Deep-Learning Climate Emulation
Learned models of the coupled Earth system that run for years to centuries, for seasonal and climate questions rather than daily weather.
Tier Grades
Tier A: Frontier AI labs
feasible
University of Washington and NVIDIA publish long-run emulators.
Tier B: Indonesian AI labs
stretch
Emulation is described as efficient, but no compute figure is given.
Tier C: Institution units such as ITB
stretch
Running a released emulator is possible, but no compute figure is given.
Tier D: Organisations such as SEMAI (PoT, PoC, PoV)
not feasible
Century-scale climate simulation is outside proof-of-concept scope.
Where this comes from
Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value. Each rationale cites the anchors below.
Anchors
- ComputeDLESyM: described as parsimonious and highly efficient; compute not stated in the abstract.
- DataDLESyM: historical reanalysis plus satellite observations; stable 1000-year simulations without drift.
- Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 7 authors on the DLESyM paper (arXiv 2409.16247).
- Time horizonMulti-year.
- Shelf-life signalsDLESyM matches or exceeds four CMIP6 models on key variability metrics, including the Indian Summer monsoon.
- Pipeline roleTier A and universities build emulators; Tier B and Tier C study regional climate with them; Tier D has no observed role.
Where this comes from
Team size is a proxy (author counts of exemplar papers), not a headcount.
Who and What
Links
R&D TopicDeep-Learning Climate Emulation
Models
DLESyMUniversity of Washington, weather
Climate in a Bottle (cBottle)NVIDIA, weather
NeuralGCM (0.7 and ENS)Google, weather
DLESyMUniversity of Washington, weatherClimate in a Bottle (cBottle)NVIDIA, weatherNeuralGCM (0.7 and ENS)Google, weatherUse Casesnone recorded
Who Runs Itnone recorded
Papers
- WeatherBench: A benchmark dataset for data-driven weather forecastingStephan Rasp et al., arXiv preprint, 2020 founding
- ACE2: Accurately learning subseasonal to decadal atmospheric variability and forced responsesOliver Watt-Meyer et al., arXiv preprint, 2024 state of the art
- LUCIE: A Lightweight Uncoupled ClImate Emulator with long-term stability and physical consistency for O(1000)-member ensemblesHaiwen Guan et al., arXiv preprint, 2024 state of the art
- ClimaX: A foundation model for weather and climateTung Nguyen et al., arXiv preprint, 2023 state of the art
- Spherical Fourier Neural Operators: Learning Stable Dynamics on the SphereBoris Bonev et al., arXiv preprint, 2023 state of the art
Where this comes from
The papers to read first: where the direction began, its current best, and a survey when one exists.
Sources
- arxiv.org/abs/2409.16247 https://arxiv.org/abs/2409.16247
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Deep-Learning Climate Emulation", https://aurion.semai.id/topics/deep-learning-climate-emulation/, records of October 2026.