Long-Term and Seasonal Trends in Global Wave Height Extremes Derived from ERA-5 Reanalysis Data

Alicia Takbash,Ian Young

doi:10.3390/jmse8121015

Alicia Takbash, Ian Young

Open Access

https://doi.org/10.3390/jmse8121015

Copy DOI

Abstract

A non-stationary extreme value analysis of 41 years (1979–2019) of global ERA5 (European Centre for Medium-Range Weather Forecasts Reanalysis) significant wave height data is undertaken to investigate trends in the values of 100-year significant wave height, Hs100. The analysis shows that there has been a statistically significant increase in the value of Hs100 over large regions of the Southern Hemisphere. There have also been smaller decreases in Hs100 in the Northern Hemisphere, although the related trends are generally not statistically significant. The increases in the Southern Hemisphere are a result of an increase in either the frequency or intensity of winter storms, particularly in the Southern Ocean.

Highlights

The investigation of temporal and spatial variabilities of metocean parameters, such as the significant wave height, Hs, is of significant importance, especially in times of climate change
The transformed stationary (TS) Extreme Value Analysis (EVA) was applied at 1◦ × 1◦ resolution as noted in Section 3, and the trend was evaluated at each grid point
This is a location in the Pacific basin of the Southern Ocean, west of South America

Summary

Introduction

The investigation of temporal (long-term and seasonal) and spatial variabilities of metocean parameters, such as the significant wave height, Hs, is of significant importance, especially in times of climate change. In addition to violating the requirement for independent data, as the pdf is fitted to all data rather than only extremes, the resulting estimates tend to largely follow spatial distributions of mean values rather than true extremes. Despite these limitations, the approach has been widely applied for studies of the global distribution of wave height extremes [11,20]. That both satellite and model datasets span more than 30 years, the need to rely on such approximations no longer exists

Objectives

Methods

Results

Discussion

Conclusion