8. Using EO Earth Observation Datacubes for Understanding the Earth’s Rapid Changes [Hackathon, 4.5 hrs + preparation before BiDS]

The Earth is constantly changing and reliably understanding these changes is key to supporting today’s most pressing issues such as environmental degradation, biodiversity loss, and natural disasters. However, reliably understanding and tracking changes as they occur has been a historically difficult task due to the challenges of working with large remote-sensing data sets from different missions. With the emergence of many machine learning-based tools, platforms able to make data normalization easier, and the cloud to manage Big Data volumes and processing, it is now feasible to develop effective tools to understand the rapid changes of the Earth and take action for environmental protection and promoting sustainability.

Participants will work together in a hackathon to develop automated tools to classify and understand land changes in order to better understand our evolving planet. Examples include automated monitoring of rivers and water bodies to quantify the recession and expansion, detecting deforestation in near-real-time, assessing forest fire impact during and after fires, or evolution of green landscape infrastructure like hedges supporting biodiversity. Study sites will be provided to participants to explore, analyze and develop a suite of innovative open tool to automatically understand the changing landscape and support today’s environmental challenges.

Participants should be comfortable using Jupyter notebooks and will have knowledge in one of the following areas: remote sensing, data science, data processing, or data visualization. Participants will work together to develop innovative solutions to automatically detect changes and/or visualize those changes in a way to facilitate Big Data analysis.

9. Working with EUMETSAT and Copernicus marine and atmospheric data across multiple cloud-based systems [Tutorial, half day]

EUMETSAT has developed a prototype of a multi-cloud framework which allows the scalable access and processing of Copernicus and EUMETSAT data for marine and atmospheric composition applications. The framework is deployed across different cloud systems, including the European Weather Cloud and the Copernicus WEkEO DIAS. This hands-on tutorial session will give an opportunity to explore the use and benefits of such a framework and will consist of three main parts. The first part will provide an introduction of common terminologies used in the context of cloud computing and Earth observation applications, including an overview of the current Earth observation cloud landscape and cloud-native geospatial tools. The second part will practically introduce the data-proximate, multi-cloud framework developed by EUMETSAT, ECMWF, and UKMO, with a specific focus on how the Python library for distributed and parallel computing, Dask, can be extended from a single- to a multi-cloud framework using a mesh-network architecture. The third part will then highlight two concrete application examples focused on EUMETSAT marine and atmospheric composition data.

This tutorial has a strong hands-on component during which participants will have ample time to explore the developed framework based on application case studies. It is suitable for a diverse audience, including beginners as well as more advanced users of cloud- based systems. A basic knowledge of Python programming will be required.

Organisers, expected Instructors and Contributors: Ben Loveday (EUMETSAT/Innoflair UG), Julia Wagemann (EUMETSAT/MEEO SRL) Hayley Evers-King (EUMETSAT), Federico Fierli (EUMETSAT), Mike Grant (EUMETSAT), Roope Tervo (EUMETSAT), Joana Miguens (EUMETSAT), Anna-Lena Erdmann (EUMETSAT), Armagan Karatosun (ECMWF)

Resources will be made available at: https://github.com/wekeo/multi_cloud_exploitation 

10. Exploring the FuseTS Toolbox: Fusing and Analysing Multi-Source EO Time Series Data [Tutorial, half day]

We introduce a comprehensive training program aimed at enhancing Earth observation data analysis skills. This session includes a hands-on open-source FuseTS toolbox developed as part of the AI4FOOD project (ESA).

This training program aims at a deep dive into the capabilities of the FuseTS toolbox. Through its user-friendly Python implementation, participants will learn to seamlessly fuse optical and SAR data, unlocking the ability to conduct advanced time series analytics.

Our training program combines valuable insights with practical application through hands-on exercises, ensuring a well-rounded learning experience. Real-world case studies will be presented, showcasing the successful implementation of the toolbox.

This training program is specifically tailored for professionals in EO observation who want to enhance their data analysis capabilities. Regardless of your expertise level, this program provides the necessary insights to leverage multi-temporal fusion for real-world applications.

We look forward to exploring the FuseTS toolbox with you and helping you take your data analysis to the next level.

Planned Contributors: Matic Lubej (Sinergise), Darius Couchard (VITO), Jochem Verrelst (University of Valencia), Jeroen Dries (VITO) and Bram Janssen (VITO). 

Please find more information on the website: https://open-eo.github.io/FuseTS/                                                                and in the repository: https://github.com/Open-EO/FuseTS

11. Mastering Earth Observation Application Packaging with CWL [Tutorial, 1.5 hrs]

Join us for the "Mastering Earth Observation Application Packaging with CWL" tutorial event, where we will dive into the world of EO Application Packages and explore how to effectively package, share, and execute Earth observation workflows using the Common Workflow Language (CWL) standard.

This tutorial event is designed for developers, scientists, and Earth observation enthusiasts who want to enhance their skills in creating and sharing EO Application Packages. Whether you are new to CWL or already have some experience, this event will provide valuable insights and practical knowledge to boost your expertise.

During the event, you will learn:

• The fundamentals of EO Application Packages and their role in the Earth observation domain.
• How to leverage CWL to describe, package, and share workflows.
• Techniques for incorporating data, code, configuration files, and documentation into an EO Application Package.
• Best practices for creating portable and reproducible Earth observation workflows.
• Hands-on exercises to reinforce your understanding and gain practical experience.

This tutorial will guide you through step-by-step tutorials, demonstrating the process of creating EO Application Packages using CWL. Don't miss this opportunity to enhance your skills in Earth observation application packaging and embark on a journey towards mastering EO Application Packages with CWL.

12. Big Earth observation data analysis using satellite image time series [Tutorial, 4.5 hrs]

This tutorial presents a hands-on introduction to Big Earth observation (EO) data analysis using satellite image time series, using the open-source R package sits that provides an end-to-end solution for land classification, with a broad functionality:

• Retrieval of big EO data for STAC-compliant cloud provides, including; AWS, USGS, Planetary Computer, NASA HSL, Digital Earth Africa, Swiss Data Cube and Brazil Data Cube.
• Regularization of data cubes from Sentinel-2, Landsat, MODIS imagery (Sentinel-1 data from June 2023).
• Extraction of time series samples from data cubes;
• Band operations and mixture models;
• Training data quality control using an innovative SOM-based approach;
• Methods for correcting imbalance in training samples;
• Machine learning methods for time series, including random forest, SVM, xgboost, multilayer perceptrons, temporal CNN, residual networks, and temporal self-attention encoder;
• Tuning methods for deep learning hyperparameter optimization;
• Efficient parallel processing for large-scale data ML classification, including GPU support;
• Validation and accuracy support FAO-recommended best practices for classification evaluation;
• Spatial uncertainty estimation of classification results;
• Active learning to reduce bias and errors in sample selection;
• Ensemble prediction from multiple models.

The package has reached NASA TRL level 9, and is used operationally by Brazil’s National Institute for Space Research to produce large-scale land use and land cover maps for the Amazonia and Cerrado biomes. This tutorial is based on the on-line book available at https://e-sensing.github.io/sitsbook and on the Kaggle notebooks available at www.kaggle.com/esensing/code.

Contributor: Prof. Dr. Gilberto Camara (INPE). 

The tutorial will be based on the on-line book, "sits: Satellite Image Time Series Analysis on Earth Observation Data Cubes”, available at https://e-sensing.github.io/sitsbook/ 

14. AI4Copernicus tools and methods for bridging AI and EO [Tutorial, half day]

Artificial Intelligence (AI) represents a collection of tools and methodologies that have the potential of transforming virtually all aspects of human activity. Earth observation (EO) data, including satellite and in-situ, are essential for a number of applications, covering high-impact domains as diverse as security, agriculture and health. The H2020 AI4Copernicus project (https://ai4copernicus-project.eu) delivers a technological framework for developers and businesses to combine AI-infused tools and EO data and services in order to create high impact applications. This is further facilitated by providing a bridge between DIAS platforms and the European AI-on-Demand platform.

This tutorial will present the main technological assets AI4Copernicus brings to the table as well as its methodology and tools for linking the DIAS and AI-on-Demand platforms, through appropriately selected use-cases during a hands-on session. Indicatively, technological assets AI4Copernicus contributes to the community include the following: Sentinel-1 and Sentinel-2 pre-processing chains, Deep network for pixel-level classification of S2 patches, Probabilistic downscaling of CAMS air quality model data, and many others. In addition, AI4Copernicus contributed semantics-based tools for visualisation and discovery of complex EO data. More information can be found in the AI4Copernicus Technical Documentation.

The choice of the use cases for the hands-on applications will come from AI4Copernicus' rich collaborations with 3rd parties who specialise in EO applications in the fields of Security, Health, Agriculture, Climate and others, and will be consolidated nearer the time of the tutorial.

This tutorial will target EO and AI specialists and will require little IT technical background. The hands-on session will take place on the CREODIAS infrastructure and will involve the application of AI techniques on available datasets.

Organisers, Instructors and Contributors: Antonis Troumpoukis (National Centre for Scientific Research Demokritos), Despina-Athanasia Pantazi (National and Kapodistrian University of Athens), Omar Barrilero (European Union Satellite Centre), Giulio Weikmann (University of Trento), Mohanad Albughdadi (ECMWF), Vasileios Baousis (ECMWF), Iraklis Klampanos (National Centre for Scientific Research Demokritos), Jacek Tokarski (CloudFerro), Lorenzo Bruzzone (University of Trento), Michele Lazzarini (European Union Satellite Centre), George Stamoulis (National and Kapodistrian University of Athens), Manolis Koubarakis (National and Kapodistrian University of Athens).

Please find further information at: https://ai4copernicus-bids2023.github.io/