Enhancing impact monitoring by using computational text analyses

Climate extremes, such as floods, heatwaves, and droughts, have myriad impacts across natural and social systems. However, traditional methods used for monitoring impacts tend to focus on single hazards or indicators (e.g., fatalities), address only quantitative consequences (e.g., economic losses), and frequently overlook indirect and social consequences (e.g., conflicts, mental health). Here, we show how text data can be used to measure the societal impacts of climate extremes across diverse text sources, including newspapers, social media, and Wikipedia articles.

First, we analyze over 26,000 newspaper articles on the July 2021 river floods in Germany to reveal cascading impacts across sectors like infrastructure, water quality, mental health, and tourism. Second, Twitter data from the 2022 drought in Italy is used to map public concern and perceived consequences, which align with observed socioeconomic indicators. Finally, we scale our analysis globally with Wikimpacts 1.0, a database of climate impacts extracted from 3,368 Wikipedia articles covering 2,928 events from 1034 to 2024, providing national and sub-national records of deaths, injuries, displacements, damaged buildings, and economic losses.

Together, these case studies illustrate the value of text-derived impact datasets for complementing traditional monitoring approaches. We also discuss the challenges of using such datasets, including representational biases, uneven temporal and spatial coverage, and differences in how impacts are reported. We conclude by discussing how the field can move towards shared standards and best practices, enabling more comparable and transparent use of text data for monitoring the impacts of climate extremes.