2022 Shortlist

Engolindo Fumaça (Inhaling Smoke)

Country/area: Brazil

Organisation: InfoAmazonia

Organisation size: Small

Publication date: 23/8/2021

Credit: Juliana Mori, Renata Hirota, Eduardo Geraque, Felipe Barros, Sonaira Silva, Tatiane Moraes, Guilherme Guerreiro Neto, Juliana Arini, Leandro Chaves, Camilo Estevam, Rebeca Navarro, Lucas Lobo, Leandro Amorim, André Hanauer, Erlan De Almeida Carvalho, Erico Rosa, Guilherme Lobo, Robson Klein Ramon Aquim, Dell Pinheiro, Laiza Lopes, Laura Sanchez, Tony Gross

Biography: Juliana Mori is a journalist specialized in audiovisual productions and visualization of geospatial data. Co-founder and editorial director of InfoAmazonia, an independent media outlet that uses maps, data, and geolocalized reports to tell stories about the tropical forest over the nine Amazon countries. Graduated in journalism at Pontifícia Universidade Católica (PUC), in São Paulo, Master in Digital Arts at Universitat Pompeu Fabra (UPF), in Barcelona.

Project description:

Inhaling Smoke is a special project that investigates the effects of air pollution caused by wildfires on the health of the Brazilian Amazon population during the pandemic.

This toxic synergy was the object of an unprecedented data analysis carried out by a multidisciplinary team of journalists, geographers, statisticians and scientists. We analysed satellite data to determine which locations were most affected by air pollution during the 2020 wildfires and how it impacted the health of amazonians. Smoke was related to an 18% increase in severe cases of Covid and 24% increase in hospitalizations for respiratory syndromes in the 5 most affected states.

Impact reached:

The relevance of this project is to show the relationship between two seemingly disconnected events (fire and Covid), highlighting how environmental and public health issues are closely linked.

The data reveals how the pollution from the Amazon fires has a perverse effect on the population that is explained by a specific geography of fire – the most affected municipalities, in different states, indicate the expansion of the arc of deforestation.

Through extensive data analysis, the project was able to quantify the impact of fire-attributable pollution on the worsening of Covid cases, providing subsidies both for the reader to understand the gravity of the environmental crisis and for the government to make decisions based on the data.

The importance of a project like this in 2021 is similar to a large-scale post-mortem examination, bringing evidence and proof of the devastation of the Amazon and how this affects its citizens, even those in urban areas far from where the environmental crimes occur.

To tell the stories that the data was revealing, we assembled a team of local reporters, distributed throughout the most impacted states according to our analysis.

Five reports were published and gained nationwide attention, driven also by a publishing partnership with the largest national daily newspaper in Brazil and with the two institutions that were partners in the analyses, the Ufac and the Oswaldo Cruz Foundation (Fiocruz), the federal institution of science and technology.

The articles were republished in local newspapers and by research institutes that had been a reference for our project. We have participated in several meetings and interviews to talk about the results, an academic article with Fiocruz is being written based on our data, and a technical note from the Acre Public Ministry also refers to our data to stress the importance of monitoring air quality.

Techniques/technologies used:

As a primarily data-driven journalism project, data investigation was the core and starting point of the project. Given the absence of regional air quality data (no Amazonian city had fixed air quality monitoring stations), we processed satellite information to calculate air pollution in the region, generating open data (accessible and documented) for all municipalities in the Legal Amazon.

This data was then, through statistical analysis, cross-checked with the respiratory illness hospitalization (SARS) database, and specifically the hospitalization cases classified as Covid-19, and we were able to prove the hypothesis that particulate matter from smoke aggravated Covid cases in the 2020 burning season.

The work with the data was extensive but can be summarized in two main steps: geoprocessing and statistical modeling. The InfoAmazonia analysis processed the various estimates per day from CAMS to arrive at the daily average concentration of fine particulate matter (PM 2.5) for all municipalities in the Legal Amazon.

The statistical model built specially for the analysis tested several scenarios – including wildfires, deforestation alerts, population, and precipitation – and found significance mainly between cumulative pollution and official numbers of hospitalizations for both SARS and Covid-19.

Most of the code used to download, tidy and analyze the data was written in R, besides QGis and Google Earth Engine for geoprocessing.

Besides this, the main characteristic of the work was to join data with locally told stories.

To support the stories, a project was developed that combines Editorial Design and Information Design, combining visual impact with data visualization. It helped to tell the story of Inhaling Smoke with interactive graphics, visual effects, colors, and typographical choices that contributed to the reader’s immersion in the special.

What was the hardest part of this project?

The hardest part of the project was to work with the satellite data and defining which pollution air databases would most fit our purposes.

The immediate data cross-referencing of Covid cases and fire hotspots soon showed that the relationship with health was not the fire itself, but the pollution that it generates. We then looked at air quality datasets and realized there were a multitude of variables and models, and that it required a lot of processing to get the data we needed to analyze the impact of fires on human health.

An initial survey, through interviews with experts, scientists consulting and documentation of academic studies, was done to identify the key air pollution datasets that would be of interest to the project. At this stage, we understood that fine Particulate Matter (PM 2.5, up to 2.5 micrometers in diameter) would be the main variable to be.

Still at this stage, we also understood that PM 2.5 cannot be observed directly from satellites, and therefore there is a need to translate it from the observed Aerosol Optical Depth (AOD). This can be done in several ways and with different models. A series of comparative tests were done with the main datasets (as explained in this documentation), and we decided to use the near real time estimates of CAMS-NRT, from the European Centre for Weather Forecasts (ECMWF). This remote sensing data processed by the InfoAmazonia team was validated with the data measured on the ground by air pollution sensors in Acre.

Working with Covid-19 data was also not easy, because we were working with data from a pandemic still ongoing, so a totally dynamic knowledge about the disease, in addition to the high underreporting of Covid-19 cases (43% of SARS admissions had undefined causes).

What can others learn from this project?

Knowing the health impact of air pollution related to fires gives you important knowledge to tell stories of the deforestation fires that affect not only the rural area where the fires are emitted but also the urban population hundreds kilometers away, helping to show the public, and possibly decision-makers, how the environmental and public health issues are closely related.

The difficulty of obtaining regional data on air pollution (none of the Brazilian Amazonian cities have permanent monitoring stations, for instance) can be overcome using global satellite images which, despite showing the plumes of pollution from burning and providing a good research alternative, do not allow for detailed analysis at the more local level.

For the stories we’re telling it is then interesting to confront the data observed through remote sensing with more granular information that reflects local complexity and helps to validate the results obtained in numbers with real stories. In addition to the local population affected by the problem, it is important to listen to local health authorities, doctors, and frontline professionals who can tell us if what we observed from space was really felt on the ground.

Project links: