Global Earth Monitor Webinar

The H2020 Global Earth Monitoring (GEM) project, whose aim is addressing the challenge of continuous monitoring of large areas in a sustainable cost-effective way, has entered in its last phase. The webinar aspired to gather different actors from the Earth Observation domain and beyond to show the results of the project (with a particular focus on the outcome of its use cases). This remote and free event was held on 14 February 2023.

Morning Session: Project outcome and user feedback

The EC Project Officer and the Project Coordinator provided an overview of GEM and its position in the EO R & I scenario. Successively, the project partners resumed the pilot outcomes and the audience was be able to provide feedback through interactive polls on the proposed solutions; the feedback is considered by the consortium in the definition of the way forward.

Watch the recording of the webinar (for an easier navigation through the webinar, open the recording in Youtube and follow the time indications in the video description):

General agenda:

Presentation Speaker
Welcome – GEM in the EC ecosystem Massimo Ciscato, EC, Project Officer 
The GEM Project Matej Batič, Sinergise, Project Coordinator
GEM Processing Framework Matej Batič, Sinergise
Conflict Pre-Warning Map Use Case Michele Lazzarini, SatCen
Map Making Use Case and Land Classification Use Cases Francesco Perfetto and Ashish Dhiman, TomTom
Efficient access to weather data for ML meteoblue team
The use of AI in GEM Michael Engel, TUM
Conclusions and way forward Matej Batič, Sinergise 

 

Afternoon Sessions: Hands-on GEM

The step-by-step exercises featured the use of GEM resources. The exercises, for which preparatory material and requirements were circulated in advance, are targeting students, Software developers, Data scientists, EO-developers, and anyone with an interest in the topic.

The afternoon session was dedicated to guided exercises to show GEM outcomes and to exploit the developed tools by the different partners.

Guided exercises (GEM consortium)

Sinergise

How do we apply EO workflows to larger scale? Working with EO data is made easy by the eo-learn package, while the eo-grow package takes care of running the solutions at a large scale. In this session we will have a look at the GEM processing framework:

  • First, we show how we have used eo-grow to create two static data cubes, Sentinel-2 data and aggregated weather data, both with monthly cadence at 120 m resolution.
  • Next, we will guide the audience through a slimmed-down EO workflow using a Jupyter notebook and eo-grow, which is small enough to be run on a laptop. You will learn about the main eo-grow structural blocks (pipelines, managers, configuration schemas), that drive the execution and allow for the reproducibility of the runs. The example will show the typical EO pipeline, from data retrieval, predictions with preexisting model on the features, and finally to uploading results as data cubes on Sentinel Hub.

meteoblue

  • Weather data access examples
    Learn how to load a large variety of weather data for your geography and period of choice, at the spatial and temporal resolutions and aggregations that you need. A high performance API gives rapid access to top quality data with minimal effort. An interactive configurator allows you to compose your query, preview results, and generate the appropriate API call(s).
     
  • Walk-through of creating urban heat maps
    Learn how to extract the structure of a city's temperature field from satellite data. Combine it with weather data to generate an urban heat island map.

TUM

  • HowTo use GEM ML framework
    Learn how to use the GEM ML Framework for synergetic interaction of PyTorch and eo-learn for ML in EO at scale. We will have a look at a segmentation pipeline for the detection of deforested areas. In the process, you will learn how to properly normalize your larger-than-memory datasets, how to feed them to your model of choice and how our generic pipelines can be used (for your own research). In the end, you will be able to integrate your trained models into eo-learn workflows for inference pipelines.
    Please note that you need a running version of Python, a CUDA-enabled device and a good understanding of programming to fully enjoy the experience.

SatCen

  • During the demo session, different datasets of interest for understanding climate security issues will be explored.
    The demo session will be executed in Binder, using Python and Jupyter notebooks. There is no need of previous software or library installation, basic knowledge of Python would be desirable.

    • Exercise 1 will focus on the visualization and analysis of socio economic data of relevance for climate security in the Sahel region. The exercise will use open data from World Bank (e.g. GDP, Unemployment rate, mortality, birth rate, …).

    • Exercise 2 will focus on the visualization and analysis of soil moisture levels in Nigeria, specifically in the northwest region of the country that was affected by floods in 2019. The exercise will use Sentinel-2 data provided by Sinergise via Sentinel-Hub and meteo data provided by meteoblue.

TomTom

  • In this hands-on workshop session on Land Cover - Continuous Monitoring Service (LC-CMS), you will learn about:
    • An overview and firsthand experience of a LC-CMS pipeline prototype, with results generated for the region of Paris.
    • How the eo-grow and eo-learn libraries could be used to build a pipeline for your requirements.
    • The concept of Change Detection and its application in monitoring land cover evolution and change monitoring, with results generated by LC-CMS in demonstration area.
  • Prerequisites:
    • Before the session please clone/download the github repository from this link and follow the prerequisite steps in the ReadMe.md.
    • A computer with Docker installation to build & run the Docker image for this workshop, some familiarity with Python, QGIS installation on your machine for visualizing the final results.
  • A session by Sinergise on how to use eo-grow is highly recommended prior to this.