In the highECS project we have worked to gather evidence regarding Earth's climate sensitivity from:
- Pre-historic climates
- The instrumental record of warming
- Satellite observations
as well as participated in multiple international assessments that combined all available evidence of both ECS, and the closely related anthropogenic aerosol cooling, to widely accepted and far narrower ranges than was the case before (Sherwood et al. 2020, Bellouin et al. 2020, Forster et al. 2021, Arias et al. 2021). In addition, we have developed new reconstructions of the paleoclimates during the Last Glacial Maximum (LGM) and the Pliocene epoch. These new reconstructions are unbiased relative to the sparse proxy temperature records and take into account more sources of errors than earlier reconstructions, and so are therefore particularly suited for estimating ECS (Annan et al. 2022, 2023).
A series of studies have focused on using climate models to interpret observed historical warming (e.g. Jimenez-de-la-Cuesta and Mauritsen 2019, Gregory et al. 2019). Here an important focus area of several studies have been the influence on transient global warming by inhomogeneous patterns of sea surface warming that may temporarily dampen warming, thereby potentially masking a large ECS. These studies have focused both on understanding the origin of these patterns, as well as quantifying and bounding their influence on climate (Lewis and Mauritsen 2021, Andrews et al. 2022, Modak and Mauritsen 2023). The presence of such pattern effects, as well as uncertainty on the cooling induced by anthropogenic aerosols (Flynn and Mauritsen 2020, Huusko et al. 2022, Flynn et al. 2023), means that long term historical warming by itself is not a very strong constraint on ECS at the high end, even if substantial progress was made during the highECS project. For instance in Modak and Mauritsen (2023) we were able to constrain ECS to 1.8-11.0 K (5-95 percentiles), which is not great, but an advance over prior unbounded estimates (e.g. Forster et al. 2021).
Variations in the historical record can, however, be exploited to learn about ECS. For example the prominent 'pause' in global warming form 1998 to 2012, known as the hiatus. Such natural variations, on top of global warming, can occur due to internal variability mainly driven by El Nino and La Nina. We showed, contrary to prior beliefs, that the probability of observing such a long period of slow warming is extremely low if ECS exceeds 5-6 K (Modak and Mauritsen 2021). But also shorter variations from month to month, or year to year, can be used in conjunction with the satellite records to constrain feedback, and thereby potentially ECS (Uribe et al. 2022). However, we show in a follow-on study that the records are still too short for this. This work, in turn, has motivated me to apply for a new Earth explorer mission by the European Space Agency to measure radiation balance directly from space which will be new and unique.
Looking instead at pre-historic climates offers a better chance at bounding ECS. This is because these climates are usually close to equilibrium with the given boundary conditions, but the drawback is that the state, and also the boundary conditions, are less well known. In Renoult et al. (2020) we use the emergent constraint technique to establish the relationship between cooling in the Last Glacial Maximum (LGM) about 20 thousand years ago, warming during the Pliocene epoch about 3 million years ago and ECS using climate models. This relationship can then be used to determining ECS given estimates of the actual cooling and warming during these periods. There are, however, several challenges with using the cold LGM to predict a warm future (Renoult et al. 2023).
I have been involved in the writing of the Sixth IPCC Assessment Report, and in particular was responsible for the assessments of Earth's climate sensitivities, both ECS and also the transient climate response (TCR) to doubled CO2. This new assessment builds on Sherwood et al. (2020), but uses a simpler, more understandable statistical approach to combining evidence, and also integrates more information than what could previously be done. Thereby uncertainty regarding ECS has been reduced by a factor of 2 to 3 compared with the fifth assessment report from 2013. This is a major breakthrough given that uncertainty has essentially remained unchanged between 1979 and 2013. Another important invention is the use of emulators, rather than complex climate models, to translate the improved understanding of ECS into future model projections (Rohrschneider et al. 2019).
This has mainly been possible through two major community-driven assessments which the PI contributed to (Sherwood et al. 2020, Bellouin et al. 2020). In the former we applied a novel idea which is to assess as well as possible ECS based on three lines of independent evidence, namely process-understanding, historical warming and paleo climates, and then combining these through Bayesian updating, a kind of machine learning technique. By looking at the problem from multiple angles an increasing confidence is achieved. Here in particular the other assessment by Bellouin et al. (2020) contributed an updated understanding of anthropogenic aerosol forcing, which is one of the most important sources of uncertainty in interpreting historical warming. Aerosols cool climate, and therefore mask part of the greenhouse gas warming.
All in all, the highECS project and research conducted in the community has resulted in a leap forward in predicting Earth's future. It is very telling that politicians and the public now discuss for instance that current policies will lead to 2.9 degree global warming, and implementing all pledges leads 2.5 degree warming. Using such accuracy would not have been possible a decade ago, and it shows the strong impact the research has had for society and for quantifying what will be needed to achieve the Paris Agreement of staying well below 2 degrees.
