The association of climate-induced stressors on risk of negative sentiment: An analysis from 462 million geotagged tweets in Europe

This study examines how climate-induced health risks influence negative sentiments on European social tweets from 2015 to 2022. Analyzing over 400 million tweets using NLP tools (NLTK, LIWC22) and spatial-temporal aggregation at the NUTS2 weekly level, we applied a Poisson generalized additive model (GAM) with integrated nested Laplace approximation (INLA) and fused lasso regularization to capture sentiment fluctuations. Results show negative sentiments rise by up to 0.36% when maximum temperatures exceed 26.9°C and by 0.49% during severe droughts (SPI < -3.72). Elevated alder pollen counts (>135 grains/m³) also increase risk of negative sentiment by 0.21%, while temperatures below 2.9°C reduce it by 0.63%. No significant association was found with heat-related mortality or West Nile virus incidence. These findings suggest specific climate-related health factors-high temperatures, droughts, and pollen-trigger negative social media reactions, whereas others, such as mortality and infectious outbreaks, appear unnoticed in public sentiment.