SCIPCC Dashboard: Visualizing Climate Security Evidence in the IPCC WGII AR6 report
The Scientific Uncertainty in IPCC dashboard (SCIPCC) gathers and classifies IPCC statements based on likelihood and confidence levels. The dashboard provides a visualization of the evidence presented in IPCC reports and offers a search tool that allows to identify statements, assess their confidence levels, and compare them.
In this post, I provide a first application of the dashboard in regards to statements on climate security made in the Working Group II (WGII) contribution to the Sixth Assessment Report (AR6).
How to Read the Scatter Plot:
The scatter plot compares two climate security sub-corpora: statements with high to very high confidence levels, and statements with low to medium levels. The y-axis plots the frequency of words contained in the former, and the x-axis plots the frequency of the latter. Words near the y-axis are from statements that are backed only by strong evidence while words near the x-axis are from statements with incomplete or medium evidence and consensus. Words in the middle are from statements with mixed confidence levels. The further the words are from the origin point, the more frequent they are in the climate security corpus.
How to Use the Dashboard:
You can browse the climate security statements in the AR6 WGII report either by typing words in the search bar below or by clicking on words in the scatter plot. Once a word has been selected, the dashboard will show a list of sentences categorized by confidence levels. Note that only sentences containing confidence statements are included in the corpus. If you wish to cite an IPCC statement, I recommend you to search it in the original PDF as the text in this dashboard was processed and might slightly vary from the original document.
The dashboard is not responsive, please access it using a desktop computer or a tablet.
How the Dashboard Was Made:
The text contained in the WGII AR6 report was tokenized into a list of sentences using Python’s NLTK library. The sentences were then selected for inclusion in the corpus if they contained words related to climate security (i.e., conflict, violen, tension, peace, security) and scientific uncertainty (i.e., likelihood, confidence, evidence, agreement). The confidence levels were automatically extracted and converted into metadata. The final HTML dashboard and scatter plot visualization were made using Python’s Scattertext library.