Extreme sea levels threaten coastal communities around the world. Understanding, preparing for, and ultimately mitigating against these hazards is of vital importance to these communities. To inform decision makers and stakeholders, extreme value analysis can be used estimate the probability of rare events. One such method is regional frequency analysis (RFA), whereby data from across a region is pooled to inform the extremal characteristics of the entire region. In this study, we present an update to RFA to global extreme sea levels, building on the work from Collings et al., (2024) (RFA v1). This framework draws upon numerous data sources to estimate extreme sea levels driven by storm surge, astronomical tides, wave setup and regional mean sea level at any point along the global coastline (excluding Antarctica). Improvements have been made across a number of areas, including: the spatial discretisation of record locations, implementation of the novel TAILS automated threshold selection method (Collings et al., 2025b) and the introduction of the a calibrated index flood at record locations. Input data have also been updated where newer versions have been made available. We show that each of the methodological updates leads to improvements in the resulting local extreme water levels output by the RFA. Globally, mean bias at the 1 in 100-year return levels across all tide gauge locations decreases by 66% compared with RFA v1. We also show that RFA v1 typically underestimates coastal flood risk, as the magnitude of return levels increases across much of the global coastline. This is primarily caused by updated datum correction data and tide model data, and the improvements made to the threshold selection and calibration of the index flood. Overall, the updates made to this method result in a more robust estimation of extreme sea levels at the coastline, enabling better decision making and risk management for vulnerable stakeholders. • An updated and improved regional frequency analysis of global extreme sea levels, including contributions from tides, storm surge, wave setup and regional mean sea level. • Application of novel methods such as enhanced spatial discretisation, TAILS automated threshold selection and calibrating the index flood, as well as updating input data. • Overall reduction of 66% (7 cm) in mean bias at GESLA tide gauge locations for the 1 in 100-year return period, when compared against a previous global regional frequency analysis approach.
Collings et al. (Sun,) studied this question.