Loading

Mapped: Inside Carbon Brief’s Cosmos database of 1.8 million climate studies

This is the vast “cosmos” of academic literature and evidence that underpins humanity’s knowledge of climate change.

Every “star” – all 1.8m of them – represents one of the studies inside Carbon Brief’s Cosmos database.

The coloured “nebulae” and “galaxies” within this cosmos illustrate where clusters of studies share similar citations and, hence, areas of common academic focus.

For example, these two large circled areas resembling “nebulae” represent, at the broadest level, studies centred around the “physical sciences” (blue) and (yellow).

Zoom in closer and distinct “galaxy” clusters begin to emerge more clearly.

The large green area represents medical studies cited by climate studies. The smaller red area captures the fields of immunology and microbiology.

Both circles represent groups of studies relating to tropical diseases, such as malaria, but from the perspective of different scientific disciplines.

The larger coloured “stars” show the location of all the key studies contained within the “Cosmos 500” ranking of most-cited publications.

This one, for example, is the famous Stern Review published in 2006. Authored by the economist Nicholas Stern, the 700-page report was commissioned by the UK government to investigate and summarise the effects of global warming on the world economy.

As Carbon Brief’s analysis of the Cosmos database shows, it is the most-cited publication across all the reports ever published by the Intergovernmental Panel on Climate Change (IPCC).

Below is an interactive version of the “cosmos”.

Zoom in and out to view the various clusters, using the colour key to determine the topics.

Click on any of the Cosmos 500 “stars” to reveal OpenAlex’s primary metadata about the study.

This visualised “cosmos” of 1.8m data points is what is known as a “network graph”.

It is a way to illustrate the “strength” of the connections – namely, citations and references – between the various nodes within any network. In this case, the network is the Cosmos database, the nodes are the 1.8m studies inside it and the connections are the citations and references.

Studies are pulled towards each other – their “gravitational attraction” – based on their citation relationships. So, if paper A cites paper B (or they are both cited by the same papers), they are treated as similar.

More widely, an algorithm then calculates the diverse range – and strength – of relationships across all the nodes in the network. Carbon Brief estimates that there are roughly 40m relationships between the 1.8m studies.

This technique is called “uniform manifold approximation and projection” (UMAP).

It allows a vast database containing a high-dimensional network of connections to be compressed down into a flat, readable map – while preserving as much of the original structure as possible.

In the cosmos above, studies that form a tight cluster will have many citation links between them and relatively few links to other groups. These clusters might correspond to distinct research communities or topics.

Studies on the edge of clusters, or within near-empty spaces between clusters, represent cross-disciplinary publications that connect otherwise separate fields.

In contrast, studies within the centre of densely packed clusters share connections across the whole common area.

Unlike a normal chart or map, the x and y axes and positional coordinates have no meaning, as such. All that matters is spacing; two papers close together share strong ties; two papers far apart share few ties, or none.

Back to the top