These simulations were only possible by a university researcher like me with a lot of technical help (and computer time) from @NCAR_Science , especially from co-authors Adam Herrington & Isla Simpson. This project has been a slow cooker we've been working on since late 2020 (6/6)
Promoting papers feels a bit strange this week considering everything going on, but I am excited about our new paper showing a larger influence of Gulf Stream anomalies on the atmospheric circulation over Europe in models that resolve weather fronts (1/6)
https://t.co/5tVgIBGuJr
Our paper and its implications for ongoing high-resolution modeling efforts is nicely summarized in an Editor's Highlight by Hannah Chistensen. (5/6)
https://t.co/f0kcie4Yz1
Using AI to understand the Earth System response:
The Forced Component Estimation Statistical Method Intercomparison Project (ForceSMIP) uses statistical and machine learning methods to try to figure out the true forced response, as reflected in the evolving pattern of surface temperature anomalies.
From the presentation by @ClimateAnomaly in the April 2024 ECS & Climate Sensitivity Symposium. https://t.co/uV2AAnJjXy
(3) discussion of the tradeoffs of high resolution and the lack of a specific scale where all of the relevant processes are resolved, (4) discussion of new ML approaches to optimize parameters directly, and (5) discussion of a balanced approach to bring this all together. (3/3)
To optimize climate models, how do we prioritize between resolution & ensemble size, between physical param. development & incorporating AI? In a perspective paper out this week, Tapio Schneider, Ruby Leung, and myself argue we need a balanced approach: https://t.co/Vm87lbjrPz
Included is (1) an evaluation of how climate skill has improved over the last 20 years and is continuing to (mostly) improve with efforts to increase resolution, (2) considerations for successful process-informed parameterizations, ... (2/3)
@CColose I suspect the quicker reversal in CESM1-HR is instead due to aerosols, ozone, etc. (and timing supports this). Peter's shot noise explanation of the low-frequency discrepancy provides a reason there might not be a real mismatch, but it warrants caution and work on improving OBS
Is the La-Niña-like warming pattern over the past 40+ years forced or unforced?
In this seminar, I argue it is forced & show statistical + hi-res model forced response estimates. Peter Huybers discusses obs. uncertainty & evidence models have too little low-freq. variability
Recording and chat transcript of April's ECS-Cloud Feedback symposium are now available:
https://t.co/ahqRHpiwue
Really fun "debate" between Peter Huybers and @ClimateAnomaly on "Can we rule out internal variability as the main driver of recent tropical SST trends?"
@CColose Maher et al. 2023 (https://t.co/kBehCudw5Z) Fig. 7 shows that the few models that show an ocean themostat response show it until 2100 (note, different boxes than Wills et al. 2022). The case of GFDL-ESM2M is discussed in Kohyama et al. 2017 (https://t.co/Ssg7qsDuoo) (1/2)
"#Climate models have done well, but they also show some biases. It is essential to improve them in order to understand the impact of #GlobalChange on regional weather", explains @ClimateAnomaly (@usys_ethzh). Statistical and machine learning methods offer one solution 🌎 #SGCD24
Join us next month for a panel discussion on
"Can we rule out internal variability as the main driver of recent tropical SST trends?"
w/Peter Huybers & @ClimateAnomaly
New paper in @PNASNews led with @cristiproist shows that a weird spatial pattern of temperature change has slowed global-mean warming since 1980. Because the pattern could evolve in the future, observed warming doesn’t help us constrain long-term warming.
https://t.co/Mnrs7V8RzK
Meet our Speaker Robert Jnglin Wills @ClimateAnomaly@usys_ethzh at the #SGCD24 and discuss what leads some climate change impacts to be robust and others uncertain.
➡️Register now: https://t.co/UfUlsNAjhp
Really excited to share our new paper on climate-invariant machine learning https://t.co/t2yg4pxeV3 to solve extrapolation issues under climate change, led by the great Tom Beucler