Beyond carbon budgets

The remaining carbon budget consistent with limiting warming to 1.5 °C allows 20 more years of current emissions according to one study, but is already exhausted according to another. Both are defensible. We need to move on from a unique carbon budget, and face the nuances.

The world woke up to some stunning news on an otherwise quiet September day in 2017. A new study1 on carbon budgets found that it was “not yet a geophysical impossibility” to limit warming to 1.5 °C. These findings have since been confirmed, also by combining observations with Earth system models2,3. The remaining carbon budget for 1.5 °C may consequently be several times larger than previously suggested by the Intergovernmental Panel on Climate Change (IPCC)4.

With these updates, the 1.5 °C carbon budget resembles the original 2 °C carbon budget, and carbon budgets for higher temperature levels become larger accordingly. Meeting a 1.5 °C target is still profoundly challenging, but at least possible. The updated 2 °C carbon budgets may even mean that the nationally determined contributions (emission pledges) to the Paris Agreement are consistent with the ambition of the Paris Agreement, something that was not previously the case5. This is science with high political stakes.

Almost in parallel, integrated assessment modellers have generated a new set of 1.5 °C emission scenarios6. The remaining carbon budget according to these scenarios, taken as the cumulative CO2 emissions over the period 2016 to 2100, ranges from −100 to 475 Gt CO2; the scenario with −100 Gt CO2 would mean that between 2016 and 2100, more CO2 must be removed from the atmosphere than is put in. These carbon budgets are far smaller than the updated 1.5 °C carbon budgets based on observations and Earth system models1,2,3. The diverse remaining carbon budgets are all scientifically defensible, but the apparent uncertainty may be used to justify further political inaction7.

The discrepancies between various carbon budget calculations arise from a variety of known differences in the assumptions of the studies, for example, regarding the type of model used and associated shortcomings; the amount of warming that has already happened since pre-industrial times; how the Paris temperature goal is defined; when negative emissions allow it to be temporarily exceeded; the treatment of human-induced climate influences other than CO2; and how uncertainties are expressed.

I argue that carbon budgets can confuse rather than clarify, and that — given the mismatch of their global scale with decision making at the country level — they are not as relevant for policy making as it may have originally seemed. There is no single magic number to describe the mitigation challenge. Instead of oversimplifying, the scientific community should seek to discuss and emphasize the persistent uncertainties — those associated with our incomplete understanding of the carbon cycle as well as those associated with user and societal choices.

The logic of carbon budgets

Climate policy discussions are often framed around carbon budgets, since the relationship between cumulative emissions and peak warming has been shown to be largely insensitive to the emission pathway8. However, it is not clear how this framing can be reconciled with carbon budget estimates across studies ranging from −100 to about 800 Gt CO2. It seems that the carbon budget concept has either been oversold or that the uncertainties are so large that it lacks policy utility9.

It is probably a bit of both: the uncertainties in the climate system will lead to uncertainties in carbon budgets9, but there are also uncertainties related to our choices. User choices, such as how the carbon budget is defined, create large differences10. Societal choices, such as the pathways for future emissions of non-CO2 greenhouse gases and aerosols, make the outcome dependent of the emission pathway and lead to large differences across carbon budgets10. In fact, uncertainties related to user and societal choices can explain a large part of the difference between recent carbon budget estimates.

Current warming levels

The recent larger carbon budget estimates have been based on methods that better align observations with Earth system models1,2,3. This approach makes sense, as the ebbs and flows of natural variability in observations and models do not necessarily align. These approaches can also deal with the large and persistent uncertainties in the carbon cycle11,12. Carbon budgets estimated with integrated assessment models6 use simple climate models13, which are essentially already calibrated to observations and Earth system models and therefore partially circumvent the challenges with observations and Earth system models1,2,3.

The various trade-offs between observations, simple climate models and Earth system models is perhaps beyond what policy makers need to consider, but the discussion highlights unresolved issues. We need to make progress on how to interpret the Paris Agreement14: the point in time that we define as the pre-industrial temperature baseline15, or whether we reference to a more recent point in time16; the definition and application of globally average temperature17,18; and how success or failure of meeting the Paris Agreement temperature goal is defined9. Indeed, a key reason for the difference between carbon budgets estimated by Millar et al.1 and Rogelj et al.6 is a different definition of the temperature level (see the Supplementary Information of Rogelj et al.6).

Exceed or avoid

Most, if not all, 1.5 °C scenarios first exceed 1.5 °C before returning to 1.5 °C or below (Fig. 1a) by including removal of CO2 from the atmosphere at scale6. If global emissions become negative due to CO2 removal, the cumulative emissions get smaller over time making it ambiguous how to define a fixed carbon budget (Fig. 1b). An emission scenario can both exceed and avoid a given temperature threshold at different times14 (Fig. 1), and thus one emission scenario can have multiple carbon budgets depending on the desired definition. A carbon budget defined when a temperature threshold is exceeded is generally larger than a carbon budget defined when the temperature has returned below (avoided) the same temperature threshold at a later point in time (for example, in 2100)10.

Fig. 1: Temperature and cumulative emissions pathways.
Fig. 1

a,b, The temperature response (a) and cumulative emissions (b) for scenarios that reach 1.9 W m–2 in 2100 and limit warming below 1.5 °C. The emission scenarios are based on integrated assessment models6 and the temperature response using MAGICC26. The historic temperature is based on NASA GISS27,28, although noting different datasets may give slight differences17. The box plot and grey shading around 2030 indicate a carbon budget where 1.5 °C is exceeded (a definition often used with Earth system models), while the box plot and grey shading after 2100 indicate the cumulative emissions from 2016–2100 (a definition more common with integrated assessment models).

Carbon budgets defined at the time when the temperature threshold is first crossed are easier to define but are arguably less useful for policy. Carbon budgets that ensure the temperature threshold is avoided at a given point in time are more useful for policy, but since the cumulative emissions get smaller with time, the carbon budget is difficult to define.

Temperature pathways and cumulative emissions across comparable scenarios (Fig. 1) can be used to highlight the differences between carbon budget definitions. Approaches based on observations and Earth system models1,2,3 generally define carbon budgets when a temperature threshold is first exceeded. Since the Earth is already about 1.1 °C above pre-industrial levels, depending on definition17, the inertia in physical and social systems means that there is a small window of cumulative emissions before 1.5 °C is exceeded. Emission scenarios that keep the temperature increase below 1.5 °C in 2100 (ref. 6) all initially exceed 1.5 °C in the early 2030s, when a total of 400 to 600 Gt CO2 carbon has been emitted from 2016 to the year of exceedance (Fig. 1). The same 1.5 °C emission scenarios give total cumulative emissions of −100 to 475 Gt CO2 from 2016 to 2100, an equally valid definition for the carbon budget10, with the larger range due to non-CO2 greenhouse gas and aerosol emissions. These different definitions can explain another large part of the differences in carbon budgets10 originating from observations and Earth system models1,2,3 and integrated assessment models6.

Different studies therefore work with a confusing array of different carbon budgets (Fig. 1). If the goal is to know at which point in time 1.5 °C will be first exceeded1,2,3, then combining observations with Earth system models may be a useful approach. If the goal is to keep below 1.5 °C, even after a period of temperature overshoot, then integrated assessment models may give a much lower carbon budget10. The much critiqued1,2,3 and much more stringent IPCC carbon budget4 for 1.5 °C may coincidently be correct, if the goal is to avoid 1.5 °C of warming6.

Non-CO2 emission pathways

Carbon budget estimates using combinations of observations and Earth system models generally use a very limited number of scenarios1,2,3. The Representative Concentration Pathway consistent with 2 °C of warming19 (RCP2.6) has strong non-CO2 mitigation, and all else being equal, this is expected to make the carbon budget larger1,6. This could further exacerbate the differences between avoidance and exceedance budgets10.

Non-CO2 greenhouse gas and aerosol emissions have a critical role in 1.5 °C scenarios6, which affects the size of the carbon budgets. Figure 2 shows the cumulative emissions for the scenarios in Fig. 1, but separated into time periods. There is very little variation in carbon budgets to exceed 1.5 °C (grey), variation increases for the maximum cumulative emissions (grey plus blue), and the variation is largest when the carbon budget is defined for cumulative emissions from 2016 to 2100 (black dots).

Fig. 2: Model differences.
Fig. 2

The cumulative emissions for scenarios6 that keep warming below 1.5 °C, broken into three time periods: (1) from 2016 to the time 1.5 °C is exceeded (grey); (2) until the maximum cumulative emissions (grey plus blue); and (3) the net negative emissions from the maximum to 2100 (green). Black dots represent net cumulative emissions from 2017 to 2100 (sum of the three components). The grey bars and black dots correspond to the exceedance and 2100 carbon budgets in Fig. 1.

The cumulative net negative emissions (green), play an important role in offsetting the non-CO2 greenhouse gas and aerosol emissions. The higher the non-CO2 emissions, the lower the carbon budget6, and consequently the higher the need for cumulative net negative emissions. Conversely, more aggressive non-CO2 mitigation allows a bigger carbon budget6 and may also reduce reliance on negative emissions20. The obsession with cumulative carbon emissions and carbon budgets has perhaps diverted attention from gains through non-CO2 mitigation21.

Expressions of uncertainties

The way uncertainties are expressed also needs to be reconsidered. Early on, a choice was made to express carbon budgets probabilistically. For example, studies provide a carbon budget for a 66% chance to stay below 2 °C. This is problematic. First, different approaches to carbon budgets (based on Earth system models versus integrated assessment models), use different probabilistic definitions. This makes their carbon budgets incomparable and leads to confusion10. Second, probabilistic expressions require independent and sufficient input data to derive reliable distributions, whereas many of the samples are too small for reliable statistics, whether across Earth system model ensembles or across scenarios. Third, not many have sufficient understanding of statistics or realize that the probabilistic carbon budget is a way to summarize a statistical distribution.

The equilibrium climate sensitivity is generally expressed by its uncertainty range or distribution, and not as the probability that the cumulative distribution lies to one side of a given value (as done for carbon budgets). Analogously, the median remaining carbon budget in one study was expressed3 as 760 Gt CO2 (33–66% range of 475–930 Gt CO2), rather than stating the same conclusion as a remaining carbon budget of 475 Gt CO2 for a 66% chance of staying below 1.5 °C. Layered on top of this climate system uncertainty is the need to additionally include uncertainty related to alternative non-CO2 emission pathways, generally expressed as a range across independent carbon budgets from alternative scenarios10.

The policy connection

Carbon budgets are uncertain. There is no magic number that describes the mitigation challenge. Emissions pathways are important. Nuances are important. And the uncertainties, both physical and those due to societal and user choices, may be irreducible9. Yes, the carbon budget is messy, but there are limited gains to be made by continued oversimplification.

In light of this, the carbon budget is useful for an elevator pitch to summarize the climate challenge in a few sentences and to explain the importance of net zero emissions22. Beyond that, the uncertainties need to be brought out into the open, particularly the role that societal choices have on the carbon budget. Additionally, there is a huge need to understand and harmonize definitions, and the terminology that follows10.

Despite perceptions to the contrary7, carbon budgets are not all that relevant for policy. Carbon budgets are global and need translation to country pathways23. Regardless of the carbon budget, emissions need to reach zero between 2050 in 2100 (as specified by the Paris Agreement). An earlier achievement of this goal will lead to lower temperature24. And equity requires rich countries to reach zero before poor countries25. The carbon budget concept has perhaps served its purpose, time is short. To enact policy, a carbon budget is woefully too simplified.


  1. 1.

    Millar, R. J. et al. Nat. Geosci. 10, 741–747 (2017).

  2. 2.

    Goodwin, P. et al. Nat. Geosci. 11, 102–107 (2018).

  3. 3.

    Tokarska, K. B. & Gillett, N. P. Nat. Clim. Change 8, 296–299 (2018).

  4. 4.

    IPCC Climate Change 2014: Synthesis Report (eds Core Writing Team, Pachauri, R. K. & Meyer, L. A.) (IPCC, 2014).

  5. 5.

    Rogelj, J. et al. Nature 534, 631–639 (2016).

  6. 6.

    Rogelj, J. et al. Nat. Clim. Change 8, 325–332 (2018).

  7. 7.

    Geden, O. Nat. Geosci. (2018).

  8. 8.

    Allen, M. R. et al. Nature 458, 1163–1166 (2009).

  9. 9.

    Peters, G. P. Nat. Clim. Change 6, 646–649 (2016).

  10. 10.

    Rogelj, J. et al. Nat. Clim. Change 6, 245–252 (2016).

  11. 11.

    Friedlingstein, P. et al. J. Clim. 19, 3337–3353 (2006).

  12. 12.

    Jones, C. et al. J. Clim. 26, 4398–4413 (2013).

  13. 13.

    Meinshausen, M., Raper, S. C. B. & Wigley, T. M. L. Atmos. Chem. Phys. 11, 1417–1456 (2011).

  14. 14.

    Geden, O. & Löschel, A. Nat. Geosci. 10, 881–882 (2017).

  15. 15.

    Schurer, A. P., Mann, M. E., Hawkins, E., Tett, S. F. B. & Hegerl, G. C. Nat. Clim. Change 7, 563 (2017).

  16. 16.

    Hawkins, E. et al. Bull. Am. Met. Soc. 98, 1841–1856 (2017).

  17. 17.

    Schurer, A. P. et al. Nat. Geosci. 11, 220–221 (2018).

  18. 18.

    Millar, R. J. et al. Reply to ‘Interpretations of the Paris climate target’. Nat. Geosci. 11, 222–222 (2018).

  19. 19.

    van Vuuren, D. P. et al. Clim. Change 109, 95–116 (2011).

  20. 20.

    van Vuuren, D. P. et al. Nat. Clim. Change (2018).

  21. 21.

    Shindell, D. et al. Science 335, 183–189 (2012).

  22. 22.

    Geden, O. Nat. Geosci. 9, 340–342 (2016).

  23. 23.

    Peters, G. P., Andrew, R. M., Solomon, S. & Friedlingstein, P. Environ. Res. Lett. 10, 105004 (2015).

  24. 24.

    Tanaka, K. & O’Neill, B. C. Nat. Clim. Change 8, 319–324 (2018).

  25. 25.

    Robiou du Pont, Y. et al. Nat. Clim. Change 7, 38–43 (2016).

  26. 26.

    Meinshausen, M. et al. Clim. Change 109, 213–241 (2011).

  27. 27.

    Hansen, J., Ruedy, R., Sato, M. & Lo, K. Rev. Geophys. 48, RG4004 (2010).

  28. 28.

    GISS Surface Temperature Analysis (GISTEMP) (NASA GISS, accessed 27 March 2018);

Download references

Author information


  1. CICERO Center for International Climate Research, Oslo, Norway

    • Glen P. Peters


  1. Search for Glen P. Peters in:

Corresponding author

Correspondence to Glen P. Peters.