Blog

Coda: what it took to build a new Earth system model

By Tapio Schneider and Raffaele Ferrari A coda to our tour of the CliMA software stack. The series began with why we built a new Earth system model; last week we covered how it all came together. Over thirteen posts, we have toured the CliMA software stack: packages that, linked together, form a new Earth system model, the first built from scratch in decades. The tour went from sunlight and the thermodynamics of moist air, through radiation, clouds, and the algorithms that calibrate a model with data, to the numerical grid, the land, the ocean, and the atmosphere, and ended…
Read More

Putting it all together: ClimaCoupler.jl

By Henry Peterson, Akshay Sridhar, Tapio Schneider, and Raffaele Ferrari Part 13 of our tour of the CliMA software stack. The series began with why we built a new Earth system model; last week went from Hadley cells to atmospheric rivers. Using the weather app on your phone or computer, you have likely noticed that atmospheric forecast models, like the one presented last week, cannot predict the weather more than a week or two in advance. This is a property of chaotic systems such as the atmosphere: beyond some finite time, the evolution of such systems cannot be predicted. How,…
Read More

From Hadley cells to atmospheric rivers: ClimaAtmos.jl

By Costa Christopoulos, Zhaoyi Shen, and Tapio Schneider Part 12 of our tour of the CliMA software stack. The series began with why we built a new Earth system model; two weeks ago we covered why the Gulf Stream hugs the coast. If all the water vapor in Earth’s atmosphere rained out, it would cover the planet with less than an inch of water, as we saw when we met cloud microphysics. What matters for weather is where and when it rains. An atmospheric river, a band of moist air 400 to 600 km wide and thousands of kilometers long,…
Read More

Why the Gulf Stream hugs the coast: Oceananigans.jl

By Xin Kai Lee and Raffaele Ferrari Part 11 of our tour of the CliMA software stack. The series began with why we built a new Earth system model; last week covered how the atmosphere exchanges heat, water, and momentum with the surface. The ocean shapes Earth’s climate and weather. It is well known that the ocean absorbs about a quarter of our carbon dioxide emissions and most of the excess heat from global warming; other effects are less familiar. The climate of England and Portugal, for example, is much milder than that of the US coast at the same…
Read More

Why driving fast does not clean your car: SurfaceFluxes.jl

Part 10 of our tour of the CliMA software stack. The series began with why we built a new Earth system model; last week covered a land model that learns from data. No matter how fast you drive, the fine dust on your car will stay put. Air is viscous, and a viscous fluid sticks to a solid surface: right at the surface, the air moves with the car. This is the no-slip boundary condition of the Navier–Stokes equations. Even at 160 km/h (100 mph), the air within a fraction of a millimeter of the car’s finish is nearly at…
Read More

A land model that learns from data: ClimaLand.jl

By Alexandre A. Renchon and Katherine Deck Part 9 of our tour of the CliMA software stack. The series began with why we built a new Earth system model. Two weeks ago we covered how we put a planet on a grid; last week’s post was an intermezzo on AI and understanding in science. About 40% of the rain that falls on land returns to the atmosphere after passing through the leaves of plants, in a process called transpiration. Plants transpire through their stomata, small openings on their leaves through which they also absorb CO2 for photosynthesis; stomata in a…
Read More

Headlines and inside stories: understanding and trust in AI for mathematics, science, and engineering

[This post was prompted by the recent discussions about the Navier-Stokes Millennium Prize problem and is cross-posted from Terry Tao’s blog. It lays out the reasoning behind how we use AI at CliMA.] The apparent proof of finite-time blow-up of the forced Navier-Stokes equation, announced by OpenAI on September 8, has brought into focus a debate about the role of AI in mathematics. The proof was produced with a system of some 10,000 AI agents that explored many approaches in parallel; it was then formalized and verified in Lean. The formal verification supports its correctness, but mathematicians are still working…
Read More

Putting a planet on a grid: ClimaCore.jl and ClimaTimeSteppers.jl

Part 8 of our tour of the CliMA software stack. The series began with why we built a new Earth system model; last week covered how we learn a climate model’s parameters from data. Simulating a physical system in space and time requires a discretization: a grid in four-dimensional space-time, on whose nodes we predict, in the case of the atmosphere, temperatures, winds, humidity, rainfall, and other variables. CliMA’s discretization is split across two packages. ClimaCore.jl handles space; ClimaTimeSteppers.jl handles time. Four requirements shaped their design: The globe must be tiled into elements of roughly equal size, so that all…
Read More

From tuning by hand to learning from data: ClimaParams.jl and ClimaCalibrate.jl

By Nat Efrat-Henrici and Tapio Schneider. Part 7 of our tour of the CliMA software stack. The series began with why we built a new Earth system model; last week covered how to calibrate a model of a chaotic system. In the summer of 1968, a field campaign measured wind profiles over a field of wheat stubble in Kansas. Functional forms were fitted to those measured profiles, including a coefficient that sets how strongly stable temperature stratification damps turbulence near the ground. The fitted profiles with the coefficient, 4.7, have since been used to represent stable boundary layers in models…
Read More

How can we fit a climate model to observations, when no one can predict the weather a month from now?

Part 6 of our tour of the CliMA software stack. The series began with why we built a new Earth system model; last week covered how cloud droplets grow up to become raindrops. Climate models such as CliMA’s contain hundreds of adjustable parameters that shape the behavior of a simulation. They increase a model’s expressiveness where process understanding is incomplete, and they range from physically interpretable parameters in process models to the weights and biases of embedded machine learning (ML) models, all calibrated so that the model is as consistent with observed data as possible. But the climate system is…
Read More