Banca de DEFESA: Luana Oliveira Sales

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : Luana Oliveira Sales
DATE: 29/08/2023
TIME: 09:00
LOCAL: Sala do PPGTARH com Transmissão pela Plataforma MS Teams
TITLE:

NON-STATIONARY MODELING OF FLOOD FREQUENCY AT LOCAL AND REGIONAL LEVELS

KEY WORDS:

Floods, nonstationarity, climate indices.

PAGES: 152
BIG AREA: Engenharias
AREA: Engenharia Sanitária
SUBÁREA: Recursos Hídricos
SUMMARY:

Every year, floods are responsible for causing natural disasters that have impacts on society. Changes in flood behavior are responses to conditions in the hydrographic basin, which can be altered by mechanisms of natural and/or anthropogenic origin. In flood frequency analysis, it is common to assume that historical series are stationary. However, significant changes in flood behavior increase interest in non-stationarity, including investigation of its existence, possible causes, and the degree of intensity of trends in the series. This interest leads to a debate on the incorporation and modeling of changes in flood analyses. This study presents a methodological approach to detect, model, and evaluate the non-stationary behavior of floods at local and regional levels. The approach involves analyzing the quantile frequency of floods using the Generalized Extreme Value (GEV) distribution applied to a stationary statistical model and three non-stationary models. Uncertainties associated with both flood quantiles and parameters of the non-stationary models, which are functions of time and other influential covariates in the study area, are evaluated. In the local analysis, a model selection among the four proposed models is made using the Bayesian Information Criterion (BIC) that best fits the historical series. To control the Type I error (false positives) at the regional level, the False Discovery Rate (FDR) is applied. In this study, an example of applying this methodological approach to 33 series of annual maximum streamflow in the Uruguay River Basin (RH), located in Brazil, is presented. By modeling the flood quantiles associated with potential change mechanisms (covariates) in the RH, in addition to Time, with a focus on large-scale climatic mechanisms such as El Niño-Southern Oscillation (ENSO) and Pacific Decadal Oscillation (PDO), high rates of series that best fit the proposed non-stationary models were identified at the local level (21 to 91%). These rates also exist at the regional level, but in smaller proportions (0 to 88%), with changes in the 50-year flood quantiles ranging from 6% to 184% and exceedance probabilities ranging from 0% to 24%. These results contradict the stationarity scenario, highlighting the importance of nonstationary flood analysis at both local and regional levels, considering the associated uncertainties.

COMMITTEE MEMBERS:
Interno - 2764698 - CARLOS HENRIQUE RIBEIRO LIMA
Presidente - 1809020 - DIRCEU SILVEIRA REIS JUNIOR
Externo à Instituição - PEDRO LUIZ BORGES CHAFFE - UFSC
Interno - 404772 - SERGIO KOIDE