During the session on “Confronting Models with Observations: Persistent Biases, Recent Trends, and Earth’s Energy Imbalance”, the ability of Earth System Models (ESMs) to represent historical trends and recent events was discussed. While it was clear that there are some significant issues in the ability of ESM’s to represent historical trends that need to be addressed, the presentations collectively demonstrated the broad range of tools and approaches that are being used to make progress in this realm, in order to understand where and why models are going wrong and also to constraining projections in the face of known model deficiencies.
Historical trends and uncertain futures
While there are a wide range of successes and failures when it comes to the ability of ESM’s to represent historical trends, it is clear that there are a handful of features that are receiving significant attention from the research community. The most dominant of these is the apparent inability of models to represent historical trends in the tropical Pacific. In short, the observational record depicts a relative cooling in the East Pacific (a “La Nina-like pattern”) while ESM’s tend to show the opposite, with the observed trends lying at the very edge or outside of the ESM distribution, depending on which ESM is being considered. This issue also manifests in historical sea-level trends. A presentation by John Fasullo demonstrated that there are considerable improvements in the representation of historical sea-level trends globally, including in the tropical Pacific at higher resolution (~0.1 degree ocean and ~0.25 degree atmosphere) in the CESM1 MESACLIP project. This work aligns with other presentations at the conference that demonstrated similar improvements in sea surface temperature trends. The presentation by Robb Jnglin Wills further elucidated a potential mechanism behind this trend bias in models, related to the representation of the Intertropical Convergence Zone (ITCZ). A common deficiency among models is that they tend to place a spurious secondary ITCZ in the south East Pacific – a feature not seen in observations. Robb’s work argues for a connection between this ITCZ bias and the inability of models to simulate the cooling in the East Pacific via the ITCZ bias acting to block the WES feedback associated with anomalies at higher latitudes from reaching the tropics. In alignment with the aforementioned sea level work, this work finds that higher resolution models tend to have a reduced ITCZ bias, which may be one reason for their improved representation of East Pacific SST and sea level trends. Both this work and the work of Hung-Yi Tseng make it clear that accurately simulating connections between the extra-tropics and the tropics will likely be important for representing tropical Pacific trends. Hung-Yi’s work demonstrated the significant impact that radiative perturbations in the extra-tropics can have on the tropical Pacific through idealized experiments.
Changes in precipitation extremes are an important impact of climate change and the ability of models to represent trends in precipitation extremes was discussed in the presentation by Chad Thackeray. For the most part models successfully represent historical trends in precipitation extremes and this work highlighted the utility of the single forcing simulations provided through CMIP in understanding these trends and the role that aerosols have played over the historical record in modulating the changes associated with greenhouse gas-induced warming. While coupled models are generally in agreement with observations, this study indicated some potential concerns as to whether this was entirely for the right reasons given that simulations with prescribed observation-based SST trends tend to underestimate extreme precipitation trends.
For some features of the climate system, substantial changes are yet to come. The work of Anne Kristin Close highlighted that the coming decades will be critical for the Antarctic Ice Sheet. Currently, models of the Antarctic Ice Sheets show large uncertainties on whether a collapse of the West Antarctic ice sheet will happen and on what timescale. This work highlighted not only the uncertainties on the ice sheet collapse itself, but also the uncertainties this introduces into future projections of sea level rise
Using recent events to evaluate process representation
The planet has provided us with a number of recent, relatively extreme, events that can be used to test the representation of processes in ESMs. One event is the Hunga Tonga Ha’apai Volcanic eruption in 2022 which injected a large amount of water vapor into the stratosphere with associated impacts on ozone and aerosols. In response to this event, the research community developed a protocol for “Tonga-MIP” which allowed for an intercomparison of the ability of models to respond to this eruption. The work of Margot Clyne presented in this session went far beyond simply evaluating the ability of models to represent this eruption. She delved into inner workings of two sectional aerosol models that responded to this event very differently, revealing the impacts of algorithmic issues within one of the schemes that contributed to sedimentation of aerosol particles that was too rapid.
Another recent event that was discussed in the presentation of Ko Tsuchida was an extreme increase in the Earth’s Energy Imbalance (EEI) that occurred in 2023. This work demonstrated that similar increases in EEI can be sampled from ESM’s and this approach was used to reveal the important role of a transition from a prolonged La Nina state into an El Nino state combined with external forcing in producing the one observed.
Both these works serves as a prime examples of how recent events in the observational record can be used to evaluate process-representation in ESM’s.
Novel approaches to constraining projection and improving understanding
While considerable progress is being made in improving models, in the meantime, approaches are being developed to constrain future projections with the current generation of models. A number of presentations in the session investigated methods of constraining future projections via the representation of observationally constrained features in the present climate. The work of Herve Douville discussed the “Kriging for Climate” approach – a Bayesian approach to emergent constraints – and its utility in constraining Arctic warming, relative humidity trends over land, Earth’s Energy Imbalance, and the Atlantic Meridional Overturning Circulation (AMOC). The work of Yiqun Tian demonstrated a link between the climatology of the Southern Ocean and global mean temperature trends in the sense that a model with warmer climatological Southern Ocean SST trends exhibits a weaker global mean SST warming, and the work of Michael Sigmond argued for a connection between future projections of the Northern Hemisphere jet stream and the representation of the climatological zonal winds in the “neck region” between the tropospheric mid-latitude jet and the polar vortex. Using this relationship, Michael reasoned that the jet stream projections are likely not biased given that models do not exhibit a significant bias in the neck winds. This result contradicts prior work on emergent constraints for the Northern Hemisphere jet stream related to the model response to sea-ice loss.
Path’s forward
The presentation by Lisa Bock demonstrated using ESMVal tool that ESM’s have improved throughout the ages in going from CMIP3 to CMIP6 with significant improvements arising when moving towards higher resolution. So ESM’s have been improving, in general, and there are many features and processes that are well represented. But as we continue to scrutinize our models and watch the forced signal start to emerge in our observational record, discrepancies are emerging. There are, however, exciting new directions underway to both improve our models and to improve our ability to constrain projections in the presence of model deficiencies.
Our enhanced ability to run ESM’s at higher resolution is clearly leading to new understanding and potentially significant improvements in process representation. Initial results presented in this session and in other parts of the conference demonstrate great promise that increasing resolution and starting to resolve features such as ocean eddies can help to reduce some model biases and the inability of models to represent trends in certain features of the climate system, in particular the Tropical Pacific. Studies at higher resolution have so far been limited to a small number of models, so it’ll be important to test robustness with a greater number of models and simulations.
In the meantime, emergent constraint approaches are clearly a powerful tool to constrain future projections but must be accompanied by accurate quantification of uncertainties and a solid mechanistic understanding of the origins of the constraint to ensure confidence.
Underpinning all of this work, however, is the presence of a reliable observational record as well as continued advances in ESMs and the availability of coordinated ESM simulations to the research community through efforts like CMIP.
