Small changes, big impact: the role of threshold selection in extreme risk assessment

Extreme Value Theory (EVT) provides a critical framework for modeling rare, high-impact events like heavy precipitation or financial losses, primarily through the Peaks Over Threshold (POT) approach. Central to this method is the selection of an appropriate threshold to isolate extreme observations; however, this choice presents a significant trade-off between bias and stability. A threshold set too low introduces non-extreme data that biases results, while one set too high limits the sample size, leading to unreliable, unstable estimates. Consequently, careful threshold selection is essential for ensuring the accuracy and reliability of risk assessments in strategic planning.