the necessary bits is not easy at that level. However, in the end it contains quantitative target information and is well structured, following the recommendations of the Katowice climate package (COP24 in 2018). The guidelines on how NDCs are to be presented have been developed after the PA (The Katowice climate package: Making The Paris Agreement Work For All) to improve the reporting and help the review process. Unfortunately countries could only agree on applying the new standards for NDCs submitted for the 2025 review process. However, some countries decided to follow this guidance already, which is nice. Going through those NDCs shows that this leads to some repetitions, but also gives structure.
Inconsistencies. As a second example, let’s look at the Bahamas’ 2016 NDC (submitted as non machine searchable pdf …) that is difficult to interpret as it actually includes two different specifications of emission reductions. Firstly, it states “an economy-wide reduction GHG emission of 30% when compared to its Business as Usual (BAU) scenario by 2030”, but secondly it also says that “The Bahamas has set a target of 30 percent emissions reductions, below 2010 levels by 2030” (p. 4 vs. 11 - as I said, the document is not machine searchable, so better to have the page numbers, no?). If the country additionally stated that the BAU emissions in 2030 are expected to be at 2010 levels it would not really be inconsistent. But nevertheless, it is not very clear whether that is really meant as it could have been communicated much simpler. Well, in the end we decided for the first BAU-based specification in our assessment being aware that the second specification may imply a slightly different emissions trajectory.
This is not the only NDC whose interpretation is ambiguous. However, in most of the cases the different interpretations do not lead to large discrepancies on a global scale and do not matter that much in the end when estimating the global warming associated with the NDC-based global emissions pathway. What has turned out to matter most is the uncertainty associated with the NDCs of the largest emitters (biggest three emitters in 2017: China 27.2%, USA 13.7%, India 6.3%; based on PRIMAP-hist v2.1 HISTCR, excluding LULUCF and bunkers fuels): generally speaking, the more unknown parameters are in the equation (e.g., projected GHG emissions, population, or GDP) and the higher the global emissions share, the higher the uncertainty and possibly the resulting emissions and temperature ranges. While the USA’s target is a relative reduction against historical base year emissions and hence well defined, China and India specify their mitigation targets in terms of relative reductions in emissions intensity per unit of GDP, i.e. it needs assumptions about GDP development, which adds quite substantially to the uncertainties in global aggregates of target emissions (China’s NDC from 2016: “China has nationally determined its actions by 2030 as follows: To achieve the peaking of carbon dioxide emissions around 2030 and making best efforts to peak early; To lower carbon dioxide emissions per unit of GDP by 60% to 65% from the 2005 level; …”; India’s NDC from 2016: “… reduce the emissions intensity of its GDP by 33 to 35 percent by 2030 from 2005 level …”). From these three big emitters, only the USA has by now submitted an updated NDC.
The USA is a special case with its back and forth in the PA. It joined the PA under president Obama (USA’s NDC from 2016: 26-28% reduction against 2005 levels in 2025), but under former President Trump it withdrew from the Agreement. Under President Biden it rejoined and submitted an updated contribution in April 2021 with “an economy-wide target of reducing its net greenhouse gas emissions by 50-52 percent below 2005 levels in 2030”. As our current emissions and temperature assessments are based on all submissions by 31 December 2020, we do not consider this update yet, but consider the 2016 NDC.
Our baseline assumptions
If a country chose a target referring to a historical base year, e.g., “a reduction of 50% in 2025 against our 1990 emissions level”, we need to know the 1990 emissions to estimate the target. That is relatively easy to deal with, thanks to national emissions inventories and f.ex., the PRIMAP-hist dataset. If a country chose a target referring to its Business As Usual (BAU) emissions level, e.g., “a reduction of 25% compared to BAU in 2030”, we need an estimate of the 2030 BAU. Some countries indicate their estimated BAU level in their NDC, which then can be used as a reference. When this information is missing, however, and additionally for comparison purposes, we depend on other emissions projections. Here, we can use the Shared Socio-Economic Pathways (SSPs).
The SSPs describe certain narratives of socio-economic developments and were, as an example, used to derive greenhouse gas emissions scenarios that correspond to these developments of, i.a., population and GDP. We use the “Middle-of-the-Road” baseline marker scenario SSP2. But what does that stand for? SSP2 is a narrative with little shifts in socio-economic patterns compared to historical ones, and is connected to medium socio-economic challenges for both climate mitigation and adaptation. Different models were used to model each storyline (SSPs1-5), but per SSP, one model was chosen to provide a representative “marker scenario”. We are dealing with national mitigation targets here, but the emissions projections are not readily available on a country-level. Luckily, a downscaled version of the pathways to national levels is available from J. Gütschow et al. (2021) and can be used in NDC analyses. From this data set, we can use the national emissions projections, and the corresponding population and GDP scenarios. The latter are needed if a country chose an emissions intensity target, like the big emitters China and India did (e.g., “50% reduction in emissions intensity per unit of GDP by 2030, compared to 1990 levels”, in which case we need to know the GDP growth from the year 1990 to 2030 to estimate the target emissions).
Assumptions we add when a country does not cover its entire emissions
Countries do not always cover all of their emissions sectors and emitted gases with their targets. An example would be that only CO2 and CH4 from the Energy and Agriculture sectors are targeted, leaving the rest aside. In terms of Kyoto GHGs this means that N2O, HFCs, PFCs, SF6, and NF3 are not targeted, while in terms of emissions sectors, this leaves Industrial Processes and Product Use (IPPU), and Waste not targeted. Not only CO2 emissions are important when we talk about global warming, why over time, countries are expected to cover the total of their national emissions.
From the NDC information on covered sectors and gases, combined with historical national emissions time series resolved into the contributions by different sectors and gases (PRIMAP-hist composite data set), we derived estimates of the covered share of emissions per country and included this information in our quantifications. Even though we do consider the targeted share of emissions while deriving target estimates, all of our quantified pledges stand for countries’ total emissions (from all Kyoto GHGs and sectors, excluding LULUCF). In Figure 2, you can find estimates of the covered share of emissions per country for the year 2017. But as you can imagine, a country’s covered share of emissions is not a constant value, just as sectoral and gas shares are not constant and evolve over time, e.g., when a country gets more industrialised and agricultural production is reduced.
Share of national emissions covered by an NDC, for the year 2017. The legend reads as follows: e.g., “> 99% (#124, 57.4%)” means that 124 countries cover more than 99% of their national emissions and these countries represented 57.4% of global 2017’ emissions. All data presented here exclude LULUCF and bunkers emissions. Current state of this figure: only NDCs submitted by 31 December 2020 are accounted for, and we are working on an update (nonetheless, the shown information is generally valid).
FIG 2 / Image source: A. Günther, Gütschow, and Jeffery (2021)
How to derive temperature estimates from emissions pathways?
How do we get from the NDCs’ GHG mitigation targets to the corresponding temperature estimates?
After quantifying the NDCs, we are not done. The main question is: do we meet the Paris Agreement’s temperature limit of 1.5-2°C with the countries’ mitigation targets set out in their NDCs? Logically, we need to get temperature estimates in line with the quantified pledges. There are several intermittent steps necessary, as outlined below and in Figure 3:
- National pathways and global aggregates: the national targets are quantified, e.g., we estimate that in 2030 country X plans to emit Y megatonnes of CO2 equivalent. Now we need to aggregate these targets to get global values. However, the targets are generally point values, for the year 2025 or often 2030. To sum the targets up we depend on an estimate for each country in the same year. Else there would be “holes”. So we construct national mitigated pathways to the year 2030, around the targets, and sum those up!
- Pathway extension to 2100: now we have a pathway up to the year 2030 for each country. However, to review the NDC targets in context of the global mean temperature limits agreed on in the PA we have to estimate global warming across the 21st century at least, which needs an associated emissions pathway. So we must extend the NDC pathway(s) in a senseful manner. To this end we draw from a huge pool of emissions trajectories generated by Integrated Assessment Models (IAMs) for different purposes and collected in a common database (AR5 Scenario Database). These models calculate economic and technological scenarios and can prescribe, i.a., population, GDP, CO2 prices, etc., while other entities are modelled in a least cost optimisation approach. One challenge in defining a 2100 emissions pathway compatible with the 2030 NDC pathway is that scenarios in the database are crossing in the near future. This means that one cannot simply choose the scenario closest to the 2030 NDC estimate. Instead, we can use database quantiles over all scenarios. We can check for the NDC scenario quantile for the year 2030 and hold this quantile constant to get the emissions pathway to 2100. This method is called Constant Quantile Extension and works on five world regions, similar to the underlying database. In Figure 4 we explain the method in a schematic way, as well as what this “quantile” is we are talking about.
- Splitting up the gases: can we calculate the temperatures now? Not yet. Our pathways for the five world groups represent not only CO2 or CH4 emissions, but all Kyoto GHGs are combined in these pathways. Contributions from Land Use, Land-Use Change and Forestry (LULUCF) are not included, as they are often very uncertain. The “problem” is that what we have until now is not sufficient as an input to a climate model that can translate the emissions into temperature estimates. We need to add a global LULUCF scenario (available from models). And even though countries do not account for international transport (shipping and aviation, emitting bunkers fuels), the emissions exist and are anthropogenic and we need to add a bunkers fuels scenario (available from models). Climate models need pathways for individual climate forcers, which we do not have from quantifying the NDC targets. We need to split up the 2100 emissions pathway somehow to get reasonable input for the climate model. The approach is similar to the pathway extension above. We go on the “Equal Quantile Walk” (Meinshausen et al. 2006), walking along quantiles in a scenario base. Instead of checking for the 2030 quantile and keeping it constant (Fig. 4), here one analyses for each year which quantile fits the Kyoto GHG emissions scenario we have. And this time series of quantiles is then used for all individual gases, and also for aerosols and non-Kyoto GHGs. We need scenario databases which provide scenarios of the Kyoto GHGs and the individual climate forcers (SSP Database, AR5 Scenario Database). As the IAMs with which the scenarios were created account for the different emissions sources in a consistent way they are assumed to describe plausible combinations of all GHG and aerosol emissions.
- Temperatures: with the previous step performed, finally we have the NDC-based input to run a climate model and answer the question whether the NDCs’ mitigation targets are ambitious enough to meet the PA temperature goal. We use the very fast running Model for the Assessment of Greenhouse Gas Induced Climate Change (MAGICC). It is a simple climate model that can emulate complex models by tuning parameters.
Schematic example of the NDCs’ pathway extension to the year 2100 (no real scenarios). From the scenario database, the “blue” scenarios are available. For these scenarios we calculate the “cyan” quantiles with a certain step (e.g., 1% step). But what are quantiles? They are similar to thresholds, e.g. the 1% quantile means that 1% of the scenarios lie below this threshold of X megatonnes of CO2 equivalent, while 99% lie above this threshold. The median means that 50% of scenarios lie below and 50% above this value. For 2030, we check which quantile the estimated NDC-based pathway corresponds to. In our example, this is 45%. So in 2030 the NDC pathway lies in the lower half of the entire ensemble provided by the IAMs. Only 45% of the simulations have lower values in 2030, while 55% of the ensemble members have higher values in 2030. The NDC pathway is extended by keeping this quantile constant at 45% until 2100.
Results: global emissions and temperature estimates
To wrap things up: it is a lot of work to quantify the NDCs’ mitigation targets, construct emissions pathways in line with the targets, and translate the global trajectories into temperature estimates. Based on the assumptions we described above, current NDCs (submissions until 31 December 2020) are not sufficient to keep global warming under the PA temperature limit. Instead, median warming is estimated to reach about +2.6°C at the end of the century, compared to pre-industrial times. But let’s look at this a bit closer.
As described above, we first quantified the NDC targets and defined national mitigated emissions pathways. These are aggregated and by joining different analysis tools they are finally translated into temperature estimates (pathway extension to 2100; split-up into single climate forcers; climate model simulations). In Figure 5, you can see the outcome of the first step, the globally aggregated NDC-based emissions pathways up to the year 2030. As we quantified the pathways per conditionality & range, you see four pathways (un-/conditional best/worst). And as we performed quantifications with different set-ups, these pathways indicate the median per set of quantifications (median: separates the higher half from the lower half of the ensemble). Additionally, you can see the scenario range, from our minimal to maximal estimates, together with the 10% and 90% quantiles (e.g., 10% quantile: 10% of pathways below and 90% above the shown value), to get an impression of how much the scenarios differ. In general, e.g., if the distance between the minimum and 10% quantile is big, the minimum might be more of an outlier, and the same is true for 90% vs. maximum. Here, this is not the case.
We do not want to go too much into detail, but only shortly mention different quantification options we chose per conditionality & range case to capture target uncertainties: prioritising the emissions data provided in an NDC if available, or prioritising the SSP baseline information to calculate the target even if the NDC provides data; setting the covered share of emissions to 100% for each nation, or using the percentages we calculated; using two different data sets for LULUCF emissions, which are needed in case a country covers LULUCF in its target, but does not provide LULUCF emissions estimates; instead of using the quantified target, use the baseline emissions, if the target lies above the chosen baseline; use the baseline growth rate after 2025 if a country has a 2025 target, but no 2030 target, or keep the relative difference to the baseline constant. You might be wondering why we always say “excluding contributions from LULUCF”, and in the list of options there is something written about “LULUCF data sets”. Well, we need an approach to deal with a country that includes LULUCF in its target: we first calculate the target including LULUCF, and then derive the target estimate excluding LULUCF based on the national LULUCF emissions.
Per conditionality & range case, the 2030 estimates span an emissions range of about 6.6-7.7 Gt CO2eq (Figure 5, vertical lines). This is quite a difference compared to the total amount being around 50 Gt CO2eq. Conditional targets can be less likely to be met, as they depend on, e.g., international financial support. If it is not provided by developed nations, the conditional target might not be implemented. As generally, only lower emitters have conditional targets, the difference between the conditional best and worst pathways are marginal, while for the unconditional pathways, the difference between the best and worst case is rather large.
One point that we are not showing here but want to mention is the influence of choosing different SSPs as baselines. They differ regarding the emissions, population, and GDP pathways. As China and India are big emitters and chose emissions intensity targets with a reduction relative to the national emissions per unit of GDP, the growth rate of the underlying GDP scenario between the base and target years shows quite an impact on the global scale. We did not focus on policy analyses and do not have the means to include renewable energy targets in our assessment. Both countries have such targets, which can decrease the influence of the assumed GDP scenario, but we cannot estimate by how much.
Globally aggregated NDC-based mitigated emissions pathways. Emissions exclude contributions from LULUCF and bunker fuel emissions, and analysed are NDC submissions until 31 December 2020 (including the USA’s 2016 NDC). The baseline scenario is SSP2 (in black, solid), while the “baseline: prio NDCs” includes baseline data from within the NDCs, where available (so it is a mix between SSP2 and NDC-emissions). Colour-coded are the un-/conditional best/worst pathways. Several quantification options were used (i.a., differing assumptions on covered share of emissions and emissions data sources), and here you see the ensemble minimum, 10% quantile, median (50%), 90% quantile, and maximum for the year 2030 (vertical lines). For the example “conditional best”, the minimum and maximum pathways are shown as a filled area.
In order to be able to say, whether these emissions estimates (Figure 5, median per conditionality & range case) are putting the world on track to reach the 2100 temperature goal of 1.5-2°C set out in the PA, we used the emissions pathways from Figure 5 to fuel the simple climate model MAGICC (using median per conditionality & range case; and with the intermittent steps described above performed in between). The outcome (Figure 6) ranges between 2.4°C and 2.8°C for the conditional to unconditional targets. Which is not yet enough, not even if all conditional targets are met in addition to the unconditional pledges. Two things have to be kept in mind here, which would probably lead to better results in terms of keeping the PA temperature limits: several countries submitted long-term targets by now that are not included in our current assessment, pledging, e.g., climate neutrality by 2050; and since 31 December 2021, the cut-off date for the presented emissions and temperature estimates, 18 countries have submitted updated NDCs. So be aware that it might be worth coming back here, as next time you check this page, we might have updated our numbers. On this website, you will also find the temperature estimates from Figure 6 such as information in impact assessments relative to different levels of warming for crop failure, flooding, and droughts.
Temperature estimates based on the median NDC-based mitigated emissions pathways from Figure 5. Analysed NDCs are documents submitted until 31 December 2020, including the USA’s 2016 NDC. As the conditional best and worst pathways are very similar but more uncertain to be implemented, the presented mean and median are computed from the unconditional best & unconditional worst temperature pathways, and the mean over the conditional best and worst pathways.
1 Potsdam Institute for Climate Impact Research (PIK), 14473 Potsdam, Germany