Description |
An ICTP Hybrid meeting
Applications of ML/AI to weather and climate modelling have experienced an exponential growth. Weather forecasts based on AI models trained on re-analysis data are now possible with forecasting skills comparable to those of physical numerical models.
However, to what extent the AI approach can be extended to longer time scales is a research topic of increasing importance. At the same time, traditional atmospheric and ocean models based on numerical solutions of PDEs have increased their resolution to a few kilometers, which makes their application on climate time scale very challenging. In order to test the validity of these methodologies on climate time scales in a statistically sound way, experiments on several multi-decadal simulations are required. In this context, intermediate-complexity climate models provide convenient 'numerical laboratories'. The goal of this workshop is to present state-of-the-art applications of AI/ML, advanced numerical methods and innovative hardware/software solutions tested on intermediate-complexity models, such as the SPEEDY global model developed at ICTP. The workshop will also review the state-of-the-art of AI development.
Topics:
Speakers and Panelists include:
Tom Beucler, University of Lausanne, Switzerland
Matthew Chantry, ECMWF, UK
Erika Coppola, ICTP, Italy
Alban Farchi, ECMWF, UK
Shruti Nath, University of Oxford, UKPierre Gentine, Columbia University, USA Graziano Giuliani, ICTP, Italy Scott Martin, University of Washington, USA
Tim Palmer, University of Oxford, UK
Erik van Sebille, Utrecht University, Netherlands Jakob Schloer, ECMWF, UK Research abstracts: In the application form, all applicants are requested to submit a brief research abstract for a poster presentation and/or contributed talk. A limited number of abstracts will be selected for the poster session or to give a talk. Please use ICTP templates available for download here and below under 'Material'.
Grants: A limited number of grants are available to support the attendance of selected participants, with priority given to participants from developing countries. There is no registration fee.
|
6th Summer School on Theory, Mechanisms and Hierarchical Modelling of Climate Dynamics: Artificial Intelligence and Climate Modelling | (smr 4067)
Go to day
-
-
08:30 - 09:00
Registration and administrative formalities
-
09:00 - 17:30
-
09:00
Opening
1h0'
Speaker: Fred Kucharski (ICTP), Serafina Di Gioia (ICTP), Riccardo Farneti (ICTP), Franco Molteni (Agenzia Italia Meteo), Adrian Tompkins (ICTP) Material: Slides
-
10:00
Testing Leonardo Accounts and Configuration
30'
Speaker: Serafina Di Gioia (ICTP, Italy) - 10:30 Coffee Break 30'
-
11:00
Introduction to Atmospheric Modelling
1h30'
Speaker: Franco Molteni (Agenzia Italiana Meteo, Italy) Material: Slides
- 12:30 Lunch 1h30'
-
14:00
Introduction to Ocean Modelling
1h0'
Speaker: Erik van Sebille (Utrecht University, Netherlands) Material: Slides
-
15:00
Intro to Machine Learning
1h0'
Speaker: Serafina Di Gioia (ICTP, Italy) Material: Slides
-
16:00
Introduction to Deep Learning
1h30'
Speaker: Serafina Di Gioia (ICTP, Italy) Material: Slides
-
09:00
Opening
1h0'
-
08:30 - 09:00
Registration and administrative formalities
-
-
09:00 - 17:30
-
09:00
Convolutional Neural Network (CNN) for Classification and Regression
1h30'
- 10:30 Coffee Break 1h0'
-
11:30
Deep Learning for Time Series (E.G. RNN, Long-Short Term Memory NN)
1h0'
- 12:30 Lunch 1h30'
-
14:00
Lab activity: Typical Structure and Input-Output of Climate Data
1h30' (
AGH - Informatics Laboratory
)
only for participants 'in person'
Speaker: Graziano Giuliani (ICTP, Italy) - 15:30 Coffee Break 30'
-
16:00
Lab activity: Simple Deep Learning Model to Have Temperature Forecast from Other Input Fields
1h30' (
AGH - Informatics Laboratory
)
only for participants 'in person'
Speaker: Serafina Di Gioia (ICTP, Italy)
-
09:00
Convolutional Neural Network (CNN) for Classification and Regression
1h30'
-
09:00 - 17:30
-
-
09:00 - 20:00
-
09:00
Introduction To Physics-Informed Neural Networks (PINN)
1h30'
- 10:30 Coffee Break 30'
-
11:00
Lab activity
1h30'
only for participants 'in person'
- 12:30 Lunch 1h30'
-
14:00
Lab activity: Comparison Between Traditional Numerical Schemes For Climate Simulations (PINN)
1h30' (
AGH - Informatics Laboratory
)
only for participants 'in person'
Speakers: Serafina Di Gioia (ICTP, Italy), Graziano Giuliani (ICTP, Italy) - 15:30 Coffee Break 30'
-
16:00
Lab activity: Comparison Between Traditional Numerical Schemes For Climate Simulations (PINN)
1h0' (
AGH - Informatics Laboratory
)
only for participants 'in person'
Speakers: Serafina Di Gioia (ICTP, Italy), Graziano Giuliani (ICTP, Italy) -
17:00
Poster session
1h30'
Poster presenters P01 - P24 only for participants 'in person'
Material: Booklet
- 18:30 Welcome reception 1h30'
-
09:00
Introduction To Physics-Informed Neural Networks (PINN)
1h30'
-
09:00 - 20:00
-
-
09:00 - 18:00
-
09:00
Introduction to Regional Climate Modelling and the Need of AI
1h30'
Speaker: Erika Coppola (ICTP, Italy) Material: Slides
- 10:30 Coffee Break 30'
-
11:00
Introduction to Graph NN (GNN; Difference between Static and Dynamic GNN)
1h0'
- 12:00 Lunch 1h0'
-
13:00
AI Physical Routines in Physical Numerical Models
1h0'
Remote talk
Speaker: Pierre Gentine (Columbia University, USA) Material: Slides
-
14:00
Online Poster Session in Breakout Rooms in Zoom
1h15'
BO room 1: AI for downscaling and regional climate modeling BO room 2: AI for Monsoon patterns BO room 3: Hybrid Climate modeling BO room 4: Extreme events and renewable energy BO room 5: Ocean modeling and satellite data analysis BO room 6: Climate change/climate resilience
Material: Booklet
-
15:15
Introducing PINA
45'
Speaker: Nicola Demo (MATHLAB, SISSA, Italy) - 16:00 Coffee Break 30'
-
16:30
Lab Activity On GNN
1h30' (
AGH - Informatics Laboratory
)
only for participants 'in person'
-
09:00
Introduction to Regional Climate Modelling and the Need of AI
1h30'
-
09:00 - 18:00
-
-
09:00 - 17:30
-
09:00
Review of State-of-the Art of AI Developments In Weather and Climate Modelling
1h30'
Remote talk
Speaker: Matthew Chantry (ECMWF, UK) -
10:30
Group photo
2'
for remote participants
- 10:30 Coffee Break 30'
-
10:50
Group photo
5'
for participants 'in person'
-
11:00
Towards Reliable Hybrid AI-Climate Modeling
1h30'
Speaker: Tom Beucler (University of Lausanne, Switzerland) Material: Slides
Slides with the blackboard discussion
- 12:30 Lunch 1h30'
-
14:00
Towards Reliable Hybrid AI-Climate Modeling
1h30'
Speaker: Tom Beucler (University of Lausanne, Switzerland) - 15:30 Coffee Break 30'
-
16:00
Lab Activity
1h30' (
AGH - Informatics Laboratory
)
only for participants 'in person'
Speaker: Tom Beucler (University of Lausanne, Switzerland)
-
09:00
Review of State-of-the Art of AI Developments In Weather and Climate Modelling
1h30'
-
09:00 - 17:30
-
-
09:00 - 19:00
-
09:00
Internal Variability in Decadal Ensemble Simulations
30'
Speaker: Franco Molteni (Agenzia Italiana Meteo, Italy) -
09:30
Round Table Discussion on Uncertainty Evaluation
1h0'
Speaker: Tom Beucler (University of Lausanne, Switzerland), Dario Coscia (SISSA, Italy), Franco Molteni (Agenzia Italiana Meteo, Italy), Jakob Schloer (ECMWF, Germany) - 10:30 Coffee break 30'
-
11:00
Integrating AI, Ocean Models, and Satellite Observations for Eddy-Resolving Surface Ocean State Estimation
1h30'
Speaker: Scott Martin (University of Washington, USA) - 12:30 Lunch 1h30'
-
14:00
Fine-Tuning of the Prithvi Foundational Model
1h0'
Speaker: Shruti Nath (University of Oxford, UK) - 15:00 Coffee Break 30'
-
15:30
Open Problem Description for Hackathon, and Creating Groups to Resolve Them
1h30'
Speaker: Shruti Nath (University of Oxford, UK) -
17:00
Poster session
2h0'
Poster presenters P25 - P49 only for participants 'in person'
Material: Booklet
-
09:00
Internal Variability in Decadal Ensemble Simulations
30'
-
09:00 - 19:00
-
-
09:00 - 16:30
-
09:00
Hybrid models at ECMWF, from spectral nudging to online trained model error corrections
1h30'
Speaker: Alban Farchi (ECMWF, UK) - 10:30 Coffee Break 30'
-
11:00
Date-Driven Sub-Seasonal to Seasonal Forecasts: From Domain-Specific to Global Models
1h30'
Remote talk
Speaker: Jakob Schloer (ECMWF, UK) - 12:30 Lunch 1h30'
-
14:00
Investigating Extreme Precipitation Event Attribution Through Deep-Learning Approaches
15'
Specialized Talk
Speaker: Jeremy Malcom Corner (University of Virginia, USA) Material: Abstract
-
14:15
Enhancing Seasonal Predictions with Machine Learning: A Global Perspective on SST Influence in Early Winter
15'
Specialized Talk
Speaker: Victor Galvan Fraile (University Complutense of Madrid, Spain) Material: Abstract
Slides
-
14:30
Retrieving tropical cyclone intensity from climate output: AI approaches and challenges
15'
Specialized Talk
Speaker: Chanh Quoc Kieu (Indiana University, USA) Material: Abstract
- 14:45 Coffee Break 30'
-
15:15
Enhancing Local Climate Zone Classification in Jakarta Using AI: Machine Learning Approach with Landsat 9 and Land Surface Temperature Integration
15'
Specialized Talk
Speaker: Rezky Yunita (Indonesian Agency for Meteorology, Indonesia) Material: Abstract
-
15:30
On the application of generative modelling for seasonal climate predictions
15'
Specialized Talk
Speaker: Lluis Palma I Garcia (Barcelona Supercomputing Center, Spain) Material: Abstract
Slides
-
15:45
CNN-based forecasting of early winter NAO using sea surface temperature
15'
Specialized Talk
Speaker: Elena Provenzano (Sorbonne University, France) Material: Abstract
Slides
-
16:00
Fine-tuning a global weather model for superior subseasonal forecasting
15'
Specialized Talk
Speaker: Vateanui Francois Ernest Sansine (French National Institute of Research for Sustainable Development, French Polynesia) Material: Abstract
-
16:15
Climate projections in the Euro-Mediterranean region: atmospheric circulation patterns and temperature and rainfall future changes
15'
Specialized Talk
Speaker: Matias Ezequiel Olmo (Barcelona Supercomputing Center, Spain) Material: Abstract
Slides
-
09:00
Hybrid models at ECMWF, from spectral nudging to online trained model error corrections
1h30'
-
09:00 - 16:30
-
-
09:30 - 16:45
-
09:30
Linking Research to Action: Exploring the Value of AI-Driven Rainfall Forecasts over the Greater Horn of Africa
1h30'
Speaker: Shruti Nath (University of Oxford, UK) - 11:00 Coffee Break 30'
-
11:30
AI And Regional Climate Downscaling
1h0'
Speaker: Valentina Blasone (University of Trieste, Italy) Material: Slides
- 12:30 Lunch 1h30'
-
14:00
AI-Driven reconstruction of oxygen profiles in the North Atlantic Oxygen Minimum Zone
15'
Specialized Talk
Speaker: Domitille Tihaine Coron (University of Western Britanny, France) Material: Abstract
-
14:15
Towards a Physics-Informed Neural Network for Ocean Circulation Modelling
15'
Specialized Talk
Speaker: Iuri Gorenstein (Universidade de São Paulo, Brazil) Material: Abstract
-
14:30
Learning the Ocean: Leveraging AI to Improve Ocean Data Assimilation
15'
Specialized Talk
Speaker: Davide Grande (IUSS - Istituto Universitario di Studi Superiori, Italy) Material: Abstract
-
14:45
Reconstruction of the Surface Geostrophic Current in the Mediterranean Sea from Absolute Dynamic Topography Generated by Generative Adversarial Network (Adtgan)
15'
Specialized Talk
Speaker: Gasana Elysee Manimpire (University of Trieste, Italy) Material: Abstract
- 15:00 Coffee Break 30'
-
15:30
A machine learning method is used to identify sources of long-term ENSO predictability in the ocean (sea surface temperature (SST) and heat content) and the atmosphere (near-surface zonal wind (U10)). Tropical 10m wind sseem to represents an inreresting source of ENSO predictability
15'
Specialized Talk
Speaker: Ioana Colfescu (National Centre for Atmospheric Science, UK) Material: Abstract
-
15:45
Data-Informed Inversion Model (DIIM): a framework to retrieve marine optical constituents in the BOUSSOLE site using a three-stream irradiance model
15'
Specialized Talk
Speaker: Carlos Enmanuel Soto Lopez (University of Trieste, Italy) Material: Abstract
-
16:00
A Probabilistic Forecast for Multi-year ENSO Using Bayesian Convolutional Neural Network
15'
Specialized Talk
Speaker: Balaji Baduru (Indian National Centre for Ocean Information Services (INCOIS), India) Material: Abstract
-
16:15
Detecting Localized Density Anomalies in Climate Data via Coin-Flip Statistics
30'
Speaker: Sebastian Springer (SISSA, Italy)
-
09:30
Linking Research to Action: Exploring the Value of AI-Driven Rainfall Forecasts over the Greater Horn of Africa
1h30'
-
09:30 - 16:45
-
-
09:30 - 17:30
-
09:30
From observations to realistic simulations, the mechanisms that develop and prevent convective organization
15'
Specialized Talk
Speaker: Alejandro Casallas Garcia (Institute of Science and Technology Austria, Austria) Material: Abstract
-
09:45
Investigating deep convective systems with Machine Learning and causal inference across a hierarchy of datasets
15'
Specialized Talk
Speaker: Andrea Polesello (Institute of Science and Technology Austria, Austria) Material: Abstract
-
10:00
Data-Driven Equation Discovery of In-Cloud Vertical Velocity
15'
Specialized Talk
Speaker: Alzbeta Pechacova (Institute of Science and Technology Austria, Austria) Material: Abstract
-
10:15
Exploring windows of opportunity at sub-seasonal time scales for extreme precipitation events: the Valencia DANA case study
15'
Specialized Talk
Speaker: Nuria Perez Zanon (Barcelona Supercomputing Center, Spain) Material: Abstract
Slides
- 10:30 Coffee Break 30'
-
11:00
Circulation Regimes and The Signal-to-Noise Paradox in AI Prediction Models
30'
Speaker: Tim Palmer (University of Oxford, UK) -
11:30
Round Table Discussion: Can AI Models Fully Replace Physical Models for Long-Range Forecasting and Climate Simulations
1h0'
Speaker: Erika Coppola (ICTP, Italy), Francisco Doblas-Reyes (BSC, Spain), Franco Molteni (Agenzia Italiana Meteo, Italy), Tim Palmer (University of Oxford, UK), Jakob Schloer (ECMWF, UK) - 12:30 Lunch 1h30'
-
14:00
Hackathon
1h0'
only for participants 'in person'
- 15:00 Coffee Break 30'
-
15:30
Hackathon
1h30'
only for participants 'in person'
-
09:30
From observations to realistic simulations, the mechanisms that develop and prevent convective organization
15'
-
09:30 - 17:30