&lt;b&gt;Extreme value theory applied to the standardized precipitation index&lt;/b&gt; - doi: 10.4025/actascitechnol.v36i1.17475

Gabriel Constantino Blain

doi:10.4025/17475

Abstract

The Standardized Precipitation Index (SPI) is a mathematical algorithm developed for detecting and characterizing precipitation departures with regard to an expected regional climate condition. Thus, this study aimed to verify the possibility of using the time-independent general extreme value distribution (GEV) for modeling the probability of occurrence of both SPI annual maxima (the maximum monthly SPI value; SPImax) and SPI annual minima (the minimum monthly SPI value; SPImim) obtained from the weather station of Campinas, State of São Paulo, Brazil (1891-2011) and to evaluate the presence of trends, temporal persistence and periodical components in these two datasets. The goodness-of-fit tests used in this study quantify the agreement between the empirical cumulative distribution and the GEV cumulative function. Our results have indicated that such parametric function can be used to assess the probability of occurrence of SPImin and SPImax values. No significant serial correlation and no trend were detected in both series. For the SPImim, the wavelet analysis has detected a dominant mode in the 4-8 year band. Future studies should focus on the development of a GEV model capable of accounting for such feature. No dominant mode was found for the annual monthly SPI maximums.

Highlights

The Standardized Precipitation Index (SPI)1 is a mathematical algorithm developed for detecting and characterizing precipitation departures with regard to an expected regional climate condition
The KolmogorovSmirnov test (KS)-L, Anderson-Darling test (AD), AU and AL tests quantify the agreement between the empirical cumulative distribution and the theoretical cumulative function (SHIN et al, 2012; WILKS, 2011)
All the goodness-of-fit tests have indicated that the time independent general extreme value distribution (GEV) model may be used to evaluate the probability of occurrence of the SPImax as well as the SPImim obtained from the weather station of Campinas

Summary

Introduction

The Standardized Precipitation Index (SPI) is a mathematical algorithm developed for detecting and characterizing precipitation departures with regard to an expected regional climate condition. According to Bordi et al (2007) and Hayes et al (2011), the occurrence of dry or wet periods does not depend entirely on low or high precipitation totals. It depends on a negative or a positive anomaly of these totals with respect to an expected regional climate condition. After Bordi et al (2007) have evaluated the nature of the tails of both precipitation and SPI distributions they indicated that this drought index is better than the precipitation for representing extreme wet and dry (monthly) events

Objectives

Methods

Results

Conclusion