Regression models for the full distribution to exceedance data.

Area of Science:

• Environmental Science
• Statistics
• Climate Science

Background:

• Increasing occurrences of climate change-linked events necessitate improved analytical methods.
• Extreme Value Theory (EVT) is crucial for analyzing the tail of probability distributions.
• Existing extensions to EVT models include regression for tail parameters and bulk distribution modeling.

Purpose of the Study:

• To present a novel extension to exceedance models for analyzing extreme climate events.
• To incorporate covariates like location and seasonality into the bulk distribution of exceedance models.
• To improve the estimation of extreme quantiles and enhance predictive capabilities for climate-related data.

Main Methods:

• Developed a new extension to exceedance models, integrating covariate effects into the bulk distribution.
• Employed a Bayesian inference approach for parameter estimation.
• Utilized Markov Chain Monte Carlo (MCMC) methods for computational implementation.

Main Results:

• The proposed model effectively captures the influence of covariates such as location and seasonality.
• Demonstrated efficient estimation of extreme quantiles using temperature data (maximum and minimum).
• Showcased a predictive advantage over previously established models in literature.

Conclusions:

• The novel exceedance model extension provides a robust framework for analyzing climate change impacts.
• The Bayesian approach with MCMC ensures reliable parameter estimation and uncertainty quantification.
• The model's ability to incorporate covariates enhances its applicability for real-world climate risk assessment.