Visual Analytics for Sustainability and Climate Change/Gitanjali Yadav

Interview to Gitanjali Yadav by Chiara Somajni, Giovanni Profeta and Iolanda Pensa, 4 June 2025, CC BY 4.0.

OSF Open Science Framework: https://osf.io/y9pmr/
ORCID Gitanjali Yadav. https://orcid.org/0000-0001-6591-9964

Gitanjali Yadav has been working at NIPGR-National Institute of Plant Genome Research, New Delhi, since 2006. Her research is in Data Science for Climate, Conservation and Food Security. During this time, she spent five years teaching at the University of Cambridge. She returned to India on the last flight before the global lockdown, with a heavier focus on Climate and Policy. This was also when she co-founded semanticClimate with Cambridge and EU Colleagues: a citizen movement to liberate Climate Data from locked literature. She is an advisory board member of the Delhi City Knowledge Cluster, managed by Delhi Research Implementation and Innovation. This City Cluster (and four others across India) were established by the Office of the Principal Scientific Advisor (PSA) to the Government of India. Apart from Climate, her interest is in Invasive Species and the use of AI to address Human-Wildlife Conflict. She works with CODATA and CoARA to advocate FAIR data principles and open science in the Global South.

Notes from the interview

Research focus

Climate change and crops (climate change resilient crops; soils; photosynthesis/carbon capture and carbon sequestration; biodiversity conservation, e.g. invasive species).
As an activist scientist, liberate climate data to empower any person to take climate action locally.

#semanticClimate – is a global citizen science movement. Co-founded by Yadav with Peter Murray-Rust (chemist, professor emeritus, Cambridge University, UK, https://orcid.org/0000-0003-3386-3972) and Simon Worthington (researcher in future publishing, member of HsH JointLab and Open Science Lab, TIB, Leibniz University, Hannover, Germany, http://orcid.org/0000-0002-8579-9717).

semanticClimate purpose – empower young scientists, women in science, and rural and suburban communities of youth. Making climate change knowledge accessible, facilitating action.

semanticClimate activities – tech development and research to liberate knowledge; outreach work; volunteer internship.

semanticClimate /process –

pick one IPCC AR6 report chapter (based on individual interest),
convert chapter to machine-readable file (html),
read to identify and manually underline difficult words (to laymen),
run code to extract words and phrases with difficult words from the whole chapter, erasing the ones that are not relevant,
link difficult terms to wikipedia articles where present,
create knowledge graphs through Obsidian,
embed knowledge graphs back to Wikidata,
list difficult terms that are not present in wikipedia.

Scripts and code used, and more details on process:

Jupiter notebooks and Google Colab, written in Python
Started using Hugging Face's transformer models, incorporated in Google Colab
Recently acquired an Nvidia chip to run more powerful GPU transformer models (not allowed in Google Colab)
Incorporate scripts and code into Jupiter notebooks where the Nvidia chip will be used.

download paper from OpenAlex
create a datatable with metadata extracted from articles
perform named entities recognition via transformers
perform statistical pairwise correlation to create co-occurrence maps and clustering
visualisable graphs: word cloud, co-occurrence maps, heat map (clustering analysis)

semanticClimate /languages – Focus on articles in English (working language within the team). Some students from rural India suggested implementing other languages. Since 2020, the project has been multilingual, though not perfectly implemented: words are extracted in English, dictionaries translated into multiple languages.

semanticClimate /accessibility – All work is open on GitHub, the meeting notes are on Slack. We are allowed to use the list of articles and the title of chapters of the IPCC as clusters, and the other clusters Yadav created. The IPCC is a reference. Work on the IPCC AR6 report is not finished yet, and more reports are being produced.

Visualisations run by Yadav

Atomic 3d structure visualisation,
Atomic network 3d visualisation for molecules.
For large genomes: data visualised as heatmaps to show expressed genes under different circumstances, under different conditions of climate change, soil parameters.
Interested in the xillion civilisation of microbes, communities in the soil, kinds of fungal networks in the forest - from the fungi spreading for kilometres, to the microbe communities on one single leaf.

Visualisation related to semanticClimate based on each article's metadata:

bibliometry graphs: (co-)authorship networks, graphs of * what species are in which countries
word clouds
co-occurrance maps of terms
statistical pairwise frequency of terms
clustered frequencies.

Requirements

Based on keywords, visualise what work has been done either inside or outside wiki on that topic.
What are the terms that are occurring more frequently
Which are the terms that are co-occuring more significantly than other pairs of terms
What are the clusters of terms that are constantly recurring together
Connecting publications, topics/subjects starting from a keyword, problems...

Examples:

what kind of frogs are associated with what kind of parameters,
what kinds of papers have been published on frogs which have anything to do with a third variable.

Yadav would like to map multiple variables: in semanticClimate right now, you can only map one variable at a time, like location, species, chemistry.

Gaps

Yadav's group uses Wikidata to create lists of terms (geographies including small localities, hotspots, species…), but the lists are never complete.
Who's been funding organisations?

Tools /comment on prototypes The project's prototypes are very visual. Popularity is not an important parameter for Yadav's work.

AI – Yadav's group uses Hugging Face, more to come. No special recommendation on AI use.

Further information shared – Wikipedia is not the place to go for information in Yadav's work.

Dehli has a terrible problem of pollution. Working at a local level in different regions, community members suggested creating a penalty of a small amount of money to correct harmful habits (e.g. burning of tyres). Showing that local problems are linked to global ones and addressed globally, too, is reassuring and empowering.

Collaboration opportunities – Links shared by Gitanjali Yadav:

Event (invitation to present the project): https://semanticclimate.github.io/p/en/events/ALR_June25/
Funding opportunity: The Navigation Fund https://os.nav.fund/meeting-fund/ supports an international meeting this year.

Links shared by Iolanda:

Requirements suggested

Downloading the list of articles
Links to papers
Connecting articles with other content (papers, different subjects, problems)
Geographic coordinates of content
Who funds organisations working on climate change
Based on keywords, visualise what work has been done either in wiki or outside on that topic
Terms occurring more frequently within one paper/article
Significant co-occurring terms
Clusters of terms constantly recurring together
Mapping multiple variables (e.g. location, species, chemistry)

Knowledge gaps about climate change on Wikipedia

Content in languages other than English
On Wikidata, lists (e.g. of species, locations) are never complete