ABOUT EVENT

This workshop focuses on applying extreme value statistics to estimate the probability and magnitude of rare events in time-series datasets. Using wave heights as a case study, participants will learn to apply the Generalized Extreme Value distribution and Generalized Pareto Distribution through the Annual Maximum Method and Peaks Over Threshold approach, with additional discussion on machine learning approaches for modeling extremes in non-stationary datasets. Using Python, attendees will apply two common approaches for extreme value modeling to a real-world environmental time-series dataset. The Annual Maximum Method will be applied using the Generalized Extreme Value distribution, while the Peaks Over Threshold approach will be applied using the Generalized Pareto Distribution. Attendees will also estimate the probability and magnitude of extreme events and compare results from the two methods.

Prerequisites

• Workshop materials, including Python notebooks and datasets, will be provided through the workshop GitHub repository. Materials will need to be downloaded prior to the workshop. https://github.com/lydia-rice/extreme-value-modeling-workshop

Learning Objectives

Participants will learn to apply extreme value statistics to estimate the probability and magnitude of extreme events in time-series datasets. They will be able to:

• Use the Generalized Extreme Value distribution and Generalized Pareto Distribution to model extremes.
• Apply the Annual Maximum Method and Peaks Over Threshold approach to a wave-height case study.
• Discuss the potential application of machine learning approaches for modeling extremes in environmental datasets.

Tools Used

• GitHub (workshop materials and datasets); Python and Jupyter Notebooks (coding exercises); Pandas, SciPy, and Matplotlib (data analysis and visualization).