As a first-ever initiative, the political initiative FAIR Forward – Artificial Intelligence for All, held its training program for practitioners in the field of Earth Observation in South Africa. The program, called ML4EO-2022 (Machine Learning for Earth Observation), empowers participants to improve food production efficiency and quality and ultimately use these skills and knowledge in their work. As countries face the challenge of climate change, this know-how is key to developing climate-smart agricultural practices.
Partners to ML4EO 2022 were Dept. of Science and Innovation, South African National Space Agency, Council for Scientific and Industrial Research, Wits University, and Agricultural Research Council of South Africa. Move Beyond Consulting was contracted to develop and implement this training.
234 applications were received globally, with 133 coming from South Africa. A rigorous selection process saw 20 professionals (9 female, 11 male) chosen. Participant backgrounds ranged from remote sensing and ML application to smart agriculture and actuarial science. From April to July 2022, they engaged in intense training and a field trip to Limpopo province, where they experienced collecting ground truth data, drone management, agricultural instrumentation usage and learnt data analysis. They also engaged local farmers for insights into the challenges and opportunities in local farming.
The program included six modules, covering topics such as introduction to remote sensing in agriculture, classification-based remote sensing applications, and machine learning modelling of important agricultural crop parameters. The learning was designed to be interactive and practical. In order to successfully complete the program, participants were grouped into pairs and had to conclude a case study with a research report, based on their activities in class and the field work. The final project presentations were done at an award and closing ceremony, held in Pretoria, South Africa.
The projects presented ranged from detecting tomato stressors, weed detection, and crop mapping, to developing a new type of agricultural insurance. Overall there were ten project presentations. Most participants indicated that they could apply the acquired knowledge and adapt the training into their own teaching and work environments.
“This was quite an exciting journey for me. I did not have any background in Remote Sensing, GIS or Earth Observation. My background is agriculture. It was fascinating to see available opportunities – merging the industry in terms of the agriculture industry, RS, and EO – and putting it all together.”
Account from one of the participants
By equipping expert practitioners with the necessary AI skills and knowledge, the program has the potential to lead to more adoption of precision agriculture. That way, the partnership has set a new standard for capacity building on Earth Observation and machine learning that actors across the globe may adopt.
The full report of the program provides extensive information, including details on planning and delivery, student/alumni and facilitator feedback, a study of the current Earth Observation and Machine Learning skills landscape and suggestions for sustaining the program. A copy of the report as well as further information on the programme can be obtained from Deshni Govender at FAIR Forward in South Africa.
In addition, a replication kit including the datasets and project documents can be made available as an open educational resource.
