Description

The articles scraped from the web, while talking about climate change with words, described the controversy also with images. This visualization aims to show how climate change, a scientific phenomenon that most people reading these articles don't experience first hand, is visually represented by the internet news outlets.

Each year, climate change is visually represented by different categories of images: in 2014, the most recurring categories were images of pollution, animals and visualizations, in 2015 there were mostly pollution, disasters and depiction of the Earth, and in 2016 the most recurring images were from the categories of industries, pollution and disasters. In the below visualization, it's possible to see how the borderline cases change through the years: Industry is the category that sees the most significat growth, while Animals decreases the most.

To better help the reader, sub-categories were visualized to give a sense of the images used, where the size of each image represents the number of photos that depict the same subject.

The images were also analyzed by their color, to see if each category helped shape a different kind of collective imagery of climate change. With this analysis, that was executed with Kromotology and Clarifai, it was noticed that while the quantity of the categories changes through time, the same cannot be said for their quality: the images share very similar color pattern and do not define a distinct chromatic imagery.

Protocol

Starting from the corpus of articles gathered from Google News, each link was scraped for its best image, usually the one used as the hero image for the article. This gathering was made with Aylien, with the result of a list of links that were then downloaded with the Firefox extension DownThemAll. This corpus of images was then tagged manually in the categories shown in the above visualization by concept. This led to a corpus of images that were distributed to the different categories, sometimes more than one. The sub-categorization was done with the help of IBM’s Watson, in particular with the image analysis tool. The result was a list of visual correspondences with a certainty index: the result with the highest score was used to define the sub-category. The result was then organized in a general treemap of the year, and for each category a smaller treemap that visualized the quantities through the years.

Data

Data source: Google News, Aylien, IBM Watson Image Analysis

Download data

The 1122 images are divided in a series of categories, that were organized in folders and sub-folders. Images that belonged to different categories were put in all the folders, duplicating them.
The second part of the visualization is based on the quantities of these categories, organized in another series of sub-folders that was counted to create the series of treemaps.