Training School on Statistical Modelling of Compound Events

Abstract

In this project, we aim to extract detailed information on the impacts of extreme events (e.g., floods, storms, and droughts) based on Red Cross reports. By analyzing these reports, we will gather comprehensive data on both direct and indirect impacts on society and the environment. This includes metrics like fatalities, economic losses, and increases in migration. Since the Red Cross focuses especially on the Global South, we expect to address spatial biases common in global impact databases (e.g., EM-DAT focuses on Europe and the US). We will use natural language processing (NLP) tools to extract structured information from the reports. Our approach will incorporate both supervised classification methods (e.g., Sodoge et al., 2023) and large language models (LLMs) (e.g., Carvalho et al., 2024).

Taís Carvalho

Postdoc researcher

I like to explore unusual data sources to support water resources planning and management and natural hazards research.