ESA-ECMWF Workshop 2021 Machine Learning for Earth System Observation and Prediction

Monday, 15 November
Opening
10:30 10:45
Welcome and introduction ESA
Speaker: Pierre-Philippe Mathieu (ESA)

Virtual

15m
10:45 11:00
Welcome and introduction ECMWF
Speaker: Andy Brown (ECMWF)

Virtual

15m
Keynote: ML for the ESOP - Setting the scene
Chair: Rochelle Schneider (ESA)
11:00 11:45
Machine learning on the Earth System with remote sensing: towards machines that we can understand and interact with
Speaker: Devis Tuia (EPFL)

Virtual

45m
11:45 12:30
Combining data assimilation and machine learning to extract more information from earth observations
Speaker: Alan Geer (ECMWF)

Virtual

45m
12:30 14:00
Poster session

Virtual

1h 30m
Session 1.1: Enhancing Satellite Observation with ML
Chair: Bertrand Le Saux (ESA)
14:00 14:30
Benefits and opportunities of Explainable Machine Learning in the environmental sciences
Speaker: Ribana Roscher (Technical University of Munich and University of Bonn)

Virtual

30m
14:30 15:00
Atmospheric Retrievals in a Machine Learning Context: A Radiometric Story Over the Ocean
Speaker: Mario Echeverri Bautista (KNMI)

Virtual

30m
15:00 15:30
Permutation invariance and uncertainty in multitemporal image super-resolution
Speaker: Diego Valsesia (Politecnico di Torino)

Virtual

30m
15:30 16:00
Coffee break

Virtual

30m
Session 1.2: Enhancing Satellite Observation with ML
Chair: Rochelle Schneider (ESA)
16:00 16:30
CNN-based Detection of Tiny Objects in Remote Areas Using Spatiotemporal Earth Observation Data
Speaker: Roman Pflugfelder (AIT - Austrian Institute of Technology )

Virtual

30m
16:30 17:00
Estimate of XCO2 from OCO-2 Observations Using a Neural Network Approach
Speaker: Francois-Marie Breon (CEA/LSCE)

Virtual

30m
17:00 17:30
Autonomous Robotic Teams, Machine Learning, and the Next Generation of Earth Observing System
Speaker: David Lary (Hanson Center for Space Sciences, University of Texas at Dallas)

Virtual

30m
Tuesday, 16 November
Session 1.3: Enhancing Satellite Observation with ML
Chair: Begüm Demir (TU Berlin)
09:00 09:30
Free High Resolution Imagery - Open-Source Models and Package for Sentinel 2 Super-Resolution
Speaker: Julien Cornebise (UCL)

Virtual

30m
09:30 10:00
Deep Learning Implementations to Facilitate the Assimilation of Satellite Observations. A Case Study for LST and SST Using IASI Observations
Speaker: Eulalie Boucher (Lerma - Observatoire de Paris)

Virtual

30m
10:00 10:30
Machine Learning Techniques for Automated ULF Wave Recognition in Swarm Time Series
Speaker: Georgios Balasis (National Observatory of Athens)

Virtual

30m
10:30 11:00
Coffee break

Virtual

30m
Session 2.1: Hybrid Data Assimilation - ML Approaches
Chair: Alan Geer (ECMWF)
11:00 11:30
Data Assimilation + Machine Learning = Data Learning
Speaker: Rossella Arcucci (Imperial College London)

Virtual

30m
11:30 12:00
Use of AI to facilitate the NWP assimilation of EOs: retrieval, radiative transfer and physical integration of satellite products
Speaker: Filipe Aires (CNRS)

Virtual

30m
12:00 12:30
Gaussian Assimilation of non-Gaussian Image Data via Pre-Processing by Variational Auto-Encoder (VAE)
Speaker: Daisuke Hotta (JMA)

Virtual

30m
12:30 13:00
Correcting model error with an online Artificial Neural Network
Speaker: Marcin Chrust (ECMWF)

Virtual

30m
13:00 14:00
Lunch break

Virtual

1h
Session 2.2: Hybrid Data Assimilation - ML Approaches
Chair: Rossella Arcucci (Imperial College London)
14:00 14:30
Model error correction with data assimilation and machine learning
Speaker: Alban Farchi (CEREA, ENPC)

Virtual

30m
14:30 15:00
Towards the Direct Assimilation of Scatterometer Backscatter Triplet
Speaker: Sean Healy (ECMWF)

Virtual

30m
15:00 15:30
Data-Driven Surrogate Model with Latent Data assimilation for Wildfire Forecasting
Speaker: Sibo Cheng (Imperial College London)

Virtual

30m
15:30 16:00
Coffee break

Virtual

30m
Session 3.1: Geophysical Forecasting with ML and Hybrid Models
Chair: Anca Anghelea (ESA)
16:00 16:30
Neural-Network Parametrization of Subgrid Momentum Transport Learned from a High-Resolution Simulation
Speaker: Janni Yuval (MIT)

Virtual

30m
16:30 17:00
Machine learning based climate modelling and analysis
Speaker: Veronika Eyring (DLR and University of Bremen)

Virtual

30m
17:00 17:30
Atmospheric Physics-Guided Machine Learning: Towards Physically-Consistent, Data-Driven, and Interpretable Models of Convection
Speaker: Tom Beucler (University of Lausanne)

Virtual

30m
17:30 18:00
Seamless rainfall forecast using encoder-decoder networks: bridging weather and interannual forecast horizons
Speaker: Gabriel Martins Palma Perez (University of Reading)

Virtual

30m
Wednesday, 17 November
Session 3.2: Geophysical Forecasting with ML and Hybrid Models
Chair: Marc Bocquet (ENPC)
09:00 09:30
Learning parameters of a numerical model from observations
Speaker: Tijana Janjic

Virtual

30m
09:30 10:00
Machine Learning for Seamless Thunderstorm Nowcasting from Multiple Data Sources
Speaker: Jussi Leinonen (MeteoSwiss)

Virtual

30m
10:00 10:30
Smartriver: Artificial Intelligence for Effective Water Resoures Forecast and Management
Speaker: Stefano Bagli (Gecosistema)

Virtual

30m
10:30 11:00
Coffee break

Virtual

30m
Session 4.1: ML for Post-Processing and Dissemination
Chair: Gustau Camps Vals (IPL - Universitat de Valencia)
11:00 11:30
Post-processing of precipitation with Bernstein quantile distribution networks
Speaker: John Bjørnar Bremnes (Norwegian Meteorological Institute)

Virtual

30m
11:30 12:00
Machine Learning-Based Post-Process Correction of the High-Resoulution Multi-Wavelength Sentinel-3 Synergy Aerosol Product
Speaker: Antti Lipponen (Finnish Meteorological Institute)

Virtual

30m
12:00 12:30
Integrating Reanalysis and Satellite Cloud Information to Estimate Downward Long-wave Radiation Fluxes Using Multivariate Adaptive Regression Splines: Application to EUMETSAT LSA-SAF
Speaker: Francisco Lopes (Instituto Dom Luis)

Virtual

30m
12:30 13:00
A spatiotemporal ensemble machine learning framework for predictive mapping: One model to rule them all?
Speaker: Tom Hengl (OpenGeoHub Foundation)

Virtual

30m
13:00 14:00
Lunch break

Virtual

1h
Session 4.2: ML for Post-Processing and Dissemination
Chair: Claudia Vitolo (ESA)
14:00 14:30
Using Convolutional Neural Networks to Detect Emissions Plumes from TROPOMI Data
Speaker: Douglas Finch (University of Edinburgh)

Virtual

30m
14:30 15:00
Downscaling air pollution levels by fusing Geospatial Vector Data and Sentinel 5P observations over Europe
Speaker: Marvin Mc Cutchan (IARAI and TU Wien)

Virtual

30m
15:00 15:30
The Self-Attentive Ensemble Transformer: Representing Ensemble Interactions in Neural Networks
Speaker: Tobias Finn (Universität Hamburg)

Virtual

30m
15:30 16:00
Coffee break

Virtual

30m
Session 4.3: ML for Post-Processing and Dissemination
Chair: Peter Dueben (ECMWF)
16:00 16:30
How Earth Observations and Machine Learning are Supporting Agricultural Monitoring and Food Security Globally
Speaker: Catherine Nakalembe (University of Maryland and NASA Harvest)

Virtual

30m
16:30 17:00
Flood Segmentation on Sentinel-1 SAR Imagery with Semi-Supervised Learning
Speaker: Siddha Ganju (NVIDIA, SpaceML, FDL)

Virtual

30m
17:00 17:30
Description of Working Group
Speaker: Massimo Bonavita (ECMWF)

Virtual

30m
Thursday, 18 November
Working Groups - Thematic areas
09:00 10:30
WG1 - Enhancing Satellite Observations with ML - chaired by Begüm Demir and Bertrand Le Saux

Virtual

1h 30m
09:00 10:30
WG2 - Hybrid Data Assimilation - ML Approaches - chaired by Rossella Arcucci and Alan Geer

Virtual

1h 30m
09:00 10:30
WG3 - Geophysical Forecasting with ML and Hybrid Models - chaired by Claudia Vitolo and Peter Dueben

Virtual

1h 30m
09:00 10:30
WG4 - ML for Post-Processing and Dissemination - chaired by Rochelle Schneider and Massimo Bonavita

Virtual

1h 30m
10:30 11:00
Coffee break

Virtual

30m
11:00 12:00
WG1 - Enhancing Satellite Observations with ML - chaired by Begüm Demir and Bertrand Le Saux

Virtual

1h
11:00 12:00
WG2 - Hybrid Data Assimilation - ML Approaches - chaired by Rossella Arcucci and Alan Geer

Virtual

1h
11:00 12:00
WG3 - Geophysical Forecasting with ML and Hybrid Models - chaired by Claudia Vitolo and Peter Dueben

Virtual

1h
11:00 12:00
WG4 - ML for Post-Processing and Dissemination - chaired by Rochelle Schneider and Massimo Bonavita

Virtual

1h
12:00 13:00
WG Chairs finalise reports

Virtual

1h
13:00 14:00
Lunch break

Virtual

1h
14:00 15:30
Session 5: Working Groups plenary discussion and close
Chair: Massimo Bonavita (ECMWF)
1h 30m

Please select a timezone to show adjusted event times accordingly.