Robust fitting of generalized extreme value distribution for extreme wind speeds: a case study using d4PDF data in Japan
This study proposes a robust technique for parameter estimation in the generalized extreme value (GEV) distribution, utilizing the probability integral transform statistics. This estimator is tailored to improve robustness against outliers, which are commonly encountered in extreme value analysis. W...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Springer
2026
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/122897/1/122897.pdf http://psasir.upm.edu.my/id/eprint/122897/ https://link.springer.com/article/10.1007/s10651-025-00691-5?error=cookies_not_supported&code=367205eb-8de2-4370-82fd-fdb2df769d76 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study proposes a robust technique for parameter estimation in the generalized extreme value (GEV) distribution, utilizing the probability integral transform statistics. This estimator is tailored to improve robustness against outliers, which are commonly encountered in extreme value analysis. We examine the properties of probability integral transform estimator (PITE) by evaluating its efficiency metrics, robustness through its score function, and breakdown point. This demonstrates its capacity to manage extreme datasets with low susceptibility to outlier effects. Additionally, PITE is computationally straightforward, facilitating its use in real-world settings. Results from Monte Carlo simulations highlight the remarkable performance of PITE over conventional methods, particularly under conditions of significant data variability and contamination, establishing it as a viable alternative for modeling extreme values. Applied alongside the block maxima approach, the GEV model with the PITE effectively captures the behavior of extreme wind speeds in Japan. The method is applied to analyze extreme wind speed data from the database for policy decision-making for future climate changes (d4PDF), which includes both historical records and future projections (4°C global warming scenario) across the region. This analysis extends to estimating return levels, providing crucial insights into the intensity of extreme wind speeds expected in Japan. The return level estimation quantifies the wind speeds likely to be exceeded once every 5, 10, 50, and 100 years, delivering essential data for infrastructure planning and risk assessment. By analyzing historical and future d4PDF datasets, this approach allows for a comparison of past and projected wind speed extremes. These insights are vital for regions in Japan prone to severe meteorological events, ensuring they are better prepared and resilient against future extreme wind scenarios. |
|---|