Project description:

Latin America has become a dumping ground for plastics from other countries. For more than six months, we analyzed how thousands of tons of this waste arrive in Mexico, Peru, Ecuador, Chile and Colombia from the United States and other European Union countries, with the promise of being recycled. The garbage enters customs through a commercial exchange that lacks transparency. Poor oversight by the authorities prevents the final destination of this waste from being known.

Impact reached:

It helped the public to know, to understand and measure the problem of plastic waste importations. Unlike the reports on the subject published by some organizations, with this work we share information about the companies responsible for this trade, we talk about the omissions of the authorities and the gaps that can give rise to organized crime.

Techniques/technologies used:

The website was made using Angular as it’s framework. The most complex part of the development however was on the data analysis and data explorer. We had to query a private registry of global shipping entries. Then proceeded to analyze the resulting datasets with various tools ranging from excel sheets to more advanced tools like pandas, NumPy and others. The final dataset was comprised of 59110 entries.

We didn’t want to just present the data as a table or some charts. We wanted the user to explore each entry individually. After looking for alternatives, we found inspiration for our data explorer in Flavio Gortana’s Coin museum. The making of the data explorer was challenging. We decided against other data oriented toolsets as D3.js performance concerns. Just like the coin museum, our data explorer uses Pixi.js which is a 2d WebGl renderer. This means we had to code the whole thing from scratch e.g. the positions of each data-point, their animations, mouse interactions, etc.

We tried to offset most of the heavy calculations from the browser by calculating positions, categories, and the circle-area chart data ahead of time using Pandas and NumPy. During this step we also cleaned the datasets and translated them to spanish using a meta-translator available for python.

As for the graphical assets, we used both StableDiffusionV1-4 and MidJourney for image generation. We used MidJourney to generate representative images of each country filled with pollution. This was an interesting process that was not as straight forward as we initially thought it’d be. As for StableDiffusion we ran the AI in our machines to generate a collection of 996 different landscapes of a future filled with trash and pollution. We used this gallery to illustrate each data point in our data explorer.

Context about the project:

For six months, more than 100,000 shipments registered on an international trade platform were reviewed. This analysis had to be done one by one, since it includes the information that each exporter registers and is not standardized. In the end, we only used half of the information, since there were containers that included other materials, which made it almost impossible to determine how many kilos corresponded to plastics. But that review allowed us to see the large amount of garbage that enters our countries. Another problem is that governments do not provide information in such detail and tend to report fewer kilos. And the companies, not being obliged to provide this information publicly, simply did not provide it, nor did they respond to our interview requests.

What can other journalists learn from this project?

In this work we combine the use of technologies, but we also resorted to a data-by-data review carried out by two people, which helped us to have a complete picture.
We also believe that it is important to look at these practices that are harmful to the environment and health.

Project links:

https://ojo-publico.com/especiales/latin-america-the-repository-for-other-peoples-garbage/

https://open.spotify.com/episode/0uQQf0SOnUOR6wD56EfNZf?si=ae7c98d053ee4dbb