Adapting the regional frequency analysis based on L-moments to improve the standardized precipitationevapotranspiration index

ABSTRACT

The standardized precipitation-evapotranspiration index (SPEI) is a widely used probability-based method that categorizes drought and wet events based on their expected frequency of occurrence. Regional frequency analysis based on L-moments (RFA-Lmom) is often employed to enhance probabilistic assessments of extreme hydrological data, which typically assume positive values. This study investigated whether applying the RFA-Lmom regionalization technique to the difference between precipitation and potential evapotranspiration (P-PE), the input variable for SPEI, can improve the index’s ability to meet the normality assumption. We conducted analyses using Monte Carlo experiments, accounting for distinct climate conditions worldwide, and a case study in the watershed of the Piracicaba, Capivari and Jundiaí rivers, situated in the states of São Paulo and Minas Gerais, Brazil. In this region, P-PE frequency distributions may exhibit negative and positive sample means. Our findings suggested that modifying the RFA-Lmom method by replacing the L-moment ratio L-CV with the L-moment scale measure and using an additive model (instead of the original multiplicative procedure) for SPEI calculation allows the application of this regionalization technique to P-PE amounts. This adapted RFA-Lmom consistently enhanced the ability of SPEI frequency distributions to meet their normality assumption, thereby improving the quality of drought assessments based on this standardized index.

Key words drought; additive approach; normality assumption; L-moments