“It is really gratifying for us at SAEON to be able to pull this kind of thing together using government-based data sources; it shows us what’s possible. We spend so much time and money working on climate change with government and non-government partners, but no one has done anything like this before.”
Dr Amelia Hilgart is talking about the recently-published African air quality dataset developed in partnership with Zindi, created using a model developed by Zindi users as part of a 60-hour hackathon earlier this year. As a data scientist at the uLwazi Node of the South African Earth Observation Network, Hilgart’s job is to collect, collate and share data like this to inform local government decision-making in the complex field of global change.
Urban Air Pollution Challenge
You may have seen recent news articles stating that air quality has improved due to COVID-19. This is true for some…
“We can’t decouple climate change from climate policy and good governance, which is in turn tied to active democracy, education, public health, and so on,” Hilgart explains. “For instance, climate change could cause drought, which reduces food security and impacts the economy. With less water for sanitation, we see an increase in disease and public health burdens. We must think about the system as a whole and how the pieces fit together in order to actually make progress.”
Build it and they will come
It was in looking to fill in missing data gaps that Hilgart stumbled across Zindi’s early efforts at an air quality prediction model using publicly-available Sentinel-5P satellite data. Zindi competitors had created the model as part of a #ZindiWeekendz hackathon during the early stages of the COVID-19 pandemic, in recognition of the fact that air quality data across Africa is poor to non-existent.
Zindi solutions: A useful open-source model of urban air quality for Africa
In just a few days, the Zindi community has built a model that can accurately predict air quality in cities and towns…
“We know that air quality has a significant negative impact on public health outcomes, especially around respiratory diseases like asthma or indeed, COVID-19,” says Jonathan Whitaker, data scientist at Zindi and lead on the project. “So we challenged our community to build a machine learning model to fill in the gaps on air quality in African cities and towns. And they delivered brilliantly.”
Zindi verified and streamlined the winning models, and published them under a CC BY-SA 4.0 license, inviting anyone who wanted to use them to get in touch. SAEON saw the potential in this model and the data it produced, and reached out.
Implementation of an air quality model based on satellite data adapted from a Zindi winning solution In this project…
From model to dataset
Whitaker was joined by Yasin Ayami, Zindi Ambassador for Zambia, and following guidance from SAEON, the two of them developed the preliminary model and used it to make predictions of historical air quality in African cities. SAEON has now shared datasets for Africa and South Africa on their Open Data Platform as well as on partner sites like Africa GeoPortal and Esri’s Living Atlas project. SAEON also ensured that the dataset has a doi (digital object identifier) so that it is searchable through tools like Google’s dataset search engine.
This predicted air quality dataset is now ready to be used by researchers, local governments and other actors to better understand air quality in urban areas across Africa for the first time.
It will also be put to use in the new air quality page for the South Africa Risk and Vulnerability Atlas (SARVA), one of SAEON’s decision support tools for local government.
“This work will also go into the next release of SARVA,” Hilgart says. “Our focus is not only identifying datasets but also data gaps, so we can show government that if they start collecting certain data, we can better understand these chains of events.”
“We have lists of missing data that we reach out to boundary organisations for assistance with. That’s where our relationship with Zindi is — helping us fill data gaps in our picture of global change in South Africa.”
See the project on SAEON
View and downoad the models on GitHub