Geospatial Design of Energy Systems for Africa: Citizen Science

Geospatial Design of Energy Systems for Africa (GeoDESA) Citizen Science aims to integrate citizen science in rural electrical grid planning for developing regions. As over 800 million people still lack reliable access to electricity, and the UN is targeting affordable clean energy access for all by 2030, innovative electrification strategies are needed to close the gap.

Data-driven geospatial system planning can be an effective way to accelerate electrification. High-resolution geographic data and modern computing power can be leveraged to plan community-specific grid development cheaply and quickly, reducing costs by an order of magnitude compared to traditional methods.

One key challenge in this approach is the need for detailed home location data. Home locations are necessary to design grid topologies and architectures. This level of specificity enables better grid cost estimates and community-appropriate design. Currently available home location datasets, such as OpenStreetMap, are typically incomplete, particularly in the hardest-to-reach rural poor areas.

High-resolution satellite imagery and modern computer vision algorithms can fill the gaps in these datasets. While many algorithms have been trained to detect housing in the past, they are typically based on rich urban contexts. The full diversity of rural and remote housing styles must be taken into account to enable the dwelling detection needed for electrification design.

By engaging citizen scientists fluent in local rural housing styles in the labeling of homes in satellite imagery, local knowledge can be incorporated in a scalable data-driven approach. With accurate and context-informed labelled satellite data, computer vision algorithms can be trained to reliably locate rural dwellings in the hardest to reach areas, allowing grids to be efficiently designed to suit community needs.

This proposal complements past and ongoing work undertaken at the University of Oxford in rural Africa electrification from the SONG and RELCON projects. The methods developed and data generated will be useful for electrical system planning in many rural and remote contexts, and can cross-apply into other geographic planning disciplines, such as urban planning, migration tracking, and geospatial poverty estimation.

This research will involve outreach engagement with citizen scientists in sub-Saharan Africa. The intended focus for engagement is Ethiopia. 56% of Ethiopia's population still lack access to electricity. The government of Ethiopia has a strong interest in geospatial electrical system planning. Existing connections at universities in Bahir Dar and Addis Ababa will enable citizen science engagement. However, should there be any challenges in the political environment in Ethiopia which make on-the-ground engagement problematic, we also have connections in Kenya through past and ongoing projects, particularly in universities in Nairobi and Mombasa, which will enable us to work there.

GeoDESA Citizen Science will bring on-the-ground citizen science engagement to rural grid design through rural home identification. Workshops will be arranged on sustainable grid planning using geospatial methods highlighting the importance of local knowledge to this problem. An accompanying Zooniverse-hosted citizen science platform will be made available to participants. This platform, called "Home Sweet Home" (currently viewable in pilot: www.bit.do/geodesa) will allow citizen scientists to label homes in high-resolution satellite imagery. Imagery is provided by Earth-i through an ongoing Oxford collaboration with Satellite Applications Catapult. The citizen science platform will be disseminated through Zooniverse's participant email lists, as well as through academic networks. Many participants reached through Zooniverse communications will likely be from England; this will provide us with internationally-produced labels to compare with the locally-produced labels resulting from on-the-ground outreach. The data labelled by citizen scientists will be used in the training of object detection computer vision algorithms for dwelling location. It will be analysed how the data labeling varies between local citizen scientists and international citizen scientists using label metadata.

This project has the potential for far-reaching impact, with cross-sector utility and insights. The citizen science methodologies, if successful, can be used to accelerate myriad data-driven research efforts. In this particular project, the outputs of the work have the potential to create electrical planning tools useful in diverse rural and remote developing contexts. If adopted widely, these could accelerate electrification and provide more contextually appropriate system designs, with the potential benefit of improved energy access to millions of people worldwide.

The University of Oxford Energy and Power Group has ties with relevant Ethiopian and Kenyan policy-makers in energy system planning, who we anticipate will find this work of interest. The methods produced have the potential to be used widely by multilateral organisations, government agencies, and private-sector developers.

The data and methods developed in this project will be made openly available. This is both in the spirit of open citizen science and to avoid gate-keeping resources which could have great impact in public- and private-sector electrical development. There is a precedent for open-source geographic models being used widely by multilateral organisations, as shown by the use of the OnSSET tool by the World Bank. This work could create a similarly beneficial tool to be used widely and openly.

Through work on this project, the University of Oxford's Energy and Power Group will familiarize themselves with best practices for citizen science engagement. This will equip them to integrate citizen science methods in further projects and set a precedent for citizen science engagement in rural grid design.