Applying the Generalized Pareto Principle to Predict Peak Temperatures in Northeast Algeria

Abstract

This study employs the Generalized Pareto Distribution (GPD) to model high temperature events at the Batna station, strategically  applying varying thresholds to enhance accuracy and mitigate tail distribution bias. This was done by analyzing historical rainfall data  from 1981 to 202. Using maximum likelihood estimation and an innovative approach to fitting multi-threshold GPDs, stable thresholds  were established, particularly identifying the Pareto II type as the most suitable model for temperatures exceeding 25°C. Analysis revealed a consistent average temperature of 27.52°C, with a negative skewness indicating a bias toward lower values within extreme  temperatures. The study also predicts return levels for different periods, providing critical temperature thresholds expected to be  exceeded every 2, 20, 50, and 100 years. These findings contribute to understanding extreme temperature dynamics in the region,  supporting urban planning, infrastructure design, and climate resilience strategies.