Village Data Analytics: Satellite imagery analysis for mini-grid site selection

How do you find a good site for a mini-grid? What are the decisive factors developers look for when they pick a village? How much of this can be seen from space?

Satellite imagery is getting ever more accessible and affordable

We have been through the surveying and site selection process many times in various countries in Asia and Africa and have come across many of the pitfalls, from designing survey questions that consider the local context, understanding the various biases that arise to the challenge of error-free data recording in off-grid areas.
If we are to build more than 200'000 mini-grids by 2030 to achieve SDG-7, we need to remove the bottleneck of slow and often imprecise site selection.
With our desire to make the site selection process more digital, scalable and data-driven, we were excited to partner with the European Space Agency (ESA) in 2018 to tackle this challenge.

The result is Village Data Analytics (VIDA), a piece of software which helps developers locate and characterize the most promising villages in their area of interest, making sure that when they get on the ground they already have a detailed fact-sheet about the place. The analysis is based on satellite imagery from ESA’s Copernicus fleet as well as other publicly available datasets such as Open Street Map (OSM). VIDA uses machine learning to work at scale. Unlike first generation GIS tools VIDA can make extensive use of all-important temporal factors such as how fast the village is growing.

Many useful characteristics about a village can be identified on satellite images. The challenge lies in identifying which of these features deliver the most useful predictions of how a commerical mini-grid will perform and how to extract these characteristics at scale. To find out, we reached out to the mini-grid community in an online survey. We’ve received 24 responses from across the world from well-known energy access practitioners. Here’s what we found:

The most useful satellite imagery based factors according to the survey respondents. Grid location as well as building statistics top the list, followed by agricultural analysis and road network analysis.

• 20 developers named grid location as the key factor. Not knowing the location of the grid has been a persistent challenge in the off-grid energy access sector. However, satellite imagery increasingly adds insights. A good example is the recently published dataset by Facebook. It is based on nightlight imagery and is integrated in VIDA.

Grid lines estimated from NASA nightlight imagery (black dots, data from Facebook). Actual VIDA screengrab with settlements in purple.

• Developers went on to cite the size of the settlement and building statistics as paramount: Settlement size is related to the number of connections, building statistics provide an estimate of connection costs, types of customers (residential, public facilities, businesses) and can hint at the type of economic activity in a village.
  There are valuable open layers such as the humanitarian open street map which provide some of this information and can reinforce the findings from the satellite imagery analysis.
• A set of other factors were put forward by developers, with their importance depending on the specific local context: For some developers, learning about agriculture was interesting: Is there cropland? If so, how much? What are people growing? For other developers, road access and proximity to larger settlements are important. Finally, some developers work in flood-prone areas. For them, the analysis of historic satellite imagery to detect the extent of floods in the past helps in their risk assessment.

Classification of agricultural field sizes

Digital tools for site identification
Most of the developers stated that they use Google Earth and QGIS along with freely available geospatial layers to analyze their area of interest before they send out teams to survey the villages. While this shows the power of freely available datasets and software, we feel that this also reflects a lack of established, professional tools tailored to the needs of mini-grid developers. Contrast this to remote O&M for mini-grids, where specialized companies offer a variety of proprietary software tools to remotely monitor and control mini-grids.
We believe that the site selection process should be similarly stratified and standardized using satellite imagery, large-scale socio-economic datasets as well as on-ground surveys. This is the key motivation behind VIDA.

Speeding up site selection
With VIDA, our goal is to make it easier to select good sites whilst significantly shortening project planning time. To achieve this, we offer some of the most popular geospatial layers in our current VIDA service. We are on track to further develop VIDA and applying it to promising areas for mini-grid development worldwide.

We would like to thank all the people who took part in the survey. Please reach out to us (mail to: tfe@tfe.energy ) in case you want to know more about VIDA.