The Many Possible Futures of Climate Models Kilometer-scale models, machine learning and model diversity

Wrote about the new approaches people are taking to building climate models: km-resolution, automated tuning, ML emulators…

Climate has benefited from having 10s of models to test and compare, now we have new dimensions of model diversity to learn from

notesonclimate.substack.com/p/the-many-p...

Record high ocean temperatures off southern California raise fears of prolonged marine heatwave Researchers warn the high-pressure conditions could disrupt marine life and ecosystems if it continues

Recorded warm temperatures off Scripps pier

www.theguardian.com/us-news/2026...

I think they are confusing changes in extremes with day-to-day predictability

Why You Hate Your Weather App As the weather becomes less predictable, we need forecasts and apps that are better at telling us what we don’t know.

This is a fun piece about the design challenge of communicating uncertainty in weather apps (though I’d quibble with the claim that climate change is making weather more unpredictable)

www.newyorker.com/culture/infi...

Emulating Natural Climate Variability With InVERT (Internal Variability Emulator for Regional Temperature) We present a method for low-cost emulation of spatially resolved natural temperature variability at monthly resolution The emulator output's statistical, spatial, and temporal characteristics res...

This looks useful: 'Emulating Natural Climate Variability With InVERT (Internal Variability Emulator for Regional Temperature)' agupubs.onlinelibrary.wiley.com/doi/10.1029/...

Back to basics: if we want to understand the pattern effect we need to think about wave dynamics

Hi Ray, I agree with the sentiment (though they claim no extra fuel was burned on these flights). For me the real advance here is the ability to accurately forecast contrails, which could be a useful tool for learning about ice clouds

Most long-lived contrails form within cirrus clouds with uncertain climate impact - Nature Communications The climate impact of aviation results largely from contrail-cirrus clouds. This study shows that most contrail-cirrus exist within natural cirrus clouds. Depending on cloud properties, their climate ...

Eg www.nature.com/articles/s41...

Still, the reduction in contrails suggests the forecasting system can accurately predict when&where clear sky contrails (at least as picked up by GOES) will form. Given all the unknowns about contrails this is an exciting tool that we could learn a lot from

Our new study explores how AI can reduce the climate impact of air travel. Our new study shows what happens when contrail avoidance is built directly into the tools airlines already use.

There’s a fun new study by Google & American Airlines optimizing flight paths to reduce contrails

blog.google/innovation-a...

These are clear-sky contrails tho & some recent papers suggest in-cloud contrails are much more common (with uncertain warming effects), similar to the ship track debates

🌍 We're hiring a postdoc at the intersection of cutting-edge #AI and #Earth system science. Help us redefine how we build and evaluate the next generation of climate models. Based at @ucsandiego.bsky.social , funded by Google.org. 🧵

Thanks for the great thread Cristi! Shows the danger of fooling ourselves by trusting the paleo data too much

Even if we have to be wary of using the historical record for emergent constraints, we may be able to use it to grind through process uncertainties as more signals emerge

We can also combine this model w/the TOA models to get the atmospheric forcing ~= the direct precip response to CO2. This is surprisingly climate-dependent: weakly negative today, but more negative in warmer/humid climates and positive in cooler/drier climates.

An Analytical Model for CO2 $ Surface Forcing, With Application to the Direct Precipitation Response An analytical model for CO2 ${\text{CO}}_{2}$’s surface forcing is developed, incorporating the 667 and 1000 cm−1 bands Surface forcing is non-monotonic and peaks at relatively cool surface tempe...

In a new paper led by Yue Xu, w/Daniel Koll, we present an analytic model for CO2 surface forcing

agupubs.onlinelibrary.wiley.com/doi/10.1029/...

There’s been much recent work building simple spectral models of TOA forcing, we apply similar ideas to the surface

For a broad group Kang et al 2009 (energy fluxes and ITCZ shifts) and Armour et al 2013 (pattern effect) would be good reads

…how high cloud CREs evolve after detainment.

Lots of implications for high cloud feedbacks and parameterizations

An Analytical Model of the Lifecycle of Tropical Anvil Cloud Radiative Effects An analytical model is presented for the lifecycle of tropical anvil cloud radiative effects, including for cloud spreading The model reproduces the cloud radiative effect lifecycle seen in a clo...

New paper with Casey Wall and. @blazgaspa.bsky.social developing an “Analytical Model of the Lifecycle of Tropical Anvil CREs”

agupubs.onlinelibrary.wiley.com/doi/10.1029/...

We combine a simple radiative transfer param, with cloud spreading and microphysics models to get a pen-and-part model for…

White Moons, Blue Sunsets Atmospheric composition and color

Trying something a little different: why do we think of the Sun as yellow and the Moon as white? And why are Martian sunsets blue?

notesonclimate.substack.com/p/white-moon...

My first sole-author paper is now published! 🎉🎉🎉

It's the culmination of a lot of thinking, so I hope you enjoy it.

Thanks to the friends, colleagues, and reviewers who helped make this happen 😊

agupubs.onlinelibrary.wiley.com/doi/10.1029/...

...but if ice loss accelerates and makes SSTs more Pliocene-like, effective sensitivity could be high

Interesting new paper by @vtcoop.bsky.social argues high Pliocene sensitivity is mostly a pattern effect from non-CO2 factors, esp ice sheets:

www.pnas.org/doi/10.1073/...

Interpretation seems subtle: if ice sheets ~stable Pliocene may be a poor analogue for 21st-century warming (too hot)...

Share your comments on NCAR with NSF Tell NSF the future that you want for NCAR by 13 March

@agu.org has a tool to help you submit a comment to NSF on the importance of NCAR: agu.quorum.us/campaign/154...

#SaveNCAR

Thanks to @andrewilwilliams.bsky.social for reminding me about the H2O continuum

New post on the shape of the water vapor feedback:

notesonclimate.substack.com/p/what-contr...

Surprisingly, the WV feedback is controlled by surface temperature, not how much moisture is in the air

Bridging Physics & AI in Climate Modeling 🌍🤖

Introducing JCM v1.0: A fully differentiable, intermediate-complexity atmospheric model built in Python/JAX.

Training hybrid models is hard because legacy code lacks gradients. JCM solves this.

Do you like charts? Oh yes you do. I've just published hundreds of them, as I do every year. www.nathanielbullard.com/presentations

(Compute might be the least important factor currently, but will be more and more important)

Thanks to @raspstephan.bsky.social , @janniyuval.bsky.social and Tom Beucler for very helpful discussions

Why is Weather Forecasting such a Good Problem for ML? Infrastructure, Uncertainties and Compute

New piece trying to understand why it took <5 years for ML models to not just match, but beat traditional forecasts

notesonclimate.substack.com/p/why-is-wea...

My guess is ML's success comes from 3 factors: (1) infrastructure, (2) how it handles uncertainties inherent to forecasting, (3) compute

A Unified Theory for the Global Thunderstorm Distribution and Land–Sea Contrast Global thunderstorm hotspots occur where CAPE is large, and where thunderstorms efficiently convert CAPE into kinetic energy Kinetic energy conversion is most efficient when entrainment effects a...

and Peters et al used entraining CAPE to understand the distribution of thunderstorms

agupubs.onlinelibrary.wiley.com/doi/10.1029/...

State‐Dependence of Polar Amplification in an Idealized GCM Polar amplification (PA) is simulated in a wide range of climate states using an idealized moist atmospheric general circulation model In the absence of ice-albedo feedbacks, PA increases with in...

@andrewilwilliams.bsky.social and Tim Merlis looked at the state-dependence of polar amplification in an idealized GCM, which peaks in warm climates with large temperature gradients (hence strong energy transport):

agupubs.onlinelibrary.wiley.com/doi/10.1029/...