The last day of the excellent AoIR 2025 conference in Rio de Janeiro starts with our panel on climate change communication in Australia and Brazil – two countries which have quite a few similarities in this context: they’re both highly exposed to extreme weather events, have stressed environments in the Amazon and Great Barrier Reef, and and are major fossil fuel exporters. We start with my excellent colleague Tariq Choucair, whose focus is on the discussion of extreme climate events in social media environments.
Extreme weather experiences affect and are affected by individual perceptions and attitudes towards climate change. Individuals who have experienced such events may be more likely to show heightened climate change awareness and engagement, but this works differently for recent and more temporally distant events, and also depends on the type of climate event experience.
How does this in turn affect public discussions of climate change, then? How does increased public attention to climate change during extreme events entangle with underlying beliefs, disputes over responsibility, false claims, and media framing of such events in the context of climate change? This study therefore addresses the question of whether experience of extreme events increase or decrease discussion and/or polarisation on climate change.
This works with a dataset of public discussion of climate events on Facebook between January 2018 and July 2024, comprising some 8 million posts gathered via the now defunct CrowdTangle platform. These posts were then classified for their relevance to climate change discussions, using 500 manually classified posts to train the Portuguese-language BERTimbau model that classified the full dataset.
Peaks in relevant climate change discussions align sometimes, but not always, with the overall volume of posts over time; there is a broad relationship between these activity levels, though. Posts about extreme temperatures in Paraná coincided with Jair Bolsonaro’s cancelling of Brazil’s bid for the COP25 summit.