In July 2021, Germany experienced its costliest riverine floods in history, with over 189 fatalities and a staggering \euro33 billion in damages. Following this event, news outlets widely disseminated information on the flood’s aftermath. Here, we demonstrate how newspaper data can be instrumental in the assessment of flood socioeconomic impacts often overlooked by conventional methods. Using natural language processing tools on 14,888 unique newspaper articles, we estimate the impacts of the 2021 flood on various sectors and critical infrastructure, including water contamination, mental health, and tourism. Our results revealed severe and lasting impacts in the Ahr Valley, even months after the event. At the same time, we identified smaller-scale yet widespread impacts across Germany, which are typically overlooked by existing impact databases. Our approach advances current research by systematically examining indirect and intangible flood impacts over large areas. This underscores the value of leveraging complementary text data to provide a more comprehensive picture of flood impacts.

Hydrological extremes, such as droughts and floods, can trigger a complex web of compound and cascading impacts due to interdependencies between coupled natural and social systems. However, current decision-making processes typically only consider one impact and disaster event at a time, ignoring causal chains, feedback loops, and conditional dependencies between impacts. Analyses capturing these complex patterns across space and time are thus needed to better inform effective adaptation planning. This perspective paper aims to bridge this critical gap by presenting methods for assessing the dynamics of the multi-sector compound and cascading impacts (CCI) of hydrological extremes. We discuss existing challenges, good practices, and potential ways forward. Rather than pursuing a single methodological approach, we advocate for methodological pluralism. We see complementary roles for analyses building on quantitative (e.g. data-mining, systems modeling) and qualitative methods (e.g. mental models, qualitative storylines). We believe the data-driven and knowledge-driven methods provided here can serve as a useful starting point for understanding the dynamics of both high-frequency CCI and low-likelihood but high-impact CCI. With this perspective, we hope to foster research on CCI to improve the development of adaptation strategies for reducing the risk of hydrological extremes.