Regional Frequency Analysis applied to extreme rainfall events: Evaluating its conceptual assumptions and...

Area of Science:

• Hydrology
• Meteorology
• Statistical Analysis

Background:

• Regional Frequency Analysis (RFA) is crucial for assessing meteorological data quality and enhancing fitting processes.
• RFA assumes the formation of homogeneous groups with independent and identically distributed data.
• Its application in tropical-subtropical regions requires validation of existing homogeneity and distribution selection criteria.

Purpose of the Study:

• To apply RFA for assessing one-day annual maximum rainfall probability in São Paulo, Brazil.
• To evaluate the critical limits for homogeneity (H≤1.00) and distribution selection (|Z|≤1.64) using Monte Carlo simulations.
• To develop a computational algorithm for selecting critical limits based on specified probabilities of rejecting a true null hypothesis.

Main Methods:

• Regional Frequency Analysis (RFA) applied to rainfall data from São Paulo, Brazil.
• Monte Carlo simulations to test critical limits for homogeneity (H) and distribution selection (Z).
• Development of a novel algorithm for adaptive critical limit selection.

Main Results:

• The H≤1 limit is deemed appropriate for homogeneity assessment.
• The |Z|≤1.64 limit may excessively reject true null hypotheses, especially for the general logistic distribution.
• A new algorithm was provided for selecting critical limits tailored to desired Type I error rates.
• Four homogeneous groups were identified within the São Paulo weather station network using the refined criteria.

Conclusions:

• The study validates and refines RFA methods for rainfall analysis in tropical-subtropical climates.
• Generalized logistic and extreme value distributions are recommended for probabilistic assessments in the identified homogeneous groups.
• The developed algorithm enhances the reliability of RFA by allowing precise control over hypothesis testing error rates.