diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml index 586679fc..585c7cad 100644 --- a/.pre-commit-config.yaml +++ b/.pre-commit-config.yaml @@ -1,19 +1,19 @@ repos: - repo: https://github.com/pre-commit/pre-commit-hooks - rev: v4.4.0 + rev: v6.0.0 hooks: - id: end-of-file-fixer - id: trailing-whitespace -- repo: https://github.com/psf/black - rev: 23.3.0 +- repo: https://github.com/psf/black-pre-commit-mirror + rev: 26.1.0 hooks: - id: black - repo: https://github.com/kynan/nbstripout - rev: 0.6.1 + rev: 0.9.0 hooks: - id: nbstripout - repo: https://github.com/nbQA-dev/nbQA - rev: 1.7.0 + rev: 1.9.1 hooks: - id: nbqa-black #- id: nbqa-isort diff --git a/content/news/2511Blanke.md b/content/news/2511Blanke.md index 5a812002..1a4cb0e1 100644 --- a/content/news/2511Blanke.md +++ b/content/news/2511Blanke.md @@ -9,4 +9,4 @@ images: ['images/news/2511Blanke.png'] link: 'https://doi.org/10.48550/arXiv.2505.18017' --- -Deep generative models, AI systems that can learn to create realistic data, are increasingly used to represent complex physical systems. However, these models often produce results that break basic physical laws, such as conservation of energy or mass. This **LEAP [study](https://doi.org/10.48550/arXiv.2505.18017)**, led by **Matthieu Blanke**, introduces a new method, called **Split Augmented Langevin (SAL)**, that ensures AI-generated outputs obey these fundamental constraints. By enforcing physical constraints in the sampling algorithm of pre-trained diffusion models, the approach makes **AI-based simulations and forecasts more accurate and reliable**. The method shows promising results in climate science applications, paving the way for AI tools that better respect the laws of nature. \ No newline at end of file +Deep generative models, AI systems that can learn to create realistic data, are increasingly used to represent complex physical systems. However, these models often produce results that break basic physical laws, such as conservation of energy or mass. This **LEAP [study](https://doi.org/10.48550/arXiv.2505.18017)**, led by **Matthieu Blanke**, introduces a new method, called **Split Augmented Langevin (SAL)**, that ensures AI-generated outputs obey these fundamental constraints. By enforcing physical constraints in the sampling algorithm of pre-trained diffusion models, the approach makes **AI-based simulations and forecasts more accurate and reliable**. The method shows promising results in climate science applications, paving the way for AI tools that better respect the laws of nature. diff --git a/content/news/2511Danni.md b/content/news/2511Danni.md index 7cb810c7..1237fa4a 100644 --- a/content/news/2511Danni.md +++ b/content/news/2511Danni.md @@ -9,4 +9,4 @@ images: ['images/news/2511DanniDu.png'] link: 'https://doi.org/10.22541/essoar.176083747.76188196/v2' --- -Ocean models often struggle to represent how water mixes vertically, leading to persistent temperature and circulation biases. In this [preprint](https://doi.org/10.22541/essoar.176083747.76188196/v2), **Danni Du** and colleagues use machine learning (ML) to correct those biases by **learning directly from data assimilation outputs in NOAA’s GFDL SPEAR system**. When integrated into the ocean model, the ML corrections improved temperature and mixing accuracy, outperforming existing correction methods. Combining ML with traditional approaches produced even better results, leading to **more realistic sea surface temperatures and ocean structure**. This approach can be applied to other climate models, offering a powerful new way to make ocean simulations more accurate. \ No newline at end of file +Ocean models often struggle to represent how water mixes vertically, leading to persistent temperature and circulation biases. In this [preprint](https://doi.org/10.22541/essoar.176083747.76188196/v2), **Danni Du** and colleagues use machine learning (ML) to correct those biases by **learning directly from data assimilation outputs in NOAA’s GFDL SPEAR system**. When integrated into the ocean model, the ML corrections improved temperature and mixing accuracy, outperforming existing correction methods. Combining ML with traditional approaches produced even better results, leading to **more realistic sea surface temperatures and ocean structure**. This approach can be applied to other climate models, offering a powerful new way to make ocean simulations more accurate. diff --git a/content/news/2511Samudra.md b/content/news/2511Samudra.md index c7f78372..c97457c9 100644 --- a/content/news/2511Samudra.md +++ b/content/news/2511Samudra.md @@ -21,4 +21,4 @@ M²LInES emulator Samudra was recently featured on 2 platforms: 🎬 **Prof Grace Lindsay Youtube channel** - 5 Minute Papers AI for the Planet: How AI can speed up our study of the ocean {{< youtube ijyF16uy0Hk >}} -
\ No newline at end of file +
diff --git a/content/news/2511Samudrace.md b/content/news/2511Samudrace.md index 31121807..dd2cf047 100644 --- a/content/news/2511Samudrace.md +++ b/content/news/2511Samudrace.md @@ -11,4 +11,4 @@ link: 'https://medium.com/@lz1955/samudrace-a-fast-accurate-efficient-3d-coupled Our latest **[blogpost](https://medium.com/@lz1955/samudrace-a-fast-accurate-efficient-3d-coupled-climate-ai-emulator-fcef3c60b079) dives into the story behind SamudrACE**, the first 3D AI ocean–atmosphere–sea-ice climate emulator. Developed in collaboration with **M²LInES, AI2, and NOAA GFDL**, SamudrACE marks a major milestone in the use of AI for climate science. -The post explores how the team built a model capable of simulating 1500 years of climate in just one day on a single GPU, making state-of-the-art climate modeling accessible to anyone, without the need for supercomputers or deep expertise in numerical modeling. \ No newline at end of file +The post explores how the team built a model capable of simulating 1500 years of climate in just one day on a single GPU, making state-of-the-art climate modeling accessible to anyone, without the need for supercomputers or deep expertise in numerical modeling. diff --git a/content/news/2511Sane.md b/content/news/2511Sane.md index aa09e2a8..e253d2f6 100644 --- a/content/news/2511Sane.md +++ b/content/news/2511Sane.md @@ -9,4 +9,4 @@ images: ['images/news/2511Sane.png'] link: 'https://doi.org/10.31219/osf.io/uab7v_v2' --- -A new [study](https://doi.org/10.31219/osf.io/uab7v_v2), led by **Aakash Sane**, introduces **a two step method to improve how ocean surface mixing is represented in models**. First, neural networks predict vertical diffusivity while respecting key physical constraints. Then, symbolic regression converts these predictions into simple equations that match the neural network accuracy but are easier to interpret. The resulting formulas reveal how friction velocity, buoyancy flux, Earth’s rotation and boundary layer depth shape mixing and expose a flaw in the standard physics based scheme. This approach provides **a transparent, efficient and physically grounded way to model ocean vertical mixing.** \ No newline at end of file +A new [study](https://doi.org/10.31219/osf.io/uab7v_v2), led by **Aakash Sane**, introduces **a two step method to improve how ocean surface mixing is represented in models**. First, neural networks predict vertical diffusivity while respecting key physical constraints. Then, symbolic regression converts these predictions into simple equations that match the neural network accuracy but are easier to interpret. The resulting formulas reveal how friction velocity, buoyancy flux, Earth’s rotation and boundary layer depth shape mixing and expose a flaw in the standard physics based scheme. This approach provides **a transparent, efficient and physically grounded way to model ocean vertical mixing.** diff --git a/content/news/2512AGU.md b/content/news/2512AGU.md index 6205fabb..e93539dc 100644 --- a/content/news/2512AGU.md +++ b/content/news/2512AGU.md @@ -5,78 +5,78 @@ heroHeading: '' heroSubHeading: 'AGU 2025 – M²LInES team members and affiliates Schedule' heroBackground: '' thumbnail: 'images/news/agu25.jpg' -images: +images: link: '' --- ### 📅 Monday, 15 December 2025 -Mitch Bushuk — [Antarctic Sea Ice Trends Across a High-Resolution Coupled Model Hierarchy](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1950675) +Mitch Bushuk — [Antarctic Sea Ice Trends Across a High-Resolution Coupled Model Hierarchy](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1950675) 🖼️ Poster | 14:15–17:45 | Hall EFG (Poster Hall), NOLA CC --- ### 📅 Tuesday, 16 December 2025 -Sara Shamekh — [Precipitation Intensity Sensitivity to Large-Scale Thermodynamic State](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1952465) +Sara Shamekh — [Precipitation Intensity Sensitivity to Large-Scale Thermodynamic State](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1952465) 🖼️ Poster | 08:30–12:00 | Hall EFG (Poster Hall), NOLA CC -Nathanael Zhixin Wong — [Investigating how Different Large-Scale Environmental Conditions impact the Shallow-to-Deep Transition of Convection](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1934844) +Nathanael Zhixin Wong — [Investigating how Different Large-Scale Environmental Conditions impact the Shallow-to-Deep Transition of Convection](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1934844) 🖼️ Poster | 08:30–12:00 | Hall EFG (Poster Hall), NOLA CC -Pavel Perezhogin — [NG23A-05 Generalizable Neural-Network Parameterization of Mesoscale Eddies in Idealized and Global Ocean Models](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1896408) +Pavel Perezhogin — [NG23A-05 Generalizable Neural-Network Parameterization of Mesoscale Eddies in Idealized and Global Ocean Models](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1896408) 🎤 Oral presentation | 14:57–15:07 | Room 298–299, NOLA CC -Renaud Falga — [NG23A-06 Towards a Unified Data-Driven Boundary Layer Parameterization for Ocean and Atmosphere](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1959465) +Renaud Falga — [NG23A-06 Towards a Unified Data-Driven Boundary Layer Parameterization for Ocean and Atmosphere](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1959465) 🎤 Oral presentation | 15:07–15:17 | Room 298–299, NOLA CC -Griffin Mooers — [NG23A-08 First Coupled gSAM - Neural Network Simulations to Improve Representation of Precipitation in Climate Simulations](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1980619) +Griffin Mooers — [NG23A-08 First Coupled gSAM - Neural Network Simulations to Improve Representation of Precipitation in Climate Simulations](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1980619) 🎤 Oral presentation | 15:27–15:37 | Room 298–299, NOLA CC -Adam Subel — [NG24A-03 Probing the Dynamical Response of Ocean Climate Emulators (Invited)](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1859610) +Adam Subel — [NG24A-03 Probing the Dynamical Response of Ocean Climate Emulators (Invited)](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1859610) 🎤 Oral presentation | 16:35–16:45 | Room 298–299, NOLA CC -Shuchang Liu — [NG24A-07 CERA: A Framework for Improved Generalization of Machine Learning Models to Changed Climates](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1940186) +Shuchang Liu — [NG24A-07 CERA: A Framework for Improved Generalization of Machine Learning Models to Changed Climates](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1940186) 🎤 Oral presentation | 17:15–17:25 | Room 298–299, NOLA CC --- ### 📅 Wednesday, 17 December 2025 -Alex Connolly — [Data-driven models of a coefficient in a higher-order closure of atmospheric boundary layer turbulence](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1988373) +Alex Connolly — [Data-driven models of a coefficient in a higher-order closure of atmospheric boundary layer turbulence](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1988373) 🖼️ Poster | 08:30–12:00 | Hall EFG (Poster Hall), NOLA CC -Fabrizio Falasca — [Toward Causally-Constrained, Reduced Stochastic Neural Emulators of the Full Ocean](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1945761) +Fabrizio Falasca — [Toward Causally-Constrained, Reduced Stochastic Neural Emulators of the Full Ocean](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1945761) 🖼️ Poster | 08:30–12:00 | Hall EFG (Poster Hall), NOLA CC -Danni Du — [Reducing Coupled Model Biases with Machine Learning Corrections from Ocean Data Assimilation Increments](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1896240) +Danni Du — [Reducing Coupled Model Biases with Machine Learning Corrections from Ocean Data Assimilation Increments](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1896240) 🖼️ Poster | 08:30–12:00 | Hall EFG (Poster Hall), NOLA CC -Pierre Gentine — [B31B-05 Global model data fusion to unravel land carbon sinks and their changes (Invited)](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1859477) +Pierre Gentine — [B31B-05 Global model data fusion to unravel land carbon sinks and their changes (Invited)](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1859477) 🎤 Oral presentation | 09:10–09:20 | Room 261–262, NOLA CC -Pierre Gentine — [B34A-03 Parsimony versus complexity (Invited)](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1859486) +Pierre Gentine — [B34A-03 Parsimony versus complexity (Invited)](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1859486) 🎤 Oral presentation | 16:35–16:45 | Room 265–266, NOLA CC --- ### 📅 Thursday, 18 December 2025 -Anurup Naskar — [Multivariate Estimation of Vertical Profiles to Better Understand the Shallow-to-Deep Transition of Convection in the Bankhead National Forest](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1954999) +Anurup Naskar — [Multivariate Estimation of Vertical Profiles to Better Understand the Shallow-to-Deep Transition of Convection in the Bankhead National Forest](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1954999) 🖼️ Poster | 08:30–12:00 | Hall EFG (Poster Hall), NOLA CC -Matthieu Blanke — [GC42A-04 Physically-Constrained Deep Generative Modeling](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1992938) +Matthieu Blanke — [GC42A-04 Physically-Constrained Deep Generative Modeling](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1992938) 🎤 Oral presentation | 11:04–11:14 | New Orleans Theater C, NOLA CC -Pierre Gentine — [Emulating climate variability and extremes with a diffusion-based model trained on CESM2 and finetuned on ERA5](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1910646) +Pierre Gentine — [Emulating climate variability and extremes with a diffusion-based model trained on CESM2 and finetuned on ERA5](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1910646) 🖼️ Poster | 14:15–17:45 | Hall EFG (Poster Hall), NOLA CC --- ### 📅 Friday, 19 December 2025 -Fabrizio Falasca — [GC51A-04 Neural models of multiscale systems: conceptual limitations, stochastic parametrizations, and a climate application](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1946408) -🎤 Oral presentation | 09:04–09:14 | New Orleans Theater C, NOLA CC \ No newline at end of file +Fabrizio Falasca — [GC51A-04 Neural models of multiscale systems: conceptual limitations, stochastic parametrizations, and a climate application](https://agu.confex.com/agu/agu25/meetingapp.cgi/Paper/1946408) +🎤 Oral presentation | 09:04–09:14 | New Orleans Theater C, NOLA CC diff --git a/content/news/2512Otness.md b/content/news/2512Otness.md index abea4c5a..5f7107af 100644 --- a/content/news/2512Otness.md +++ b/content/news/2512Otness.md @@ -9,4 +9,4 @@ images: ['images/news/2512Otness.png'] link: 'https://doi.org/10.48550/arXiv.2510.22676' --- -In this [article](https://doi.org/10.1088/2632-2153/ae1a36), Karl Otness and co-authors present a new multiscale machine-learning approach designed to **improve predictions in dynamical systems**. The method captures information moving both from fine to coarse scales and from coarse to fine, **boosting model accuracy and stability**, with only **minimal added computational cost** compared to standard architectures. The team evaluates the approach on an idealized fluid-dynamics closure task, where the multiscale networks learn to correct a chaotic model by representing unresolved small-scale processes. The work highlights the **potential of multiscale AI architectures to enhance the reliability of physical system modeling.** \ No newline at end of file +In this [article](https://doi.org/10.1088/2632-2153/ae1a36), Karl Otness and co-authors present a new multiscale machine-learning approach designed to **improve predictions in dynamical systems**. The method captures information moving both from fine to coarse scales and from coarse to fine, **boosting model accuracy and stability**, with only **minimal added computational cost** compared to standard architectures. The team evaluates the approach on an idealized fluid-dynamics closure task, where the multiscale networks learn to correct a chaotic model by representing unresolved small-scale processes. The work highlights the **potential of multiscale AI architectures to enhance the reliability of physical system modeling.** diff --git a/content/news/2512Zanna.md b/content/news/2512Zanna.md index 72c33fb8..3d60727a 100644 --- a/content/news/2512Zanna.md +++ b/content/news/2512Zanna.md @@ -9,4 +9,4 @@ images: ['images/news/2512framework.png'] link: 'https://doi.org/10.48550/arXiv.2510.22676' --- -In this [preprint](https://doi.org/10.48550/arXiv.2510.22676), M²LInES demonstrates the power of **AI driven methods in producing reliable climate simulations.** We introduce a new framework that brings physics- and scale-aware machine learning into climate models. Traditional parameterizations of physical processes often produce significant biases, but AI can now learn these processes directly from data. Our team **implements a suite of data-driven parameterizations in the ocean and sea-ice components of a state-of-the-art model**, ranging from deep learning to interpretable equation-based methods. Our results demonstrate that AI-driven parameterizations can run effectively in operational climate simulations, enabling **hybrid atmosphere–ocean–sea-ice modeling. All tools are open source and available to the community.** \ No newline at end of file +In this [preprint](https://doi.org/10.48550/arXiv.2510.22676), M²LInES demonstrates the power of **AI driven methods in producing reliable climate simulations.** We introduce a new framework that brings physics- and scale-aware machine learning into climate models. Traditional parameterizations of physical processes often produce significant biases, but AI can now learn these processes directly from data. Our team **implements a suite of data-driven parameterizations in the ocean and sea-ice components of a state-of-the-art model**, ranging from deep learning to interpretable equation-based methods. Our results demonstrate that AI-driven parameterizations can run effectively in operational climate simulations, enabling **hybrid atmosphere–ocean–sea-ice modeling. All tools are open source and available to the community.** diff --git a/content/news/Newsletters/_index.md b/content/news/Newsletters/_index.md index 2f85c074..15636dc5 100644 --- a/content/news/Newsletters/_index.md +++ b/content/news/Newsletters/_index.md @@ -9,7 +9,7 @@ tags: Links to our past newsletters are below. - + ### 2025 * 12/02/2025 - [M²LInES newsletter - December 2025](https://mailchi.mp/14605e5ed14c/m2lines-dec2025) diff --git a/content/team/AnurupNaskar.md b/content/team/AnurupNaskar.md index 2c15537d..d8abf89d 100644 --- a/content/team/AnurupNaskar.md +++ b/content/team/AnurupNaskar.md @@ -5,7 +5,7 @@ image: "/images/team/AnurupNaskar.png" jobtitle: "Affiliate, Graduate Student" promoted: true weight: 26 -Website: +Website: Position: Climate Informatics tags: [Atmosphere, Machine Learning, Climate Model Development] --- diff --git a/content/team/DiajengWulandariAtmojo.md b/content/team/DiajengWulandariAtmojo.md index 66e6fe95..eccb53c9 100644 --- a/content/team/DiajengWulandariAtmojo.md +++ b/content/team/DiajengWulandariAtmojo.md @@ -6,7 +6,7 @@ jobtitle: "Affiliate, Graduate Student" promoted: true weight: 26 Website: -Position: +Position: tags: [Sea-Ice, Machine Learning, Climate Model Development] ---