Changes in IPCC Scenario Assessment Emulators Between SR1.5 and AR6 Unraveled

Abstract The IPCC's scientific assessment of the timing of net‐zero emissions and 2030 emission reduction targets consistent with limiting warming to 1.5°C or 2°C rests on large scenario databases. Updates to this assessment, such as between the IPCC's Special Report on Global Warming of 1.5°C (SR1.5) of warming and the Sixth Assessment Report (AR6), are the result of intertwined, sometimes opaque, factors. Here we isolate one factor: the Earth System Model emulators used to estimate the global warming implications of scenarios. We show that warming projections using AR6‐calibrated emulators are consistent, to within around 0.1°C, with projections made by the emulators used in SR1.5. The consistency is due to two almost compensating changes: the increase in assessed historical warming between SR1.5 (based on AR5) and AR6, and a reduction in projected warming due to improved agreement between the emulators' response to emissions and the assessment to which it is calibrated.

Here we discuss the reasons for the higher warming projections in MAGICC6 compared to the projections from AR5-like MAGICCv7.5.3 (Supplementary Figure S13). MAG-ICC6 generally has lower anthropogenic effective radiative forcing than AR5-like MAG-ICCv7.5.3 (Supplementary Figure S7). Therefore, differences in the parameterisations that link emissions and effective radiative forcing are not the reason for MAGICC6's higher warming projections and we hence conclude that differences in model calibration, particularly transient climate response (TCR), explain the difference instead. We next discuss some of the changes in effective radiative forcing of different species and why these changes are approximately zero. As the AR5-like calibration of MAGICCv7.5.3 reflects literature which is more recent than AR5, this comparison provides an insight into some, but not all, the changes in effective radiative forcing between AR5 and AR6.
CO 2 effective radiative forcing is again relatively unchanged between the two emulators (Supplementary Figure S14), reflecting the close agreement between the CO 2 ERF estimates used for the RCMIP Phase 2 tuning (Smith et al., 2020) and the AR5 assessment Taking all the changes together, we see that CO 2 effective radiative forcing is largely unchanged, aerosol effective radiative forcing is initially more negative before being more positive and methane effective radiative forcing is more positive. In sum, the combination of these and other changes leads to an increase of 0.0 -0.2 W m -2 in anthropogenic September 9, 2022, 9:12pm NICHOLLS ET AL.: CHANGES IN IPCC SCENARIO ASSESSMENT EMULATORS X -5 effective radiative forcing. We note again that these changes are scenario dependent (Supplementary Figure S10) and bespoke analysis is required to understand each specific scenario's drivers.
Here we consider the changes in effective radiative forcing of different species and why these changes cancel out to approximately zero change (with a slight skew towards an increase) in total anthropogenic effective radiative forcing between MAGICC6 and MAG-ICCv7.5.3. CO 2 effective radiative forcing has a median change of approximately zero (Supplementary Figure S18). This almost zero change is the result of a slight increase in the assessment of CO 2 effective radiative forcing (for a given atmospheric CO 2 concentration) and a decrease in projected atmospheric CO 2 concentrations (Supplementary Figure S19).
It is also worth noting that, while the median is unchanged, there is a range of differences due to differing climate-carbon feedbacks (particularly those driven by the different temperature projections).
Aerosol effective radiative forcing is more sensitive to changes in aerosol emissions. As a result, the negative aerosol effective radiative forcing is stronger today but reduced in the future as aerosol emissions drop (differences are calculated as MAGICCv7.5.3 minus Finally, methane effective radiative forcing increases in the short-term before returning to a median difference of zero with a range of -0.075 W m -2 to 0.1 W m -2 (Supplementary Figure S21). The increase follows the upwards revision presented in Etminan et al. (2016), tempered by AR6's inclusion of rapid adjustments in the assessment of methane effective radiative forcing in line with Smith et al. (2020). The increased sensitivity to methane emissions (via methane concentrations) also leads to a more pronounced reduction in methane effective radiative forcing after its peak.
In summary, CO 2 effective radiative forcing has a median change of around zero with a slight skew towards a decrease, aerosol effective radiative forcing is initially more negative before being more positive and methane effective radiative forcing is initially more positive before being around zero or slightly more negative. In general, the combination of these and other changes lead to the increase in anthropogenic effective radiative forcing.
However, the scenario dependence of these changes should once again be noted as individ-    Figure S1. Difference between the emulator output underlying the SR1.5 scenario categorisation and alternate emulator output for exceedance probabilities of 1.5°C (top row) and 2°C warming (bottom row) for 2100 (left column) and peak (i.e. maximum between 2010 and 2100, right column). The data is the same as Figure 1 but has been plotted as a histogram of differences rather than a scatter plot. Scenarios with exceedance probabilities of close to 100% cannot have large changes in exceedance probability (because exceedance probability is capped at 100%), hence the large number of results with no change (zero is marked by the dashed grey line). The difference from MAGICCv7.5.3 emulator is shown in dark blue, from FaIRv1.6.2 is shown in red and from FaIR1.3 is shown in grey.   Figure S3. Difference between the emulator output underlying the SR1.5 scenario categorisation and alternate emulator output for median (top row) and 67 th -percentile (bottom row) projections for 2100 (left column) and peak (right column) temperatures. The data is the same as Supplementary Figure S2 but has been plotted as a histogram rather than a scatter plot. The difference from MAGICCv7.5.3 is shown in dark blue, from FaIRv1.6.2 is shown in red and from  Figure S4. Relationship between exceedance probabilities and temperature projections across multiple emulators. By definition, when the median temperature is 1.5°C (2.0°C), the 1.5°C (2.0°C) exceedance probability is 50%. For every 0.01°C increase in median temperature, 1.5°C exceedance probability increases by around 1.4% and 2.0°C exceedance probability increases by around 0.9%. The difference in gradient is because the uncertainty increases as the median temperature projection increases i.e. we have wider distributions once we get to around 2.0°C warming. The relationship is remarkably consistent across the emulators, with some small variations that become more noticeable as we get into the tails of the distributions i.e. as the median temperature moves away from the exceedance threshold of interest.   Figure S6.
Difference between the emulator output underlying the SR1.5 scenario categorisation and AR5-like MAGICCv7.5.3 for median (top row) and 67 th -percentile (bottom row) projections for 2100 (left column) and peak (right column) temperatures. The data is the same as Supplementary Figure S5 but has been plotted as a histogram rather than a scatter plot. projects higher concentrations (in general) than FaIRv1.6.2 which is part of the reason for differences in emissions-driven runs like those performed in WG3.  The data is the same as Supplementary Figure S11 but has been plotted as a histogram rather than a scatter plot.