A training initiative organised by the South African Radio Astronomy Observatory (SARAO)

5th Big Data Africa School

Cape Town, 10 – 15 March 2025
Registration Deadline: 31 August 2024

Theme: Applications of Machine/Deep Learning to Earth observation data in solving African challenges

The Big Data Africa School aims to introduce fundamental data science tools & techniques to talented young science and engineering graduates across a range of disciplines, who have an interest to develop their skills and knowledge in working efficiently on extremely large datasets in any research environment.

The 5th Big Data Africa School will allow students to work on real-life data sets (e.g. satellite imagery, remote sensing, aerial photography and/or localised ground station data) in the area of Earth observation by applying the most advanced machine learning techniques in trying to solve some of the biggest challenges facing the African continent.

Who can apply?

Eligible countries* – South Africa, Botswana, Ghana, Namibia, Kenya, Mauritius, Madagascar, Mozambique, Zambia

  1. Students currently undertaking their 4th year BSc Honours or final year BEng degree, Master of Science or PhD degree in: Earth observation/ Geo-information science/ Environmental science/ Geographical Information Systems/ Remote sensing/ Computer Science/ any degree in the Physical, Natural and Applied Sciences.
  2. South African students and students having nationality from any of the 8 Square Kilometre Array (SKA) Africa partner countries* are eligible for this opportunity.
  3. Students must be currently registered at a university in Africa.
  4. Intermediate to advanced programming skills will be advantageous to the applicants. Python will be the programming language used in most projects at the Big Data Africa school.

Application enquiries can be emailed to bigdataschool@ska.ac.za

Partners and Contributors

The 5th Big Data Africa School is funded by the UK’s International Science Partnerships Fund through the Development in Africa with Radio Astronomy (DARA) project.

View and download the Brochure below

View and download the Digital Booklet below

For more information contact:

Dr Bonita de Swardt
Programme Manager: Strategic Partnerships for Human Capital Development
Email: bonita@sarao.ac.za

Big Data Africa School – Projects

  • Crop Mapping – Banana Plantations in Africa

  • Mapping Flood Extent

  • Mapping and Monitoring of Urban Development

  • Monitoring Forests, Peatland and Mangroves in Africa

  • Designing Deep Learning Weather Forecasting Models

Project 1: Crop Mapping – Banana Plantations in Africa

This project aims to accurately map banana plantations in Africa to enable near real-time monitoring. Given the genetic uniformity of cultivated bananas and their vulnerability to disease outbreaks, precise mapping is essential for effective disease control. This project will utilize fused high-resolution optical satellite Earth Observation (EO) images and Synthetic Aperture Radar (SAR) images, along with elevation models. Results from classical machine learning methods such as Random Forests (RF) will be compared with deep learning techniques like Convolutional Neural Networks (CNN) based a U-Net architecture for semantic segmentation.

Labelled area showing banana plantation

Plot showing mapped landcover types

Project 2: Mapping Flood Extent

Accurate post-event flood mapping is vital for emergency services to enable rapid relief efforts. The goal of this project is to map flood extents in urban and suburban areas, which are increasingly affected by climate change and extreme weather conditions. The project employs high-resolution SAR satellite image data and utilizes deep Convolutional Neural Networks along a simple pixel thresholding approach. The results from both approaches will be evaluated to determine the best.

Project 3: Mapping and Monitoring of Urban Development

Map showing mapped trees in urban areas

The objective of this project is to monitor changes in urban areas, including vegetation and infrastructure, to support urban planning and development. The project will use high-resolution optical satellite and aerial images with object detection or segmentation models train model that can help detect changes in land use and land cover in urban areas. The models will be based on deep learning techniques like YOLO (You Only Look Once), RetinaNet and SegNet. The results from will be evaluated to determine which one enhances the accuracy of urban monitoring.

Project 4: Monitoring Forests, Peatland and Mangroves in Africa

Figure: Mapped woodland and forests.

This project is dedicated to monitoring forests, peatlands, and mangroves, which are crucial for climate regulation, carbon capture, and coastal protection. The project uses high-resolution optical satellite or aerial images to identify different land cover types through object detection or segmentation approaches. Classical machine learning algorithms such as k-Means Clustering and Random Forest (RF) will be used along with deep learning methods like ResNet and DeepLab for precise land cover mapping and monitoring.

Project 5: Designing Deep Learning Weather Forecasting Models

This project aims to implement and benchmark deep learning weather forecasting models across several data-scarce regions in Africa, such as the Sahel, Horn of Africa, and Southern Africa. With provided datasets, including historical weather records, and ground station data, we will evaluate the accuracy and reliability of these models in environments with limited meteorological data. The project will involve building a deep learning model that combines geo-spatial with observation data and benchmarking it against actual weather events to determine its effectiveness in predicting key parameters like temperature, precipitation, and extreme weather events. Special emphasis will be placed on understanding how data scarcity impacts forecast accuracy and the adaptability of these models to regions with sparse meteorological infrastructure.

Recommendations will be provided to enhance model performance through improved data collection and regional collaboration. This initiative aims to strengthen weather prediction capabilities in Africa, supporting better preparedness and response strategies to mitigate the impacts of adverse weather and climate change.

Contacts

Dr Bonita de Swardt

Programme Manager: Strategic Partnerships for Human Capital Development

South African Radio Astronomy Observatory (SARAO)

Liesbeek House
River Park, Glouchester Road,
Mowbray, 7700,
Cape Town, South Africa

Email: bonita@sarao.ac.za

Website: https://www.sarao.ac.za

Dr Edward Salakpi

Remote Sensing Scientist

South African Radio Astronomy Observatory (SARAO)

University of Stirling
Stirling
FK9 4LA
Scotland UK

Email: edward.salakpi@stir.ac.uk

Website: https://www.stir.ac.uk

Last Updated on July 24, 2024

Share this
Page Reader Press Enter to Read Page Content Out Loud Press Enter to Pause or Restart Reading Page Content Out Loud Press Enter to Stop Reading Page Content Out Loud Screen Reader Support