Observational-based estimates of climate sensitivity: impacts of aerosol evolution, natural variability and the recent temperature records

Climate sensitivity and aerosol forcing are two of the most central, but uncertain, quantities in climate science - crucial for understanding past climate changes and future projections. In addition, both historical and future climate evolution has been and will be influenced by natural variability. In this study, we estimate inferred climate sensitivity (ECS_inf) and aerosol forcing using observations of surface temperature and ocean heat content (OHC) combined with prior knowledge of effective radiative forcing over the industrial period, within a Bayesian framework. The global mean surface temperature set new records in 2023 and 2024. Including these years had little influence on the estimated ECS_inf - due to the steadily increasing OHC - compared to previous estimates using shorter observational records. In earlier studies, where observations up to the year 2010, 2014, 2019 and 2022 were included, the ECS_inf remained stable with best estimates from 1.9 to 2.2 K and the transient climate response best estimates from 1.4 to 1.6 K. A limitation in observational based estimates of climate sensitivity is the large uncertainty in the forcing of the Earth system, primarily due to the uncertain cooling effect from aerosols. The aerosol precursor emissions have declined over the past decade, but the evolution of aerosol forcing throughout the industrial period remains poorly constrained. Allowing aerosol forcing to vary more freely tends to stretch the upper tail of the ECS_inf distribution toward larger values. Another limitation of observational-based estimates of climate sensitivity is that it only captures the feedbacks that have occurred over the historical period - and the historical climate is only a single realization of the Earth’s climate. To assess this limitation, the method is tested using climate model results. We use the transient ocean heat content and temperature response from fully coupled historical simulations of four CMIP6 models – with substantial differences among ensemble members – and ERF time series calculated from the individual models to estimate ECS_inf. As expected, most ensemble members give posterior mean ECS_inf lower than the models ECS as only feedbacks over the historical period are captured. For the individual climate models, the posterior mean ECS_inf varies by 0.6 K, 1.2 K, 2.1 K and as much as 4.1 K across ensemble members. Although there are limitations within Earth System models, particularly in reproducing observed temperature patterns, this highlights the importance of natural variability in observational-based estimates of climate sensitivity.