The Variation of Design Floods: A Case Study of the Castro Alves Hydropower Plant

Código

I-EBHE0118

Autores

João Marcos Carvalho, Ailton João Gonçalves Moreira, Júlia Mella Massing, ANAXIMANDRO STECKLING MULLER, Diego David Baptista de Souza

Tema

WG 1.10: Hydrologic Design - Solutions & Communication

Resumo

Given the recent series of floods and the disaster that they caused in the state of Rio Grande do Sul (south of Brazil), in in 2023 and 2024, discussions concerning return periods, maximum floods, and climate change has gained prominence in the national context. The central topic of discussion resides on two main questions: (i) how climate change is and will be affecting the design floods of hydraulic infrastructure, and (ii) whether methods currently adopted have become outdated, as they have not kept up with the technological and scientific developments that occurred after the 2000s. For large dams, the main guidelines developed by ELETROBRAS in 1987 are still adopted. The method consists in using frequency analysis (FA) based on Gumbel or Exponential distributions, to extrapolate the maximum annual flood with a return period equal or greater than 10,000 years. During this process, usually, there is no consideration regarding nonstationary time series or FA confidence intervals. However, given FA?s statistical and qualitative nature, and the effects of nonstationary, the results for extrapolated floods can greatly diverge with every new sample and/or year introduced to the initial analysis, resulting in confusion and debate on which value to safely apply. Faced with the difficulties in defining design floods for large structures, and the extreme situation that occurred in Rio Grande do Sul in 2024, we present a methodology to better characterize and understand the changes in the 10,000-year return period flood, using data from the HPP Castro Alves as a case study. The HPP was selected based on its location in Antas?s River (Rio Grande do Sul), and both floods that occurred in September 2023 and May 2024 were estimated to have values close to the maximum designed discharge of the HPP (9,011 m³/s). To carry out the study and evaluate the uncertainty and difficulty in characterizing the flood, FA was conducted every five years. During the analysis, data?s stationarity was verified - and corrected if necessary - and through Monte Carlo and Markov Chain Monte Carlo (MCMC) simulations, the 95% and 5% confidence intervals were estimated. With this strategy, we are able to observe not just a single flood value, but its development over time, how well the confidence intervals are respected, and at which points the design flood had more significant variability. Although the interpretation of the results may be more complex than the classical method from ELETROBRAS (1987), through this analysis we can better incorporate the variability present in the FA, creating a more suitable way to understand the designed flood for large structures. Regarding the HPP Castro Alves, we demonstrated the efficiency of the spillway?s original discharge capacity compared to the changes over time in the designed flood.