Estimated human-induced warming from a linear temperature and atmospheric CO2 relationship

Assessing compliance with the human-induced warming goal in the Paris Agreement requires transparent, robust and timely metrics. Linearity between increases in atmospheric CO2 and temperature offers a framework that appears to satisfy these criteria, producing human-induced warming estimates that are at least 30% more certain than alternative methods. Here, for 2023, we estimate humans have caused a global increase of 1.49 ± 0.11 °C relative to a pre-1700 baseline.

Given the importance of being able to assess compliance with the temperature objectives set out in the Paris Agreement1, suitable methods for specifying human-induced warming (HIW) in near real time are urgently needed2. However, the magnitude of HIW is not directly observed but is, instead, estimated from global mean surface temperature anomaly data. There are two elements to this estimation. The first involves specifying a suitable pre-industrial baseline for the temperature anomaly data to produce estimates of global mean surface temperature (GMST) change. The second involves removing the effects of natural variability from the GMST change data to leave just HIW. The Intergovernmental Panel on Climate Change (IPCC) have made the pragmatic choice to use the mean of the 1850–1900 global temperature anomaly data as the pre-industrial baseline condition3, although it is known that both emissions4 and the atmospheric burden5 (Fig. 1b) were rising well before this period. Furthermore, the 1850–1900 data are the most uncertain in the global temperature anomaly series6 (Fig. 1a), and this uncertainty is currently not accounted for when applying baseline adjustments. A range of methods have emerged for filtering out natural variability; however, the associated HIW estimates either incur lag as a byproduct of the filtering of GMST data7 or require model forecasts to make HIW estimates independent of natural variability2,8,9,10. Clearly, lag is unwelcome when evaluating climate policy as it introduces delay in policy responses to observed change, something that is particularly problematic in the context of the risks presented by, for example, climate tipping points. Although employing climate modelling approaches to avoid lag effects appears sensible, this introduces significant and often difficult-to-quantify uncertainties.

a, HadCRUT5 global temperature anomalies relative to their 1961–1990 mean6, estimated GMST change and HIW (Methods). b, Law Dome ice core5 and Mauna Loa18 atmospheric CO2 concentrations. c, The relationship between increases in atmospheric CO2 concentration above its pre-1700 baseline and the HadCRUT5 global temperature anomaly, WLS regression fit and HIW estimates (Methods). d, Median HIW estimates for 2023 from ref. 16 along with the regression-based estimates from this study. Here the regression-based estimates are also shown baselined to 1850–1900 and without the CO2 baseline uncertainty to be comparable to those in ref. 16. All uncertainties are expressed as 95th percentile ranges. Regression uncertainties are estimated from N = 104 samples (Methods).

Leave a Comment