PyData Tel Aviv 2022

Extreme Value Analysis
12-13, 15:45–16:15 (Asia/Jerusalem), Track 2

How to analyze the extremes of a distribution to predict probabilities and magnitudes of events outside the observed range. Presenting the theoretical framework of Extreme Value Theory with examples from a case study - prediction of the 50-year wind across Israel.

How do you use data from 15 years of observations to predict the magnitude of a “once in 50 years” storm? How can we build a network infrastructure that can handle the maximum traffic over a decade, using just one year of data?
Extreme value Analysis (EVA) provides a statistical and technical framework for the analysis of extreme deviation from the median of probability distributions. It is used in multiple fields to predict the probability of the recurrence of extreme outliers in data or even of the occurrence of heretofore unobserved phenomena.
This talk aims to provide the listener with a basic understanding of the analysis framework and the mathematical justification for its correctness. Numerous alternative routes, pitfalls and decision points encountered during analysis will be presented.
Currently, available libraries in Python allow only very rudimentary EVA, and many useful and sometimes necessary operations lack implementations. So while this talk is aimed at people hearing about EVA for the first time, it is also a call for contributors to implement those necessary features.

Prior Knowledge Expected

No previous knowledge expected

Former Climate Researcher in the Israeli Meteorological Service, lead author of the Wind Energy Potential Atlas of Israel and the 50-year Base Wind map for construction safety. Currently working as a Data Scientist at Pagaya.